<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.1 20151215//EN" "https://jats.nlm.nih.gov/publishing/1.1/JATS-journalpublishing1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML" dtd-version="1.1" specific-use="sps-1.9" article-type="research-article" xml:lang="en">
<front>
<journal-meta>
<journal-id journal-id-type="publisher-id">dic</journal-id>
<journal-title-group>
<journal-title>ABC Imagem Cardiovascular</journal-title>
<abbrev-journal-title abbrev-type="publisher">Arq Bras Cardiol: Imagem cardiovasc.</abbrev-journal-title></journal-title-group>
<issn pub-type="epub">2675-312X</issn>
<publisher>
<publisher-name>Sociedade Brasileira de Cardiologia</publisher-name></publisher>
</journal-meta>
<article-meta>
<article-id pub-id-type="publisher-id">abcimg.20230085i</article-id>
<article-id pub-id-type="doi">10.36660/abcimg.20230085i</article-id>
<article-categories>
<subj-group subj-group-type="heading">
<subject>Original Article</subject>
</subj-group>
</article-categories>
<title-group>
<article-title>Use of Artificial Intelligence to Assess Cardiac Function by Echocardiography: Systematic Review of the State of the Art</article-title>
</title-group>
<contrib-group>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0000-0003-0988-5282</contrib-id>
<name><surname>Neves</surname><given-names>Henrique Alexsander Ferreira</given-names></name>
<role>Conception and design of the research</role>
<role>acquisition of data and analysis and interpretation of the data</role>
<role>writing of the manuscript</role>
<xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0000-0001-9735-4590</contrib-id>
<name><surname>Yuasa</surname><given-names>Bruna Sadae</given-names></name>
<role>Conception and design of the research</role>
<role>acquisition of data and analysis and interpretation of the data</role>
<role>writing of the manuscript</role>
<xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0009-0003-8756-8116</contrib-id>
<name><surname>Costa</surname><given-names>Thamires Hadassa Leite Pereira</given-names></name>
<role>Conception and design of the research</role>
<role>acquisition of data and analysis and interpretation of the data</role>
<role>writing of the manuscript</role>
<xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0009-0009-1438-2947</contrib-id>
<name><surname>Santos</surname><given-names>Isabela Ertes</given-names></name>
<role>Conception and design of the research</role>
<role>acquisition of data and analysis and interpretation of the data</role>
<role>writing of the manuscript</role>
<xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0009-0001-3556-5218</contrib-id>
<name><surname>Benavides</surname><given-names>Yannir&#xEA; Milagros Roman</given-names></name>
<role>Conception and design of the research</role>
<role>acquisition of data and analysis and interpretation of the data</role>
<role>writing of the manuscript</role>
<xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0000-0003-0794-7339</contrib-id>
<name><surname>Lofrano-Alves</surname><given-names>Marco Stephan</given-names></name>
<role>Conception and design of the research</role>
<role>acquisition of data and analysis and interpretation of the data</role>
<role>critical revision of the manuscript for intellectual content</role>
<xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
<xref ref-type="corresp" rid="c1"/>
</contrib>
<aff id="aff1">
<label>1</label>
<institution content-type="orgname">Universidade Federal do Paran&#xE1;</institution>
<addr-line>
<named-content content-type="city">Curitiba</named-content>
<named-content content-type="state">PR</named-content>
</addr-line>
<country country="BR">Brazil</country>
<institution content-type="original">Universidade Federal do Paran&#xE1;, Curitiba, PR &#x2013; Brazil</institution>
</aff>
</contrib-group>
<author-notes>
<corresp id="c1"><bold>Mailing Address: Marco Stephan Lofrano-Alves</bold> &#x2022; Universidade Federal do Paran&#xE1;, Departamento de Cl&#xED;nica M&#xE9;dica. Rua Mateus Leme, 3945, apt 504. Postal code: <postal-code>82200-000</postal-code>. Curitiba, PR &#x2013; Brazil. E-mail: <email>mslalves@hotmail.com</email></corresp>
<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>
<pub-date publication-format="electronic" date-type="pub">
<day>18</day>
<month>12</month>
<year>2023</year></pub-date>
<pub-date publication-format="electronic" date-type="collection">
<year>2023</year></pub-date>
<volume>36</volume>
<issue>4</issue>
<elocation-id>e20230085</elocation-id>
<history>
<date date-type="received"><day>28</day><month>10</month><year>2023</year></date>
<date date-type="rev-recd"><day>30</day><month>10</month><year>2023</year></date>
<date date-type="accepted"><day>31</day><month>10</month><year>2023</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>
<abstract>
<title>Abstract</title>
<sec>
<title>Background:</title>
<p>Echocardiography is part of the routine investigation of most heart diseases. However, its operator-dependent nature and the quality of the image obtained can cause significant variability in measurements and impair their interpretation. The method has advanced on its path of standardization and incorporation of new technologies in an attempt to reduce its deficiencies. Recently, part of this advancement was due to the incorporation of artificial intelligence (AI).</p>
</sec>
<sec>
<title>Objectives:</title>
<p>The objective of this work was to investigate the use of AI in the assessment of cardiac function through echocardiography, discussing its potential to improve the method.</p>
</sec>
<sec>
<title>Methods:</title>
<p>We conducted a systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), searching for articles within the period of 2000 to 2023.</p>
</sec>
<sec>
<title>Results:</title>
<p>A total of 31 articles were included, classified into four categories of studies that used AI: 1) for image analysis and echocardiographic parameters (N = 14); 2) in the image acquisition process (N = 11); 3) for integration with clinical data (N = 3); and 4) for application to the exam flow (N = 3).</p>
</sec>
<sec>
<title>Conclusion:</title>
<p>AI provides automation and standardization of images and respective measurements that result in more accurate diagnoses, overcoming the limitations encountered by the conventional method. However, the use of AI is still an evolving reality within echocardiography.</p>
</sec>
</abstract>
<kwd-group xml:lang="en">
<title>Keywords:</title>
<kwd>Echocardiography</kwd>
<kwd>Artificial Intelligence</kwd>
<kwd>Machine Learning</kwd>
<kwd>Deep Learning</kwd>
<kwd>Global Longitudinal Strain</kwd>
</kwd-group>
<counts>
<fig-count count="4"/>
<table-count count="8"/>
<equation-count count="0"/>
<ref-count count="36"/>
</counts>
</article-meta>
</front>
<body>
<sec sec-type="intro">
<title>Introduction</title>
<p>Echocardiography is the most requested imaging method and is part of the routine investigation of most heart diseases. Due to its non-invasive, non-ionizing nature, high availability, and portability, the method is present in various clinical scenarios - outpatient, hospital, urgency and emergency units, intensive care units, and as a guide for interventional and surgical procedures - and normally is the first option to access cardiac anatomy and function. However, its operator-dependent nature and image quality can influence echocardiographic measurements. Since its development, the method has advanced in its path of standardization and incorporation of new technologies in an attempt to reduce its variability.</p>
<p>This advance is partly due to the incorporation of artificial intelligence (AI) into the method. &#x201C;AI&#x201D; is a broad term coined by Arthur Samuel in the 1960s. It applies to any computer program, such as algorithms and models, which imitate human logic and intelligence.<sup><xref ref-type="bibr" rid="B1">1</xref></sup> In echocardiography, the first use was in the 70s, when Fourier analysis was used to evaluate the mitral valve with the aid of M-mode.<sup><xref ref-type="bibr" rid="B2">2</xref></sup> Semi-automated measurements are already part of today&#x27;s echocardiography laboratory practice. In their workflow, modern devices come equipped with software packages that assist in measurements and calculations, such as left ventricular ejection fraction (LVEF), myocardial deformation (strain), and the Doppler study for hemodynamic assessment.<sup><xref ref-type="bibr" rid="B1">1</xref></sup></p>
<p>Machine learning is one of the subfields of AI. The computer receives the data, learns from it, predicts a result, and, using statistical methods, can even improve its performance the next time it is presented with similar data. This allows, for example, a trained machine to identify structures within a new image and label them accurately. Another subfield, known as deep learning, later emerged. This is applied when a large amount of data must be processed. In this case, the data is processed as large subsets. Just like in a biological neural network, the processing is organized in a hidden chain made up of thousands of consecutive artificial neural layers until a final result is reached.<sup><xref ref-type="bibr" rid="B1">1</xref></sup> An artificial neural network can use convolution operations, which allow a third to be generated from two input data called Convolutional Neural Networks.<sup><xref ref-type="bibr" rid="B3">3</xref></sup></p>
<p>In this sense, AI in echocardiography can be extremely advantageous for identifying, quantifying, and interpreting images while minimizing the variance between assessments and providing reproducible measurements and an accurate diagnosis. The examiner may find significant support in this technology, which would enable a reduction in time spent on data extraction and integration and, in turn, present the potential to make the healthcare system more sustainable.<sup><xref ref-type="bibr" rid="B4">4</xref></sup> Furthermore, AI-driven echocardiographic practice would not only streamline clinical decision-making but also provide learning methods with interactive feedback for training less-experienced examiners.<sup><xref ref-type="bibr" rid="B5">5</xref></sup></p>
<fig id="f2">
<caption>
<title>Flowchart of search and selection of studies on the use of AI in echocardiography. AI: artificial intelligence.</title></caption>
<graphic xlink:href="2675-312X-dic-36-04-e20230085-gf02.tif"/>
</fig>
<p>Thus, the use of AI in echocardiography can be divided into four main scenarios: education, image acquisition, image analysis, and integration with clinical data.<sup><xref ref-type="bibr" rid="B6">6</xref></sup> Although there is evidence that positively affirms the role of AI technology in echocardiography diagnosis, there are still challenges to be overcome, such as insufficient standardization of echocardiography protocols and insufficient generalization of models in clinical applications.<sup><xref ref-type="bibr" rid="B6">6</xref></sup></p>
<p>Given the limitations of the technique and the need to investigate future trends, this systematic review was conducted based on several studies to externally validate the performance of AI technology in echocardiography. Although similar works are found in the literature that investigated the application of AI in the automated analysis of echocardiographic images, this study sought to conduct a survey over the last five years on innovations in the area, with the aim of exploring the use of AI in echocardiography, analyzing current evidence for its effectiveness and discussing its potential to improve cardiac care.</p>
</sec>
<sec sec-type="methods">
<title>Methods</title>
<p>This is a systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).<sup><xref ref-type="bibr" rid="B7">7</xref></sup> We searched for articles within the MEDLINE, EMBASE, SCOPUS, and Cochrane Central databases from January 2000 to March 2023, without language restrictions, with our target method being echocardiography, whose intervention is the use of AI during some stage of the exam, with full texts and descriptors present in the title or summary of the article. The following descriptors in English were used for the research: (&#x201C;AI&#x201D;) AND (&#x201C;Echocardiography&#x201D;) AND (&#x201C;Function&#x201D; OR &#x201C;Strain&#x201D;). We also checked these databases for the presence of previous or ongoing systematic reviews on the subject. We combined results from different databases and discarded repeated results using Zotero software. Two investigators (H.A.F.N and B.S.Y) performed the search independently and five investigators reached an agreement on study eligibility (H.A.F.N, B.S.Y, T.H.L.P.C, I.E.S, Y.M.R.B). Initially, a first screening was conducted to identify studies that evaluated the use of any type of AI during the echocardiography exam. In a second step, we selected the articles to be read in full through the Rayyan platform and assessed the relevance of the data, as well as whether there was more than one article evaluating the same cohort of patients. A third step consisted of reviewing the references of these selected articles to obtain other sources not revealed in the digital search. We classified the studies according to their design (case-control, randomized, retrospective cohort, prospective cohort, and case report), regarding the primary objective of using AI and the main result found with the study. For each study, we extracted the following information: first author; study institution; start and end date; year of publication; single center or multicenter; characteristics of the studied population (average age and distribution by sex); echocardiographic parameters evaluated; methods and windows used to assess cardiac function; AI model; study design; and study conclusions.</p>
