<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.1 20151215//EN" "http://jats.nlm.nih.gov/publishing/1.1/JATS-journalpublishing1.dtd">
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" article-type="review-article" dtd-version="1.1" specific-use="sps-1.9" xml:lang="pt">
  <front>
    <journal-meta>
      <journal-id journal-id-type="nlm-ta">Arq Bras Cardiol</journal-id>
      <journal-id journal-id-type="publisher-id">abc</journal-id>
      <journal-title-group>
        <journal-title>Arquivos Brasileiros de Cardiologia</journal-title>
        <abbrev-journal-title abbrev-type="publisher">Arq. Bras. Cardiol.</abbrev-journal-title>
      </journal-title-group>
      <issn pub-type="ppub">0066-782X</issn>
      <issn pub-type="epub">1678-4170</issn>
      <publisher>
        <publisher-name>Sociedade Brasileira de Cardiologia - SBC</publisher-name>
      </publisher>
    </journal-meta>
    <article-meta>
      <article-id specific-use="scielo-v3" pub-id-type="publisher-id">WMgVngCLbYfJrkmC65VFCkp</article-id>
      <article-id specific-use="scielo-v2" pub-id-type="publisher-id">S0066-782X2022000100095</article-id>
      <article-id pub-id-type="publisher-id">abc.20200596</article-id>
      <article-id pub-id-type="doi">10.36660/abc.20200596</article-id>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Artigo de Revis&#227;o</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Machine Learning na Medicina: Revis&#227;o e Aplicabilidade</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0003-1349-1745</contrib-id>
          <name>
            <surname>Paix&#227;o</surname>
            <given-names>Gabriela Miana de Mattos</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
          <xref ref-type="corresp" rid="c1"/>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0003-4606-6454</contrib-id>
          <name>
            <surname>Santos</surname>
            <given-names>Bruno Campos</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-6747-5444</contrib-id>
          <name>
            <surname>Araujo</surname>
            <given-names>Rodrigo Martins de</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Ribeiro</surname>
            <given-names>Manoel Horta</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-8510-0013</contrib-id>
          <name>
            <surname>Moraes</surname>
            <given-names>Jermana Lopes de</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
          <xref ref-type="aff" rid="aff2">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-2740-0042</contrib-id>
          <name>
            <surname>Ribeiro</surname>
            <given-names>Antonio L.</given-names>
          </name>
          <xref ref-type="aff" rid="aff1">
            <sup>1</sup>
          </xref>
        </contrib>
        <aff id="aff1">
          <label>1</label>
          <institution content-type="orgname">Universidade Federal de Minas Gerais</institution>
          <addr-line>
            <named-content content-type="city">Belo Horizonte</named-content>
            <named-content content-type="state">MG</named-content>
          </addr-line>
          <country country="BR">Brasil</country>
          <institution content-type="original">Universidade Federal de Minas Gerais, Belo Horizonte, MG &#8211; Brasil</institution>
        </aff>
        <aff id="aff2">
          <label>2</label>
          <institution content-type="orgname">Universidade Federal do Cear&#225;</institution>
          <addr-line>
            <named-content content-type="city">Sobral</named-content>
            <named-content content-type="state">CE</named-content>
          </addr-line>
          <country country="BR">Brasil</country>
          <institution content-type="original">Universidade Federal do Cear&#225;, Sobral, CE &#8211; Brasil</institution>
        </aff>
      </contrib-group>
      <author-notes>
        <corresp id="c1"><bold>Correspond&#234;ncia: Gabriela Miana de Mattos Paix&#227;o</bold> &#8226; Universidade Federal de Minas Gerais &#8211; Avenida Alfredo Balena, 110. CEP <postal-code>31270&#8211;901</postal-code>, Belo Horizonte, MG &#8211; Brasil. E-mail: <email>gabimiana@gmail.com</email> </corresp>
        <fn fn-type="con">
          <p>
            <bold>Contribui&#231;&#227;o dos autores</bold>
          </p>
          <p>Concep&#231;&#227;o e desenho da pesquisa: Paix&#227;o GMM, Ribeiro MH, Moares JL, Ribeiro AL; Obten&#231;&#227;o de dados: Paix&#227;o GMM, Santos BC, Araujo RM, Moares JL; An&#225;lise e interpreta&#231;&#227;o dos dados: Paix&#227;o GMM, Moares JL; An&#225;lise estat&#237;stica: Paix&#227;o GMM; Reda&#231;&#227;o do manuscrito: Paix&#227;o GMM, Santos BC, Araujo RM, Ribeiro MH, Moares JL; Revis&#227;o cr&#237;tica do manuscrito quanto ao conte&#250;do intelectual importante: Paix&#227;o GMM, Moares JL, Ribeiro AL.</p>
        </fn>
        <fn fn-type="conflict">
          <p>
            <bold>Potencial conflito de interesse</bold>
          </p>
          <p>N&#227;o h&#225; conflito com o presente artigo</p>
        </fn>
      </author-notes>
      <pub-date date-type="pub" publication-format="electronic">
        <day>21</day>
        <month>02</month>
        <year>2022</year>
      </pub-date>
      <pub-date date-type="collection" publication-format="electronic">
        <month>01</month>
        <year>2022</year>
      </pub-date>
      <volume>118</volume>
      <issue>1</issue>
      <fpage>95</fpage>
      <lpage>102</lpage>
      <history>
        <date date-type="received">
          <day>02</day>
          <month>09</month>
          <year>2019</year>
        </date>
        <date date-type="rev-recd">
          <day>23</day>
          <month>09</month>
          <year>2020</year>
        </date>
        <date date-type="accepted">
          <day>02</day>
          <month>12</month>
          <year>2020</year>
        </date>
      </history>
      <permissions>
        <license license-type="open-access" xlink:href="https://creativecommons.org/licenses/by/4.0/" xml:lang="en">
          <license-p>This is an open-access article distributed under the terms of the Creative Commons Attribution License</license-p>
        </license>
      </permissions>
      <kwd-group xml:lang="pt">
        <title>Palavras-chave</title>
        <kwd>Aprendizado de M&#225;quina</kwd>
        <kwd>Medicina</kwd>
        <kwd>Cardiologia</kwd>
      </kwd-group>
      <funding-group>
        <award-group>
          <funding-source>CNPq</funding-source>
          <award-id>310679/2016&#8211;8</award-id>
          <award-id>465518/2014&#8211;1</award-id>
        </award-group>
        <award-group>
          <funding-source>FAPEMIG</funding-source>
          <award-id>PPM&#8211;00428&#8211;17</award-id>
          <award-id>RED&#8211;00081&#8211;16</award-id>
        </award-group>
        <award-group>
          <funding-source>CNPq</funding-source>
          <award-id>141286/2021&#8211;0</award-id>
        </award-group>
        <funding-statement>O autor Ribeiro AL &#233; parcialmente apoiado pelo CNPq (310679/2016&#8211;8 e 465518/2014&#8211;1) e pela FAPEMIG (PPM&#8211;00428&#8211;17 e RED&#8211;00081&#8211;16). A autora Moraes JL &#233; apoiada pelo CNPq (141286/2021&#8211;0).</funding-statement>
      </funding-group>
      <counts>
        <fig-count count="6"/>
        <table-count count="4"/>
        <equation-count count="0"/>
        <ref-count count="64"/>
        <page-count count="8"/>
      </counts>
    </article-meta>
  </front>
  <body>
    <sec sec-type="intro">
      <title>Introdu&#231;&#227;o</title>
      <p>O aprendizado de m&#225;quina, ou <italic>machine learning</italic> (ML), &#233; um ramo da intelig&#234;ncia artificial (IA) que explora o estudo e a constru&#231;&#227;o de algoritmos computacionais a partir do aprendizado por dados,<sup><xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref></sup> ao inv&#233;s de instru&#231;&#245;es pr&#233;&#8211;programadas.<sup><xref ref-type="bibr" rid="B3">3</xref></sup> O objetivo principal de um modelo de ML &#233; construir um sistema de computador que aprenda com um banco de dados pr&#233;&#8211;definido e gere, ao final, um modelo de predi&#231;&#227;o, classifica&#231;&#227;o ou detec&#231;&#227;o.</p>
      <p>A aplica&#231;&#227;o de ML na pr&#225;tica &#233; voltada principalmente para o manuseio de bases de dados consolidadas com informa&#231;&#245;es heterog&#234;neas, para as quais h&#225; uma limita&#231;&#227;o do uso das t&#233;cnicas de estat&#237;stica convencionais.<sup><xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref></sup> Os algoritmos de ML j&#225; est&#227;o difundidos em diversas &#225;reas, como sistemas banc&#225;rios para detec&#231;&#227;o de fraudes, mecanismos de busca na internet, sistemas de vigil&#226;ncia em v&#237;deo, seguran&#231;a de dados, log&#237;stica de empresas, rob&#243;tica e, na medicina, para diagn&#243;stico e progn&#243;stico.<sup><xref ref-type="bibr" rid="B6">6</xref></sup> Com a digitaliza&#231;&#227;o dos prontu&#225;rios m&#233;dicos, exames laboratoriais e de imagem, houve um crescimento dos bancos de dados. Esses s&#227;o fontes para a aplica&#231;&#227;o de t&#233;cnicas de ML, visando a preven&#231;&#227;o, diagn&#243;stico precoce e o tratamento das doen&#231;as.</p>
      <p>Este artigo de revis&#227;o aborda uma introdu&#231;&#227;o sobre ML dividida em: defini&#231;&#227;o, modelos de aprendizagem e uma revis&#227;o sistem&#225;tica de artigos sobre a sua aplicabilidade na medicina e, principalmente, na cardiologia. O objetivo &#233; apresentar ML para m&#233;dicos e profissionais de sa&#250;de como uma ferramenta de aux&#237;lio para a pr&#225;tica cl&#237;nica.</p>
      <p>Para a estrutura&#231;&#227;o deste artigo de revis&#227;o foram pesquisadas duas bases de dados: PubMed (NCBI) e Medline, os seguintes descritores na l&#237;ngua inglesa: &#8220;<italic>machine learning</italic>&#8221;, &#8220;<italic>artificial intelligence</italic>&#8221;, &#8220;<italic>unsupervised learning</italic>&#8221;, &#8220;<italic>supervised learning</italic>&#8221;, &#8220;<italic>neural networks</italic>&#8221; e &#8220;<italic>cardiology</italic>&#8221;. Foram inclu&#237;dos: estudos prospectivos e retrospectivos, exclu&#237;dos: casos cl&#237;nicos e resumos apresentados em congressos (n&#227;o publicados sob a forma de artigo). A elegibilidade de cada estudo foi avaliada por dois investigadores. As opini&#245;es divergentes relativamente &#224; relev&#226;ncia dos artigos foram abordadas por consenso entre os autores.</p>
      <sec>
        <title>Machine learning</title>
        <p>O aprendizado de m&#225;quina &#233; um subcampo da ci&#234;ncia da computa&#231;&#227;o que busca uma interse&#231;&#227;o de t&#233;cnicas matem&#225;ticas e estat&#237;sticas com algoritmos computacionais.<sup><xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B7">7</xref></sup> ML utiliza algoritmos com o conceito de IA e &#233; aplicada em determinadas situa&#231;&#245;es em que se busca padr&#245;es em um conjunto de vari&#225;veis com o intuito de prever um resultado espec&#237;fico de interesse.<sup><xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref></sup></p>
        <p>A maioria das t&#233;cnicas convencionais usadas em sistemas computacionais aplicados &#224; medicina empregam o conceito de algoritmos baseados em regras, chamados de &#8220;sistemas especialistas&#8221;. Assim, o desenvolvedor codifica os conhecimentos m&#233;dicos sobre um determinado assunto para esses sistemas, utilizando regras j&#225; conhecidas. J&#225; as t&#233;cnicas de ML manuseiam um grande n&#250;mero de vari&#225;veis, buscando uma variedade de novas combina&#231;&#245;es que possam prever um resultado com confiabilidade, muitas vezes, em uma grande quantidade de dados, tais como <italic>big data</italic>.<sup><xref ref-type="bibr" rid="B7">7</xref></sup></p>
        <p>Em 2001, Doug Laney definiu um modelo de &#8220;3 Vs&#8221; para conceituar o termo <italic>big data</italic>: grande volume, alta velocidade e alta variedade de informa&#231;&#245;es exigem novas t&#233;cnicas de processamento de forma a permitir descobertas e otimizar processos.<sup><xref ref-type="bibr" rid="B10">10</xref></sup> O termo <italic>big data</italic> pode ser tanto um conjunto de dados de tamanho enorme, que nenhuma das ferramentas tradicionais de gerenciamento de dados &#233; capaz de armazen&#225;&#8211;los ou process&#225;&#8211;los com efici&#234;ncia, como tamb&#233;m pode se referir a um tipo de tecnologia (como instala&#231;&#245;es de armazenamento, ferramentas e processos).<sup><xref ref-type="bibr" rid="B11">11</xref></sup></p>
        <p>O processo de desenvolvimento de um algoritmo de ML &#233; dividido em tr&#234;s fases: pr&#233;&#8211;processamento, treinamento e avalia&#231;&#227;o do modelo (<xref ref-type="fig" rid="f1">Figura 1</xref>). A primeira fase consiste em organizar o banco de dados, definir a pergunta de pesquisa e dividir os dados em treinamento e teste. No treinamento, o aprendizado pode ocorrer de forma supervisionada ou n&#227;o supervisionada.<sup><xref ref-type="bibr" rid="B12">12</xref>&#8211;<xref ref-type="bibr" rid="B15">15</xref></sup> O aprendizado supervisionado &#233; baseado no treinamento de uma amostra de dados com a classifica&#231;&#227;o correta j&#225; atribu&#237;da, enquanto o n&#227;o supervisionado se refere &#224; capacidade de aprender e organizar informa&#231;&#245;es sem a atribui&#231;&#227;o da classifica&#231;&#227;o correta.<sup><xref ref-type="bibr" rid="B14">14</xref></sup> Na fase de avalia&#231;&#227;o, o modelo &#233; comparado com os dados de teste e os resultados s&#227;o gerados. Portanto, os algoritmos de ML aprendem atrav&#233;s de repetidas observa&#231;&#245;es e estabelecem um padr&#227;o de mapeamento com o intuito de rotular os dados e criar um modelo que generaliza as informa&#231;&#245;es, de modo que novos dados (jamais analisados pelo algoritmo) possam ser rotulados com precis&#227;o e confiabilidade.<sup><xref ref-type="bibr" rid="B15">15</xref></sup></p>
        <fig id="f1">
          <label>Figura 1</label>
          <caption>
            <title><italic>Fases para o desenvolvimento de algoritmos de machine learning</italic>.<sup><xref ref-type="bibr" rid="B15"><italic>15</italic></xref></sup></title>
          </caption>
          <graphic xlink:href="1678-4170-abc-118-01-0095-00596-gf01.jpg"/></fig>
        <p>&#201; importante salientar que o processo de desenvolvimento de um algoritmo de ML deve ser realizado com uma base de dados consolidada e validada, pois modelos de ML desenvolvidos com dados n&#227;o consolidados podem gerar resultados enganosos.<sup><xref ref-type="bibr" rid="B5">5</xref></sup></p>
      </sec>
      <sec>
        <title><italic>Machine learning</italic> supervisionada e n&#227;o supervisionada</title>
        <p>A principal diferen&#231;a entre os modelos de aprendizagem supervisionado e n&#227;o supervisionado est&#225; no algoritmo de treinamento. No aprendizado n&#227;o supervisionado, o modelo de ML extrai as caracter&#237;sticas dos dados e constr&#243;i uma representa&#231;&#227;o sem o conhecimento pr&#233;vio dos r&#243;tulos de cada dado, ou seja, identifica o padr&#227;o das informa&#231;&#245;es de classe heuristicamente. Essa falta de supervis&#227;o para o algoritmo pode ser vantajosa, pois permite que o algoritmo analise os padr&#245;es que n&#227;o foram considerados anteriormente.<sup><xref ref-type="bibr" rid="B12">12</xref>&#8211;<xref ref-type="bibr" rid="B14">14</xref></sup></p>
        <p>No aprendizado supervisionado, o modelo do ML tem o conhecimento do r&#243;tulo dos dados, ou seja, as amostras est&#227;o corretamente classificadas. O treinamento &#233; baseado na compara&#231;&#227;o entre o resultado obtido do modelo e o r&#243;tulo previamente classificado. Esse processo &#233; repetido at&#233; se obter um erro m&#237;nimo.<sup><xref ref-type="bibr" rid="B14">14</xref></sup></p>
        <p>A <xref ref-type="table" rid="t1">Tabela 1</xref> resume as principais caracter&#237;sticas de cada tipo de modelo de aprendizado, bem como suas vantagens e desvantagens e aplicabilidade pr&#225;tica.</p>
        <table-wrap id="t1">
          <label>Tabela 1</label>
          <caption>
            <title>Compara&#231;&#227;o entre processo de aprendizagem supervisionado e n&#227;o supervisionado</title>
          </caption>
          <table frame="hsides" rules="rows">
            <colgroup width="33%">
              <col/>
              <col/>
              <col/>
            </colgroup>
            <thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
              <tr>
                <th align="left" valign="middle"/>
                <th align="center" valign="middle">Aprendizado supervisionado</th>
                <th align="center" valign="middle">Aprendizado n&#227;o supervisionado</th>
              </tr>
            </thead>
            <tbody style="border-bottom: thin solid; border-color: #000000">
              <tr>
                <td align="left" valign="middle">Defini&#231;&#227;o</td>
                <td align="center" valign="middle">Algoritmos que aprendem rela&#231;&#245;es entre atributos de entrada e de sa&#237;da a partir de conjunto de exemplos rotulados</td>
                <td align="center" valign="middle">Algoritmos que buscam encontrar padr&#245;es em agrupamentos de dados com caracter&#237;sticas semelhantes, em busca de categorias e desfechos ainda n&#227;o identificados ou n&#227;o informados</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Vantagens</td>
                <td align="center" valign="middle">An&#225;lise de m&#250;ltiplos par&#226;metros. Solu&#231;&#227;o r&#225;pida e autom&#225;tica para quest&#245;es de grande escala e elevada acur&#225;cia</td>
                <td align="center" valign="middle">Menor interfer&#234;ncia humana na an&#225;lise dos dados. Excelente excelente para fontes de dados multimodais ou multidimensionais. Permite identifica&#231;&#227;o de novos desfechos</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Desvantagens</td>
                <td align="center" valign="middle">Necessidade dos dados serem rotulados, o que para grandes volumes de dados pode ser impratic&#225;vel. Tend&#234;ncia ao sobreajuste dos dados</td>
                <td align="center" valign="middle">Custo elevado e t&#233;cnicas complexas. Necessita grande quantidade de dados para elabora&#231;&#227;o do algoritmo. Interpreta&#231;&#227;o dos resultados pode ser desafiadora</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Principais tarefas</td>
                <td align="center" valign="middle">Regress&#227;o, classifica&#231;&#227;o, modelo progn&#243;stico e an&#225;lise de sobreviv&#234;ncia</td>
                <td align="center" valign="middle">Redu&#231;&#227;o da dimensionalidade do problema e agrupamento</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Exemplos de algoritmos</td>
                <td align="center" valign="middle">Regress&#227;o log&#237;stica, &#225;rvores de decis&#227;o, <italic>random forests</italic> e redes neurais artificiais</td>
                <td align="center" valign="middle">An&#225;lise das componentes principais, agrupamento hier&#225;rquico, <italic>autoencoders</italic>, an&#225;lise linear de discriminantes</td>
              </tr>
            </tbody>
          </table>
        </table-wrap>
      </sec>
      <sec>
        <title>T&#233;cnicas de <italic>machine learning</italic></title>