</sec>
<sec sec-type="results">
<title>Results</title>
<p>We identified 729 references in searches of electronic databases, discarding 336 duplicates and another 359 excluded by title and/or abstract. We obtained 33 articles for full reading, of which three were discarded because they did not fit the scope of the review according to the authors. Only one additional article was included after manually searching the references (<xref ref-type="fig" rid="f1">Figure 1</xref>). A total of 31 articles were included in the qualitative analysis of this review, being classified into four main categories: 1) studies that use AI for image analysis and echocardiographic parameters (N = 14, <xref ref-type="table" rid="t1">Table 1</xref>); 2) studies that use AI in the image acquisition process (N = 11, <xref ref-type="table" rid="t2">Table 2</xref>); 3) studies that use AI to integrate with clinical data (N = 3, <xref ref-type="table" rid="t3">Table 3</xref>); and <xref ref-type="table" rid="t4">4</xref>) studies that trained AI software for application to the exam flow (N = 3, <xref ref-type="table" rid="t4">Table 4</xref>); All articles were published in English. Twenty-six articles are from a single center, with 7 (%) from the United States, 2 (%) from Israel, 2 (%) from Norway, 2 (%) from Italy, 2 (%) from England, 1 (%) from Germany, 1 (%) from Greece, 1 (%) from France, 1 (%) from Uganda, 1 (%) from Portugal, 1 (%) from Ireland, 1 (%) from Austria, 1 (%) from Netherlands, 1 (%) from China, 1 (%) from Canada, and 1 (%) from Japan. A single article was developed in several centers (13 different centers). Four articles did not inform of the country where the study took place.</p>
<fig id="f1">
<label>Figure 1</label>
<caption>
<title>PRISMA flow diagram of study screening and selection</title></caption>
<graphic xlink:href="2675-312X-dic-36-04-e20230085-gf01.tif"/>
</fig>
<table-wrap id="t1">
<label>Table 1</label>
<caption>
<title>Studies that use AI for image analysis</title></caption>
<table frame="hsides" rules="groups">
<colgroup width="14%">
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
</colgroup>
<thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
<tr style="background-color: #C58874;">
<th align="left" valign="middle">Author/year</th>
<th align="center" valign="middle">Country</th>
<th align="center" valign="middle">Design and objective</th>
<th align="center" valign="middle">N</th>
<th align="center" valign="middle">Evaluated parameters</th>
<th align="center" valign="middle">AI model</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">Cottela <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B8">8</xref></sup></td>
<td align="left" valign="middle"/>
<td align="center" valign="middle">Retrospective cohort Evaluate AI calculation of LVEF and GLS in patients with amyloidosis</td>
<td align="center" valign="middle">51</td>
<td align="center" valign="middle">LVEF, GLS</td>
<td align="center" valign="middle">EchoGo Core 2.0, Ultromics</td>
<td align="center" valign="middle">Fully automated AI-calculated LVEF and GLS are comparable to manual measurements in patients with cardiac amyloidosis.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Sveric <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B9">9</xref></sup></td>
<td align="center" valign="middle">Germany</td>
<td align="center" valign="middle">Prospective cohort Investigate AI viability for the calculation of LVEF in the workflow</td>
<td align="center" valign="middle">889</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">LVivo Seamless<sup>TM</sup></td>
<td align="center" valign="middle">The AI-based application automatically identified standard apical LV views and measured LVEF. It yielded equivalent results to the currently used MBS method and has the potential to improve LVEF quantification.</td>
</tr>
<tr>
<td align="left" valign="middle">He <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B10">10</xref></sup></td>
<td align="center" valign="middle">USA</td>
<td align="center" valign="middle">Randomized clinical trial Compare AI with the sonographer&apos;s initial assessment of LVEF</td>
<td align="center" valign="middle">3495</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">EchoNet-Dynamic</td>
<td align="center" valign="middle">Cardiologists were not able to distinguish between the initial assessments by AI versus the sonographer. The initial assessment of LVEF by AI was non-inferior to the assessment by sonographers.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Wang <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B11">11</xref></sup></td>
<td align="center" valign="middle">USA</td>
<td align="center" valign="middle">Retrospective cohort Evaluate AI software usability in patients with COVID-19</td>
<td align="center" valign="middle">50</td>
<td align="center" valign="middle">LVEF, GLS, EDV, ESF</td>
<td align="center" valign="middle">EchoGo, Ultromics</td>
<td align="center" valign="middle">EchoGo software output had high correlations with left ventricular volumes and moderate correlations with LVEF and GLS compared with standard TTE measurements.</td>
</tr>
<tr>
<td align="left" valign="middle">Liel-Cohen <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B12">12</xref></sup></td>
<td align="center" valign="middle">Israel</td>
<td align="center" valign="middle">Retrospective cohort Evaluate AI&#x27;s capability of identifying views and compare GLS results with conventional methods</td>
<td align="center" valign="middle">100</td>
<td align="center" valign="middle">GLS</td>
<td align="center" valign="middle">LVivo Seamless<sup>TM</sup></td>
<td align="center" valign="middle">The results provided by the AI-based LVivo Seamless solution showed excellent capability to automatically identify 4CH, 2CH, and 3CH views, along with excellent GLS results compared to automated functional imaging.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Sachpekidis <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B13">13</xref></sup></td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Retrospective cohort Evaluate the accuracy of a novel HUD with AI-assisted algorithm to automatically calculate LVEF</td>
<td align="center" valign="middle">100</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">Auto EF, Vivid&#x2122;</td>
<td align="center" valign="middle">The AI-assisted autoEF algorithm in a novel HUD can accurately calculate LVEF in real-time as compared to the recommended manual biplane Simpson&#x27;s method on cart-based systems in an &#x201C;all-comers&#x201D; patient population.</td>
</tr>
<tr>
<td align="left" valign="middle">Salte <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B14">14</xref></sup></td>
<td align="center" valign="middle">Norway</td>
<td align="center" valign="middle">Retrospective cohort Examine automated measurements of GLS using deep learning and AI compared with conventional methods</td>
<td align="center" valign="middle">200</td>
<td align="center" valign="middle">GLS</td>
<td align="center" valign="middle">EchoPAC v.202, GE Ultrasound</td>
<td align="center" valign="middle">Fully automated measurements of GLS using a novel deep learning AI-based technology for motion estimation are feasible and fast and they yield results comparable with the most widely used semiautomatic software.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Hanif <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B15">15</xref></sup></td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Transversal Test whether AI measurement of strain can be more reliable compared to human interpretation</td>
<td align="center" valign="middle">52</td>
<td align="center" valign="middle">GLS</td>
<td align="center" valign="middle">EchoGo, Ultromics</td>
<td align="center" valign="middle">AI-based longitudinal strain analysis was feasible on most echocardiograms without any operator intervention. The bias between EchoGo&#x27;s longitudinal strain and conventional software appears small.</td>
</tr>
<tr>
<td align="left" valign="middle">Penso <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B16">16</xref></sup></td>
<td align="center" valign="middle">Italy</td>
<td align="center" valign="middle">Transversal Feasibility, comparison, and assessment of AI strain measurement compared with conventional methods</td>
<td align="center" valign="middle">203</td>
<td align="center" valign="middle">RVEF, RVFWLS</td>
<td align="center" valign="middle">3D Auto RV, Philips Medical Systems</td>
<td align="center" valign="middle">The rapid analysis and excellent reproducibility of AI-based 3DE RV analysis supported the routine adoption of the system.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Di Candia <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B17">17</xref></sup></td>
<td align="center" valign="middle">Italy</td>
<td align="center" valign="middle">Transversal Compare results of a fully automated analysis and manual tracing analysis using a new AI software</td>
<td align="center" valign="middle">28</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">AI Simpson e AI Wall Motion Tracking</td>
<td align="center" valign="middle">AI fully automated analysis was found to be technically simple, less time-consuming, and highly reproducible. This reduces the variability of manual measurements between different sonographers and at various times.</td>
</tr>
<tr>
<td align="left" valign="middle">Papadopoulou <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B18">18</xref></sup></td>
<td align="center" valign="middle">Greece</td>
<td align="center" valign="middle">Transversal Evaluate the reliability and diagnostic accuracy of a HUD-assisted AI to automatically calculate ejection fraction</td>
<td align="center" valign="middle">100</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">Kosmos autoEF</td>
<td align="center" valign="middle">The use of a novel HUD with AI-enabled capabilities provided similar LVEF results to those derived by manual biplane Simpson&#x27;s method on cart-based systems and shows clinical potential.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Salte <italic>et al.</italic>, 2020<sup><xref ref-type="bibr" rid="B19">19</xref></sup></td>
<td align="center" valign="middle">Norway</td>
<td align="center" valign="middle">Multicentric Prospective Examine the feasibility of automated measurements of GLS using AI compared to conventional speckle-tracking application</td>
<td align="center" valign="middle">100</td>
<td align="center" valign="middle">GLS</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Fully automated measurements of GLS using a novel deep learning AI-based technology for motion estimation are feasible and fast and they yield results comparable with the most widely used semiautomatic software.</td>
</tr>
<tr>
<td align="left" valign="middle">Popoff <italic>et al.</italic>, 2020<sup><xref ref-type="bibr" rid="B20">20</xref></sup></td>
<td align="center" valign="middle">France</td>
<td align="center" valign="middle">Transversal Evaluate the accuracy and reproducibility of AI tool to compute LV volumes and EF, in comparison with cardiac magnetic resonance as reference</td>
<td align="center" valign="middle">114</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">A deep convolutional neural networks model</td>
<td align="center" valign="middle">Compared to manual contouring, LV volumes, and EF by AI showed comparable or improved accuracy and higher reproducibility. These findings demonstrate the value of AI-based tools, with the potential for full automation, for objective assessment and follow-up of cardiac function.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Asch <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B21">21</xref></sup></td>
<td align="center" valign="middle">USA</td>
<td align="center" valign="middle">Prospective cohort Determine whether AI-automated estimates could be used to accurately classify LV function</td>
<td align="center" valign="middle">233</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">A machine learning algorithm, Caption Health</td>
<td align="center" valign="middle">The new machine-learning algorithm allows accurate automated evaluation of LV function from echocardiographic views commonly used in the POC setting. This approach will enable more POC personnel to accurately assess LV function.</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<fn id="TFN1">
<p><italic>EDV: left ventricular end-diastolic volume; ESF: left ventricular end-systolic volume; GLS: global longitudinal strain; HUD: handheld ultrasound device; LVEF: left ventricle ejection fraction; POC: point-of-care; RVEF: right ventricle ejection fraction; RVFWLS: right ventricular free wall longitudinal strain; AI: artificial intelligence; AI: artificial intelligence; TTE: transthoracic echocardiogram; AutoEF: automatic ejection fraction; EF: ejection fraction; LV: left ventricle; 4CH: four chamber view; 2CH: two chamber view; 3CH: three chamber view.</italic></p></fn>
</table-wrap-foot>
</table-wrap>
<table-wrap id="t2">
<label>Table 2</label>
<caption>
<title>Studies that use AI for image acquisition.</title></caption>
<table frame="hsides" rules="groups">
<colgroup width="14%">
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
</colgroup>
<thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
<tr style="background-color: #C58874;">
<th align="left" valign="middle">Author/year</th>
<th align="center" valign="middle">Country</th>
<th align="center" valign="middle">Study design and objective</th>
<th align="center" valign="middle">N</th>
<th align="center" valign="middle">Evaluated parameters</th>
<th align="center" valign="middle">AI model</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">Peck <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B22">22</xref></sup></td>
<td align="center" valign="middle">Uganda</td>
<td align="center" valign="middle">Cross-sectional. Evaluate the ability of nonexperts to obtain diagnostic-quality images in patients with RHD using AI guidance with color Doppler</td>
<td align="center" valign="middle">50</td>
<td align="center" valign="middle">Presence/absence of RHD, abnormal mitral valve morphology, and mitral regurgitation</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">AI guidance with color Doppler is feasible to enable RHD screening by nonexperts, performing significantly better for assessment of the mitral than aortic valve.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Klempfner <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B5">5</xref></sup></td>
<td align="center" valign="middle">USA</td>
<td align="center" valign="middle">Prospective cohort. Determine the diagnostic acumen of echocardiographic studies obtained by inexperienced users guided by an AI algorithm</td>
<td align="center" valign="middle">240</td>
<td align="center" valign="middle">Left and right ventricular size and function, pericardial effusion, valve morphology, left atrial and inferior vena cava size</td>
<td align="center" valign="middle">UltraSight algorithm</td>
<td align="center" valign="middle">After minimal training with novel AI-guidance software, novice users can acquire images in ten standard cardiac echocardiographic views, with diagnostic quality approaching that of expert sonographers.</td>
</tr>
<tr>
<td align="left" valign="middle">Dadon, Ziv <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B23">23</xref></sup></td>
<td align="center" valign="middle">Israel</td>
<td align="center" valign="middle">Randomized clinical trial Evaluate AI use as a didactic tool to improve noncardiologist clinicians&#x2019; assessment of LVEF from the A4C view</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">LVivo EF</td>
<td align="center" valign="middle">The introduction of an AI-based tool to clinicians working in the emergency department improved the assessment accuracy of LVEF as compared to the control group.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Varudo <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B24">24</xref></sup></td>
<td align="center" valign="middle">Portugal</td>
<td align="center" valign="middle">Prospective cohort. To compare real-time measurements taken by novices and experts with LVEF reference measurements taken manually by echo experts</td>
<td align="center" valign="middle">95</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">Real-Time EF algorithm</td>
<td align="center" valign="middle">Machine learning-enabled real-time measurements of LVEF were strongly correlated with manual measurements obtained by experts.</td>
</tr>
<tr>
<td align="left" valign="middle">Burke <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B25">25</xref></sup></td>
<td align="center" valign="middle">Ireland</td>
<td align="center" valign="middle">Cross-sectional. Evaluate study quality by novice scanners with AI guidance</td>
<td align="center" valign="middle">120</td>
<td align="center" valign="middle">Image quality to determine LVEF, pericardial effusion</td>
<td align="center" valign="middle">Caption AI System</td>
<td align="center" valign="middle">The Caption AI technology safely allows novice users to provide efficient and accurate point-of-care echo in differing clinical settings to a standard comparable to expert scanners.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Narang <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B26">26</xref></sup></td>
<td align="center" valign="middle">USA</td>
<td align="center" valign="middle">Prospective cohort. Test whether novice users could obtain 10-view transthoracic echocardiographic studies of diagnostic quality using a DL-based software</td>
<td align="center" valign="middle">240</td>
<td align="center" valign="middle">Left ventricular size, function, pericardial effusion, and right ventricular size</td>
<td align="center" valign="middle">Software Caption Guidance</td>
<td align="center" valign="middle">The DL algorithm allows novices without experience in ultrasonography to obtain diagnostic transthoracic echocardiographic studies for evaluation of left ventricular size and function, right ventricular size, and the presence of a nontrivial pericardial effusion.</td>
</tr>
<tr>
<td align="left" valign="middle">Surette <italic>et al.</italic>, 2020<sup><xref ref-type="bibr" rid="B27">27</xref></sup></td>
<td align="center" valign="middle">USA</td>
<td align="center" valign="middle">Prospective multicentric clinical trial. Evaluate whether novice-acquired images guided by the software were of diagnostic quality in patients with and without implanted electrophysiological devices</td>
<td align="center" valign="middle">240</td>
<td align="center" valign="middle">Right ventricular size and function</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">New DL software can guide novices to obtain TTEs that enable qualitative assessment of RV size even in the presence of implanted EP devices.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Schneider <italic>et al.</italic>, 2020<sup><xref ref-type="bibr" rid="B28">28</xref></sup></td>
<td align="center" valign="middle">Austria</td>
<td align="center" valign="middle">Cross-sectional. Evaluate if the new machine learning algorithm can help novice operators to acquire diagnostic image loops</td>
<td align="center" valign="middle">14</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Machine-learning algorithms can guide ultrasound novices to acquire diagnostic echo loops and provide an automated LVEF calculation that agrees with a human expert.</td>
</tr>
<tr>
<td align="left" valign="middle">Motazedian <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B29">29</xref></sup></td>
<td align="center" valign="middle">Netherlands</td>
<td align="center" valign="middle">Prospective cohort. Determine the accuracy of POCUS bedside AI-assisted LVEF compared to formal TTE</td>
<td align="center" valign="middle">449</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">EchoNous Kosmos</td>
<td align="center" valign="middle">AI assisted POCUS performed by both novice and experienced scanners can accurately determine LVEF compared to a comprehensive TTE.</td>
</tr>
<tr>
<td align="left" valign="middle">Zhai <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B30">30</xref></sup></td>
<td align="center" valign="middle">China</td>
<td align="center" valign="middle">Cross-sectional. Identify the degree of correlation between auto LVOT&#x2010;VTI and the manual LVOT&#x2010;VTI acquired by POCUS-trained doctors</td>
<td align="center" valign="middle">46</td>
<td align="center" valign="middle">LVOT&#x2010;VTI</td>
<td align="center" valign="middle">AI software tool</td>
<td align="center" valign="middle">AI might provide real&#x2010;time guidance among novice operators who lack the expertise to acquire the ideal standard view.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Fung <italic>et al.</italic>, 2019<sup><xref ref-type="bibr" rid="B31">31</xref></sup></td>
<td align="center" valign="middle">Canada</td>
<td align="center" valign="middle">Prospective cohort. Evaluate the accuracy of the AI system for assessing image quality, as compared to subjective expert analysis</td>
<td align="center" valign="middle">53</td>
<td align="center" valign="middle">PLAX, PSAX, A2C, A4C and subcostal IVC</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">AI systems can assess image quality with modest accuracy.</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<fn id="TFN2">
<p><italic>A2C: apical two-chamber; A4C: apical four-chamber; DL: deep learning; IVC: inferior vena cava; LVEF: left ventricle ejection fraction; LVOT-VTI: left ventricular outflow tract velocity time integral; PLAX: parasternal long-axis; POCUS: point-of-care ultrasound; PSAX: parasternal short-axis; RHD: rheumatic heart disease; RV: right ventricle; TTE: transthoracic echocardiogram; AI: artificial intelligence; EP: electrophysiological.</italic></p></fn>
</table-wrap-foot>
</table-wrap>
<table-wrap id="t3">
<label>Table 3</label>
<caption>
<title>Studies that use AI for integration with clinical data</title></caption>
<table frame="hsides" rules="groups">
<colgroup width="14%">
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
</colgroup>
<thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
<tr style="background-color: #C58874;">
<th align="left" valign="middle">Author/ year</th>
<th align="center" valign="middle">Country</th>
<th align="center" valign="middle">Study design and objective</th>
<th align="center" valign="middle">N</th>
<th align="center" valign="middle">Evaluated parameters</th>
<th align="center" valign="middle">AI model</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">Kusunose <italic>et al.</italic>, 2019<sup><xref ref-type="bibr" rid="B32">32</xref></sup></td>
<td align="center" valign="middle">Japan</td>
<td align="center" valign="middle">Prospective cohort. Investigate whether a DCNN could provide improved detection of RWMAs</td>
<td align="center" valign="middle">300</td>
<td align="center" valign="middle">Presence of wall motion abnormalities</td>
<td align="center" valign="middle">ResNet, DenseNet, Inception-ResNet, Inception, and Xception</td>
<td align="center" valign="middle">The results support the possibility of using DCNN for automated diagnosis of RWMAs in the field of echocardiography.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">O&#x27;Driscoll <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B33">33</xref></sup></td>
<td align="center" valign="middle">England</td>
<td align="center" valign="middle">Retrospective cohort. Evaluate whether LVEF and GLS, automatically calculated by AI, increases the diagnostic performance of SE for CAD detection</td>
<td align="center" valign="middle">512</td>
<td align="center" valign="middle">LVEF and GLS at rest and stress</td>
<td align="center" valign="middle">EchoGo Core 1.0, Ultromics</td>
<td align="center" valign="middle">AI calculation of LVEF and GLS by contouring of contrast-enhanced and unenhanced SEs at rest and stress is feasible and independently improves the identification of obstructive CAD beyond conventional methods.</td>
</tr>
<tr>
<td align="left" valign="middle">Asch, F.M. <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B34">34</xref></sup></td>
<td align="center" valign="middle">Multicentric</td>
<td align="center" valign="middle">Prospective cohort. Assess the hypothesis that AI-based analysis of echocardiographic images could predict mortality more accurately than conventional analysis by a human expert</td>
<td align="center" valign="middle">870</td>
<td align="center" valign="middle">LVEF, LVLS</td>
<td align="center" valign="middle">EchoGo Core, Ultromics</td>
<td align="center" valign="middle">AI-based analysis of LVEF and LVLS had similar feasibility as manual analysis and AI-based, but not manual, analyses were a significant predictor of in-hospital and follow-up mortality.</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<fn id="TFN3">
<p>CAD: coronary artery disease; DCNN: deep convolutional neural network; GLS: global longitudinal strain; LVEF: left ventricle ejection fraction; LVLS: left ventricle longitudinal strain; RWMAs: regional wall motion abnormalities; SE: stress echocardiography.</p>
</fn>
</table-wrap-foot>
</table-wrap>
<table-wrap id="t4">
<label>Table 4</label>
<caption>
<title>Studies that developed an AI</title></caption>
<table frame="hsides" rules="groups">
<colgroup width="14%">
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
</colgroup>
<thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