        <p>Diversas t&#233;cnicas de ML t&#234;m sido aplicadas como forma de sistemas de diagn&#243;stico auxiliado por computador, tais como: redes neurais artificiais (RNAs), regress&#227;o log&#237;stica, &#225;rvore de decis&#227;o, <italic>random forests</italic>, rede bayesiana, <italic>deep learning, support vector machine</italic> (SVM), entre outros.<sup><xref ref-type="bibr" rid="B16">16</xref>&#8211;<xref ref-type="bibr" rid="B21">21</xref></sup> Algumas t&#233;cnicas utilizam modelos matem&#225;ticos por meio dos dados para aprendizagem e/ou organiza&#231;&#227;o das informa&#231;&#245;es.<sup><xref ref-type="bibr" rid="B12">12</xref></sup> Outras utilizam representa&#231;&#245;es matem&#225;ticas com alto grau de abstra&#231;&#227;o (modelos matem&#225;ticos complexos). Neste caso, n&#227;o &#233; poss&#237;vel decifrar ou interpretar os m&#233;todos utilizados para obten&#231;&#227;o dos resultados de predi&#231;&#227;o, detec&#231;&#227;o ou classifica&#231;&#227;o, de modo que tais modelos de ML s&#227;o chamados de &#8220;caixa preta&#8221;.<sup><xref ref-type="bibr" rid="B22">22</xref></sup></p>
        <p>Uma RNA &#233; um modelo computacional e matem&#225;tico desenvolvido para funcionar como o c&#233;rebro humano. Uma RNA possui diversos elementos de interconex&#245;es (camada de preditores, camada oculta e camada de resultados). A rela&#231;&#227;o entre essas camadas &#233; inspirada nas conex&#245;es sin&#225;pticas entre os neur&#244;nios (<xref ref-type="fig" rid="f2">Figura 2</xref>).<sup><xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B23">23</xref></sup></p>
        <fig id="f2">
          <label>Figura 2</label>
          <caption>
            <title><italic>Estrutura do funcionamento de uma rede neural artificial</italic>.<sup><xref ref-type="bibr" rid="B19"><italic>19</italic></xref></sup></title>
          </caption>
          <graphic xlink:href="1678-4170-abc-118-01-0095-00596-gf02.jpg"/></fig>
        <p>Uma RNA &#8220;aprende&#8221; atrav&#233;s dessas conex&#245;es entre as camadas (preditores, oculta e resultados) e os pesos associados a cada camada. Sendo assim, um dado de entrada &#233; apresentado na camada de preditores, sendo esse enviado camada a camada. O processamento matem&#225;tico ocorre no envio de dados de uma camada a outra e os pesos dessas conex&#245;es s&#227;o atualizados de acordo com o erro da camada de resultados, ou seja, a rela&#231;&#227;o do resultado esperado e o resultado obtido. Esse processo &#233; repetido at&#233; o valor do erro ser m&#237;nimo ou um valor especificado de intera&#231;&#245;es.<sup><xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref></sup></p>
        <p><italic>Deep learning</italic> difere o seu aprendizado das t&#233;cnicas mais tradicionais de ML, pois processa modelos computacionais mais robustos e com m&#250;ltiplas camadas de processamento baseadas em RNAs. Sendo assim, a t&#233;cnica de <italic>deep learning</italic> funciona semelhante a uma RNA, mas possuindo um n&#250;mero maior de camadas ocultas e, consequentemente mais conex&#245;es sin&#225;pticas. Cada camada reproduz uma representa&#231;&#227;o dos dados oriundos da camada anterior e seu algoritmo de aprendizado pode ser tanto supervisionado como n&#227;o supervisionado.<sup><xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref></sup></p>
        <p>Com o grande volume e a complexidade dos dados que envolvem trabalhar com <italic>big data</italic>, o algoritmo do <italic>autoencoder</italic> &#233; um tipo de RNA que reduz a dimensionalidade dos dados. Para isso, esse algoritmo utiliza modelos matem&#225;ticos com alto grau de abstra&#231;&#227;o para gerar um novo conjunto de dados reduzidos em dimensionalidade com representa&#231;&#227;o o mais pr&#243;xima poss&#237;vel dos dados de entrada. A diferen&#231;a fundamental entre a RNA e o <italic>autoencoder</italic> &#233; que o &#250;ltimo utiliza em sua fase de treinamento dados n&#227;o&#8211;rotulados.<sup><xref ref-type="bibr" rid="B27">27</xref></sup></p>
        <p>O algoritmo da &#225;rvore de decis&#227;o &#233; o mais utilizado quando o conjunto de dados &#233; relativamente pequeno e &#233; desenvolvido com uma s&#233;rie de perguntas de sim/n&#227;o para classificar os dados em categorias. Esse algoritmo utiliza um modelo estat&#237;stico para classifica&#231;&#227;o ou predi&#231;&#227;o de dados. Cada pergunta se divide em poss&#237;veis resultados e esses se ramificam em outras possibilidades; isso se repete at&#233; um desfecho final.<sup><xref ref-type="bibr" rid="B16">16</xref></sup> As principais vantagens deste algoritmo s&#227;o sua simplicidade e interpreta&#231;&#227;o intuitiva.<sup><xref ref-type="bibr" rid="B28">28</xref></sup></p>
        <p><italic>Random forests</italic> s&#227;o uma amplia&#231;&#227;o do algoritmo da &#225;rvore de decis&#227;o, sendo bastante utilizado para resolu&#231;&#227;o de problemas de classifica&#231;&#227;o e regress&#227;o. As &#225;rvores de decis&#227;o s&#227;o combinadas e cada uma &#233; treinada independentemente. Suas principais caracter&#237;sticas s&#227;o: teoria simples, velocidade r&#225;pida na an&#225;lise dos dados, estabilidade com a presen&#231;a de excesso de ru&#237;do e mecanismo de compensa&#231;&#227;o autom&#225;tica em amostras tendenciosas dos dados.<sup><xref ref-type="bibr" rid="B29">29</xref></sup></p>
        <p>A rede bayesiana &#233; outra t&#233;cnica muito aplicada &#224; medicina. M&#233;todos estat&#237;sticos bayesianos com uma fundamenta&#231;&#227;o te&#243;rica que cren&#231;as subjetivas coerentes a especialistas de uma determinada &#225;rea podem ser expressas em uma estrutura probabil&#237;stica.<sup><xref ref-type="bibr" rid="B17">17</xref></sup></p>
        <p>O SVM &#233; um m&#233;todo de ML com aprendizado supervisionado, amplamente utilizado em bioinform&#225;tica. Este algoritmo utiliza a ideia de minimiza&#231;&#227;o do erro e trabalha com teoria estat&#237;stica do aprendizado e da otimiza&#231;&#227;o. Al&#233;m da classifica&#231;&#227;o bin&#225;ria, o SVM pode ser usado na regress&#227;o de dados cont&#237;nuos, chamado de regress&#227;o do vetor de suporte. Os resultados obtidos com o uso do SVM s&#227;o compar&#225;veis aos de RNAs, apresentando processo de treinamento f&#225;cil e trabalhando com alta dimensionalidade de dados. Portanto, esse encontra um compromisso entre menor complexidade e erro.<sup><xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref></sup></p>
        <p>Dessa forma, cada algoritmo utiliza t&#233;cnicas distintas de como aprender com observa&#231;&#245;es e como realizar um mapeamento do conjunto de preditores para o resultado final. Esse deve generalizar as informa&#231;&#245;es, de modo que uma tarefa possa ser executada corretamente com entradas novas, n&#227;o analisadas anteriormente pelo modelo.<sup><xref ref-type="bibr" rid="B14">14</xref></sup></p>
      </sec>
      <sec>
        <title><italic>Machine learning</italic> na medicina</title>
        <p>Desde o s&#233;culo passado, os pesquisadores exploram as diversas aplica&#231;&#245;es das t&#233;cnicas de ML em todos os campos da medicina.<sup><xref ref-type="bibr" rid="B32">32</xref></sup> A pesquisa m&#233;dica envolvendo ML tem crescido exponencialmente ao longo das &#250;ltimas d&#233;cadas. Os dados do PubMed (NCBI) e Medline, envolvendo os descritores &#8220;<italic>machine learning</italic>&#8221;, &#8220;<italic>artificial intelligence</italic>&#8221;, &#8220;<italic>unsupervised learning</italic>&#8221;, &#8220;<italic>supervised learning</italic>&#8221; e &#8220;<italic>neural networks</italic>&#8221;, revelou 113.127 artigos publicados entre 1951 e 2019 (<xref ref-type="fig" rid="f3">Figura 3</xref>). Ao acrescentar&#8211;se o descritor &#8220;<italic>cardiology</italic>&#8221; como condi&#231;&#227;o obrigat&#243;ria na pesquisa dos demais termos, 888 trabalhos retornam com distribui&#231;&#227;o semelhante &#224; anterior, entre os anos de 1986 e 2019.</p>
        <fig id="f3">
          <label>Figura 3</label>
          <caption>
            <title><italic>N&#250;mero de artigos por ano e acumulado durante o per&#237;odo de 1951 a 2019 no PubMed e Medline</italic>.</title>
          </caption>
          <graphic xlink:href="1678-4170-abc-118-01-0095-00596-gf03.jpg"/></fig>
        <p>A capacidade dos algoritmos de ML de reconhecer padr&#245;es e predizer diagn&#243;stico tem sido amplamente aplicada &#224;s diversas &#225;reas de aten&#231;&#227;o &#224; sa&#250;de.<sup><xref ref-type="bibr" rid="B33">33</xref>&#8211;<xref ref-type="bibr" rid="B36">36</xref></sup> Na dermatologia, uma RNA foi capaz de diferenciar les&#245;es dermatol&#243;gicas em benignas versus malignas, a partir de mais de 129.000 casos, com resultados similares a um comit&#234; de 21 dermatologistas.<sup><xref ref-type="bibr" rid="B35">35</xref></sup> No campo da psiquiatria, o estudo com t&#233;cnicas de ML reduziu o n&#250;mero de crit&#233;rios diagn&#243;sticos de 29 para 8 com 100% de acur&#225;cia em 612 pacientes com diagn&#243;stico firmado de transtorno do espectro autista.<sup><xref ref-type="bibr" rid="B36">36</xref></sup></p>
        <p>A adi&#231;&#227;o de tecnologias m&#243;veis, tais como: <italic>smartphone</italic> e <italic>smartwatches</italic>, aplicadas &#224; &#225;rea da sa&#250;de acrescentou mais uma dimens&#227;o ao ML, permitindo a leitura de grandes quantidades de dados pessoais em algoritmos de aprendizado.<sup><xref ref-type="bibr" rid="B37">37</xref></sup> Dentro de sistemas de <italic>feedback</italic>, a tecnologia m&#243;vel consegue ser um dispositivo biom&#233;trico (por exemplo, medir os n&#237;veis de glicose no sangue) com capacidade de direcionamento para interven&#231;&#245;es cl&#237;nicas em tempo real, baseadas em algoritmos que atualizam continuamente as informa&#231;&#245;es pessoais do paciente.<sup><xref ref-type="bibr" rid="B38">38</xref></sup> A tecnologia pode simplificar os processos diagn&#243;sticos e facilitar a pr&#225;tica cl&#237;nica.</p>
      </sec>
      <sec>
        <title><italic>Machine learning</italic> na cardiologia</title>
        <p>O avan&#231;o na capacidade computacional nas &#250;ltimas d&#233;cadas impactou especialmente o campo da detec&#231;&#227;o e predi&#231;&#227;o de doen&#231;as cardiovasculares por meio da interpreta&#231;&#227;o de dados, como: estudos dos prontu&#225;rios m&#233;dicos, exames de imagem, banco de dados biol&#243;gicos e gen&#244;micos e de avalia&#231;&#227;o molecular.<sup><xref ref-type="bibr" rid="B32">32</xref></sup> A cardiologia &#233; uma das &#225;reas de maior impacto na produ&#231;&#227;o cient&#237;fica usando t&#233;cnicas de ML (<xref ref-type="table" rid="t2">Tabela 2</xref>). Desde a predi&#231;&#227;o de eventos cardiovasculares<sup><xref ref-type="bibr" rid="B39">39</xref></sup> &#224; melhoria dos diagn&#243;sticos eletrocardiogr&#225;ficos,<sup><xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref></sup> a IA tem sido ferramenta importante para a pesquisa cientifica.</p>
        <table-wrap id="t2">
          <label>Tabela 2</label>
          <caption>
            <title>Artigos com o uso das t&#233;cnicas <italic>machine learning</italic> na cardiologia</title>
          </caption>
          <table frame="hsides" rules="rows">
            <colgroup width="50%">
              <col/>
              <col/>
            </colgroup>
            <thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
              <tr>
                <th align="left" valign="middle">Artigo</th>
                <th align="center" valign="middle">Principais resultados</th>
              </tr>
            </thead>
            <tbody style="border-bottom: thin solid; border-color: #000000">
              <tr>
                <td align="left" valign="middle">Can machine-learning improve cardiovascular risk prediction using routine clinical data?<sup><xref ref-type="bibr" rid="B38">38</xref></sup></td>
                <td align="center" valign="middle">O algoritmo foi capaz de predizer 4998 de 7404 casos positivos (sensibilidade 67,5%, VPP 18,4%) e 53458 de 75585 casos negativos (especificidade 70,7% e VPN 95,7%), com ganho de 355 pacientes (+7,6%) que desenvolveram doen&#231;as cardiovasculares em rela&#231;&#227;o ao m&#233;todo tradicional.</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Deep neural networks can predict mortality from 12-lead electrocardiogram voltage data<sup><xref ref-type="bibr" rid="B43">43</xref></sup></td>
                <td align="center" valign="middle">Por meio da an&#225;lise isolada do ECG por algoritmo de ML, foi poss&#237;vel predizer mortalidade por todas as causas em um ano com AUC = 0,84 e p &lt; 0,05.</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Phenomapping for the Identification of Hypertensive Patients with the Myocardial Substrate for Heart Failure with Preserved Ejection Fraction<sup><xref ref-type="bibr" rid="B56">56</xref></sup></td>
                <td align="center" valign="middle">Um grupo de 1273 pacientes hipertensos foi avaliado por meio de t&#233;cnicas de ML, utilizando dados cl&#237;nicos, laboratoriais e ecocardiogr&#225;ficos. Foi poss&#237;vel identificar um grupo de pacientes com maior risco de desenvolver insufici&#234;ncia card&#237;aca de fra&#231;&#227;o preservada que, provavelmente, devem ser beneficiar de tratamento cl&#237;nico mais intensivo.</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Cognitive Machine-Learning Algorithm for Cardiac Imaging: A Pilot Study for Differentiating Constrictive Pericarditis From Restrictive Cardiomyopathy<sup><xref ref-type="bibr" rid="B57">57</xref></sup></td>
                <td align="center" valign="middle">Utilizaram t&#233;cnicas de ML para diferenciar pericardite constritiva de cardiomiopatia restritiva com uma curva ROC de 96,2% e acur&#225;cia superior a 90%.</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Structured learning algorithm for detection of nonobstructive and obstructive coronary plaque lesions from computed tomography angiography<sup><xref ref-type="bibr" rid="B58">58</xref></sup></td>
                <td align="center" valign="middle">O algoritmo de ML foi capaz de detectar les&#245;es coronarianas superiores ou iguais a 25% com uma sensibilidade 93%, especificidade 95% e acur&#225;cia de 94% em 42 angiografias coron&#225;rias.</td>
              </tr>
              <tr>
                <td align="left" valign="middle">A deep neural network learning algorithm outperforms a conventional algorithm for emergency department electrocardiogram interpretation<sup><xref ref-type="bibr" rid="B54">54</xref></sup></td>
                <td align="center" valign="middle">A an&#225;lise autom&#225;tica pelo m&#233;todo de ML para a leitura de ECG em um departamento de emerg&#234;ncia obteve sensibilidade (88,7% versus 92,0%, p &lt; 0,086), especificidade (94% versus 84,7%, p &lt; 0,0001), VPP (88,2% versus 75,4%, p &lt; 0,0001) e acur&#225;cia (92,2% versus 87,2%, p &lt; 0,0001) em rela&#231;&#227;o ao m&#233;todo autom&#225;tico convencional.</td>
              </tr>
              <tr>
                <td align="left" valign="middle">Automatic Diagnosis of the Short-Duration 12-Lead ECG using a Deep Neural Network: the CODE Study<sup><xref ref-type="bibr" rid="B53">53</xref></sup></td>
                <td align="center" valign="middle">Uma rede neural treinada foi capaz de detectar 6 classes de anormalidades eletrocardiogr&#225;ficas com especificidade superior a 99% e performance superior a 80%, comparada com residentes de cardiologia do &#250;ltimo ano.</td>
              </tr>
              <tr>
                <td align="left" valign="middle">An artificial intelligence-enabled ECG algorithm for the identification of patients with atrial fibrillation during sinus rhythm: a retrospective analysis of outcome prediction<sup><xref ref-type="bibr" rid="B55">55</xref></sup></td>
                <td align="center" valign="middle">Um software de ML foi capaz de detectar pacientes portadores de fibrila&#231;&#227;o atrial, a partir de ECG em ritmo sinusal com uma sensibilidade de 79%, especificidade 79,5% e acur&#225;cia de 79,4%.</td>
              </tr>
            </tbody>
          </table>
          <table-wrap-foot>
            <fn id="TFN1">
              <p><italic>ECG: eletrocardiograma; ML: machine learning; VPN: valor preditivo negativo; VPP: valor preditivo positivo</italic>.</p>
            </fn>
          </table-wrap-foot>
        </table-wrap>
      </sec>
      <sec>
        <title>Progn&#243;stico</title>
        <p>Diversos escores de risco cardiovasculares foram desenvolvidos no intuito de predizer eventos cardiovasculares e identificar os indiv&#237;duos com maior risco card&#237;aco para a preven&#231;&#227;o prim&#225;ria.<sup><xref ref-type="bibr" rid="B42">42</xref></sup> No entanto, a despeito de todo o avan&#231;o proped&#234;utico e terap&#234;utico na cardiologia, ainda h&#225; uma popula&#231;&#227;o em risco n&#227;o identificada pelos m&#233;todos tradicionais.<sup><xref ref-type="bibr" rid="B43">43</xref></sup> O reconhecimento de potenciais fatores de risco n&#227;o tradicionais &#233; desej&#225;vel e o uso de novas tecnologias, como a IA, torna&#8211;se m&#233;todo promissor nessa busca.</p>
        <p>A predi&#231;&#227;o de mortalidade por todas as causas no per&#237;odo de um ano, a partir da an&#225;lise isolada do eletrocardiograma (ECG), apresentou resultados promissores (AUROC 0,87; p &lt; 0,05).<sup><xref ref-type="bibr" rid="B44">44</xref></sup> &#201; interessante ressaltar que uma an&#225;lise cega destes ECG feita por tr&#234;s cardiologistas sugere que os padr&#245;es encontrados para predizer mortalidade pelo ML n&#227;o s&#227;o aparentemente vis&#237;veis pela avalia&#231;&#227;o m&#233;dica convencional.<sup><xref ref-type="bibr" rid="B44">44</xref></sup></p>
        <p>Em estudo com 2619 pacientes submetidos &#224; tomografia computadorizada com emiss&#227;o de pr&#243;tons para a predi&#231;&#227;o de risco cardiovascular, as t&#233;cnicas de ML apresentaram melhores resultados (AUROC 0,81; p &lt; 0.01) do que a an&#225;lise isolada do exame.<sup><xref ref-type="bibr" rid="B45">45</xref></sup></p>
        <p>Estudo com mais de 380.000 pacientes do Reino Unido avaliou o uso de t&#233;cnicas de ML na predi&#231;&#227;o do risco de eventos cardiovasculares em compara&#231;&#227;o com os algoritmos tradicionais propostos pelo American College of Cardiology e pela American Heart Association<sup><xref ref-type="bibr" rid="B39">39</xref></sup>. Houve melhoria de at&#233; 7,6% na predi&#231;&#227;o de eventos com uso de RNA. Algumas vari&#225;veis cl&#237;nicas que n&#227;o s&#227;o valorizadas para doen&#231;a cardiovascular pelos m&#233;todos tradicionais como depress&#227;o e uso de corticoides foram importantes para o risco cardiovascular avaliado pelas t&#233;cnicas de ML.<sup><xref ref-type="bibr" rid="B39">39</xref></sup> Este achado foi corroborado por estudo multic&#234;ntrico estadunidense em que os par&#226;metros encontrados para predi&#231;&#227;o de risco cardiovascular diferem daqueles inclu&#237;dos nas calculadoras de risco tradicionais.<sup><xref ref-type="bibr" rid="B46">46</xref></sup></p>
        <p>A IA pode contribuir na gera&#231;&#227;o de modelos preditivos mais complexos e espec&#237;ficos para cada indiv&#237;duo,<sup><xref ref-type="bibr" rid="B47">47</xref></sup> com a incorpora&#231;&#227;o dos componentes gen&#244;micos aos escores de risco cardiovascular.<sup><xref ref-type="bibr" rid="B48">48</xref>,<xref ref-type="bibr" rid="B49">49</xref></sup> A associa&#231;&#227;o dos dados cl&#237;nicos, sociais, demogr&#225;ficos e gen&#233;ticos com os exames dispon&#237;veis pode permitir uma avalia&#231;&#227;o mais individualizada, visando &#224; promo&#231;&#227;o de sa&#250;de.<sup><xref ref-type="bibr" rid="B47">47</xref></sup></p>
      </sec>
      <sec>
        <title>Diagn&#243;stico</title>
        <p>Nos exames cardiol&#243;gicos, a necessidade de uma equipe m&#233;dica altamente especializada, a variabilidade de laudos entre os m&#233;dicos, al&#233;m do tempo dispensado aos laudos motivaram o estudo das t&#233;cnicas de ML como ferramenta diagn&#243;stica.<sup><xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B50">50</xref></sup></p>
        <p>Os estudos foram promissores e as modalidades da imagem card&#237;aca como ecocardiografia, tomografia computadorizada e resson&#226;ncia nuclear magn&#233;tica apresentaram boa acur&#225;cia em correlacionar altera&#231;&#245;es estruturais com a etiologia e fisiopatologia de doen&#231;as cardiovasculares.<sup><xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref></sup> Em um estudo com 159 pacientes, o qual utilizou tr&#234;s t&#233;cnicas de ML para auxiliar na diferencia&#231;&#227;o ecocardiogr&#225;fica entre cardiomiopatia hipertr&#243;fica e hipertrofia fisiol&#243;gica de atletas. Os par&#226;metros encontrados, como: a raz&#227;o da velocidade transmitral diast&#243;lica precoce&#8211;tardia (p &lt; 0.01), velocidade diast&#243;lica precoce (e&#8217;) (p &lt; 0.01) e a an&#225;lise de <italic>strain</italic> (p &lt; 0.01), foram melhores em sensibilidade e especificidade do que os tradicionalmente usados.<sup><xref ref-type="bibr" rid="B51">51</xref></sup></p>
        <p>Um algoritmo de ML foi desenvolvido para diferenciar as estenoses coronarianas intermedi&#225;rias pela angiografia com reserva de fluxo fracionada menor que 0,80 versus maior que 0,80, a partir de dados cl&#237;nicos e angiogr&#225;ficos. Os resultados foram satisfat&#243;rios com acur&#225;cia de aproximadamente 80% para predi&#231;&#227;o de reserva de fluxo fracionada menor que 0,8 (AUROC 0,84 a 0,87, IC 95% 0,71 a 0,89). A valida&#231;&#227;o externa do modelo desenvolvido tamb&#233;m apresentou resultados similares em 79 pacientes de dois outros centros (AUROC 0,89, IC 95% 0,83 a 0,95).<sup><xref ref-type="bibr" rid="B53">53</xref></sup></p>
        <p>Em rela&#231;&#227;o &#224; eletrocardiografia, estudos est&#227;o sendo desenvolvidos para melhoria dos diagn&#243;sticos autom&#225;ticos.<sup><xref ref-type="bibr" rid="B41">41</xref></sup> Atrav&#233;s de t&#233;cnicas de ML, nosso grupo foi capaz de identificar seis classes eletrocardiogr&#225;ficas por meio da an&#225;lise do ECG de 12 deriva&#231;&#245;es com boa acur&#225;cia, compar&#225;vel ao desempenho que residentes de cardiologia do &#250;ltimo ano.<sup><xref ref-type="bibr" rid="B54">54</xref></sup> Em pacientes com emerg&#234;ncias cardiovasculares hospitalizados, ML teve uma acur&#225;cia diagn&#243;stica de cerca de 90% para altera&#231;&#245;es maiores ao ECG.<sup><xref ref-type="bibr" rid="B54">54</xref></sup> Al&#233;m disso, estudo recente foi capaz de identificar pacientes portadores de fibrila&#231;&#227;o atrial em ECGs em ritmo sinusal com uma sensibilidade de 79%, especificidade 79,5% e acur&#225;cia de 79,4%.<sup><xref ref-type="bibr" rid="B55">55</xref></sup></p>
      </sec>
      <sec>
        <title>Limites e Desafios</title>
        <p>A utiliza&#231;&#227;o das t&#233;cnicas de ML &#233; crescente, devido ao seu potencial para solucionar problemas nas diversas &#225;reas. Na medicina, os resultados s&#227;o promissores em diversas especialidades com a expectativa de que a IA possa ser ferramenta de aux&#237;lio para a pr&#225;tica cl&#237;nica.<sup><xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B59">59</xref></sup> No entanto, ainda &#233; necess&#225;ria cautela na interpreta&#231;&#227;o e incorpora&#231;&#227;o dos resultados.</p>
        <p>Os algoritmos de ML desenvolvidos devem ser reprodut&#237;veis na popula&#231;&#227;o geral. Estudos com n&#250;mero pequeno de pacientes, em popula&#231;&#245;es espec&#237;ficas ou com vieses de sele&#231;&#227;o n&#227;o permitem a generaliza&#231;&#227;o dos seus achados.<sup><xref ref-type="bibr" rid="B60">60</xref>,<xref ref-type="bibr" rid="B61">61</xref></sup> Ainda que a capta&#231;&#227;o de dados e sua interpreta&#231;&#227;o tenham valor estat&#237;stico consider&#225;vel, os melhores cen&#225;rios ainda s&#227;o incapazes de predizer o desfecho em pessoas diferentes.<sup><xref ref-type="bibr" rid="B62">62</xref></sup></p>
        <p>O erro no processo automatizado pode induzir o profissional a conclus&#245;es incorretas, como demonstrado em estudo com 30 residentes em cl&#237;nica m&#233;dica que reduziram sua acur&#225;cia diagn&#243;stica no laudo de ECG, ap&#243;s a disponibiliza&#231;&#227;o de laudos autom&#225;ticos incorretos.<sup><xref ref-type="bibr" rid="B63">63</xref></sup></p>
        <p>O avan&#231;o da IA na medicina &#233; visto com receio por alguns m&#233;dicos. A posi&#231;&#227;o alarmista de que ML possa substituir a figura do m&#233;dico na aten&#231;&#227;o &#224; sa&#250;de tem se mostrado injustific&#225;vel. Nenhum <italic>software</italic>, at&#233; o momento, foi capaz de substituir o aspecto subjetivo da experi&#234;ncia cl&#237;nica na tomada de decis&#245;es favor&#225;veis ao paciente, exatamente, pela medicina n&#227;o ser uma ci&#234;ncia exata.<sup><xref ref-type="bibr" rid="B64">64</xref></sup> A nega&#231;&#227;o ao avan&#231;o tecnol&#243;gico e &#224;s ferramentas de IA, hoje dispon&#237;veis, tem potencial t&#227;o danoso quanto a sua total depend&#234;ncia no atendimento ao paciente. A combina&#231;&#227;o entre ML e o julgamento cl&#237;nico tem apresentado melhores resultados em conjunto do que o seu uso isolado.<sup><xref ref-type="bibr" rid="B59">59</xref></sup></p>
      </sec>
    </sec>
    <sec sec-type="conclusions">
      <title>Conclus&#227;o</title>
      <p>O uso de t&#233;cnicas de ML na medicina deixou o campo te&#243;rico e se tornou uma realidade. Embora o uso do ML em medicina ainda esteja em desenvolvimento, estudos mostram a sua aplicabilidade cl&#237;nica com impacto na avalia&#231;&#227;o diagn&#243;stica e progn&#243;stica.</p>
    </sec>
  </body>
  <back>
    <fn-group>
      <fn fn-type="financial-disclosure" id="fn1">
        <p>
          <bold>Fontes de financiamento</bold>
        </p>
        <p>O autor Ribeiro AL &#233; parcialmente apoiado pelo CNPq (310679/2016&#8211;8 e 465518/2014&#8211;1) e pela FAPEMIG (PPM&#8211;00428&#8211;17 e RED&#8211;00081&#8211;16). A autora Moraes JL &#233; apoiada pelo CNPq (141286/2021&#8211;0).</p>
      </fn>
      <fn fn-type="other" id="fn2">
        <p>
          <bold>Vincula&#231;&#227;o acad&#234;mica</bold>
        </p>
        <p>Este artigo &#233; parte de disserta&#231;&#227;o de mestrado de Gabriela Miana de Mattos Paix&#227;o pela Universidade Federal de Minas Gerais.</p>
      </fn>
    </fn-group>
    <ref-list>
      <title>Refer&#234;ncias</title>
      <ref id="B1">
        <label>1</label>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Mitchell</surname>
              <given-names>TM</given-names>
            </name>
          </person-group>
          <source>The Discipline of Machine Learning</source>
          <publisher-loc>Pittsburgh</publisher-loc>
          <publisher-name>Mach Learning Department</publisher-name>
          <year>2006</year>
        </element-citation>
        <mixed-citation>1 Mitchell TM, The Discipline of Machine Learning. Pittsburgh: Mach Learning Department; 2006.</mixed-citation>
      </ref>
      <ref id="B2">
        <label>2</label>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Hastie</surname>
              <given-names>T</given-names>
            </name>
            <name>
              <surname>Tibshirani</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Friedman</surname>
              <given-names>J</given-names>
            </name>
          </person-group>
          <source>The Elements of Statistical Learning: Data Mining, Inference, and Prediction</source>
          <edition>2nd ed.</edition>
          <publisher-loc>Berlin</publisher-loc>
          <publisher-name>Springer Science &amp; Business Media</publisher-name>
          <year>2009</year>
        </element-citation>
        <mixed-citation>2 Hastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning: Data Mining, Inference, and Prediction. 2nd ed. Berlin: Springer Science &amp; Business Media; 2009.</mixed-citation>
      </ref>
      <ref id="B3">
        <label>3</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Deo</surname>
              <given-names>RC</given-names>
            </name>
          </person-group>
          <article-title>Machine Learning in Medicine</article-title>
          <source>Circulation</source>
          <year>2015</year>
          <volume>132</volume>
          <issue>20</issue>
          <fpage>1920</fpage>
          <lpage>1930</lpage>
          <pub-id pub-id-type="doi">10.1161/CIRCULATIONAHA.115.001593</pub-id>
        </element-citation>
        <mixed-citation>3 Deo RC. Machine Learning in Medicine. Circulation. 2015;132(20):1920-30. doi: 10.1161/CIRCULATIONAHA.115.001593.</mixed-citation>
      </ref>
      <ref id="B4">
        <label>4</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Jordan</surname>
              <given-names>MI</given-names>
            </name>
            <name>
              <surname>Mitchell</surname>
              <given-names>TM</given-names>
            </name>
          </person-group>
          <article-title>Machine Learning: Trends, Perspectives, and Prospects</article-title>
          <source>Science</source>
          <year>2015</year>
          <volume>349</volume>
          <issue>6245</issue>
          <fpage>255</fpage>
          <lpage>260</lpage>
          <pub-id pub-id-type="doi">10.1126/science.aaa8415</pub-id>
        </element-citation>
        <mixed-citation>4 Jordan MI, Mitchell TM. Machine Learning: Trends, Perspectives, and Prospects. Science. 2015;349(6245):255-60. doi: 10.1126/science.aaa8415.</mixed-citation>
      </ref>
      <ref id="B5">
        <label>5</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Chen</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Mao</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Liu</surname>
              <given-names>Y</given-names>
            </name>
          </person-group>
          <article-title>Big data: A survey</article-title>
          <source>Mob. Netw. Appl</source>
          <year>2014</year>
          <volume>19</volume>
          <issue>2</issue>
          <fpage>171</fpage>
          <lpage>209</lpage>
          <pub-id pub-id-type="doi">10.1007/s11036&#8211;013&#8211;0489&#8211;0</pub-id>
        </element-citation>
        <mixed-citation>5 Chen M, Mao S, Liu Y. Big data: A survey. Mob. Netw. Appl. 2014;19(2):171-209. doi:10.1007/s11036-013-0489-0.</mixed-citation>
      </ref>
      <ref id="B6">
        <label>6</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Zhou</surname>
              <given-names>L</given-names>
            </name>
            <name>
              <surname>Pan</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Vasilakos</surname>
              <given-names>AV</given-names>
            </name>
          </person-group>
          <article-title>Machine Learning on Big Data: Opportunities and Challenges</article-title>
          <source>Neurocomputing</source>
          <year>2017</year>
          <volume>237</volume>
          <fpage>350</fpage>
          <lpage>361</lpage>
          <pub-id pub-id-type="doi">10.1016/j.neucom.2017.01.026</pub-id>
        </element-citation>
        <mixed-citation>6 Zhou L, Pan S, Wang J, Vasilakos AV. Machine Learning on Big Data: Opportunities and Challenges. Neurocomputing. 2017;237:350-61. doi: 10.1016/j.neucom.2017.01.026.</mixed-citation>
      </ref>
      <ref id="B7">
        <label>7</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Obermeyer</surname>
              <given-names>Z</given-names>
            </name>
            <name>
              <surname>Emanuel</surname>
              <given-names>EJ</given-names>
            </name>
          </person-group>
          <article-title>Predicting the Future &#8211; Big Data, Machine Learning, and Clinical Medicine</article-title>
          <source>N Engl J Med</source>
          <year>2016</year>
          <volume>375</volume>
          <issue>13</issue>
          <fpage>1216</fpage>
          <lpage>1219</lpage>
          <pub-id pub-id-type="doi">10.1056/NEJMp1606181</pub-id>
        </element-citation>
        <mixed-citation>7 Obermeyer Z, Emanuel EJ. Predicting the Future - Big Data, Machine Learning, and Clinical Medicine. N Engl J Med. 2016;375(13):1216-9. doi: 10.1056/NEJMp1606181.</mixed-citation>
      </ref>
      <ref id="B8">
        <label>8</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Waljee</surname>
              <given-names>AK</given-names>
            </name>
            <name>
              <surname>Higgins</surname>
              <given-names>PD</given-names>
            </name>
          </person-group>
          <article-title>Machine Learning in Medicine: A Primer for Physicians</article-title>
          <source>Am J Gastroenterol</source>
          <year>2010</year>
          <volume>105</volume>
          <issue>6</issue>
          <fpage>1224</fpage>
          <lpage>1226</lpage>
          <pub-id pub-id-type="doi">10.1038/ajg.2010.173</pub-id>
        </element-citation>
        <mixed-citation>8 Waljee AK, Higgins PD. Machine Learning in Medicine: A Primer for Physicians. Am J Gastroenterol. 2010;105(6):1224-6. doi: 10.1038/ajg.2010.173.</mixed-citation>
      </ref>
      <ref id="B9">
        <label>9</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Darcy</surname>
              <given-names>AM</given-names>
            </name>
            <name>
              <surname>Louie</surname>
              <given-names>AK</given-names>
            </name>
            <name>
              <surname>Roberts</surname>
              <given-names>LW</given-names>
            </name>
          </person-group>
          <article-title>Machine Learning and the Profession of Medicine</article-title>
          <source>JAMA</source>
          <year>2016</year>
          <volume>315</volume>
          <issue>6</issue>
          <fpage>551</fpage>
          <lpage>552</lpage>
          <pub-id pub-id-type="doi">10.1001/jama.2015.18421</pub-id>
        </element-citation>
        <mixed-citation>9 Darcy AM, Louie AK, Roberts LW. Machine Learning and the Profession of Medicine. JAMA. 2016;315(6):551-2. doi: 10.1001/jama.2015.18421.</mixed-citation>
      </ref>
      <ref id="B10">
        <label>10</label>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Laney</surname>
              <given-names>D</given-names>
            </name>
          </person-group>
          <source>3D Data Management: Controlling Data Volume, Velocity, and Variety</source>
          <publisher-loc>Milan</publisher-loc>
          <publisher-name>META Group Research Note</publisher-name>
          <year>2001</year>
        </element-citation>
        <mixed-citation>10 Laney D. 3D Data Management: Controlling Data Volume, Velocity, and Variety. Milan: META Group Research Note; 2001.</mixed-citation>
      </ref>
      <ref id="B11">
        <label>11</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Martin&#8211;Sanchez</surname>
              <given-names>F</given-names>
            </name>
            <name>
              <surname>Verspoor</surname>
              <given-names>K</given-names>
            </name>
          </person-group>
          <article-title>Big Data in Medicine is Driving Big Changes</article-title>
          <source>Yearb Med Inform</source>
          <year>2014</year>
          <volume>9</volume>
          <issue>1</issue>
          <fpage>14</fpage>
          <lpage>20</lpage>
          <pub-id pub-id-type="doi">10.15265/IY&#8211;2014&#8211;0020</pub-id>
        </element-citation>
        <mixed-citation>11 Martin-Sanchez F, Verspoor K. Big Data in Medicine is Driving Big Changes. Yearb Med Inform. 2014;9(1):14-20. doi: 10.15265/IY-2014-0020.</mixed-citation>
      </ref>
      <ref id="B12">
        <label>12</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Barreto</surname>
              <given-names>GA</given-names>
            </name>
            <name>
              <surname>Souza</surname>
              <given-names>LG</given-names>
            </name>
          </person-group>
          <article-title>Adaptive Filtering with the Self&#8211;Organizing Map: A Performance Comparison</article-title>
          <source>Neural Netw</source>
          <year>2006</year>
          <volume>19</volume>
          <issue>6&#8211;7</issue>
          <fpage>785</fpage>
          <lpage>798</lpage>
          <pub-id pub-id-type="doi">10.1016/j.neunet.2006.05.005</pub-id>
        </element-citation>
        <mixed-citation>12 Barreto GA, Souza LG. Adaptive Filtering with the Self-Organizing Map: A Performance Comparison. Neural Netw. 2006;19(6-7):785-98. doi: 10.1016/j.neunet.2006.05.005.</mixed-citation>
      </ref>
      <ref id="B13">
        <label>13</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Kohonen</surname>
              <given-names>T</given-names>
            </name>
            <name>
              <surname>Honkela</surname>
              <given-names>T</given-names>
            </name>
          </person-group>
          <article-title>Kohonen Network</article-title>
          <source>Scholarpedia</source>
          <year>2007</year>
          <volume>2</volume>
          <issue>1</issue>
          <fpage>1568</fpage>
          <lpage>1568</lpage>
          <pub-id pub-id-type="doi">10.4249/scholarpedia.1568</pub-id>
        </element-citation>
        <mixed-citation>13 Kohonen T, Honkela T. Kohonen Network. Scholarpedia. 2007;2(1):1568. doi: 10.4249/scholarpedia.1568.</mixed-citation>
      </ref>
      <ref id="B14">
        <label>14</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Sathya</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Abraham</surname>
              <given-names>A</given-names>
            </name>
          </person-group>
          <article-title>Comparison of Supervised and Unsupervised Learning Algorithms for Pattern Classification</article-title>
          <source>Int. J. Artif. Intell</source>
          <year>2013</year>
          <volume>2</volume>
          <issue>2</issue>
          <fpage>34</fpage>
          <lpage>38</lpage>
          <pub-id pub-id-type="doi">10.14569/IJARAI.2013.020206</pub-id>
        </element-citation>
        <mixed-citation>14 Sathya R, Abraham A. Comparison of Supervised and Unsupervised Learning Algorithms for Pattern Classification. Int. J. Artif. Intell. 2013;2(2):34-8. doi: 10.14569/IJARAI.2013.020206.</mixed-citation>
      </ref>
      <ref id="B15">
        <label>15</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Rajkomar</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Dean</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Kohane</surname>
              <given-names>I</given-names>
            </name>
          </person-group>
          <article-title>Machine Learning in Medicine</article-title>
          <source>N Engl J Med</source>
          <year>2019</year>
          <volume>380</volume>
          <issue>14</issue>
          <fpage>1347</fpage>