<tr style="background-color: #C58874;">
<th align="left" valign="middle">Author and year</th>
<th align="center" valign="middle">Country</th>
<th align="center" valign="middle">Study design and objective</th>
<th align="center" valign="middle">Number of patients</th>
<th align="center" valign="middle">Evaluated parameters</th>
<th align="center" valign="middle">AI Model</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">Asch <italic>et al.</italic>, 2019<sup>38</sup></td>
<td align="center" valign="middle">USA</td>
<td align="center" valign="middle">Transversal<break/><break/> Develop an algorith and assess its feasibility and accuracy for automated quantification of LV EF by a well-trained expert computer against reference</td>
<td align="center" valign="middle">99</td>
<td align="center" valign="middle">LVEF</td>
<td align="center" valign="middle">AutoEF (BayLabs)</td>
<td align="center" valign="middle">A machine learning algorithm for the estimation of LV EF that avoids image segmentation and volume measurements is highly feasible and can yield results that are in close agreement with what experienced readers measure using conventional methodology.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Stowell <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B35">35</xref></sup></td>
<td align="center" valign="middle">England</td>
<td align="center" valign="middle">Transversal<break/><break/> Develop a method based on machine learning, trained and validated with accredited experts from AI Echocardiography Collaborative</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Left ventricle longitudinal tension</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">The open-source, vendor-independent AI-based strain measurement automatically produces values that agree with expert consensus as strongly as individual experts.</td>
</tr>
<tr>
<td align="left" valign="middle">Tromp <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B36">36</xref></sup></td>
<td align="center" valign="middle">Singapore</td>
<td align="center" valign="middle">Transversal<break/><break/> Develop a fully automated deep-learning workflow to classify, segment, and annotate two-dimensional videos and Doppler modalities in echocardiograms</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Deep learning algorithms can automatically annotate 2D videos and Doppler modalities with similar accuracy to manual measurements by expert sonographers.</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<fn id="TFN4">
<p><italic>AutoEF: automatic ejection fraction; LVEF: left ventricle ejection fraction.</italic></p>
</fn>
</table-wrap-foot>
</table-wrap>
</sec>
<sec sec-type="discussion">
<title>Discussion</title>
<p>The implementation of AI tools in echocardiographic study has the potential to improve patient care and optimize diagnosis and treatment. The studies present in this systematic review are favorable to the application of automated technology in the examination routine, due to its similarity, feasibility, usefulness, precision, accuracy, and reproducibility comparable to manual assessment. Although the use of AI in the examination is still limited and, therefore, prone to modifications aimed at advancing and improving the method, the benefit that the technology represents for echocardiography seems to outweigh any fears or disadvantages.</p>
<sec>
<title>Image analysis and echocardiographic parameters</title>
<p>AI-automated measurement of parameters such as LVEF and global longitudinal strain (GLS) can be comparable to manual measurement by experienced sonographers and incorporations of these tools have reliably maintained measurement accuracy.<sup><xref ref-type="bibr" rid="B8">8</xref>&#x2013;<xref ref-type="bibr" rid="B13">13</xref></sup> This fact is relevant due to the additional benefits they offer: they increase the operational effectiveness of the exam and the consistency quality of the analysis;<sup><xref ref-type="bibr" rid="B6">6</xref></sup> allowing the reduction of variability between different examiners,<sup><xref ref-type="bibr" rid="B14">14</xref>&#x2013;<xref ref-type="bibr" rid="B18">18</xref></sup> overcoming human limitations such as tiredness or distraction; and increasing the reproducibility of the exam.<sup><xref ref-type="bibr" rid="B1">1</xref></sup></p>
</sec>
<sec>
<title>Image Acquisition</title>
<p>Much of the literature focuses on AI interpretation or analysis of already acquired images, but there is growing evidence regarding the use of AI technology for image acquisition by non-experts.<sup><xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B19">19</xref>&#x2013;<xref ref-type="bibr" rid="B24">24</xref></sup> Automated technology can guide a non-expert examiner by recognizing incorrect images and the need for additional views or enhancements. It can also help standardize image acquisition and measurements.<sup><xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B6">6</xref></sup> In this way, it would be a tool used to optimize training and echocardiographic studies,<sup><xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref></sup> with the possibility of acquiring images of diagnostic quality comparable to those acquired manually by experts.<sup><xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B22">22</xref>&#x2013;<xref ref-type="bibr" rid="B24">24</xref></sup> The technology is applicable in clinical practice, in emergencies, and in remote areas with scarce resources and few experienced professionals.<sup><xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B27">27</xref></sup></p>
</sec>
<sec>
<title>Integration with clinical data</title>
<p>In the field of echocardiography, interpretation of exam findings is necessary to improve patient care and monitoring. Thus, training AI software to associate image data with patient clinical data allows the retention of relevant diagnostic information without the need for advanced analysis skills by an expert and without the expenditure of time and resources. Correlation with functional abnormalities, pathological states, prediction of outcomes, and risk stratification are examples of the usefulness of integrating algorithms with clinical data. The results presented by the study by Kusunose <italic>et al.</italic> support the possibility of using AI software for automated diagnosis of functional abnormalities, and the study by O&#x27;Driscoll <italic>et al.</italic> to calculate LVEF and GLS.<sup><xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref></sup> From a future perspective, the integration of AI software with clinical data could allow the creation of personalized therapies for each patient.</p>
</sec>
<sec>
<title>AI development</title>
<p>Databases of echocardiographic studies were used to train protocols and deep learning algorithms to detect cardiac dysfunction, identify anatomical abnormalities, and define workflows.<sup><xref ref-type="bibr" rid="B30">30</xref>&#x2013;<xref ref-type="bibr" rid="B32">32</xref></sup> After training, the accuracy found was similar to the accuracies using conventional methodologies. This validation is essential since, in the development of AI software, both software and images come from external suppliers. The quality of the images used is a relevant point because, if the data entered is of low quality, biased, or restricted, the interpretation by the algorithm will be as well. A future challenge will be to develop deep learning algorithms that run on different equipment, have uniform naming standards, and store and retrieve data in a vendor-independent manner.<sup><xref ref-type="bibr" rid="B1">1</xref></sup></p>
<p>The incorporation of AI into clinical practice raises ethical concerns, primarily about responsibility, as care becomes more indirect, and transparency as the extent to which the patient needs to know and agree is called into question. It is worth remembering that the use of AI is already part of people&#x27;s routine, and this type of technology is increasingly available to society. This interaction tends to be irreversible and, certainly, in medical practice - and in echocardiography - it will be no different. Even so, this is not a man versus machine competition. We cannot imagine a scenario in which one replaces the other. On the contrary, AI is in favor of the doctor and against the obscurity of human and systemic limitations, such as high demand, lack of specialists, ergonomics, and service exhaustion. Especially because, by increasing the operational effectiveness of the exam, there can be better use and direction of the specialist doctors&#x2019; skills. It is expected that this incorporation will contribute to the democratization of access to the exam, representing an advance in health care.</p>
<p>However, this systematic review has its limitations. First, the vast majority of studies have a cross-sectional design. However, as our objective was diagnostic validation, studies with this design are sufficient to show the non-inferiority of their methods. Designs that study long-term association measures may better clarify the benefits of AI for the quality of longitudinal patient follow-up. Second, AI software usually requires a minimum image quality. Sometimes, patients with poor windows were excluded.<sup><xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref></sup> In several of the studies, the images analyzed were collected in ideal scenarios, by experienced specialists, or were packages from supplier centers. Therefore, larger studies with low-quality images, especially in patients with poor windows, hemodynamically unstable, or in non-ideal settings (ICU, emergency) should be encouraged. Third, most softwares are capable of measuring a limited number of parameters. Thus, further studies can validate the benefits of using AI beyond the parameters studied in this review.</p>
</sec>
</sec>
<sec sec-type="conclusions">
<title>Conclusion</title>
<p>The application of AI in the area of echocardiography has proven to be a highly effective and accurate method in four main domains (Central Illustration). AI provides the automation and standardization of images and respective measurements, which results in more accurate diagnoses, overcoming the limitations found in manipulation by a human sonographer. Furthermore, AI has also proven to be a facilitating tool in obtaining satisfactory results even when used by non-experts. The advantages offered by AI make the exam much faster and more accessible, as the duration of training for a specialist professional is one of the challenges of exam availability in public health. Therefore, as a result of the facts presented, AI is a beneficial resource that can be widely implemented in echocardiography services, as the advantages presented by the method cover a wide range of dimensions, from the accuracy of results and data integration for clinical correlation to the prevention of ergonomic diseases related to the sonographer&#x27;s work.</p>
</sec>
</body>
<back>
<fn-group>
<fn fn-type="other" id="fn1">
<p>Editor responsible for the review: Daniela do Carmo Rassi Frota</p></fn>
<fn fn-type="supported-by" id="fn2">
<p><bold>Sources of Funding</bold></p>
<p>There were no external funding sources for this study.</p></fn>
<fn fn-type="other" id="fn3">
<p><bold>Study Association</bold></p>
<p>This study is not associated with any thesis or dissertation work.</p></fn>
<fn fn-type="other" id="fn4">
<p><bold>Ethics Approval and Consent to Participate</bold></p>
<p>This article does not contain any studies with human participants or animals performed by any of the authors.</p></fn>
</fn-group>
<ref-list>
<title>Refer&#xEA;ncias</title>
<ref id="B1">
<label>1</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Davis</surname><given-names>A</given-names></name>
<name><surname>Billick</surname><given-names>K</given-names></name>
<name><surname>Horton</surname><given-names>K</given-names></name>
<name><surname>Jankowski</surname><given-names>M</given-names></name>
<name><surname>Knoll</surname><given-names>P</given-names></name>
<name><surname>Marshall</surname><given-names>JE</given-names></name>
<etal/>
</person-group>
<article-title>Artificial Intelligence and Echocardiography: A Primer for Cardiac Sonographers</article-title>
<source>J Am Soc Echocardiogr</source>
<year>2020</year>
<volume>33</volume>
<issue>9</issue>
<fpage>1061</fpage>
<lpage>1066</lpage>
<pub-id pub-id-type="doi">10.1016/j.echo.2020.04.025</pub-id>
</element-citation>
<mixed-citation>Davis A, Billick K, Horton K, Jankowski M, Knoll P, Marshall JE, et al. Artificial Intelligence and Echocardiography: A Primer for Cardiac Sonographers. J Am Soc Echocardiogr. 2020;33(9):1061-6. doi: 10.1016/j.echo.2020.04.025.</mixed-citation>
</ref>
<ref id="B2">
<label>2</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Chu</surname><given-names>WK</given-names></name>
<name><surname>Raeside</surname><given-names>DE</given-names></name>
</person-group>
<article-title>Fourier Analysis of the Echocardiogram</article-title>
<source>Phys Med Biol</source>
<year>1978</year>
<volume>23</volume>
<issue>1</issue>