          <lpage>1358</lpage>
          <pub-id pub-id-type="doi">10.1056/NEJMra1814259</pub-id>
        </element-citation>
        <mixed-citation>15 Rajkomar A, Dean J, Kohane I. Machine Learning in Medicine. N Engl J Med. 2019;380(14):1347-58. doi: 10.1056/NEJMra1814259.</mixed-citation>
      </ref>
      <ref id="B16">
        <label>16</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Podgorelec</surname>
              <given-names>V</given-names>
            </name>
            <name>
              <surname>Kokol</surname>
              <given-names>P</given-names>
            </name>
            <name>
              <surname>Stiglic</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Rozman</surname>
              <given-names>I</given-names>
            </name>
          </person-group>
          <article-title>Decision Trees: An Overview and Their Use in Medicine</article-title>
          <source>J Med Syst</source>
          <year>2002</year>
          <volume>26</volume>
          <issue>5</issue>
          <fpage>445</fpage>
          <lpage>463</lpage>
          <pub-id pub-id-type="doi">10.1023/a:1016409317640</pub-id>
        </element-citation>
        <mixed-citation>16 Podgorelec V, Kokol P, Stiglic B, Rozman I. Decision Trees: An Overview and Their Use in Medicine. J Med Syst. 2002;26(5):445-63. doi: 10.1023/a:1016409317640.</mixed-citation>
      </ref>
      <ref id="B17">
        <label>17</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Pang</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Zhang</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Li</surname>
              <given-names>N</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>K</given-names>
            </name>
          </person-group>
          <article-title>Computerized Tongue Diagnosis Based on Bayesian Networks</article-title>
          <source>IEEE Trans Biomed Eng</source>
          <year>2004</year>
          <volume>51</volume>
          <issue>10</issue>
          <fpage>1803</fpage>
          <lpage>1810</lpage>
          <pub-id pub-id-type="doi">10.1109/TBME.2004.831534</pub-id>
        </element-citation>
        <mixed-citation>17 Pang B, Zhang D, Li N, Wang K. Computerized Tongue Diagnosis Based on Bayesian Networks. IEEE Trans Biomed Eng. 2004;51(10):1803-10. doi: 10.1109/TBME.2004.831534.</mixed-citation>
      </ref>
      <ref id="B18">
        <label>18</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Lisboa</surname>
              <given-names>PJ</given-names>
            </name>
            <name>
              <surname>Taktak</surname>
              <given-names>AF</given-names>
            </name>
          </person-group>
          <article-title>The Use of Artificial Neural Networks in Decision Support in Cancer: A Systematic Review</article-title>
          <source>Neural Netw</source>
          <year>2006</year>
          <volume>19</volume>
          <issue>4</issue>
          <fpage>408</fpage>
          <lpage>415</lpage>
          <pub-id pub-id-type="doi">10.1016/j.neunet.2005.10.007</pub-id>
        </element-citation>
        <mixed-citation>18 Lisboa PJ, Taktak AF. The Use of Artificial Neural Networks in Decision Support in Cancer: A Systematic Review. Neural Netw. 2006;19(4):408-15. doi: 10.1016/j.neunet.2005.10.007.</mixed-citation>
      </ref>
      <ref id="B19">
        <label>19</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ramesh</surname>
              <given-names>AN</given-names>
            </name>
            <name>
              <surname>Kambhampati</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Monson</surname>
              <given-names>JR</given-names>
            </name>
            <name>
              <surname>Drew</surname>
              <given-names>PJ</given-names>
            </name>
          </person-group>
          <article-title>Artificial Intelligence in Medicine</article-title>
          <source>Ann R Coll Surg Engl</source>
          <year>2004</year>
          <volume>86</volume>
          <issue>5</issue>
          <fpage>334</fpage>
          <lpage>338</lpage>
          <pub-id pub-id-type="doi">10.1308/147870804290</pub-id>
        </element-citation>
        <mixed-citation>19 Ramesh AN, Kambhampati C, Monson JR, Drew PJ. Artificial Intelligence in Medicine. Ann R Coll Surg Engl. 2004;86(5):334-8. doi: 10.1308/147870804290.</mixed-citation>
      </ref>
      <ref id="B20">
        <label>20</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Mavroforakis</surname>
              <given-names>ME</given-names>
            </name>
            <name>
              <surname>Theodoridis</surname>
              <given-names>S</given-names>
            </name>
          </person-group>
          <article-title>A Geometric Approach to Support Vector Machine (SVM) Classification</article-title>
          <source>IEEE Trans Neural Netw</source>
          <year>2006</year>
          <volume>17</volume>
          <issue>3</issue>
          <fpage>671</fpage>
          <lpage>682</lpage>
          <pub-id pub-id-type="doi">10.1109/TNN.2006.873281</pub-id>
        </element-citation>
        <mixed-citation>20 Mavroforakis ME, Theodoridis S. A Geometric Approach to Support Vector Machine (SVM) Classification. IEEE Trans Neural Netw. 2006;17(3):671-82. doi: 10.1109/TNN.2006.873281.</mixed-citation>
      </ref>
      <ref id="B21">
        <label>21</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Smith</surname>
              <given-names>SW</given-names>
            </name>
            <name>
              <surname>Walsh</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Grauer</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Rapin</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Li</surname>
              <given-names>J</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>A Deep Neural Network Learning Algorithm Outperforms a Conventional Algorithm for Emergency Department Electrocardiogram Interpretation</article-title>
          <source>J Electrocardiol</source>
          <year>2019</year>
          <volume>52</volume>
          <fpage>88</fpage>
          <lpage>95</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jelectrocard.2018.11.013</pub-id>
        </element-citation>
        <mixed-citation>21 Smith SW, Walsh B, Grauer K, Wang K, Rapin J, Li J, et al. A Deep Neural Network Learning Algorithm Outperforms a Conventional Algorithm for Emergency Department Electrocardiogram Interpretation. J Electrocardiol. 2019;52:88-95. doi: 10.1016/j.jelectrocard.2018.11.013.</mixed-citation>
      </ref>
      <ref id="B22">
        <label>22</label>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Bianchi</surname>
              <given-names>RE</given-names>
            </name>
          </person-group>
          <source>Extra&#231;&#227;o de Conhecimento Simb&#243;lico em T&#233;cnicas de Aprendizado de M&#225;quina Caixa-Preta por Similaridade de Rankings [dissertation]</source>
          <publisher-loc>S&#227;o Paulo</publisher-loc>
          <publisher-name>Universidade de S&#227;o Paulo</publisher-name>
          <year>2008</year>
        </element-citation>
        <mixed-citation>22 Bianchi RE. Extra&#231;&#227;o de Conhecimento Simb&#243;lico em T&#233;cnicas de Aprendizado de M&#225;quina Caixa-Preta por Similaridade de Rankings [dissertation]. S&#227;o Paulo: Universidade de S&#227;o Paulo; 2008.</mixed-citation>
      </ref>
      <ref id="B23">
        <label>23</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Al&#8211;Shayea</surname>
              <given-names>QK</given-names>
            </name>
          </person-group>
          <article-title>Artificial Neural Networks in Medical Diagnosis</article-title>
          <source>Int. J. Comput. Sci. Issues</source>
          <year>2011</year>
          <volume>8</volume>
          <issue>2</issue>
          <fpage>150</fpage>
          <lpage>154</lpage>
        </element-citation>
        <mixed-citation>23 Al-Shayea QK. Artificial Neural Networks in Medical Diagnosis. Int. J. Comput. Sci. Issues. 2011;8(2):150-4.</mixed-citation>
      </ref>
      <ref id="B24">
        <label>24</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Bengio</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Courville</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Vincent</surname>
              <given-names>P</given-names>
            </name>
          </person-group>
          <article-title>Representation Learning: A Review and New Perspectives</article-title>
          <source>IEEE Trans Pattern Anal Mach Intell</source>
          <year>2013</year>
          <volume>35</volume>
          <issue>8</issue>
          <fpage>1798</fpage>
          <lpage>1828</lpage>
          <pub-id pub-id-type="doi">10.1109/TPAMI.2013.50</pub-id>
        </element-citation>
        <mixed-citation>24 Bengio Y, Courville A, Vincent P. Representation Learning: A Review and New Perspectives. IEEE Trans Pattern Anal Mach Intell. 2013;35(8):1798-828. doi: 10.1109/TPAMI.2013.50.</mixed-citation>
      </ref>
      <ref id="B25">
        <label>25</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Miotto</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>F</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Jiang</surname>
              <given-names>X</given-names>
            </name>
            <name>
              <surname>Dudley</surname>
              <given-names>JT</given-names>
            </name>
          </person-group>
          <article-title>Deep Learning for Healthcare: Review, Opportunities and Challenges</article-title>
          <source>Brief Bioinform</source>
          <year>2018</year>
          <volume>19</volume>
          <issue>6</issue>
          <fpage>1236</fpage>
          <lpage>1246</lpage>
          <pub-id pub-id-type="doi">10.1093/bib/bbx044</pub-id>
        </element-citation>
        <mixed-citation>25 Miotto R, Wang F, Wang S, Jiang X, Dudley JT. Deep Learning for Healthcare: Review, Opportunities and Challenges. Brief Bioinform. 2018;19(6):1236-46. doi: 10.1093/bib/bbx044.</mixed-citation>
      </ref>
      <ref id="B26">
        <label>26</label>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Bengio</surname>
              <given-names>Y</given-names>
            </name>
          </person-group>
          <source>Learning Deep Architectures for AI</source>
          <publisher-loc>Pittsburgh</publisher-loc>
          <publisher-name>Mach Learn</publisher-name>
          <year>2009</year>
        </element-citation>
        <mixed-citation>26 Bengio Y. Learning Deep Architectures for AI. Pittsburgh: Mach Learn; 2009.</mixed-citation>
      </ref>
      <ref id="B27">
        <label>27</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Raghavendra</surname>
              <given-names>U</given-names>
            </name>
            <name>
              <surname>Gudigar</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Bhandary</surname>
              <given-names>SV</given-names>
            </name>
            <name>
              <surname>Rao</surname>
              <given-names>TN</given-names>
            </name>
            <name>
              <surname>Ciaccio</surname>
              <given-names>EJ</given-names>
            </name>
            <name>
              <surname>Acharya</surname>
              <given-names>UR</given-names>
            </name>
          </person-group>
          <article-title>A Two Layer Sparse Autoencoder for Glaucoma Identification with Fundus Images</article-title>
          <source>J Med Syst</source>
          <year>2019</year>
          <volume>43</volume>
          <issue>9</issue>
          <fpage>299</fpage>
          <lpage>299</lpage>
          <pub-id pub-id-type="doi">10.1007/s10916&#8211;019&#8211;1427&#8211;x</pub-id>
        </element-citation>
        <mixed-citation>27 Raghavendra U, Gudigar A, Bhandary SV, Rao TN, Ciaccio EJ, Acharya UR. A Two Layer Sparse Autoencoder for Glaucoma Identification with Fundus Images. J Med Syst. 2019;43(9):299. doi: 10.1007/s10916-019-1427-x.</mixed-citation>
      </ref>
      <ref id="B28">
        <label>28</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Goodman</surname>
              <given-names>KE</given-names>
            </name>
            <name>
              <surname>Lessler</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Cosgrove</surname>
              <given-names>SE</given-names>
            </name>
            <name>
              <surname>Harris</surname>
              <given-names>AD</given-names>
            </name>
            <name>
              <surname>Lautenbach</surname>
              <given-names>E</given-names>
            </name>
            <name>
              <surname>Han</surname>
              <given-names>JH</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>A Clinical Decision Tree to Predict Whether a Bacteremic Patient Is Infected with an Extended&#8211;Spectrum &#946;&#8211;Lactamase&#8211;Producing Organism</article-title>
          <source>Clin Infect Dis</source>
          <year>2016</year>
          <volume>63</volume>
          <issue>7</issue>
          <fpage>896</fpage>
          <lpage>903</lpage>
          <pub-id pub-id-type="doi">10.1093/cid/ciw425</pub-id>
        </element-citation>
        <mixed-citation>28 Goodman KE, Lessler J, Cosgrove SE, Harris AD, Lautenbach E, Han JH, et al. A Clinical Decision Tree to Predict Whether a Bacteremic Patient Is Infected with an Extended-Spectrum &#946;-Lactamase-Producing Organism. Clin Infect Dis. 2016;63(7):896-903. doi: 10.1093/cid/ciw425.</mixed-citation>
      </ref>
      <ref id="B29">
        <label>29</label>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Segal</surname>
              <given-names>MR</given-names>
            </name>
          </person-group>
          <source>Machine Learning Benchmarks and Random Forest Regression</source>
          <publisher-loc>S&#227;o Francisco</publisher-loc>
          <publisher-name>Biostatistics</publisher-name>
          <year>2004</year>
        </element-citation>
        <mixed-citation>29 Segal MR. Machine Learning Benchmarks and Random Forest Regression. S&#227;o Francisco: Biostatistics; 2004.</mixed-citation>
      </ref>
      <ref id="B30">
        <label>30</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Chen</surname>
              <given-names>KC</given-names>
            </name>
            <name>
              <surname>Chen</surname>
              <given-names>CYC</given-names>
            </name>
          </person-group>
          <article-title>Stroke Prevention by Traditional Chinese Medicine? A Genetic Algorithm, Support Vector Machine and Molecular Dynamics Approach</article-title>
          <source>Soft Matter</source>
          <year>2011</year>
          <volume>7</volume>
          <issue>8</issue>
          <fpage>4001</fpage>
          <lpage>4008</lpage>
          <pub-id pub-id-type="doi">10.1039/c0sm01548b</pub-id>
        </element-citation>
        <mixed-citation>30 Chen KC, Chen CYC. Stroke Prevention by Traditional Chinese Medicine? A Genetic Algorithm, Support Vector Machine and Molecular Dynamics Approach. Soft Matter. 2011. 7(8):4001-8. doi: 10.1039/c0sm01548b.</mixed-citation>
      </ref>
      <ref id="B31">
        <label>31</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Krittanawong</surname>
              <given-names>C</given-names>
            </name>
            <name>
              <surname>Zhang</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>Z</given-names>
            </name>
            <name>
              <surname>Aydar</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Kitai</surname>
              <given-names>T</given-names>
            </name>
          </person-group>
          <article-title>Artificial Intelligence in Precision Cardiovascular Medicine</article-title>
          <source>J Am Coll Cardiol</source>
          <year>2017</year>
          <volume>69</volume>
          <issue>21</issue>
          <fpage>2657</fpage>
          <lpage>2664</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jacc.2017.03.571</pub-id>
        </element-citation>
        <mixed-citation>31 Krittanawong C, Zhang H, Wang Z, Aydar M, Kitai T. Artificial Intelligence in Precision Cardiovascular Medicine. J Am Coll Cardiol. 2017;69(21):2657-64. doi: 10.1016/j.jacc.2017.03.571.</mixed-citation>
      </ref>
      <ref id="B32">
        <label>32</label>
        <element-citation publication-type="book">
          <person-group person-group-type="author">
            <name>
              <surname>Fan</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Shen</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Davatzikos</surname>
              <given-names>C</given-names>
            </name>
          </person-group>
          <chapter-title>Detecting Cognitive States from fMRI Images by Machine Learning and Multivariate Classification</chapter-title>
          <source>CVPRW&#8217;06 2006: Conference on Computer Vision and Pattern Recognition Workshop</source>
          <year>2006</year>
          <month>06</month>
          <day>17-22</day>
          <publisher-loc>Ney York, USA</publisher-loc>
          <publisher-name>IEEE</publisher-name>
          <comment>2006</comment>
          <fpage>89</fpage>
          <lpage>89</lpage>
        </element-citation>
        <mixed-citation>32 Fan Y, Shen D, Davatzikos C. Detecting Cognitive States from fMRI Images by Machine Learning and Multivariate Classification. CVPRW&#8217;06 2006: Conference on Computer Vision and Pattern Recognition Workshop; 2006 Jun 17-22; Ney York, USA: IEEE; 2006. p. 89.</mixed-citation>
      </ref>
      <ref id="B33">
        <label>33</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Gulshan</surname>
              <given-names>V</given-names>
            </name>
            <name>
              <surname>Peng</surname>
              <given-names>L</given-names>
            </name>
            <name>
              <surname>Coram</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Stumpe</surname>
              <given-names>MC</given-names>
            </name>
            <name>
              <surname>Wu</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Narayanaswamy</surname>
              <given-names>A</given-names>
            </name>
          </person-group>
          <article-title>Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs</article-title>
          <source>JAMA</source>
          <year>2016</year>
          <volume>316</volume>
          <issue>22</issue>
          <fpage>2402</fpage>
          <lpage>2410</lpage>
          <pub-id pub-id-type="doi">10.1001/jama.2016.17216</pub-id>
        </element-citation>
        <mixed-citation>33 Gulshan V, Peng L, Coram M, Stumpe MC, Wu D, Narayanaswamy A, et al. Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs. JAMA. 2016;316(22):2402-10. doi: 10.1001/jama.2016.17216.</mixed-citation>
      </ref>
      <ref id="B34">
        <label>34</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Esteva</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Kuprel</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Novoa</surname>
              <given-names>RA</given-names>
            </name>
            <name>
              <surname>Ko</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Swetter</surname>
              <given-names>SM</given-names>
            </name>
            <name>
              <surname>Blau</surname>
              <given-names>HM</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Dermatologist&#8211; Level Classification of Skin Cancer with Deep Neural Networks</article-title>
          <source>Nature</source>
          <year>2017</year>
          <volume>542</volume>
          <issue>7639</issue>
          <fpage>115</fpage>
          <lpage>118</lpage>
          <pub-id pub-id-type="doi">10.1038/nature21056</pub-id>
        </element-citation>
        <mixed-citation>34 Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, et al. Dermatologist-Level Classification of Skin Cancer with Deep Neural Networks. Nature. 2017;542(7639):115-8. doi: 10.1038/nature21056.</mixed-citation>
      </ref>
      <ref id="B35">
        <label>35</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Wall</surname>
              <given-names>DP</given-names>
            </name>
            <name>
              <surname>Kosmicki</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Deluca</surname>
              <given-names>TF</given-names>
            </name>
            <name>
              <surname>Harstad</surname>
              <given-names>E</given-names>
            </name>
            <name>
              <surname>Fusaro</surname>
              <given-names>VA</given-names>
            </name>
          </person-group>
          <article-title>Use of Machine Learning to Shorten Observation&#8211;Based Screening and Diagnosis of Autism</article-title>
          <source>Transl Psychiatry</source>
          <year>2012</year>
          <volume>2</volume>
          <issue>4</issue>
          <elocation-id>e100</elocation-id>
          <pub-id pub-id-type="doi">10.1038/tp.2012.10</pub-id>
        </element-citation>
        <mixed-citation>35 Wall DP, Kosmicki J, Deluca TF, Harstad E, Fusaro VA. Use of Machine Learning to Shorten Observation-Based Screening and Diagnosis of Autism. Transl Psychiatry. 2012;2(4):e100. doi: 10.1038/tp.2012.10.</mixed-citation>
      </ref>
      <ref id="B36">
        <label>36</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Chen</surname>
              <given-names>JH</given-names>
            </name>
            <name>
              <surname>Asch</surname>
              <given-names>SM</given-names>
            </name>
          </person-group>
          <article-title>Machine Learning and Prediction in Medicine &#8211; Beyond the Peak of Inflated Expectations</article-title>
          <source>N Engl J Med</source>
          <year>2017</year>
          <volume>376</volume>
          <issue>26</issue>
          <fpage>2507</fpage>
          <lpage>2509</lpage>
          <pub-id pub-id-type="doi">10.1056/NEJMp1702071</pub-id>
        </element-citation>
        <mixed-citation>36 Chen JH, Asch SM. Machine Learning and Prediction in Medicine - Beyond the Peak of Inflated Expectations. N Engl J Med. 2017;376(26):2507-9. doi: 10.1056/NEJMp1702071.</mixed-citation>
      </ref>
      <ref id="B37">
        <label>37</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Bergenstal</surname>
              <given-names>RM</given-names>
            </name>
            <name>
              <surname>Klonoff</surname>
              <given-names>DC</given-names>
            </name>
            <name>
              <surname>Garg</surname>
              <given-names>SK</given-names>
            </name>
            <name>
              <surname>Bode</surname>
              <given-names>BW</given-names>
            </name>
            <name>
              <surname>Meredith</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Slover</surname>
              <given-names>RH</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Threshold&#8211;Based Insulin&#8211;Pump Interruption for Reduction of Hypoglycemia</article-title>
          <source>N Engl J Med</source>
          <year>2013</year>
          <volume>369</volume>
          <issue>3</issue>
          <fpage>224</fpage>
          <lpage>232</lpage>
          <pub-id pub-id-type="doi">10.1056/NEJMoa1303576</pub-id>
        </element-citation>
        <mixed-citation>37 Bergenstal RM, Klonoff DC, Garg SK, Bode BW, Meredith M, Slover RH, et al. Threshold-Based Insulin-Pump Interruption for Reduction of Hypoglycemia. N Engl J Med. 2013;369(3):224-32. doi: 10.1056/NEJMoa1303576.</mixed-citation>
      </ref>
      <ref id="B38">
        <label>38</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Weng</surname>
              <given-names>SF</given-names>
            </name>
            <name>
              <surname>Reps</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Kai</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Garibaldi</surname>
              <given-names>JM</given-names>
            </name>
            <name>
              <surname>Qureshi</surname>
              <given-names>N</given-names>
            </name>
          </person-group>
          <article-title>Can Machine&#8211;Learning Improve Cardiovascular Risk Prediction Using Routine Clinical Data?</article-title>
          <source>PLoS One</source>
          <year>2017</year>
          <volume>12</volume>
          <issue>4</issue>
          <elocation-id>e0174944</elocation-id>
          <pub-id pub-id-type="doi">10.1371/journal.pone.0174944</pub-id>
        </element-citation>
        <mixed-citation>38 Weng SF, Reps J, Kai J, Garibaldi JM, Qureshi N. Can Machine-Learning Improve Cardiovascular Risk Prediction Using Routine Clinical Data? PLoS One. 2017;12(4):e0174944. doi: 10.1371/journal.pone.0174944.</mixed-citation>
      </ref>
      <ref id="B39">
        <label>39</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Slomka</surname>
              <given-names>PJ</given-names>
            </name>
            <name>
              <surname>Dey</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Sitek</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Motwani</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Berman</surname>
              <given-names>DS</given-names>
            </name>
            <name>
              <surname>Germano</surname>
              <given-names>G</given-names>
            </name>
          </person-group>
          <article-title>Cardiac Imaging: Working Towards Fully&#8211;Automated Machine Analysis &amp; Interpretation</article-title>
          <source>Expert Rev Med Devices</source>
          <year>2017</year>
          <volume>14</volume>
          <issue>3</issue>
          <fpage>197</fpage>
          <lpage>212</lpage>
          <pub-id pub-id-type="doi">10.1080/17434440.2017.1300057</pub-id>
        </element-citation>
        <mixed-citation>39 Slomka PJ, Dey D, Sitek A, Motwani M, Berman DS, Germano G. Cardiac Imaging: Working Towards Fully-Automated Machine Analysis &amp; Interpretation. Expert Rev Med Devices. 2017;14(3):197-212. doi: 10.1080/17434440.2017.1300057.</mixed-citation>
      </ref>
      <ref id="B40">
        <label>40</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Minchol&#233;</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Camps</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Lyon</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Rodr&#237;guez</surname>
              <given-names>B</given-names>
            </name>
          </person-group>
          <article-title>Machine Learning in the Electrocardiogram</article-title>
          <source>J Electrocardiol</source>
          <year>2019</year>
          <volume>57S</volume>
          <fpage>61</fpage>
          <lpage>64</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jelectrocard.2019.08.008</pub-id>
        </element-citation>
        <mixed-citation>40 Minchol&#233; A, Camps J, Lyon A, Rodr&#237;guez B. Machine Learning in the Electrocardiogram. J Electrocardiol. 2019;57S:61-4. doi: 10.1016/j.jelectrocard.2019.08.008.</mixed-citation>
      </ref>
      <ref id="B41">
        <label>41</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>D&#8217;Agostino</surname>
              <given-names>RB</given-names>
              <suffix>Sr</suffix>
            </name>
            <name>
              <surname>Pencina</surname>
              <given-names>MJ</given-names>
            </name>
            <name>
              <surname>Massaro</surname>
              <given-names>JM</given-names>
            </name>
            <name>
              <surname>Coady</surname>
              <given-names>S</given-names>
            </name>
          </person-group>
          <article-title>Cardiovascular Disease Risk Assessment: Insights from Framingham</article-title>
          <source>Glob Heart</source>
          <year>2013</year>
          <volume>8</volume>
          <issue>1</issue>
          <fpage>11</fpage>
          <lpage>23</lpage>
          <pub-id pub-id-type="doi">10.1016/j.gheart.2013.01.001</pub-id>
        </element-citation>
        <mixed-citation>41 D&#8217;Agostino RB Sr, Pencina MJ, Massaro JM, Coady S. Cardiovascular Disease Risk Assessment: Insights from Framingham. Glob Heart. 2013;8(1):11-23. doi: 10.1016/j.gheart.2013.01.001.</mixed-citation>
      </ref>
      <ref id="B42">
        <label>42</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Lin</surname>
              <given-names>JS</given-names>
            </name>
            <name>
              <surname>Evans</surname>
              <given-names>CV</given-names>
            </name>
            <name>
              <surname>Johnson</surname>
              <given-names>E</given-names>
            </name>
            <name>
              <surname>Redmond</surname>
              <given-names>N</given-names>
            </name>
            <name>
              <surname>Coppola</surname>
              <given-names>EL</given-names>
            </name>
            <name>
              <surname>Smith</surname>
              <given-names>N</given-names>
            </name>
          </person-group>
          <article-title>Nontraditional Risk Factors in Cardiovascular Disease Risk Assessment: Updated Evidence Report and Systematic Review for the US Preventive Services Task Force</article-title>
          <source>JAMA</source>
          <year>2018</year>
          <volume>320</volume>
          <issue>3</issue>
          <fpage>281</fpage>
          <lpage>297</lpage>
          <pub-id pub-id-type="doi">10.1001/jama.2018.4242</pub-id>
        </element-citation>
        <mixed-citation>42 Lin JS, Evans CV, Johnson E, Redmond N, Coppola EL, Smith N. Nontraditional Risk Factors in Cardiovascular Disease Risk Assessment: Updated Evidence Report and Systematic Review for the US Preventive Services Task Force. JAMA. 2018;320(3):281-97. doi: 10.1001/jama.2018.4242.</mixed-citation>
      </ref>
      <ref id="B43">
        <label>43</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Raghunath</surname>
              <given-names>SM</given-names>
            </name>
            <name>
              <surname>Cerna</surname>
              <given-names>AU</given-names>
            </name>
            <name>
              <surname>Jing</surname>
              <given-names>L</given-names>
            </name>
            <name>
              <surname>vanMaanen</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Stough</surname>
              <given-names>JV</given-names>
            </name>
            <name>
              <surname>Hartzel</surname>
              <given-names>D</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Deep Neural Networks Can Predict 1&#8211;Year Mortality Directly from ECG Signal, Even when Clinically Interpreted as Normal</article-title>
          <source>Circulation</source>
          <year>2019</year>
          <volume>140</volume>
          <supplement>Suppl 1</supplement>
          <comment>A14425</comment>
        </element-citation>
        <mixed-citation>43 Raghunath SM, Cerna AU, Jing L, vanMaanen D, Stough JV, Hartzel D, et al. Deep Neural Networks Can Predict 1-Year Mortality Directly from ECG Signal, Even when Clinically Interpreted as Normal. Circulation. 2019;140(Suppl 1):A14425-.</mixed-citation>
      </ref>
      <ref id="B44">
        <label>44</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Betancur</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Otaki</surname>
              <given-names>Y</given-names>
            </name>
            <name>
              <surname>Motwani</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Fish</surname>
              <given-names>MB</given-names>
            </name>
            <name>
              <surname>Lemley</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Dey</surname>
              <given-names>D</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Prognostic Value of Combined Clinical and Myocardial Perfusion Imaging Data Using Machine Learning</article-title>
          <source>JACC Cardiovasc Imaging</source>
          <year>2018</year>
          <volume>11</volume>
          <issue>7</issue>
          <fpage>1000</fpage>
          <lpage>1009</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jcmg.2017.07.024</pub-id>
        </element-citation>
        <mixed-citation>44 Betancur J, Otaki Y, Motwani M, Fish MB, Lemley M, Dey D, et al. Prognostic Value of Combined Clinical and Myocardial Perfusion Imaging Data Using Machine Learning. JACC Cardiovasc Imaging. 2018;11(7):1000-9. doi: 10.1016/j.jcmg.2017.07.024.</mixed-citation>
      </ref>
      <ref id="B45">
        <label>45</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ambale&#8211;Venkatesh</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Yang</surname>
              <given-names>X</given-names>
            </name>
            <name>
              <surname>Wu</surname>
              <given-names>CO</given-names>
            </name>
            <name>
              <surname>Liu</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Hundley</surname>
              <given-names>WG</given-names>
            </name>
            <name>
              <surname>McClelland</surname>
              <given-names>R</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Cardiovascular Event Prediction by Machine Learning: The Multi&#8211;Ethnic Study of Atherosclerosis</article-title>
          <source>Circ Res</source>
          <year>2017</year>
          <volume>121</volume>
          <issue>9</issue>
          <fpage>1092</fpage>
          <lpage>1101</lpage>
          <pub-id pub-id-type="doi">10.1161/CIRCRESAHA.117.311312</pub-id>
        </element-citation>
        <mixed-citation>45 Ambale-Venkatesh B, Yang X, Wu CO, Liu K, Hundley WG, McClelland R, et al. Cardiovascular Event Prediction by Machine Learning: The Multi-Ethnic Study of Atherosclerosis. Circ Res. 2017;121(9):1092-101. doi: 10.1161/CIRCRESAHA.117.311312.</mixed-citation>
      </ref>
      <ref id="B46">
        <label>46</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Antman</surname>
              <given-names>EM</given-names>
            </name>
            <name>
              <surname>Loscalzo</surname>
              <given-names>J</given-names>
            </name>
          </person-group>
          <article-title>Precision Medicine in Cardiology</article-title>
          <source>Nat Rev Cardiol</source>
          <year>2016</year>
          <volume>13</volume>
          <issue>10</issue>
          <fpage>591</fpage>
          <lpage>602</lpage>
          <pub-id pub-id-type="doi">10.1038/nrcardio.2016.101</pub-id>
        </element-citation>
        <mixed-citation>46 Antman EM, Loscalzo J. Precision Medicine in Cardiology. Nat Rev Cardiol. 2016;13(10):591-602. doi: 10.1038/nrcardio.2016.101.</mixed-citation>
      </ref>
      <ref id="B47">
        <label>47</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Johnson</surname>
              <given-names>KW</given-names>
            </name>
            <name>
              <surname>Shameer</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Glicksberg</surname>
              <given-names>BS</given-names>
            </name>
            <name>
              <surname>Readhead</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Sengupta</surname>
              <given-names>PP</given-names>
            </name>
            <name>
              <surname>Bj&#246;rkegren</surname>
              <given-names>JLM</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Enabling Precision Cardiology Through Multiscale Biology and Systems Medicine</article-title>
          <source>JACC Basic Transl Sci</source>
          <year>2017</year>
          <volume>2</volume>
          <issue>3</issue>
          <fpage>311</fpage>
          <lpage>327</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jacbts.2016.11.010</pub-id>
        </element-citation>
        <mixed-citation>47 Johnson KW, Shameer K, Glicksberg BS, Readhead B, Sengupta PP, Bj&#246;rkegren JLM, et al. Enabling Precision Cardiology Through Multiscale Biology and Systems Medicine. JACC Basic Transl Sci. 2017;2(3):311-27. doi: 10.1016/j.jacbts.2016.11.010.</mixed-citation>
      </ref>
      <ref id="B48">
        <label>48</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Kullo</surname>
              <given-names>IJ</given-names>
            </name>
            <name>
              <surname>Jouni</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Austin</surname>
              <given-names>EE</given-names>
            </name>
            <name>
              <surname>Brown</surname>
              <given-names>SA</given-names>
            </name>
            <name>
              <surname>Kruisselbrink</surname>
              <given-names>TM</given-names>
            </name>
            <name>
              <surname>Isseh</surname>
              <given-names>IN</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Incorporating a Genetic Risk Score into Coronary Heart Disease Risk Estimates: Effect on Low&#8211;Density Lipoprotein Cholesterol Levels (the MI&#8211; GENES Clinical Trial)</article-title>
          <source>Circulation</source>
          <year>2016</year>
          <volume>133</volume>
          <issue>12</issue>
          <fpage>1181</fpage>
          <lpage>1188</lpage>
          <pub-id pub-id-type="doi">10.1161/CIRCULATIONAHA.115.020109</pub-id>
        </element-citation>
        <mixed-citation>48 Kullo IJ, Jouni H, Austin EE, Brown SA, Kruisselbrink TM, Isseh IN, et al. Incorporating a Genetic Risk Score into Coronary Heart Disease Risk Estimates: Effect on Low-Density Lipoprotein Cholesterol Levels (the MI-GENES Clinical Trial). Circulation. 2016;133(12):1181-8. doi: 10.1161/CIRCULATIONAHA.115.020109.</mixed-citation>
      </ref>
      <ref id="B49">
        <label>49</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Johnson</surname>
              <given-names>KW</given-names>
            </name>
            <name>
              <surname>Soto</surname>
              <given-names>JT</given-names>
            </name>
            <name>
              <surname>Glicksberg</surname>
              <given-names>BS</given-names>
            </name>
            <name>
              <surname>Shameer</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Miotto</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Ali</surname>
              <given-names>M</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Artificial Intelligence in Cardiology</article-title>
          <source>J Am Coll Cardiol</source>
          <year>2018</year>
          <volume>71</volume>
          <issue>23</issue>
          <fpage>2668</fpage>
          <lpage>2679</lpage>
        </element-citation>