<fpage>100</fpage>
<lpage>105</lpage>
<pub-id pub-id-type="doi">10.1088/0031-9155/23/1/009</pub-id>
</element-citation>
<mixed-citation>Chu WK, Raeside DE. Fourier Analysis of the Echocardiogram. Phys Med Biol. 1978;23(1):100-5. doi: 10.1088/0031-9155/23/1/009.</mixed-citation>
</ref>
<ref id="B3">
<label>3</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Akkus</surname><given-names>Z</given-names></name>
<name><surname>Aly</surname><given-names>YH</given-names></name>
<name><surname>Attia</surname><given-names>IZ</given-names></name>
<name><surname>Lopez-Jimenez</surname><given-names>F</given-names></name>
<name><surname>Arruda-Olson</surname><given-names>AM</given-names></name>
<name><surname>Pellikka</surname><given-names>PA</given-names></name>
<etal/>
</person-group>
<article-title>Artificial Intelligence (AI)-Empowered Echocardiography Interpretation: A State-of-the-Art Review</article-title>
<source>J Clin Med</source>
<year>2021</year>
<volume>10</volume>
<issue>7</issue>
<fpage>1391</fpage>
<lpage>1391</lpage>
<pub-id pub-id-type="doi">10.3390/jcm10071391</pub-id>
</element-citation>
<mixed-citation>Akkus Z, Aly YH, Attia IZ, Lopez-Jimenez F, Arruda-Olson AM, Pellikka PA, et al. Artificial Intelligence (AI)-Empowered Echocardiography Interpretation: A State-of-the-Art Review. J Clin Med. 2021;10(7):1391. doi: 10.3390/jcm10071391.</mixed-citation>
</ref>
<ref id="B4">
<label>4</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Alsharqi</surname><given-names>M</given-names></name>
<name><surname>Woodward</surname><given-names>WJ</given-names></name>
<name><surname>Mumith</surname><given-names>JA</given-names></name>
<name><surname>Markham</surname><given-names>DC</given-names></name>
<name><surname>Upton</surname><given-names>R</given-names></name>
<name><surname>Leeson</surname><given-names>P</given-names></name>
</person-group>
<article-title>Artificial Intelligence and Echocardiography</article-title>
<source>Echo Res Pract</source>
<year>2018</year>
<volume>5</volume>
<issue>4</issue>
<fpage>R115</fpage>
<lpage>R125</lpage>
<pub-id pub-id-type="doi">10.1530/ERP-18-0056</pub-id>
</element-citation>
<mixed-citation>Alsharqi M, Woodward WJ, Mumith JA, Markham DC, Upton R, Leeson P. Artificial Intelligence and Echocardiography. Echo Res Pract. 2018;5(4):R115-R125. doi: 10.1530/ERP-18-0056.</mixed-citation>
</ref>
<ref id="B5">
<label>5</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Klempfner</surname><given-names>RV</given-names></name>
<name><surname>Lang</surname><given-names>RM</given-names></name>
<name><surname>Mor-Avi</surname><given-names>V</given-names></name>
<name><surname>Guile</surname><given-names>B</given-names></name>
<name><surname>Hipke</surname><given-names>K</given-names></name>
<name><surname>Cotella</surname><given-names>JI</given-names></name>
<etal/>
</person-group>
<article-title>Novel Artificial Intelligence Guidance Algorithm Enables Acquisition by Novices of Diagnostic Quality Echocardiographic Images</article-title>
<source>J Am Coll Cardiol</source>
<year>2023</year>
<volume>81</volume>
<issue>8</issue>
<fpage>1401</fpage>
<lpage>1401</lpage>
<pub-id pub-id-type="doi">10.1016/S0735-1097(23)01845-4</pub-id>
</element-citation>
<mixed-citation>Klempfner RV, Lang RM, Mor-Avi V, Guile B, Hipke K, Cotella JI, et al. Novel Artificial Intelligence Guidance Algorithm Enables Acquisition by Novices of Diagnostic Quality Echocardiographic Images. J Am Coll Cardiol. 2023;81(8):1401. doi: 10.1016/S0735-1097(23)01845-4.</mixed-citation>
</ref>
<ref id="B6">
<label>6</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Thachil</surname><given-names>R</given-names></name>
<name><surname>Hanson</surname><given-names>D</given-names></name>
</person-group>
<article-title>Artificial Intelligence in Echocardiography: A Disruptive Technology for Democratizing and Standardizing Health</article-title>
<source>J Am Soc Echocardiogr</source>
<year>2022</year>
<volume>35</volume>
<issue>8</issue>
<fpage>A14</fpage>
<lpage>A16</lpage>
<pub-id pub-id-type="doi">10.1016/j.echo.2022.06.001</pub-id>
</element-citation>
<mixed-citation>Thachil R, Hanson D. Artificial Intelligence in Echocardiography: A Disruptive Technology for Democratizing and Standardizing Health. J Am Soc Echocardiogr. 2022;35(8):A14-A16. doi: 10.1016/j.echo.2022.06.001.</mixed-citation>
</ref>
<ref id="B7">
<label>7</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Helbach</surname><given-names>J</given-names></name>
<name><surname>Hoffmann</surname><given-names>F</given-names></name>
<name><surname>Pieper</surname><given-names>D</given-names></name>
<name><surname>Allers</surname><given-names>K</given-names></name>
</person-group>
<article-title>Reporting According to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Abstracts (PRISMA-A) Depends on Abstract Length</article-title>
<source>J Clin Epidemiol</source>
<year>2023</year>
<volume>154</volume>
<fpage>167</fpage>
<lpage>177</lpage>
<pub-id pub-id-type="doi">10.1016/j.jclinepi.2022.12.019</pub-id>
</element-citation>
<mixed-citation>Helbach J, Hoffmann F, Pieper D, Allers K. Reporting According to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Abstracts (PRISMA-A) Depends on Abstract Length. J Clin Epidemiol. 2023;154:167-77. doi: 10.1016/j.jclinepi.2022.12.019.</mixed-citation>
</ref>
<ref id="B8">
<label>8</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Cotella</surname><given-names>JI</given-names></name>
<name><surname>Slivnick</surname><given-names>JA</given-names></name>
<name><surname>Sanderson</surname><given-names>E</given-names></name>
<name><surname>Singulane</surname><given-names>C</given-names></name>
<name><surname>O&#x27;Driscoll</surname><given-names>J</given-names></name>
<name><surname>Asch</surname><given-names>FM</given-names></name>
<etal/>
</person-group>
<article-title>Artificial Intelligence Based Left Ventricular Ejection Fraction and Global Longitudinal Strain in Cardiac Amyloidosis</article-title>
<source>Echocardiography</source>
<year>2023</year>
<volume>40</volume>
<issue>3</issue>
<fpage>188</fpage>
<lpage>195</lpage>
<pub-id pub-id-type="doi">10.1111/echo.15516</pub-id>
</element-citation>
<mixed-citation>Cotella JI, Slivnick JA, Sanderson E, Singulane C, O&#x27;Driscoll J, Asch FM, et al. Artificial Intelligence Based Left Ventricular Ejection Fraction and Global Longitudinal Strain in Cardiac Amyloidosis. Echocardiography. 2023;40(3):188-95. doi: 10.1111/echo.15516.</mixed-citation>
</ref>
<ref id="B9">
<label>9</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>He</surname><given-names>B</given-names></name>
<name><surname>Kwan</surname><given-names>AC</given-names></name>
<name><surname>Cho</surname><given-names>JH</given-names></name>
<name><surname>Yuan</surname><given-names>N</given-names></name>
<name><surname>Pollick</surname><given-names>C</given-names></name>
<name><surname>Shiota</surname><given-names>T</given-names></name>
<etal/>
</person-group>
<article-title>Blinded, Randomized Trial of Sonographer Versus AI Cardiac Function Assessment</article-title>
<source>Nature</source>
<year>2023</year>
<volume>616</volume>
<issue>7957</issue>
<fpage>520</fpage>
<lpage>524</lpage>
<pub-id pub-id-type="doi">10.1038/s41586-023-05947-3</pub-id>
</element-citation>
<mixed-citation>He B, Kwan AC, Cho JH, Yuan N, Pollick C, Shiota T, et al. Blinded, Randomized Trial of Sonographer Versus AI Cardiac Function Assessment. Nature. 2023;616(7957):520-4. doi: 10.1038/s41586-023-05947-3.</mixed-citation>
</ref>
<ref id="B10">
<label>10</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Papadopoulou</surname><given-names>SL</given-names></name>
<name><surname>Sachpekidis</surname><given-names>V</given-names></name>
<name><surname>Kantartzi</surname><given-names>V</given-names></name>
<name><surname>Styliadis</surname><given-names>I</given-names></name>
<name><surname>Nihoyannopoulos</surname><given-names>P</given-names></name>
</person-group>
<article-title>Clinical Validation of an Artificial Intelligence-Assisted Algorithm for Automated Quantification of Left Ventricular Ejection Fraction in Real Time by a Novel Handheld Ultrasound Device</article-title>
<source>Eur Heart J Digit Health</source>
<year>2022</year>
<volume>3</volume>
<issue>1</issue>
<fpage>29</fpage>
<lpage>37</lpage>
<pub-id pub-id-type="doi">10.1093/ehjdh/ztac001</pub-id>
</element-citation>
<mixed-citation>Papadopoulou SL, Sachpekidis V, Kantartzi V, Styliadis I, Nihoyannopoulos P. Clinical Validation of an Artificial Intelligence-Assisted Algorithm for Automated Quantification of Left Ventricular Ejection Fraction in Real Time by a Novel Handheld Ultrasound Device. Eur Heart J Digit Health. 2022;3(1):29-37. doi: 10.1093/ehjdh/ztac001.</mixed-citation>
</ref>
<ref id="B11">
<label>11</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Salte</surname><given-names>IM</given-names></name>
<name><surname>&#xD8;stvik</surname><given-names>A</given-names></name>
<name><surname>Smistad</surname><given-names>E</given-names></name>
<name><surname>Melichova</surname><given-names>D</given-names></name>
<name><surname>Nguyen</surname><given-names>TM</given-names></name>
<name><surname>Karlsen</surname><given-names>S</given-names></name>
<etal/>
</person-group>
<article-title>Artificial Intelligence for Automatic Measurement of Left Ventricular Strain in Echocardiography</article-title>
<source>JACC Cardiovasc Imaging</source>
<year>2021</year>
<volume>14</volume>
<issue>10</issue>
<fpage>1918</fpage>
<lpage>1928</lpage>
<pub-id pub-id-type="doi">10.1016/j.jcmg.2021.04.018</pub-id>
</element-citation>
<mixed-citation>Salte IM, &#xD8;stvik A, Smistad E, Melichova D, Nguyen TM, Karlsen S, et al. Artificial Intelligence for Automatic Measurement of Left Ventricular Strain in Echocardiography. JACC Cardiovasc Imaging. 2021;14(10):1918-28. doi: 10.1016/j.jcmg.2021.04.018.</mixed-citation>
</ref>
<ref id="B12">
<label>12</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Penso</surname><given-names>M</given-names></name>
<name><surname>Ranalletta</surname><given-names>RA</given-names></name>
<name><surname>Pepi</surname><given-names>M</given-names></name>
<name><surname>Garlasch&#xE8;</surname><given-names>A</given-names></name>
<name><surname>Ali</surname><given-names>SG</given-names></name>
<name><surname>Fusini</surname><given-names>L</given-names></name>
<etal/>
</person-group>
<article-title>Comparison Between Automatic and Semiautomatic System for the 3D Echocardiographic Multiparametric Evaluation of RV Function and Dimension</article-title>
<source>J Clin Med</source>
<year>2022</year>
<volume>11</volume>
<issue>15</issue>
<fpage>4528</fpage>
<lpage>4528</lpage>
<pub-id pub-id-type="doi">10.3390/jcm11154528</pub-id>
</element-citation>
<mixed-citation>Penso M, Ranalletta RA, Pepi M, Garlasch&#xE8; A, Ali SG, Fusini L, et al. Comparison Between Automatic and Semiautomatic System for the 3D Echocardiographic Multiparametric Evaluation of RV Function and Dimension. J Clin Med. 2022;11(15):4528. doi: 10.3390/jcm11154528.</mixed-citation>
</ref>
<ref id="B13">
<label>13</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Asch</surname><given-names>FM</given-names></name>
<name><surname>Mor-Avi</surname><given-names>V</given-names></name>
<name><surname>Rubenson</surname><given-names>D</given-names></name>
<name><surname>Goldstein</surname><given-names>S</given-names></name>
<name><surname>Saric</surname><given-names>M</given-names></name>
<name><surname>Mikati</surname><given-names>I</given-names></name>
<etal/>
</person-group>
<article-title>Deep Learning-Based Automated Echocardiographic Quantification of Left Ventricular Ejection Fraction: A Point-of-Care Solution</article-title>
<source>Circ Cardiovasc Imaging</source>
<year>2021</year>
<volume>14</volume>
<issue>6</issue>
<elocation-id>e012293</elocation-id>
<pub-id pub-id-type="doi">10.1161/CIRCIMAGING.120.012293</pub-id>
</element-citation>
<mixed-citation>Asch FM, Mor-Avi V, Rubenson D, Goldstein S, Saric M, Mikati I, et al. Deep Learning-Based Automated Echocardiographic Quantification of Left Ventricular Ejection Fraction: A Point-of-Care Solution. Circ Cardiovasc Imaging. 2021;14(6):e012293. doi: 10.1161/CIRCIMAGING.120.012293.</mixed-citation>
</ref>
<ref id="B14">
<label>14</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Sveric</surname><given-names>KM</given-names></name>
<name><surname>Botan</surname><given-names>R</given-names></name>
<name><surname>Dindane</surname><given-names>Z</given-names></name>
<name><surname>Winkler</surname><given-names>A</given-names></name>
<name><surname>Nowack</surname><given-names>T</given-names></name>
<name><surname>Heitmann</surname><given-names>C</given-names></name>
<etal/>
</person-group>
<article-title>Single-Site Experience with an Automated Artificial Intelligence Application for Left Ventricular Ejection Fraction Measurement in Echocardiography</article-title>
<source>Diagnostics (Basel)</source>
<year>2023</year>
<volume>13</volume>
<issue>7</issue>
<fpage>1298</fpage>
<lpage>1298</lpage>
<pub-id pub-id-type="doi">10.3390/diagnostics13071298</pub-id>
</element-citation>
<mixed-citation>Sveric KM, Botan R, Dindane Z, Winkler A, Nowack T, Heitmann C, et al. Single-Site Experience with an Automated Artificial Intelligence Application for Left Ventricular Ejection Fraction Measurement in Echocardiography. Diagnostics (Basel). 2023;13(7):1298. doi: 10.3390/diagnostics13071298.</mixed-citation>
</ref>
<ref id="B15">
<label>15</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Wang</surname><given-names>TKM</given-names></name>
<name><surname>Cremer</surname><given-names>PC</given-names></name>
<name><surname>Chan</surname><given-names>N</given-names></name>
<name><surname>Piotrowska</surname><given-names>H</given-names></name>
<name><surname>Woodward</surname><given-names>G</given-names></name>
<name><surname>Jaber</surname><given-names>WA</given-names></name>
</person-group>
<article-title>Utility of an Automated Artificial Intelligence Echocardiography Software in Risk Stratification of Hospitalized COVID-19 Patients</article-title>
<source>Life (Basel)</source>
<year>2022</year>
<volume>12</volume>
<issue>9</issue>
<fpage>1413</fpage>
<lpage>1413</lpage>
<pub-id pub-id-type="doi">10.3390/life12091413</pub-id>
</element-citation>
<mixed-citation>Wang TKM, Cremer PC, Chan N, Piotrowska H, Woodward G, Jaber WA. Utility of an Automated Artificial Intelligence Echocardiography Software in Risk Stratification of Hospitalized COVID-19 Patients. Life (Basel). 2022;12(9):1413. doi: 10.3390/life12091413.</mixed-citation>
</ref>
<ref id="B16">
<label>16</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Liel-Cohen</surname><given-names>N</given-names></name>
<name><surname>Kobal</surname><given-names>S</given-names></name>
<name><surname>Butnaru</surname><given-names>A</given-names></name>
<name><surname>Alper-Suissa</surname><given-names>L</given-names></name>
<name><surname>Feinstein</surname><given-names>S</given-names></name>
</person-group>
<article-title>A Novel Artificial Intelligence-Based System Provides Global Longitudinal Strain Automatically Behind the Scenes</article-title>
<source>Eur Heart J Cardiovasc Imaging</source>
<year>2022</year>
<volume>23</volume>
<issue>1</issue>
<fpage>i8</fpage>
<lpage>i9</lpage>
<pub-id pub-id-type="doi">10.1093/ehjci/jeab289.004</pub-id>
</element-citation>
<mixed-citation>Liel-Cohen N, Kobal S, Butnaru A, Alper-Suissa L, Feinstein S. A Novel Artificial Intelligence-Based System Provides Global Longitudinal Strain Automatically Behind the Scenes. Eur Heart J Cardiovasc Imaging. 2022;23(1):i8-i9. doi: 10.1093/ehjci/jeab289.004.</mixed-citation>
</ref>
<ref id="B17">
<label>17</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Di Candia</surname><given-names>A</given-names></name>
<name><surname>Castaldi</surname><given-names>B</given-names></name>
<name><surname>Cati</surname><given-names>A</given-names></name>
<name><surname>Basso</surname><given-names>A</given-names></name>
<name><surname>Fumanelli</surname><given-names>J</given-names></name>
<name><surname>Sabatino</surname><given-names>J</given-names></name>
<etal/>
</person-group>
<article-title>AI do it Better. Reproducibility of Fully Automated Versus Manual Analysis of Left Ventricle Function Using a New Software</article-title>
<source>Eur Heart J Cardiovasc Imaging</source>
<year>2022</year>
<volume>23</volume>
<issue>1</issue>
<fpage>i252</fpage>
<lpage>i253</lpage>
<pub-id pub-id-type="doi">10.1093/ehjci/jeab289.183</pub-id>
</element-citation>
<mixed-citation>Di Candia A, Castaldi B, Cati A, Basso A, Fumanelli J, Sabatino J, et al. AI do it Better. Reproducibility of Fully Automated Versus Manual Analysis of Left Ventricle Function Using a New Software. Eur Heart J Cardiovasc Imaging. 2022;23(1):i252-i3. doi: 10.1093/ehjci/jeab289.183.</mixed-citation>
</ref>
<ref id="B18">
<label>18</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Popoff</surname><given-names>A</given-names></name>
<name><surname>Langet</surname><given-names>H</given-names></name>
<name><surname>Piro</surname><given-names>P</given-names></name>
<name><surname>Ropert</surname><given-names>C</given-names></name>
<name><surname>Allain</surname><given-names>P</given-names></name>
<name><surname>Gauriau</surname><given-names>R</given-names></name>
<etal/>
</person-group>
<article-title>Can we Improve the Accuracy and Reproducibility of Left Ventricular Ejection Fraction from 2D Echocardiography Using Artificial Intelligence? A Validation Against Cardiac Magnetic resonance</article-title>
<source>Eur Heart J Cardiovasc Imaging</source>
<year>2020</year>
<volume>21</volume>
<issue>1</issue>
<fpage>I34</fpage>
<lpage>I35</lpage>
<pub-id pub-id-type="doi">10.1093/ehjci/jez319.032</pub-id>
</element-citation>
<mixed-citation>Popoff A, Langet H, Piro P, Ropert C, Allain P, Gauriau R, et al. Can we Improve the Accuracy and Reproducibility of Left Ventricular Ejection Fraction from 2D Echocardiography Using Artificial Intelligence? A Validation Against Cardiac Magnetic resonance. Eur Heart J Cardiovasc Imaging. 2020;21(1):I34-I5. doi: 10.1093/ehjci/jez319.032.</mixed-citation>
</ref>
<ref id="B19">
<label>19</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Varudo</surname><given-names>R</given-names></name>
<name><surname>Gonzalez</surname><given-names>FA</given-names></name>
<name><surname>Leote</surname><given-names>J</given-names></name>
<name><surname>Martins</surname><given-names>C</given-names></name>
<name><surname>Bacariza</surname><given-names>J</given-names></name>
<name><surname>Fernandes</surname><given-names>A</given-names></name>
<etal/>
</person-group>
<article-title>Machine Learning for the Real-Time Assessment of Left Ventricular Ejection Fraction in Critically Ill Patients: A Bedside Evaluation by Novices and Experts in Echocardiography</article-title>
<source>Crit Care</source>
<year>2022</year>
<volume>26</volume>
<issue>1</issue>
<fpage>386</fpage>
<lpage>386</lpage>
<pub-id pub-id-type="doi">10.1186/s13054-022-04269-6</pub-id>
</element-citation>
<mixed-citation>Varudo R, Gonzalez FA, Leote J, Martins C, Bacariza J, Fernandes A, et al. Machine Learning for the Real-Time Assessment of Left Ventricular Ejection Fraction in Critically Ill Patients: A Bedside Evaluation by Novices and Experts in Echocardiography. Crit Care. 2022;26(1):386. doi: 10.1186/s13054-022-04269-6.</mixed-citation>
</ref>
<ref id="B20">
<label>20</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Burke</surname><given-names>D</given-names></name>
<name><surname>Corrigan</surname><given-names>N</given-names></name>
<name><surname>Herlihy</surname><given-names>M</given-names></name>
<name><surname>Nasaj</surname><given-names>O</given-names></name>
<name><surname>Dickson</surname><given-names>J</given-names></name>
<name><surname>Delaney</surname><given-names>D</given-names></name>
<etal/>
</person-group>
<article-title>Real World Evaluation of Artificial Intelligence Echocardiography Image Guidance and Acquisition with Novice Scanners in Multiple Clinical Settings</article-title>
<source>Eur Heart J Cardiovasc Imaging</source>
<year>2022</year>
<volume>23</volume>
<issue>1</issue>
<fpage>i20</fpage>
<lpage>i20</lpage>
<pub-id pub-id-type="doi">10.1093/ehjci/jeab289.011</pub-id>
</element-citation>
<mixed-citation>Burke D, Corrigan N, Herlihy M, Nasaj O, Dickson J, Delaney D, et al. Real World Evaluation of Artificial Intelligence Echocardiography Image Guidance and Acquisition with Novice Scanners in Multiple Clinical Settings. Eur Heart J Cardiovasc Imaging. 2022;23(1):i20. 10.1093/ehjci/jeab289.011.</mixed-citation>
</ref>
<ref id="B21">
<label>21</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Narang</surname><given-names>A</given-names></name>
<name><surname>Bae</surname><given-names>R</given-names></name>
<name><surname>Hong</surname><given-names>H</given-names></name>
<name><surname>Thomas</surname><given-names>Y</given-names></name>
<name><surname>Surette</surname><given-names>S</given-names></name>
<name><surname>Cadieu</surname><given-names>C</given-names></name>
<etal/>
</person-group>
<article-title>Utility of a Deep-Learning Algorithm to Guide Novices to Acquire Echocardiograms for Limited Diagnostic Use</article-title>
<source>JAMA Cardiol</source>
<year>2021</year>
<volume>6</volume>
<issue>6</issue>
<fpage>624</fpage>
<lpage>632</lpage>
<pub-id pub-id-type="doi">10.1001/jamacardio.2021.0185</pub-id>
</element-citation>
<mixed-citation>Narang A, Bae R, Hong H, Thomas Y, Surette S, Cadieu C, et al. Utility of a Deep-Learning Algorithm to Guide Novices to Acquire Echocardiograms for Limited Diagnostic Use. JAMA Cardiol. 2021;6(6):624-32. doi: 10.1001/jamacardio.2021.0185.</mixed-citation>
</ref>
<ref id="B22">
<label>22</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Surette</surname><given-names>S</given-names></name>
<name><surname>Narang</surname><given-names>A</given-names></name>
<name><surname>Bae</surname><given-names>R</given-names></name>
<name><surname>Hong</surname><given-names>H</given-names></name>
<name><surname>Thomas</surname><given-names>Y</given-names></name>
<name><surname>Cadieu</surname><given-names>C</given-names></name>
<etal/>
</person-group>
<article-title>Artificial Intelligence-Guided Image Acquisition on Patients with Implanted Electrophysiological Devices: Results from a Pivotal Prospective Multi-Center Clinical Trial</article-title>
<source>Eur Heart J</source>
<year>2020</year>
<volume>41</volume>
<issue>2</issue>
<fpage>6</fpage>
<lpage>6</lpage>
<pub-id pub-id-type="doi">10.1093/ehjci/ehaa946.0006</pub-id>
</element-citation>
<mixed-citation>Surette S, Narang A, Bae R, Hong H, Thomas Y, Cadieu C, et al. Artificial Intelligence-Guided Image Acquisition on Patients with Implanted Electrophysiological Devices: Results from a Pivotal Prospective Multi-Center Clinical Trial. Eur Heart J. 2020;41(2):6. doi: 10.1093/ehjci/ehaa946.0006.</mixed-citation>
</ref>
<ref id="B23">
<label>23</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Schneider</surname><given-names>M</given-names></name>
<name><surname>Bartko</surname><given-names>P</given-names></name>
<name><surname>Geller</surname><given-names>W</given-names></name>
<name><surname>Dannenberg</surname><given-names>V</given-names></name>
<name><surname>K&#xF6;nig</surname><given-names>A</given-names></name>
<name><surname>Binder</surname><given-names>C</given-names></name>
<etal/>
</person-group>
<article-title>A machine Learning Algorithm Supports Ultrasound-Na&#xEF;ve Novices in the Acquisition of Diagnostic Echocardiography Loops and Provides Accurate Estimation of LVEF</article-title>
<source>Int J Cardiovasc Imaging</source>
<year>2021</year>
<volume>37</volume>
<issue>2</issue>
<fpage>577</fpage>
<lpage>586</lpage>
<pub-id pub-id-type="doi">10.1007/s10554-020-02046-6</pub-id>
</element-citation>
<mixed-citation>Schneider M, Bartko P, Geller W, Dannenberg V, K&#xF6;nig A, Binder C, et al. A machine Learning Algorithm Supports Ultrasound-Na&#xEF;ve Novices in the Acquisition of Diagnostic Echocardiography Loops and Provides Accurate Estimation of LVEF. Int J Cardiovasc Imaging. 2021;37(2):577-86. doi: 10.1007/s10554-020-02046-6.</mixed-citation>
</ref>
<ref id="B24">
<label>24</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Motazedian</surname><given-names>P</given-names></name>
<name><surname>Marbach</surname><given-names>J</given-names></name>
<name><surname>Prosperi-Porta</surname><given-names>G</given-names></name>
<name><surname>Parlow</surname><given-names>S</given-names></name>
<name><surname>Di Santo</surname><given-names>P</given-names></name>
<name><surname>Jung</surname><given-names>R</given-names></name>
<etal/>
</person-group>
<article-title>Point-of-Care Ultrasonography: Artificial Intelligence in the Assessment of Left Ventricular Function</article-title>
<source>Circulation</source>
<year>2022</year>
<volume>146</volume>
<issue>1</issue>
<fpage>15710</fpage>
<lpage>15710</lpage>
<pub-id pub-id-type="doi">10.1161/circ.146.suppl_1.15710Circulation</pub-id>
</element-citation>
<mixed-citation>Motazedian P, Marbach J, Prosperi-Porta G, Parlow S, Di Santo P, Jung R, et al. Point-of-Care Ultrasonography: Artificial Intelligence in the Assessment of Left Ventricular Function. Circulation. 2022;146(1):15710. doi: 10.1161/circ.146.suppl_1.15710Circulation.</mixed-citation>
</ref>
<ref id="B25">
<label>25</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Peck</surname><given-names>D</given-names></name>
<name><surname>Rwebembera</surname><given-names>J</given-names></name>
<name><surname>Nakagaayi</surname><given-names>D</given-names></name>
<name><surname>Minja</surname><given-names>NW</given-names></name>
<name><surname>Ollberding</surname><given-names>NJ</given-names></name>
<name><surname>Pulle</surname><given-names>J</given-names></name>
<etal/>
</person-group>
<article-title>The Use of Artificial Intelligence Guidance for Rheumatic Heart Disease Screening by Novices</article-title>
<source>J Am Soc Echocardiogr</source>
<year>2023</year>
<volume>36</volume>
<issue>7</issue>
<fpage>724</fpage>
<lpage>732</lpage>
<pub-id pub-id-type="doi">10.1016/j.echo.2023.03.001</pub-id>
</element-citation>