        <mixed-citation>49 Johnson KW, Soto JT, Glicksberg BS, Shameer K, Miotto R, Ali M, et al. Artificial Intelligence in Cardiology. J Am Coll Cardiol. 2018;71(23):2668-79.</mixed-citation>
      </ref>
      <ref id="B50">
        <label>50</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Narula</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Shameer</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Omar</surname>
              <given-names>AMS</given-names>
            </name>
            <name>
              <surname>Dudley</surname>
              <given-names>JT</given-names>
            </name>
            <name>
              <surname>Sengupta</surname>
              <given-names>PP</given-names>
            </name>
          </person-group>
          <article-title>Machine&#8211;Learning Algorithms to Automate Morphological and Functional Assessments in 2D Echocardiography</article-title>
          <source>J Am Coll Cardiol</source>
          <year>2016</year>
          <volume>68</volume>
          <issue>21</issue>
          <fpage>2287</fpage>
          <lpage>2295</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jacc.2016.08.062</pub-id>
        </element-citation>
        <mixed-citation>50 Narula S, Shameer K, Omar AMS, Dudley JT, Sengupta PP. Machine-Learning Algorithms to Automate Morphological and Functional Assessments in 2D Echocardiography. J Am Coll Cardiol. 2016;68(21):2287-95. doi: 10.1016/j.jacc.2016.08.062.</mixed-citation>
      </ref>
      <ref id="B51">
        <label>51</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Samad</surname>
              <given-names>MD</given-names>
            </name>
            <name>
              <surname>Ulloa</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Wehner</surname>
              <given-names>GJ</given-names>
            </name>
            <name>
              <surname>Jing</surname>
              <given-names>L</given-names>
            </name>
            <name>
              <surname>Hartzel</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Good</surname>
              <given-names>CW</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Predicting Survival From Large Echocardiography and Electronic Health Record Datasets: Optimization With Machine Learning</article-title>
          <source>JACC Cardiovasc Imaging</source>
          <year>2019</year>
          <volume>12</volume>
          <issue>4</issue>
          <fpage>681</fpage>
          <lpage>689</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jcmg.2018.04.026</pub-id>
        </element-citation>
        <mixed-citation>51 Samad MD, Ulloa A, Wehner GJ, Jing L, Hartzel D, Good CW, et al. Predicting Survival From Large Echocardiography and Electronic Health Record Datasets: Optimization With Machine Learning. JACC Cardiovasc Imaging. 2019;12(4):681-9. doi: 10.1016/j.jcmg.2018.04.026.</mixed-citation>
      </ref>
      <ref id="B52">
        <label>52</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Hae</surname>
              <given-names>H</given-names>
            </name>
            <name>
              <surname>Kang</surname>
              <given-names>SJ</given-names>
            </name>
            <name>
              <surname>Kim</surname>
              <given-names>WJ</given-names>
            </name>
            <name>
              <surname>Choi</surname>
              <given-names>SY</given-names>
            </name>
            <name>
              <surname>Lee</surname>
              <given-names>JG</given-names>
            </name>
            <name>
              <surname>Bae</surname>
              <given-names>Y</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Machine Learning Assessment of Myocardial Ischemia Using Angiography: Development and Retrospective Validation</article-title>
          <source>PLoS Med</source>
          <year>2018</year>
          <volume>15</volume>
          <issue>11</issue>
          <elocation-id>e1002693</elocation-id>
          <pub-id pub-id-type="doi">10.1371/journal.pmed.1002693</pub-id>
        </element-citation>
        <mixed-citation>52 Hae H, Kang SJ, Kim WJ, Choi SY, Lee JG, Bae Y, et al. Machine Learning Assessment of Myocardial Ischemia Using Angiography: Development and Retrospective Validation. PLoS Med. 2018;15(11):e1002693. doi: 10.1371/journal.pmed.1002693.</mixed-citation>
      </ref>
      <ref id="B53">
        <label>53</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ribeiro</surname>
              <given-names>AH</given-names>
            </name>
            <name>
              <surname>Ribeiro</surname>
              <given-names>MH</given-names>
            </name>
            <name>
              <surname>Paix&#227;o</surname>
              <given-names>GMM</given-names>
            </name>
            <name>
              <surname>Oliveira</surname>
              <given-names>DM</given-names>
            </name>
            <name>
              <surname>Gomes</surname>
              <given-names>PR</given-names>
            </name>
            <name>
              <surname>Canazart</surname>
              <given-names>JA</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Automatic Diagnosis of the 12&#8211;lead ECG Using a Deep Neural Network</article-title>
          <source>Nat Commun</source>
          <year>2020</year>
          <volume>11</volume>
          <issue>1</issue>
          <fpage>1760</fpage>
          <lpage>1760</lpage>
          <pub-id pub-id-type="doi">10.1038/s41467&#8211;020&#8211;15432&#8211;4</pub-id>
        </element-citation>
        <mixed-citation>53 Ribeiro AH, Ribeiro MH, Paix&#227;o GMM, Oliveira DM, Gomes PR, Canazart JA, et al. Automatic Diagnosis of the 12-lead ECG Using a Deep Neural Network. Nat Commun. 2020;11(1):1760. doi: 10.1038/s41467-020-15432-4.</mixed-citation>
      </ref>
      <ref id="B54">
        <label>54</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Smith</surname>
              <given-names>SW</given-names>
            </name>
            <name>
              <surname>Walsh</surname>
              <given-names>B</given-names>
            </name>
            <name>
              <surname>Grauer</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Wang</surname>
              <given-names>K</given-names>
            </name>
            <name>
              <surname>Rapin</surname>
              <given-names>J</given-names>
            </name>
            <name>
              <surname>Li</surname>
              <given-names>J</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>A Deep Neural Network Learning Algorithm Outperforms a Conventional Algorithm for Emergency Department Electrocardiogram Interpretation</article-title>
          <source>J Electrocardiol</source>
          <year>2019</year>
          <volume>52</volume>
          <fpage>88</fpage>
          <lpage>95</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jelectrocard.2018.11.013</pub-id>
        </element-citation>
        <mixed-citation>54 Smith SW, Walsh B, Grauer K, Wang K, Rapin J, Li J, et al. A Deep Neural Network Learning Algorithm Outperforms a Conventional Algorithm for Emergency Department Electrocardiogram Interpretation. J Electrocardiol. 2019;52:88-95. doi: 10.1016/j.jelectrocard.2018.11.013.</mixed-citation>
      </ref>
      <ref id="B55">
        <label>55</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Attia</surname>
              <given-names>ZI</given-names>
            </name>
            <name>
              <surname>Noseworthy</surname>
              <given-names>PA</given-names>
            </name>
            <name>
              <surname>Lopez&#8211;Jimenez</surname>
              <given-names>F</given-names>
            </name>
            <name>
              <surname>Asirvatham</surname>
              <given-names>SJ</given-names>
            </name>
            <name>
              <surname>Deshmukh</surname>
              <given-names>AJ</given-names>
            </name>
            <name>
              <surname>Gersh</surname>
              <given-names>BJ</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>An artificial Intelligence&#8211;Enabled ECG Algorithm for the Identification of Patients with Atrial Fibrillation During Sinus Rhythm: A Retrospective Analysis of Outcome Prediction</article-title>
          <source>Lancet</source>
          <year>2019</year>
          <volume>394</volume>
          <issue>10201</issue>
          <fpage>861</fpage>
          <lpage>867</lpage>
          <pub-id pub-id-type="doi">10.1016/S0140&#8211;6736(19)31721&#8211;0</pub-id>
        </element-citation>
        <mixed-citation>55 Attia ZI, Noseworthy PA, Lopez-Jimenez F, Asirvatham SJ, Deshmukh AJ, Gersh BJ, et al. An artificial Intelligence-Enabled ECG Algorithm for the Identification of Patients with Atrial Fibrillation During Sinus Rhythm: A Retrospective Analysis of Outcome Prediction. Lancet. 2019;394(10201):861-7. doi: 10.1016/S0140-6736(19)31721-0.</mixed-citation>
      </ref>
      <ref id="B56">
        <label>56</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Katz</surname>
              <given-names>DH</given-names>
            </name>
            <name>
              <surname>Deo</surname>
              <given-names>RC</given-names>
            </name>
            <name>
              <surname>Aguilar</surname>
              <given-names>FG</given-names>
            </name>
            <name>
              <surname>Selvaraj</surname>
              <given-names>S</given-names>
            </name>
            <name>
              <surname>Martinez</surname>
              <given-names>EE</given-names>
            </name>
            <name>
              <surname>Beussink&#8211;Nelson</surname>
              <given-names>L</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Phenomapping for the Identification of Hypertensive Patients with the Myocardial Substrate for Heart Failure with Preserved Ejection Fraction</article-title>
          <source>J Cardiovasc Transl Res</source>
          <year>2017</year>
          <volume>10</volume>
          <issue>3</issue>
          <fpage>275</fpage>
          <lpage>284</lpage>
          <pub-id pub-id-type="doi">10.1007/s12265&#8211;017&#8211;9739&#8211;z</pub-id>
        </element-citation>
        <mixed-citation>56 Katz DH, Deo RC, Aguilar FG, Selvaraj S, Martinez EE, Beussink-Nelson L, et al. Phenomapping for the Identification of Hypertensive Patients with the Myocardial Substrate for Heart Failure with Preserved Ejection Fraction. J Cardiovasc Transl Res. 2017;10(3):275-84. doi: 10.1007/s12265-017-9739-z.</mixed-citation>
      </ref>
      <ref id="B57">
        <label>57</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Sengupta</surname>
              <given-names>PP</given-names>
            </name>
            <name>
              <surname>Huang</surname>
              <given-names>YM</given-names>
            </name>
            <name>
              <surname>Bansal</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Ashrafi</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Fisher</surname>
              <given-names>M</given-names>
            </name>
            <name>
              <surname>Shameer</surname>
              <given-names>K</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Cognitive Machine&#8211;Learning Algorithm for Cardiac Imaging: A Pilot Study for Differentiating Constrictive Pericarditis from Restrictive Cardiomyopathy</article-title>
          <source>Circ Cardiovasc Imaging</source>
          <year>2016</year>
          <volume>9</volume>
          <issue>6</issue>
          <elocation-id>e004330</elocation-id>
          <pub-id pub-id-type="doi">10.1161/CIRCIMAGING.115.004330</pub-id>
        </element-citation>
        <mixed-citation>57 Sengupta PP, Huang YM, Bansal M, Ashrafi A, Fisher M, Shameer K, et al. Cognitive Machine-Learning Algorithm for Cardiac Imaging: A Pilot Study for Differentiating Constrictive Pericarditis from Restrictive Cardiomyopathy. Circ Cardiovasc Imaging. 2016;9(6):e004330. doi: 10.1161/CIRCIMAGING.115.004330.</mixed-citation>
      </ref>
      <ref id="B58">
        <label>58</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Kang</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Dey</surname>
              <given-names>D</given-names>
            </name>
            <name>
              <surname>Slomka</surname>
              <given-names>PJ</given-names>
            </name>
            <name>
              <surname>Arsanjani</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Nakazato</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Ko</surname>
              <given-names>H</given-names>
            </name>
            <etal/>
          </person-group>
          <article-title>Structured Learning Algorithm for Detection of Nonobstructive and Obstructive Coronary Plaque Lesions from Computed Tomography Angiography</article-title>
          <source>J Med Imaging (Bellingham)</source>
          <year>2015</year>
          <volume>2</volume>
          <issue>1</issue>
          <fpage>014003</fpage>
          <lpage>014003</lpage>
          <pub-id pub-id-type="doi">10.1117/1.JMI.2.1.014003</pub-id>
        </element-citation>
        <mixed-citation>58 Kang D, Dey D, Slomka PJ, Arsanjani R, Nakazato R, Ko H, et al. Structured Learning Algorithm for Detection of Nonobstructive and Obstructive Coronary Plaque Lesions from Computed Tomography Angiography. J Med Imaging (Bellingham). 2015;2(1):014003. doi: 10.1117/1.JMI.2.1.014003.</mixed-citation>
      </ref>
      <ref id="B59">
        <label>59</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Ribeiro</surname>
              <given-names>AL</given-names>
            </name>
            <name>
              <surname>Oliveira</surname>
              <given-names>GMM</given-names>
            </name>
          </person-group>
          <article-title>Toward a Patient&#8211;Centered, Data&#8211;Driven Cardiology</article-title>
          <source>Arq Bras Cardiol</source>
          <year>2019</year>
          <volume>112</volume>
          <issue>4</issue>
          <fpage>371</fpage>
          <lpage>373</lpage>
          <pub-id pub-id-type="doi">10.5935/abc.20190069</pub-id>
        </element-citation>
        <mixed-citation>59 Ribeiro AL, Oliveira GMM. Toward a Patient-Centered, Data-Driven Cardiology. Arq Bras Cardiol. 2019;112(4):371-3. doi: 10.5935/abc.20190069.</mixed-citation>
      </ref>
      <ref id="B60">
        <label>60</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Anderson</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Labus</surname>
              <given-names>JS</given-names>
            </name>
            <name>
              <surname>Vianna</surname>
              <given-names>EP</given-names>
            </name>
            <name>
              <surname>Mayer</surname>
              <given-names>EA</given-names>
            </name>
            <name>
              <surname>Cohen</surname>
              <given-names>MS</given-names>
            </name>
          </person-group>
          <article-title>Common Component Classification: What Can We Learn from Machine Learning?</article-title>
          <source>Neuroimage</source>
          <year>2011</year>
          <volume>56</volume>
          <issue>2</issue>
          <fpage>517</fpage>
          <lpage>524</lpage>
          <pub-id pub-id-type="doi">10.1016/j.neuroimage.2010.05.065</pub-id>
        </element-citation>
        <mixed-citation>60 Anderson A, Labus JS, Vianna EP, Mayer EA, Cohen MS. Common Component Classification: What Can We Learn from Machine Learning? Neuroimage. 2011;56(2):517-24. doi: 10.1016/j.neuroimage.2010.05.065.</mixed-citation>
      </ref>
      <ref id="B61">
        <label>61</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Halevy</surname>
              <given-names>A</given-names>
            </name>
            <name>
              <surname>Norvig</surname>
              <given-names>P</given-names>
            </name>
            <name>
              <surname>Pereira</surname>
              <given-names>F</given-names>
            </name>
          </person-group>
          <article-title>The Unreasonable Effectiveness of Data</article-title>
          <source>IEEE Intelligent Systems</source>
          <year>2009</year>
          <volume>24</volume>
          <issue>2</issue>
          <fpage>8</fpage>
          <lpage>12</lpage>
          <pub-id pub-id-type="doi">10.1109/MIS.2009.36</pub-id>
        </element-citation>
        <mixed-citation>61 Halevy A, Norvig P, Pereira F. The Unreasonable Effectiveness of Data. IEEE Intelligent Systems. 2009;24(2):8-12. doi:10.1109/MIS.2009.36.</mixed-citation>
      </ref>
      <ref id="B62">
        <label>62</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Shaw</surname>
              <given-names>LJ</given-names>
            </name>
          </person-group>
          <article-title>Can a Machine Learn Better Than Humans?</article-title>
          <source>JACC Cardiovasc Imaging</source>
          <year>2018</year>
          <volume>11</volume>
          <issue>7</issue>
          <fpage>1010</fpage>
          <lpage>1011</lpage>
          <pub-id pub-id-type="doi">10.1016/j.jcmg.2017.07.025</pub-id>
        </element-citation>
        <mixed-citation>62 Shaw LJ. Can a Machine Learn Better Than Humans? JACC Cardiovasc Imaging. 2018;11(7):1010-1. doi: 10.1016/j.jcmg.2017.07.025.</mixed-citation>
      </ref>
      <ref id="B63">
        <label>63</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Tsai</surname>
              <given-names>TL</given-names>
            </name>
            <name>
              <surname>Fridsma</surname>
              <given-names>DB</given-names>
            </name>
            <name>
              <surname>Gatti</surname>
              <given-names>G</given-names>
            </name>
          </person-group>
          <article-title>Computer Decision Support as a Source of Interpretation Error: The Case of Electrocardiograms</article-title>
          <source>J Am Med Inform Assoc</source>
          <year>2003</year>
          <volume>10</volume>
          <issue>5</issue>
          <fpage>478</fpage>
          <lpage>483</lpage>
          <pub-id pub-id-type="doi">10.1197/jamia.M1279</pub-id>
        </element-citation>
        <mixed-citation>63 Tsai TL, Fridsma DB, Gatti G. Computer Decision Support as a Source of Interpretation Error: The Case of Electrocardiograms. J Am Med Inform Assoc. 2003;10(5):478-83. doi: 10.1197/jamia.M1279.</mixed-citation>
      </ref>
      <ref id="B64">
        <label>64</label>
        <element-citation publication-type="journal">
          <person-group person-group-type="author">
            <name>
              <surname>Svensson</surname>
              <given-names>CM</given-names>
            </name>
            <name>
              <surname>H&#252;bler</surname>
              <given-names>R</given-names>
            </name>
            <name>
              <surname>Figge</surname>
              <given-names>MT</given-names>
            </name>
          </person-group>
          <article-title>Automated Classification of Circulating Tumor Cells and the Impact of Interobsever Variability on Classifier Training and Performance</article-title>
          <source>J Immunol Res</source>
          <year>2015</year>
          <volume>2015</volume>
          <fpage>573165</fpage>
          <lpage>573165</lpage>
          <pub-id pub-id-type="doi">10.1155/2015/573165</pub-id>
        </element-citation>
        <mixed-citation>64 Svensson CM, H&#252;bler R, Figge MT. Automated Classification of Circulating Tumor Cells and the Impact of Interobsever Variability on Classifier Training and Performance. J Immunol Res. 2015;2015:573165. doi: 10.1155/2015/573165.</mixed-citation>
      </ref>
    </ref-list>
  </back>
  <sub-article article-type="translation" id="S01" xml:lang="en">
    <front-stub>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Review Article</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Machine Learning in Medicine: Review and Applicability</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0003-1349-1745</contrib-id>
          <name>
            <surname>Paix&#227;o</surname>
            <given-names>Gabriela Miana de Mattos</given-names>
          </name>
          <xref ref-type="aff" rid="aff3">
            <sup>1</sup>
          </xref>
          <xref ref-type="corresp" rid="c2"/>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0003-4606-6454</contrib-id>
          <name>
            <surname>Santos</surname>
            <given-names>Bruno Campos</given-names>
          </name>
          <xref ref-type="aff" rid="aff3">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-6747-5444</contrib-id>
          <name>
            <surname>Araujo</surname>
            <given-names>Rodrigo Martins de</given-names>
          </name>
          <xref ref-type="aff" rid="aff3">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <name>
            <surname>Ribeiro</surname>
            <given-names>Manoel Horta</given-names>
          </name>
          <xref ref-type="aff" rid="aff3">
            <sup>1</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-8510-0013</contrib-id>
          <name>
            <surname>Moraes</surname>
            <given-names>Jermana Lopes de</given-names>
          </name>
          <xref ref-type="aff" rid="aff3">
            <sup>1</sup>
          </xref>
          <xref ref-type="aff" rid="aff4">
            <sup>2</sup>
          </xref>
        </contrib>
        <contrib contrib-type="author">
          <contrib-id contrib-id-type="orcid">0000-0002-2740-0042</contrib-id>
          <name>
            <surname>Ribeiro</surname>
            <given-names>Antonio L.</given-names>
          </name>
          <xref ref-type="aff" rid="aff3">
            <sup>1</sup>
          </xref>
        </contrib>
        <aff id="aff3">
          <label>1</label>
          <addr-line>
            <named-content content-type="city">Belo Horizonte</named-content>
            <named-content content-type="state">MG</named-content>
          </addr-line>
          <country country="BR">Brazil</country>
          <institution content-type="original">Universidade Federal de Minas Gerais, Belo Horizonte, MG &#8211; Brazil</institution>
        </aff>
        <aff id="aff4">
          <label>2</label>
          <addr-line>
            <named-content content-type="city">Sobral</named-content>
            <named-content content-type="state">CE</named-content>
          </addr-line>
          <country country="BR">Brazil</country>
          <institution content-type="original">Universidade Federal do Cear&#225;, Sobral, CE &#8211; Brazil</institution>
        </aff>
      </contrib-group>
      <author-notes>
        <corresp id="c2"><bold>Mailing Address: Gabriela Miana de Mattos Paix&#227;o</bold> &#8226; Universidade Federal de Minas Gerais - Avenida Alfredo Balena, 110. Postal Code <postal-code>31270-901</postal-code>, Belo Horizonte, MG &#8211; Brazil E-mail: <email>gabimiana@gmail.com</email> </corresp>
        <fn fn-type="con">
          <p>
            <bold>Author Contributions</bold>
          </p>
          <p>Conception and design of the research: Paix&#227;o GMM, Ribeiro MH, Moares JL, Ribeiro AL; Acquisition of data: Paix&#227;o GMM, Santos BC, Araujo RM, Moares JL; Analysis and interpretation of the data: Paix&#227;o GMM, Moares JL; Statistical analysis: Paix&#227;o GMM; Writing of the manuscript: Paix&#227;o GMM, Santos BC, Araujo RM, Ribeiro MH, Moares JL; Critical revision of the manuscript for intellectual content: Paix&#227;o GMM, Moares JL, Ribeiro AL.</p>
        </fn>
        <fn fn-type="conflict">
          <p>
            <bold>Potential Conflict of Interest</bold>
          </p>
          <p>No potential conflict of interest relevant to this article was reported.</p>
        </fn>
      </author-notes>
      <kwd-group xml:lang="en">
        <title>Keywords</title>
        <kwd>Machine Learning</kwd>
        <kwd>Medicine</kwd>
        <kwd>Cardiology</kwd>
      </kwd-group>
      <funding-group>
        <award-group>
          <funding-source>CNPq</funding-source>
          <award-id>310679/2016-8</award-id>
          <award-id>465518/2014-1</award-id>
        </award-group>
        <award-group>
          <funding-source>FAPEMIG</funding-source>
          <award-id>PPM-00428-17</award-id>
          <award-id>RED-00081-16</award-id>
        </award-group>
        <award-group>
          <funding-source>CNPq</funding-source>
          <award-id>141286/2021-0</award-id>
        </award-group>
        <funding-statement>Ribeiro AL is partially supported by CNPq (310679/2016-8 and 465518/2014-1) and by FAPEMIG (PPM-00428-17 and RED-00081-16). Moraes JL is supported by the CNPq (141286/2021-0).</funding-statement>
      </funding-group>
    </front-stub>
    <body>
      <sec sec-type="intro">
        <title>Introduction</title>
        <p>Machine learning (ML) is a branch of artificial intelligence (AI) that explores the study and construction of computational algorithms based on data learning,<sup><xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B2">2</xref></sup> rather than preprogrammed instructions.<sup><xref ref-type="bibr" rid="B3">3</xref></sup> The main objective of an ML model is to construct a computer system that learns from a predefined database and, in the end, generates a model for prediction, classification, or detection.</p>
        <p>The application of ML in practice is mainly aimed at handling consolidated databases with heterogeneous information where there is a limitation to the use of conventional statistical techniques.<sup><xref ref-type="bibr" rid="B4">4</xref>,<xref ref-type="bibr" rid="B5">5</xref></sup> ML algorithms are already widespread in different areas, such as banking systems for fraud detection, internet search engines, video surveillance systems, data security, business logistics, robotics, and, in medicine, diagnosis and prognosis.<sup><xref ref-type="bibr" rid="B6">6</xref></sup> With the digitalization of medical records, laboratory tests, and imaging, there has been a growth in database are sources for the application of ML techniques, with the aims of prevention, early diagnosis, and treatment of diseases.</p>
        <p>This review article provides an introduction to ML structured as follows: definition, learning models, a systematic review of articles on its applicability in medicine, especially cardiology. The objective is to introduce doctors and healthcare professionals to ML as a tool to assist clinical practice.</p>
        <p>To structure this review article, the following descriptors in English were searched in the databases PubMed (NCBI) and Medline: &#8220;machine learning,&#8221; &#8220;artificial intelligence,&#8221; &#8220;unsupervised learning,&#8221; &#8220;supervised learning,&#8221;, &#8220;neural networks&#8221; and &#8220;cardiology.&#8221; Prospective and retrospective studies were included, and clinical cases and abstracts presented at conferences (not published as articles) were excluded. Each study's eligibility was assessed by two investigators. Divergent opinions regarding the relevance of articles were addressed by consensus among the authors.</p>
        <sec>
          <title>Machine learning</title>
          <p>ML is a subfield of computer science that seeks an intersection between mathematical and statistical techniques and computational algorithms.<sup><xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B7">7</xref></sup> It uses algorithms with the concept of AI, and it is applied to certain situations to look for patterns within a set of variables in order to predict a specific result of interest.<sup><xref ref-type="bibr" rid="B8">8</xref>,<xref ref-type="bibr" rid="B9">9</xref></sup></p>
          <p>Most of the conventional techniques used in computer systems applied to medicine employ the concept of rule-based algorithms, known as &#8220;expert systems.&#8221; Thus, developers encode medical knowledge regarding a particular subject for these systems, using rules that are already known. ML techniques, on the other hand, handle a large number of variables, seeking a variety of new combinations that can reliably predict a result, in many cases, in a high volume of data, such as big data.<sup><xref ref-type="bibr" rid="B7">7</xref></sup></p>
          <p>In 2001, Doug Laney defined the &#8220;3 Vs&#8221; model to conceptualize the term big data: high volume, high velocity, and high variety of information, which require new processing techniques in order to allow discoveries and optimize processes.<sup><xref ref-type="bibr" rid="B10">10</xref></sup> The term big data may refer to either an enormous dataset that no traditional data management tools are able to store or process efficiently or to a type of technology (such as storage facilities, tools, and processes).<sup><xref ref-type="bibr" rid="B11">11</xref></sup></p>
          <p>The process of developing an ML algorithm is divided into 3 phases: preprocessing, training, and model evaluation (<xref ref-type="fig" rid="f4">Figure 1</xref>). The first phase consists of organizing the databank, defining the research question, and dividing the data into training and testing. During training, learning can take place in a supervised or unsupervised manner.<sup><xref ref-type="bibr" rid="B12">12</xref>&#8211;<xref ref-type="bibr" rid="B15">15</xref></sup> Supervised learning is based on training a data sample where the correct classification has already been assigned, whereas unsupervised learning refers to the capability to learn and organize information when the correct classification has not been assigned.<sup><xref ref-type="bibr" rid="B14">14</xref></sup> During the evaluation phase, the model is compared with test data, and the results are generated. Therefore, ML algorithms learn by means of repeated observations, and they establish a mapping pattern in order to label the data and create a model that generalizes the information so that new data (that have never been analyzed by the algorithm) can be accurately and reliably labeled.<sup><xref ref-type="bibr" rid="B15">15</xref></sup></p>
          <fig id="f4">
            <label>Figure 1</label>
            <caption>
              <title>Phases for developing machine learning algorithms.<sup><xref ref-type="bibr" rid="B15">15</xref></sup></title>
            </caption>
            <graphic xlink:href="1678-4170-abc-118-01-0095-00596-gf01-en.jpg"/></fig>
          <p>It is important to emphasize that the process of developing a ML algorithm must be carried out with a consolidated and validated database, because ML models that are developed with unconsolidated data can generate misleading results.<sup><xref ref-type="bibr" rid="B5">5</xref></sup></p>
        </sec>
        <sec>
          <title>Supervised and unsupervised machine learning</title>
          <p>The main difference between supervised and unsupervised learning models is in the training algorithm. In unsupervised learning, the ML model extracts the data characteristics and builds a representation without prior knowledge of the labels of each piece of data, that is, it identifies the information classification patterns heuristically. This lack of supervision for the algorithm may be advantageous, because it allows the algorithm to analyze patterns that have not been previously considered.<sup><xref ref-type="bibr" rid="B12">12</xref>&#8211;<xref ref-type="bibr" rid="B14">14</xref></sup></p>
          <p>In supervised learning, the ML model possesses knowledge regarding the data labels, that is, the samples are correctly classified. Training is based on the comparison between the result obtained from the model and the previously classified label. This process is repeated until minimum error is reached.<sup><xref ref-type="bibr" rid="B14">14</xref></sup></p>
          <p><xref ref-type="table" rid="t3">Table 1</xref> summarizes the main characteristics of each type of learning model, as well as their advantages, disadvantages, and practical applicability.</p>
          <table-wrap id="t3">
            <label>Table 1</label>
            <caption>
              <title>Comparison between supervised and unsupervised learning processes</title>
            </caption>
            <table frame="hsides" rules="rows">
              <colgroup width="33%">
                <col/>
                <col/>
                <col/>
              </colgroup>
              <thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
                <tr>
                  <th align="left" valign="middle"/>
                  <th align="center" valign="middle">Supervised learning</th>
                  <th align="center" valign="middle">Unsupervised learning</th>
                </tr>
              </thead>
              <tbody style="border-bottom: thin solid; border-color: #000000">
                <tr>
                  <td align="left" valign="middle">Definition</td>
                  <td align="center" valign="middle">Algorithms that learn relationships between input and output attributes based on a set of labeled examples.</td>
                  <td align="center" valign="middle">Algorithms that attempt to find patterns in data clusters with similar characteristics, looking for unidentified or uninformed categories and outcomes.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Advantages</td>
                  <td align="center" valign="middle">Analysis of multiple parameters; quick, automatic solution for large-scale questions and high accuracy.</td>
                  <td align="center" valign="middle">Less human interference in data analysis; excellent for multimodal or multidimensional data sources; allows identification of new outcomes.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Disadvantages</td>
                  <td align="center" valign="middle">Requires data to be labeled; may be impractical for large volumes of data. Tendency to overfit data.</td>
                  <td align="center" valign="middle">High cost; complex techniques. It requires a large amount of data to elaborate the algorithm, and it can be challenging to interpret the results.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Main tasks</td>
                  <td align="center" valign="middle">Regression, classification, prognostic model, and survival analysis.</td>
                  <td align="center" valign="middle">Reducing dimensionality of the problem and grouping.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Examples of algorithms</td>
                  <td align="center" valign="middle">Logistic regression, decision trees, random forests, and artificial neural networks.</td>
                  <td align="center" valign="middle">Principal component analysis, hierarchical clustering, autoencoders, and linear discriminant analysis.</td>
                </tr>
              </tbody>
            </table>
          </table-wrap>
        </sec>
        <sec>
          <title>Machine learning techniques</title>
          <p>Several ML techniques have been applied as a form of computer-assisted diagnostic systems, such as artificial neural networks (ANNs), logistic regression, decision tree, random forests, Bayesian network, deep learning, support vector machine (SVM), and others.<sup><xref ref-type="bibr" rid="B16">16</xref>&#8211;<xref ref-type="bibr" rid="B21">21</xref></sup> Some techniques use mathematical models by means of data for learning and/or organization of information.<sup><xref ref-type="bibr" rid="B12">12</xref></sup> Others apply mathematical representations with a high degree of abstraction (complex mathematical models). In this case, it is not possible to decipher or interpret the methods used to obtain the prediction, detection, or classification results; these ML models are, thus, known as &#8220;black box&#8221;.<sup><xref ref-type="bibr" rid="B22">22</xref></sup></p>
          <p>An ANN is a computational and mathematical model developed to function like the human brain. An ANN possesses several interconnecting elements (predictor layer, hidden layer, and output layer), and the relationship between these layers is inspired by the synaptic connections between neurons (<xref ref-type="fig" rid="f5">Figure 2</xref>).<sup><xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B15">15</xref>,<xref ref-type="bibr" rid="B23">23</xref></sup></p>
          <fig id="f5">
            <label>Figure 2</label>
            <caption>
              <title><italic>Functional structure of an artificial neural network</italic>.<sup><xref ref-type="bibr" rid="B19">19</xref></sup></title>
            </caption>
            <graphic xlink:href="1678-4170-abc-118-01-0095-00596-gf02-en.jpg"/></fig>
          <p>An ANN &#8220;learns&#8221; by means of these connections between layers (predictors, hidden layer, and results), as well as the weights associated with each layer. Thus, a piece of input data is introduced in the predictor layer and is sent layer by layer. Mathematical processing takes place by sending data from one layer to another, and the weights of these connections are updated according to the error in the result layer, that is, the relationship between the expected result and the obtained result. This process is repeated until the error value is minimal or until a specified interaction value.<sup><xref ref-type="bibr" rid="B12">12</xref>,<xref ref-type="bibr" rid="B23">23</xref>,<xref ref-type="bibr" rid="B24">24</xref></sup></p>
          <p>Deep learning differs from more traditional ML techniques to the extent that it processes more robust computational models with multiple processing layers based on ANNs. Thus, the technique of deep learning works in accordance with an ANN, but it possesses a greater number of hidden layers and, consequently, synaptic connections. Each layer reproduces a representation of data from the previous layer, and the learning algorithm can be either supervised or unsupervised.<sup><xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref></sup></p>
          <p>Given the high data volume and complexity involved in working with big data, the autoencoder algorithm is a type of ANN that reduces data dimensionality. In order to do this, the algorithm uses mathematical models with a high degree of abstraction to generate a new dataset with reduced dimensionality and representation as close as possible to the input data. The fundamental difference between an ANN and an autoencoder is that the latter uses unlabeled data during the training phase.<sup><xref ref-type="bibr" rid="B27">27</xref></sup></p>
          <p>The decision tree algorithm is most used when the dataset is relatively small, and it is developed with a series of yes/no questions to classify data into categories. This algorithm uses a statistical model for data classification or prediction, using the idea of nodes. Each node (question) is divided into possible outcomes, and they branch into other possibilities; this is repeated until the final outcome.<sup><xref ref-type="bibr" rid="B16">16</xref></sup> The main advantages of this algorithm are its simplicity and intuitive interpretation.<sup><xref ref-type="bibr" rid="B28">28</xref></sup></p>