<mixed-citation>Peck D, Rwebembera J, Nakagaayi D, Minja NW, Ollberding NJ, Pulle J, et al. The Use of Artificial Intelligence Guidance for Rheumatic Heart Disease Screening by Novices. J Am Soc Echocardiogr. 2023;36(7):724-32. doi: 10.1016/j.echo.2023.03.001.</mixed-citation>
</ref>
<ref id="B26">
<label>26</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Zhai</surname><given-names>S</given-names></name>
<name><surname>Wang</surname><given-names>H</given-names></name>
<name><surname>Sun</surname><given-names>L</given-names></name>
<name><surname>Zhang</surname><given-names>B</given-names></name>
<name><surname>Huo</surname><given-names>F</given-names></name>
<name><surname>Qiu</surname><given-names>S</given-names></name>
<etal/>
</person-group>
<article-title>Artificial Intelligence (AI) Versus Expert: A Comparison of Left Ventricular Outflow Tract Velocity Time Integral (LVOT-VTI) Assessment Between ICU Doctors and an AI Tool</article-title>
<source>J Appl Clin Med Phys</source>
<year>2022</year>
<volume>23</volume>
<issue>8</issue>
<elocation-id>e13724</elocation-id>
<pub-id pub-id-type="doi">10.1002/acm2.13724</pub-id>
</element-citation>
<mixed-citation>Zhai S, Wang H, Sun L, Zhang B, Huo F, Qiu S, et al. Artificial Intelligence (AI) Versus Expert: A Comparison of Left Ventricular Outflow Tract Velocity Time Integral (LVOT-VTI) Assessment Between ICU Doctors and an AI Tool. J Appl Clin Med Phys. 2022;23(8):e13724. doi: 10.1002/acm2.13724.</mixed-citation>
</ref>
<ref id="B27">
<label>27</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Dadon</surname><given-names>Z</given-names></name>
<name><surname>Butnaru</surname><given-names>A</given-names></name>
<name><surname>Rosenmann</surname><given-names>D</given-names></name>
<name><surname>Alper-Suissa</surname><given-names>L</given-names></name>
<name><surname>Glikson</surname><given-names>M</given-names></name>
<name><surname>Alpert</surname><given-names>EA</given-names></name>
</person-group>
<article-title>Use of Artificial Intelligence as a Didactic Tool to Improve Ejection Fraction Assessment in the Emergency Department: A Randomized Controlled Pilot Study</article-title>
<source>AEM Educ Train</source>
<year>2022</year>
<volume>6</volume>
<issue>2</issue>
<elocation-id>e10738</elocation-id>
<pub-id pub-id-type="doi">10.1002/aet2.10738</pub-id>
</element-citation>
<mixed-citation>Dadon Z, Butnaru A, Rosenmann D, Alper-Suissa L, Glikson M, Alpert EA. Use of Artificial Intelligence as a Didactic Tool to Improve Ejection Fraction Assessment in the Emergency Department: A Randomized Controlled Pilot Study. AEM Educ Train. 2022;6(2):e10738. doi: 10.1002/aet2.10738.</mixed-citation>
</ref>
<ref id="B28">
<label>28</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>O&#x27;Driscoll</surname><given-names>JM</given-names></name>
<name><surname>Hawkes</surname><given-names>W</given-names></name>
<name><surname>Beqiri</surname><given-names>A</given-names></name>
<name><surname>Mumith</surname><given-names>A</given-names></name>
<name><surname>Parker</surname><given-names>A</given-names></name>
<name><surname>Upton</surname><given-names>R</given-names></name>
<etal/>
</person-group>
<article-title>Left Ventricular Assessment with Artificial Intelligence Increases the Diagnostic Accuracy of Stress Echocardiography</article-title>
<source>Eur Heart J Open</source>
<year>2022</year>
<volume>2</volume>
<issue>5</issue>
<comment>oeac059</comment>
<pub-id pub-id-type="doi">10.1093/ehjopen/oeac059</pub-id>
</element-citation>
<mixed-citation>O&#x27;Driscoll JM, Hawkes W, Beqiri A, Mumith A, Parker A, Upton R, et al. Left Ventricular Assessment with Artificial Intelligence Increases the Diagnostic Accuracy of Stress Echocardiography. Eur Heart J Open. 2022;2(5):oeac059. doi: 10.1093/ehjopen/oeac059.</mixed-citation>
</ref>
<ref id="B29">
<label>29</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Kusunose</surname><given-names>K</given-names></name>
<name><surname>Abe</surname><given-names>T</given-names></name>
<name><surname>Haga</surname><given-names>A</given-names></name>
<name><surname>Fukuda</surname><given-names>D</given-names></name>
<name><surname>Yamada</surname><given-names>H</given-names></name>
<name><surname>Harada</surname><given-names>M</given-names></name>
<etal/>
</person-group>
<article-title>A Deep Learning Approach for Assessment of Regional Wall Motion Abnormality from Echocardiographic Images</article-title>
<source>JACC Cardiovasc Imaging</source>
<year>2020</year>
<volume>13</volume>
<issue>2 Pt 1</issue>
<fpage>374</fpage>
<lpage>381</lpage>
<pub-id pub-id-type="doi">10.1016/j.jcmg.2019.02.024</pub-id>
</element-citation>
<mixed-citation>Kusunose K, Abe T, Haga A, Fukuda D, Yamada H, Harada M, et al. A Deep Learning Approach for Assessment of Regional Wall Motion Abnormality from Echocardiographic Images. JACC Cardiovasc Imaging. 2020;13(2 Pt 1):374-81. doi: 10.1016/j.jcmg.2019.02.024.</mixed-citation>
</ref>
<ref id="B30">
<label>30</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Stowell</surname><given-names>C</given-names></name>
<name><surname>Howard</surname><given-names>J</given-names></name>
<name><surname>Demetrescu</surname><given-names>C</given-names></name>
<name><surname>Bhattacharyya</surname><given-names>S</given-names></name>
<name><surname>Mangion</surname><given-names>K</given-names></name>
<name><surname>Vimalesvaran</surname><given-names>K</given-names></name>
<etal/>
</person-group>
<article-title>Fully Automated Global Longitudinal Strain Assessment Using Artificial Intelligence Developed and Validated by a UK-wide Echocardiography Expert Collaborative</article-title>
<source>Eur Heart J</source>
<year>2021</year>
<volume>42</volume>
<issue>1</issue>
<fpage>2</fpage>
<lpage>2</lpage>
<pub-id pub-id-type="doi">10.1161/CIRCIMAGING.120.011951</pub-id>
</element-citation>
<mixed-citation>Stowell C, Howard J, Demetrescu C, Bhattacharyya S, Mangion K, Vimalesvaran K, et al. Fully Automated Global Longitudinal Strain Assessment Using Artificial Intelligence Developed and Validated by a UK-wide Echocardiography Expert Collaborative. Eur Heart J. 2021;42(1):2. doi: 10.1161/CIRCIMAGING.120.011951.</mixed-citation>
</ref>
<ref id="B31">
<label>31</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Tromp</surname><given-names>J</given-names></name>
<name><surname>Seekings</surname><given-names>PJ</given-names></name>
<name><surname>Hung</surname><given-names>CL</given-names></name>
<name><surname>Iversen</surname><given-names>MB</given-names></name>
<name><surname>Frost</surname><given-names>MJ</given-names></name>
<name><surname>Ouwerkerk</surname><given-names>W</given-names></name>
<etal/>
</person-group>
<article-title>Automated Interpretation of Systolic and Diastolic Function on the Echocardiogram: A Multicohort Study</article-title>
<source>Lancet Digit Health</source>
<year>2022</year>
<volume>4</volume>
<issue>1</issue>
<fpage>e46</fpage>
<lpage>e54</lpage>
<pub-id pub-id-type="doi">10.1016/S2589-7500(21)00235-1</pub-id>
</element-citation>
<mixed-citation>Tromp J, Seekings PJ, Hung CL, Iversen MB, Frost MJ, Ouwerkerk W, et al. Automated Interpretation of Systolic and Diastolic Function on the Echocardiogram: A Multicohort Study. Lancet Digit Health. 2022;4(1):e46-e54. doi: 10.1016/S2589-7500(21)00235-1.</mixed-citation>
</ref>
<ref id="B32">
<label>32</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Asch</surname><given-names>FM</given-names></name>
<name><surname>Poilvert</surname><given-names>N</given-names></name>
<name><surname>Abraham</surname><given-names>T</given-names></name>
<name><surname>Jankowski</surname><given-names>M</given-names></name>
<name><surname>Cleve</surname><given-names>J</given-names></name>
<name><surname>Adams</surname><given-names>M</given-names></name>
<etal/>
</person-group>
<article-title>Automated Echocardiographic Quantification of Left Ventricular Ejection Fraction Without Volume Measurements Using a Machine Learning Algorithm Mimicking a Human Expert</article-title>
<source>Circ Cardiovasc Imaging</source>
<year>2019</year>
<volume>12</volume>
<issue>9</issue>
<elocation-id>e009303</elocation-id>
<pub-id pub-id-type="doi">10.1161/CIRCIMAGING.119.009303</pub-id>
</element-citation>
<mixed-citation>Asch FM, Poilvert N, Abraham T, Jankowski M, Cleve J, Adams M, et al. Automated Echocardiographic Quantification of Left Ventricular Ejection Fraction Without Volume Measurements Using a Machine Learning Algorithm Mimicking a Human Expert. Circ Cardiovasc Imaging. 2019;12(9):e009303. doi: 10.1161/CIRCIMAGING.119.009303.</mixed-citation>
</ref>
<ref id="B33">
<label>33</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Hanif</surname><given-names>W</given-names></name>
<name><surname>Goldberg</surname><given-names>Y</given-names></name>
<name><surname>Taub</surname><given-names>CC</given-names></name>
<name><surname>Vorchheimer</surname><given-names>DA</given-names></name>
<name><surname>Slipczuk</surname><given-names>L</given-names></name>
<name><surname>Ho</surname><given-names>E</given-names></name>
<etal/>
</person-group>
<article-title>Automated Measurement of Global Longitudinal Strain by Speckle-Tracking Echocardiography in Cardio-Oncology Patients Using Artificial Intelligence</article-title>
<source>Circulation</source>
<year>2021</year>
<volume>144</volume>
<issue>1</issue>
<fpage>A11383</fpage>
<lpage>A11383</lpage>
<pub-id pub-id-type="doi">10.1161/circ.144.suppl_1.11383</pub-id>
</element-citation>
<mixed-citation>Hanif W, Goldberg Y, Taub CC, Vorchheimer DA, Slipczuk L, Ho E, et al. Automated Measurement of Global Longitudinal Strain by Speckle-Tracking Echocardiography in Cardio-Oncology Patients Using Artificial Intelligence. Circulation. 2021;144(1):A11383. doi: 10.1161/circ.144.suppl_1.11383.</mixed-citation>
</ref>
<ref id="B34">
<label>34</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Fung</surname><given-names>A</given-names></name>
<name><surname>Moulson</surname><given-names>N</given-names></name>
<name><surname>Balthazaar</surname><given-names>S</given-names></name>
<name><surname>Girgis</surname><given-names>H</given-names></name>
<name><surname>van Woudenberg</surname><given-names>N</given-names></name>
<name><surname>Abolmaesumi</surname><given-names>P</given-names></name>
<etal/>
</person-group>
<article-title>Can Artificial Intelligence Assess Image Quality in Point-Of-Care Ultrassound?</article-title>
<source>Canadian Journal of Cardiology</source>
<year>2019</year>
<volume>35</volume>
<issue>10</issue>
<fpage>146</fpage>
<lpage>146</lpage>
<pub-id pub-id-type="doi">10.1016/j.cjca.2019.07.566</pub-id>
</element-citation>
<mixed-citation>Fung A, Moulson N, Balthazaar S, Girgis H, van Woudenberg N, Abolmaesumi P, et al. Can Artificial Intelligence Assess Image Quality in Point-Of-Care Ultrassound? Canadian Journal of Cardiology. 2019;35(10):146. doi: 10.1016/j.cjca.2019.07.566.</mixed-citation>
</ref>
<ref id="B35">
<label>35</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Kusunose</surname><given-names>K</given-names></name>
<name><surname>Haga</surname><given-names>A</given-names></name>
<name><surname>Abe</surname><given-names>T</given-names></name>
<name><surname>Sata</surname><given-names>M</given-names></name>
<etal/>
</person-group>
<article-title>Utilization of Artificial Intelligence in Echocardiography</article-title>
<source>Circ J</source>
<year>2019</year>
<volume>83</volume>
<issue>8</issue>
<fpage>1623</fpage>
<lpage>1629</lpage>
<pub-id pub-id-type="doi">10.1253/circj.CJ-19-0420</pub-id>
</element-citation>
<mixed-citation>Kusunose K, Haga A, Abe T, Sata M. Utilization of Artificial Intelligence in Echocardiography. Circ J. 2019;83(8):1623-9. doi: 10.1253/circj.CJ-19-0420.</mixed-citation>
</ref>
<ref id="B36">
<label>36</label>
<element-citation publication-type="journal">
<person-group person-group-type="author">
<name><surname>Asch</surname><given-names>FM</given-names></name>
<name><surname>Descamps</surname><given-names>T</given-names></name>
<name><surname>Sarwar</surname><given-names>R</given-names></name>
<name><surname>Karagodin</surname><given-names>I</given-names></name>
<name><surname>Singulane</surname><given-names>CC</given-names></name>
<name><surname>Xie</surname><given-names>M</given-names></name>
<etal/>
</person-group>
<article-title>Human Versus Artificial Intelligence-Based Echocardiographic Analysis as a Predictor of Outcomes: An Analysis from the World Alliance Societies of Echocardiography COVID Study</article-title>
<source>J Am Soc Echocardiogr</source>
<year>2022</year>
<volume>35</volume>
<issue>12</issue>
<fpage>1226.e7</fpage>
<lpage>1237.e7</lpage>
<pub-id pub-id-type="doi">10.1016/j.echo.2022.07.004</pub-id>
</element-citation>
<mixed-citation>Asch FM, Descamps T, Sarwar R, Karagodin I, Singulane CC, Xie M, et al. Human Versus Artificial Intelligence-Based Echocardiographic Analysis as a Predictor of Outcomes: An Analysis from the World Alliance Societies of Echocardiography COVID Study. J Am Soc Echocardiogr. 2022;35(12):1226-37.e7. doi: 10.1016/j.echo.2022.07.004.</mixed-citation>
</ref>
</ref-list>
</back>
<sub-article article-type="translation" id="S01" xml:lang="pt">
<front-stub>