          <p>Random forests are an extension of the decision tree algorithm, and they are widely used to solve classification and regression problems. Decision trees are combined, and each one is trained independently. Its main features are simple theory, fast data analysis, stability in the presence of excessive noise, and an automatic compensation mechanism for biased data samples.<sup><xref ref-type="bibr" rid="B29">29</xref></sup></p>
          <p>A Bayesian network is another technique that is widely applied to medicine. It consists of Bayesian statistical methods based on the theoretical foundation that consistent subjective beliefs of specialists in a given area can be expressed in a probabilistic structure.<sup><xref ref-type="bibr" rid="B17">17</xref></sup></p>
          <p>SVM is a supervised ML method that is widely used in bioinformatics. This algorithm uses the idea of error minimization, working with the statistical theory of learning and optimization. In addition to binary classification, SVM can be used in continuous data regression, called support vector regression. The results obtained with the use of SVM are comparable to those of ANNs, presenting an easy training process and working with high data dimensionality. It, therefore, reaches a compromise between less complexity and error.<sup><xref ref-type="bibr" rid="B30">30</xref>,<xref ref-type="bibr" rid="B31">31</xref></sup></p>
          <p>In this manner, each algorithm applies different techniques regarding how to learn from observations and how to carry out mapping of a set of predictors for the final result. It must generalize information so that a task can be performed correctly with new inputs that have not been previously analyzed by the model.<sup><xref ref-type="bibr" rid="B14">14</xref></sup></p>
        </sec>
        <sec>
          <title>Machine learning in medicine</title>
          <p>Since the past century, researchers have been exploring different applications of ML techniques in all fields of medicine.<sup><xref ref-type="bibr" rid="B32">32</xref></sup> Medical research involving ML has grown exponentially over the past few decades. Data from PubMed (NCBI) and Medline, involving the descriptors &#8220;machine learning,&#8221; &#8220;artificial intelligence,&#8221; &#8220;unsupervised learning,&#8221; &#8220;supervised learning,&#8221; and &#8220;neural networks,&#8221; revealed 113,127 articles published between 1951 and 2019 (<xref ref-type="fig" rid="f6">Figure 3</xref>). When the descriptor &#8220;cardiology&#8221; was added as a mandatory condition in the search for the other terms, 888 studies were found, with distribution similar to the previous one, between the years 1986 and 2019.</p>
          <fig id="f6">
            <label>Figure 3</label>
            <caption>
              <title><italic>Number of articles per year and cumulative during the period from 1951 to 2019 in PubMed and Medline</italic>.</title>
            </caption>
            <graphic xlink:href="1678-4170-abc-118-01-0095-00596-gf03-en.jpg"/></fig>
          <p>The capability of ML algorithms to recognize patterns and predict diagnoses has been widely applied to different areas of healthcare.<sup><xref ref-type="bibr" rid="B33">33</xref>&#8211;<xref ref-type="bibr" rid="B36">36</xref></sup> In dermatology, an ANN was able to differentiate dermatological lesions as benign versus malignant, based on more than 129,000 cases, with results similar to those of a committee of 21 dermatologists.<sup><xref ref-type="bibr" rid="B35">35</xref></sup> In the field of psychiatry, a study with ML techniques reduced the number of diagnostic criteria from 29 to 8, with 100% accuracy in 612 patients with confirmed diagnosis of autistic spectrum disorder.<sup><xref ref-type="bibr" rid="B36">36</xref></sup></p>
          <p>The addition of mobile technologies, such as smartphones and smartwatches, applied to the area of healthcare has added another dimension to ML, making it possible to read large quantities of personal data in learning algorithms.<sup><xref ref-type="bibr" rid="B37">37</xref></sup> Within feedback systems, mobile technology is able to be a biometric device (for example, measuring blood glucose levels), capable of targeting real-time clinical interventions, based on algorithms that continuously update patients&#8217; personal information.<sup><xref ref-type="bibr" rid="B38">38</xref></sup> Technology is able to simplify diagnostic processes and facilitate clinical practice.</p>
        </sec>
        <sec>
          <title>Machine learning in cardiology</title>
          <p>Advances in computational capacity in recent decades have especially impacted the field of detection and prediction of cardiovascular diseases through interpretation of data, such as studies of medical records; imaging exams; and biological, genomic, and molecular evaluation databanks.<sup><xref ref-type="bibr" rid="B32">32</xref></sup> Cardiology is one of the areas with the greatest impact on scientific production using ML techniques (<xref ref-type="table" rid="t4">Table 2</xref>). From the prediction of cardiovascular events<sup><xref ref-type="bibr" rid="B39">39</xref></sup> to the improvement of electrocardiographic and imaging diagnoses,<sup><xref ref-type="bibr" rid="B40">40</xref>,<xref ref-type="bibr" rid="B41">41</xref></sup> AI has been an important tool for scientific research.</p>
          <table-wrap id="t4">
            <label>Table 2</label>
            <caption>
              <title>Articles on the use of machine learning in cardiology</title>
            </caption>
            <table frame="hsides" rules="rows">
              <colgroup width="50%">
                <col/>
                <col/>
              </colgroup>
              <thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
                <tr>
                  <th align="left" valign="middle">Article</th>
                  <th align="center" valign="middle">Main results</th>
                </tr>
              </thead>
              <tbody style="border-bottom: thin solid; border-color: #000000">
                <tr>
                  <td align="left" valign="middle">Can machine-learning improve cardiovascular risk prediction using routine clinical data?<sup><xref ref-type="bibr" rid="B38">38</xref></sup></td>
                  <td align="center" valign="middle">The algorithm was able to predict 4998 of 7404 positive cases (sensitivity 67.5%, PPV 18.4%) and 53,458 of 75,585 negative cases (specificity 70.7% and NPV 95.7%), with a gain of 355 patients (+7.6%) who developed cardiovascular diseases, compared to the traditional method.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Deep neural networks can predict mortality from 12-lead electrocardiogram voltage data<sup><xref ref-type="bibr" rid="B43">43</xref></sup></td>
                  <td align="center" valign="middle">By means of isolated analysis of ECG using a ML algorithm, it was possible to predict 1-year all-cause mortality with AUC = 0.84 and p &lt; 0.05.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Phenomapping for the Identification of Hypertensive Patients with the Myocardial Substrate for Heart Failure with Preserved Ejection Fraction<sup><xref ref-type="bibr" rid="B56">56</xref></sup></td>
                  <td align="center" valign="middle">A group of 1273 patients with hypertension was evaluated using ML techniques, using clinical, laboratory, and echocardiography data. It was possible to identify a group of patients at a higher risk of developing heart failure with preserved ejection fraction who were likely to benefit from more intensive medical treatment.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Cognitive Machine-Learning Algorithm for Cardiac Imaging: A Pilot Study for Differentiating Constrictive Pericarditis From Restrictive Cardiomyopathy<sup><xref ref-type="bibr" rid="B57">57</xref></sup></td>
                  <td align="center" valign="middle">They used ML techniques to differentiate constrictive pericarditis from restrictive cardiomyopathy with a ROC curve of 96.2% and accuracy greater than 90%.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Structured learning algorithm for detection of nonobstructive and obstructive coronary plaque lesions from computed tomography angiography<sup><xref ref-type="bibr" rid="B58">58</xref></sup></td>
                  <td align="center" valign="middle">The ML algorithm was able to detect coronary lesions greater than or equal to 25% with 93% sensitivity, 95% specificity, and 94% accuracy in 42 coronary angiographies.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">A deep neural network learning algorithm outperforms a conventional algorithm for emergency department electrocardiogram interpretation<sup><xref ref-type="bibr" rid="B54">54</xref></sup></td>
                  <td align="center" valign="middle">Automatic analysis using the ML method for reading ECG in an emergency department obtained sensitivity (88.7% versus 92.0%, p &lt; 0.086), specificity (94% versus 84.7%, p &lt; 0.0001), PPV (88.2% versus 75.4%, p &lt; 0.0001), and accuracy (92.2% versus 87.2%, p &lt; 0.0001), compared to the conventional automatic method.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">Automatic Diagnosis of the Short-Duration 12-Lead ECG using a Deep Neural Network: the CODE Study<sup><xref ref-type="bibr" rid="B53">53</xref></sup></td>
                  <td align="center" valign="middle">A trained neural network was able to detect 6 classes of electrocardiographic abnormalities with specificity greater than 99% and performance greater than 80%, compared to last-year cardiology residents.</td>
                </tr>
                <tr>
                  <td align="left" valign="middle">An artificial intelligence-enabled ECG algorithm for the identification of patients with atrial fibrillation during sinus rhythm: a retrospective analysis of outcome prediction<sup><xref ref-type="bibr" rid="B55">55</xref></sup></td>
                  <td align="center" valign="middle">ML software was able to detect patients with atrial fibrillation, based on ECG in sinus rhythm, with a sensitivity of 79%, specificity of 79.5%, and accuracy of 79.4%.</td>
                </tr>
              </tbody>
            </table>
            <table-wrap-foot>
              <fn id="TFN2">
                <p>
                  <italic>ECG: electrocardiogram; ML: machine learning; NPV: negative predictive value; PPV: positive predictive value.</italic>
                </p>
              </fn>
            </table-wrap-foot>
          </table-wrap>
        </sec>
        <sec>
          <title>Prognosis</title>
          <p>Several cardiovascular risk scores have been developed in order to predict cardiovascular events and identify individuals with higher cardiac risks, for primary prevention.<sup><xref ref-type="bibr" rid="B42">42</xref></sup> However, in spite of all the advances in diagnostic workup and therapy in cardiology, there is still a population at risk that is has not been identified by traditional methods.<sup><xref ref-type="bibr" rid="B43">43</xref></sup> It is desirable to recognize potential non-traditional risk factors, and the use of new technologies, such as AI, has become a promising method in this search.</p>
          <p>The prediction of all-cause mortality over a 1-year period, based on isolated analysis of electrocardiogram (ECG) has shown promising results (AUROC 0.87; p &lt; 0.05).<sup><xref ref-type="bibr" rid="B44">44</xref></sup> It is interesting to underscore that a blind analysis of the same ECGs by 3 cardiologists suggested that the patterns which ML found to predict mortality were not apparently visible on conventional medical assessment.<sup><xref ref-type="bibr" rid="B44">44</xref></sup></p>
          <p>In a study including 2619 patients who underwent computerized tomography with proton emission for prediction of cardiovascular risk, ML techniques showed better results (AUROC 0.81; p &lt; 0.01) than isolated analysis of the exam.<sup><xref ref-type="bibr" rid="B45">45</xref></sup></p>
          <p>A study with more than 380,000 patients in the United Kingdom evaluated the use of ML techniques to predict the risk of cardiovascular events in comparison with the traditional algorithms proposed by the American College of Cardiology and the American Heart Association.<sup><xref ref-type="bibr" rid="B39">39</xref></sup> There was an improvement of up to 7.6% in the prediction of events with the use of ANN. Some clinical variables that are not valued for cardiovascular disease by traditional methods, such as depression and corticosteroid use, were important to cardiovascular risk assessed by ML techniques<sup><xref ref-type="bibr" rid="B39">39</xref></sup>. This finding was corroborated by a multicenter study from the United States where the parameters found for cardiovascular risk prediction differed from those included in traditional risk calculators.<sup><xref ref-type="bibr" rid="B46">46</xref></sup></p>
          <p>AI can contribute to the generation of more complex and specific predictive models for each individual,<sup><xref ref-type="bibr" rid="B47">47</xref></sup> by incorporating genomic components in cardiovascular risk scores.<sup><xref ref-type="bibr" rid="B48">48</xref>,<xref ref-type="bibr" rid="B49">49</xref></sup> The association of clinical, social, demographic, and genetic data with the available exams can allow more individualized assessment, with the aim of health promotion.<sup><xref ref-type="bibr" rid="B47">47</xref></sup></p>
        </sec>
        <sec>
          <title>Diagnosis</title>
          <p>In cardiac exams, the need for a highly specialized medical team a variability of reports among physicians, and time spent on reports have led to the study of ML techniques as a diagnostic tool.<sup><xref ref-type="bibr" rid="B41">41</xref>,<xref ref-type="bibr" rid="B50">50</xref></sup></p>
          <p>The studies have been promising, and cardiac imaging modalities such as echocardiography, computed tomography, and nuclear magnetic resonance have shown good accuracy in correlating structural changes with the etiology and pathophysiology of cardiovascular diseases.<sup><xref ref-type="bibr" rid="B51">51</xref>,<xref ref-type="bibr" rid="B52">52</xref></sup> In a study with 159 patients, 3 ML techniques were used to aid in the echocardiographic differentiation between hypertrophic cardiomyopathy and physiological hypertrophy in athletes. The parameters found, such as early-to-late transmitral diastolic velocity ratio (p &lt; 0.01), early diastolic velocity (e&#8217;) (p &lt; 0.01), and strain analysis (p &lt; 0.01), were better in sensitivity and specificity than those that are traditionally used.<sup><xref ref-type="bibr" rid="B51">51</xref></sup></p>
          <p>A ML algorithm was developed to differentiate intermediate coronary stenosis on angiography with fractional flow reserve less than 0.80 versus greater than 0.80, based on clinical and angiography data. The results were satisfactory, with an accuracy of approximately 80% for prediction of fractional flow reserve less than 0.8 (AUROC 0.84 to 0.87, 95% CI 0.71 to 0.89). The external validation of the developed model also showed similar results in 79 patients from 2 other centers (AUROC 0.89, 95% CI 0.83 to 0.95).<sup><xref ref-type="bibr" rid="B53">53</xref></sup></p>
          <p>In relation to ECG, studies are being developed to improve automatic diagnoses.<sup><xref ref-type="bibr" rid="B41">41</xref></sup> By means of ML techniques, our group has been able to identify 6 ECG classes through 12-lead ECG analysis, with good accuracy, comparable to the performance of last-year cardiology residents.<sup><xref ref-type="bibr" rid="B54">54</xref></sup> In hospitalized patients with cardiovascular emergencies, ML had diagnostic accuracy of about 90% for major ECG changes.<sup><xref ref-type="bibr" rid="B54">54</xref></sup> Furthermore, a recent study was able to identify patients with atrial fibrillation on ECG in sinus rhythm with a sensitivity of 79%, specificity of 79.5%, and accuracy of 79.4%.<sup><xref ref-type="bibr" rid="B55">55</xref></sup></p>
        </sec>
        <sec>
          <title>Limits and challenges</title>
          <p>The use of ML techniques is growing, due to their potential to solve problems in different areas. In medicine, the results have been promising in several specialties, with the expectation that AI can be a tool to assist clinical practice.<sup><xref ref-type="bibr" rid="B3">3</xref>,<xref ref-type="bibr" rid="B59">59</xref></sup> Nevertheless, it is still necessary to be cautious when interpreting and incorporating the results.</p>
          <p>The ML algorithms developed must be reproducible in the general population. Studies with small numbers of patients, in specific populations, or with selection biases do not allow for generalization of their findings.<sup><xref ref-type="bibr" rid="B60">60</xref>,<xref ref-type="bibr" rid="B61">61</xref></sup> Even though data capture and interpretation have considerable statistical value, the best scenarios are still not capable of predicting the outcome in different people.<sup><xref ref-type="bibr" rid="B62">62</xref></sup></p>
          <p>An error in an automated process can lead professionals to incorrect conclusions, as demonstrated in a study with 30 internal medicine residents whose diagnostic accuracy for ECG reports was reduced when they were provided with incorrect automatic reports.<sup><xref ref-type="bibr" rid="B63">63</xref></sup></p>
          <p>Some doctors have viewed the advancement of AI in medicine with concern. The alarmist position that ML might replace doctors in healthcare has proved to be unjustified. No software, so far, has been able to replace the subjective aspect of clinical experience in making favorable decisions for the patient, precisely because medicine is not an exact science.<sup><xref ref-type="bibr" rid="B64">64</xref></sup> The denial of technological advancement and the AI tools that are available today is as potentially damaging as total dependence on ML for patient care. The combination of ML and clinical judgment has shown better results together than its isolated application.<sup><xref ref-type="bibr" rid="B59">59</xref></sup></p>
        </sec>
      </sec>
      <sec sec-type="conclusions">
        <title>Conclusion</title>
        <p>The use of ML techniques in medicine has left the field of theory and gone on to become a reality. Although the use of ML in medicine is still in development, studies have demonstrated its clinical applicability, with an impact on diagnostic and prognostic evaluation.</p>
      </sec>
    </body>
    <back>
      <fn-group>
        <fn fn-type="financial-disclosure" id="fn3">
          <p>
            <bold>Sources of Funding</bold>
          </p>
          <p>Ribeiro AL is partially supported by CNPq (310679/2016-8 and 465518/2014-1) and by FAPEMIG (PPM-00428-17 and RED-00081-16). Moraes JL is supported by the CNPq (141286/2021-0).</p>
        </fn>
        <fn fn-type="other" id="fn4">
          <p>
            <bold>Study Association</bold>
          </p>
          <p>This article is part of the thesis of Doctoral submitted by Gabriela Miana de Mattos Paix&#227;o, from Universidade Federal de Minas Gerais.</p>
        </fn>
      </fn-group>
    </back>
  </sub-article>
</article>