<article-id pub-id-type="doi">10.36660/abcimg.20230085</article-id>
<article-categories>
<subj-group subj-group-type="heading">
<subject>Artigo Original</subject>
</subj-group>
</article-categories>
<title-group>
<article-title>Uso da Intelig&#xEA;ncia Artificial para Avalia&#xE7;&#xE3;o da Fun&#xE7;&#xE3;o Card&#xED;aca pela Ecocardiografia: Revis&#xE3;o Sistem&#xE1;tica do Estado da Arte</article-title>
</title-group>
<contrib-group>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0000-0003-0988-5282</contrib-id>
<name><surname>Neves</surname><given-names>Henrique Alexsander Ferreira</given-names></name>
<role>Concep&#xE7;&#xE3;o e desenho da pesquisa</role>
<role>obten&#xE7;&#xE3;o de dados e an&#xE1;lise e interpreta&#xE7;&#xE3;o dos dados</role>
<role>reda&#xE7;&#xE3;o do manuscrito</role>
<xref ref-type="aff" rid="aff2"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0000-0001-9735-4590</contrib-id>
<name><surname>Yuasa</surname><given-names>Bruna Sadae</given-names></name>
<role>Concep&#xE7;&#xE3;o e desenho da pesquisa</role>
<role>obten&#xE7;&#xE3;o de dados e an&#xE1;lise e interpreta&#xE7;&#xE3;o dos dados</role>
<role>reda&#xE7;&#xE3;o do manuscrito</role>
<xref ref-type="aff" rid="aff2"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0009-0003-8756-8116</contrib-id>
<name><surname>Costa</surname><given-names>Thamires Hadassa Leite Pereira</given-names></name>
<role>Concep&#xE7;&#xE3;o e desenho da pesquisa</role>
<role>obten&#xE7;&#xE3;o de dados e an&#xE1;lise e interpreta&#xE7;&#xE3;o dos dados</role>
<role>reda&#xE7;&#xE3;o do manuscrito</role>
<xref ref-type="aff" rid="aff2"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0009-0009-1438-2947</contrib-id>
<name><surname>Santos</surname><given-names>Isabela Ertes</given-names></name>
<role>Concep&#xE7;&#xE3;o e desenho da pesquisa</role>
<role>obten&#xE7;&#xE3;o de dados e an&#xE1;lise e interpreta&#xE7;&#xE3;o dos dados</role>
<role>reda&#xE7;&#xE3;o do manuscrito</role>
<xref ref-type="aff" rid="aff2"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0009-0001-3556-5218</contrib-id>
<name><surname>Benavides</surname><given-names>Yannir&#xEA; Milagros Roman</given-names></name>
<role>Concep&#xE7;&#xE3;o e desenho da pesquisa</role>
<role>obten&#xE7;&#xE3;o de dados e an&#xE1;lise e interpreta&#xE7;&#xE3;o dos dados</role>
<role>reda&#xE7;&#xE3;o do manuscrito</role>
<xref ref-type="aff" rid="aff2"><sup>1</sup></xref>
</contrib>
<contrib contrib-type="author">
<contrib-id contrib-id-type="orcid">0000-0003-0794-7339</contrib-id>
<name><surname>Lofrano-Alves</surname><given-names>Marco Stephan</given-names></name>
<role>Concep&#xE7;&#xE3;o e desenho da pesquisa</role>
<role>obten&#xE7;&#xE3;o de dados e an&#xE1;lise e interpreta&#xE7;&#xE3;o dos dados</role>
<role>revis&#xE3;o cr&#xED;tica do manuscrito quanto ao conte&#xFA;do intelectual importante</role>
<xref ref-type="aff" rid="aff2"><sup>1</sup></xref>
<xref ref-type="corresp" rid="c2"/>
</contrib>
<aff id="aff2">
<label>1</label>
<addr-line>
<named-content content-type="city">Curitiba</named-content>
<named-content content-type="state">PR</named-content>
</addr-line>
<country country="BR">Brasil</country>
<institution content-type="original">Universidade Federal do Paran&#xE1;, Curitiba, PR &#x2013; Brasil</institution>
</aff>
</contrib-group>
<author-notes>
<corresp id="c2"><bold>Correspond&#xEA;ncia: Marco Stephan Lofrano-Alves</bold> &#x2022; Universidade Federal do Paran&#xE1;, Departamento de Cl&#xED;nica M&#xE9;dica. Rua Mateus Leme, 3945, apt 504. CEP: <postal-code>82200-000</postal-code>. Curitiba, PR &#x2013; Brasil. E-mail: <email>mslalves@hotmail.com</email></corresp>
<fn fn-type="conflict">
<p><bold>Potencial Conflito de Interesse</bold></p>
<p>Declaro n&#xE3;o haver conflito de interesses pertinentes.</p></fn>
</author-notes>
<abstract>
<title>Resumo</title>
<sec>
<title>Hist&#xF3;rico:</title>
<p>A ecocardiografia faz parte da investiga&#xE7;&#xE3;o de rotina da maioria das doen&#xE7;as card&#xED;acas. No entanto, por depender do operador e considerando a qualidade da imagem obtida, &#xE9; poss&#xED;vel haver variabilidade significativa nas medi&#xE7;&#xF5;es, o que prejudica a interpreta&#xE7;&#xE3;o. O m&#xE9;todo avan&#xE7;ou no sentido da padroniza&#xE7;&#xE3;o e incorpora&#xE7;&#xE3;o de novas tecnologias com o objetivo de reduzir suas defici&#xEA;ncias. Recentemente, parte desse avan&#xE7;o se deve ao uso da intelig&#xEA;ncia artificial (IA).</p>
</sec>
<sec>
<title>Objetivos:</title>
<p>O objetivo deste trabalho foi realizar uma investiga&#xE7;&#xE3;o sobre o uso da IA na avalia&#xE7;&#xE3;o da fun&#xE7;&#xE3;o card&#xED;aca por meio da ecocardiografia, discutindo seu potencial para aperfei&#xE7;oar o m&#xE9;todo.</p>
</sec>
<sec>
<title>M&#xE9;todos:</title>
<p>Realizou-se uma revis&#xE3;o sistem&#xE1;tica de acordo com o Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) e foram buscados artigos no per&#xED;odo de 2000 a 2023.</p>
</sec>
<sec>
<title>Resultados:</title>
<p>Foram inclu&#xED;dos 31 artigos, classificados em quatro categorias de estudos que utilizaram IA: 1) para an&#xE1;lise de imagens e par&#xE2;metros ecocardiogr&#xE1;ficos (N = 14); 2) no processo de aquisi&#xE7;&#xE3;o de imagens (N = 11); 3) para integra&#xE7;&#xE3;o com dados cl&#xED;nicos (N = 3); e 4) para aplica&#xE7;&#xE3;o ao fluxo do exame (N = 3).</p>
</sec>
<sec>
<title>Conclus&#xE3;o:</title>
<p>A IA proporciona automatiza&#xE7;&#xE3;o e padroniza&#xE7;&#xE3;o das imagens e respectivas medi&#xE7;&#xF5;es que resultam em diagn&#xF3;sticos mais precisos, superando as limita&#xE7;&#xF5;es encontradas no m&#xE9;todo convencional. No entanto, o uso da IA ainda &#xE9; uma realidade em evolu&#xE7;&#xE3;o na ecocardiografia.</p>
</sec>
</abstract>
<kwd-group xml:lang="pt">
<title>Palavras-chave:</title>
<kwd>Ecocardiografia</kwd>
<kwd>Intelig&#xEA;ncia Artificial</kwd>
<kwd>Aprendizado de M&#xE1;quina</kwd>
<kwd>Aprendizado Profundo</kwd>
<kwd>Deforma&#xE7;&#xE3;o Longitudinal Global</kwd>
</kwd-group>
</front-stub>
<body>
<fig id="f4">
<caption>
<title>Fluxograma de busca e sele&#xE7;&#xE3;o de estudos sobre o uso de IA na ecocardiografia. IA: intelig&#xEA;ncia artificial</title></caption>
<graphic xlink:href="2675-312X-dic-36-04-e20230085-gf02-pt.tif"/>
</fig>
<sec sec-type="intro">
<title>Introdu&#xE7;&#xE3;o</title>
<p>A ecocardiografia &#xE9; o m&#xE9;todo de imagem mais solicitado na Cardiologia e faz parte da investiga&#xE7;&#xE3;o de rotina da maioria das doen&#xE7;as card&#xED;acas. Devido &#xE0; sua natureza n&#xE3;o invasiva, n&#xE3;o ionizante, alta disponibilidade e portabilidade, o m&#xE9;todo est&#xE1; presente em v&#xE1;rios cen&#xE1;rios cl&#xED;nicos: ambulatoriais, hospitalares, unidades de urg&#xEA;ncia e emerg&#xEA;ncia, unidades de terapia intensiva e como guia para procedimentos intervencionistas e cir&#xFA;rgicos. Normalmente, &#xE9; a primeira op&#xE7;&#xE3;o para acessar a anatomia e fun&#xE7;&#xE3;o card&#xED;acas. No entanto, a depend&#xEA;ncia do operador e a qualidade da imagem podem influenciar as medi&#xE7;&#xF5;es ecocardiogr&#xE1;ficas. Desde seu desenvolvimento, o m&#xE9;todo tem avan&#xE7;ado no sentido da padroniza&#xE7;&#xE3;o e incorpora&#xE7;&#xE3;o de novas tecnologias com o objetivo de reduzir sua variabilidade.</p>
<p>Este avan&#xE7;o se deve, em parte, ao uso da intelig&#xEA;ncia artificial (IA). &#x201C;IA&#x201D; &#xE9; um termo amplo cunhado por Arthur Samuel na d&#xE9;cada de 1960. Aplica-se a qualquer programa de computador, como algoritmos e modelos, que imitam a l&#xF3;gica e a intelig&#xEA;ncia humanas.<sup><xref ref-type="bibr" rid="B1">1</xref></sup> Na ecocardiografia, o primeiro uso foi na d&#xE9;cada de 1970, quando a an&#xE1;lise de Fourier foi utilizada para avaliar a v&#xE1;lvula mitral com o aux&#xED;lio do modo M.<sup><xref ref-type="bibr" rid="B2">2</xref></sup> As medi&#xE7;&#xF5;es semiautomatizadas j&#xE1; fazem parte da pr&#xE1;tica laboratorial ecocardiogr&#xE1;fica de hoje. Em seu fluxo de trabalho, os dispositivos modernos v&#xEA;m equipados com pacotes de software que auxiliam nas medi&#xE7;&#xF5;es e c&#xE1;lculos, como a fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o do ventr&#xED;culo esquerdo (FEVE), a deforma&#xE7;&#xE3;o mioc&#xE1;rdica (strain) e o estudo Doppler para avalia&#xE7;&#xE3;o hemodin&#xE2;mica.<sup><xref ref-type="bibr" rid="B1">1</xref></sup></p>
<p>O aprendizado de m&#xE1;quina &#xE9; um dos subcampos da IA. O computador recebe os dados, aprende com eles, prev&#xEA; um resultado e, utilizando m&#xE9;todos estat&#xED;sticos, pode at&#xE9; melhorar o seu desempenho na pr&#xF3;xima vez que lhe forem apresentados dados semelhantes. Isso permite, por exemplo, que uma m&#xE1;quina treinada identifique estruturas dentro de uma nova imagem e as identifique com precis&#xE3;o. Outro subcampo surgiu posteriormente: o <italic>deep learning</italic>, ou aprendizado profundo. Ele &#xE9; aplicado quando se deseja processar uma grande quantidade de dados. Neste caso, os dados s&#xE3;o tratados como grandes subconjuntos. Assim como em uma rede neural biol&#xF3;gica, o processamento &#xE9; organizado em uma cadeia oculta composta por milhares de camadas neurais artificiais consecutivas at&#xE9; que um resultado final seja alcan&#xE7;ado.<sup><xref ref-type="bibr" rid="B1">1</xref></sup> Uma rede neural artificial pode usar opera&#xE7;&#xF5;es de convolu&#xE7;&#xE3;o, que permitem que um terceiro dado seja gerado a partir de dois dados de entrada, chamadas redes neurais convolucionais.<sup><xref ref-type="bibr" rid="B3">3</xref></sup></p>
<p>Nesse sentido, a IA na ecocardiografia pode ser extremamente vantajosa para identificar, quantificar e interpretar imagens, minimizar a varia&#xE7;&#xE3;o entre as avalia&#xE7;&#xF5;es, fornecer medi&#xE7;&#xF5;es reprodut&#xED;veis e fornecer um diagn&#xF3;stico preciso. O examinador pode encontrar grande suporte nesta tecnologia, o que permitiria reduzir o tempo gasto na extra&#xE7;&#xE3;o e integra&#xE7;&#xE3;o de dados, ficando evidente o potencial de tornar o sistema de sa&#xFA;de mais sustent&#xE1;vel.<sup><xref ref-type="bibr" rid="B4">4</xref></sup> Al&#xE9;m disso, a pr&#xE1;tica ecocardiogr&#xE1;fica orientada por IA n&#xE3;o apenas simplificaria a tomada de decis&#xF5;es cl&#xED;nicas, mas tamb&#xE9;m forneceria m&#xE9;todos de aprendizagem com feedback interativo para treinar examinadores menos experientes.<sup><xref ref-type="bibr" rid="B5">5</xref></sup></p>
<p>Assim, o uso da IA na ecocardiografia pode ser dividido em quatro cen&#xE1;rios principais: educa&#xE7;&#xE3;o, aquisi&#xE7;&#xE3;o de imagens, an&#xE1;lise de imagens e integra&#xE7;&#xE3;o com dados cl&#xED;nicos.<sup><xref ref-type="bibr" rid="B6">6</xref></sup> Embora existam evid&#xEA;ncias positivas sobre o papel da tecnologia de IA no diagn&#xF3;stico ecocardiogr&#xE1;fico, ainda h&#xE1; desafios a serem superados, como a padroniza&#xE7;&#xE3;o insuficiente dos protocolos ecocardiogr&#xE1;ficos e a generaliza&#xE7;&#xE3;o insuficiente dos modelos em aplica&#xE7;&#xF5;es cl&#xED;nicas.<sup><xref ref-type="bibr" rid="B6">6</xref></sup></p>
<p>Dadas as limita&#xE7;&#xF5;es da t&#xE9;cnica e a necessidade de investigar tend&#xEA;ncias futuras, a presente revis&#xE3;o sistem&#xE1;tica foi realizada com base em diversos estudos para validar externamente o desempenho da tecnologia de IA na ecocardiografia. Embora trabalhos semelhantes sejam encontrados na literatura que investigaram a aplica&#xE7;&#xE3;o da IA na an&#xE1;lise automatizada de imagens ecocardiogr&#xE1;ficas, este estudo procurou realizar um levantamento das inova&#xE7;&#xF5;es na &#xE1;rea com &#xEA;nfase nos &#xFA;ltimos cinco anos para explorar o uso da IA na ecocardiografia, analisando as evid&#xEA;ncias atuais de sua efic&#xE1;cia e discutindo seu potencial para melhorar o cuidado card&#xED;aco.</p>
</sec>
<sec sec-type="methods">
<title>M&#xE9;todos</title>
<p>Trata-se de uma revis&#xE3;o sistem&#xE1;tica de acordo com o Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).<sup><xref ref-type="bibr" rid="B7">7</xref></sup> Foram pesquisados artigos nas bases de dados MEDLINE, EMBASE, SCOPUS e Cochrane Central de janeiro de 2000 a mar&#xE7;o de 2023, sem restri&#xE7;&#xF5;es de idiomas, sendo o nosso m&#xE9;todo alvo a ecocardiografia, cuja interven&#xE7;&#xE3;o &#xE9; o uso de IA durante alguma fase do exame, com textos completos e descritores presentes no t&#xED;tulo ou resumo do artigo. Os seguintes descritores em ingl&#xEA;s foram utilizados para a pesquisa: (&#x201C;<italic>Artificial Intelligence</italic>&#x201D;) E (&#x201C;<italic>Echocardiography</italic>&#x201D;) E (&#x201C;<italic>Function</italic>&#x201D; OU &#x201C;<italic>Strain</italic>&#x201D;). Tamb&#xE9;m verificamos as referidas bases de dados quanto &#xE0; presen&#xE7;a de revis&#xF5;es sistem&#xE1;ticas anteriores ou em curso sobre o assunto. Foram combinados resultados de diferentes bases de dados e resultados duplicados foram descartados com uso do software Zotero. Dois investigadores (H.A.F.N e B.S.Y) realizaram a pesquisa de forma independente e cinco investigadores chegaram a um acordo sobre a elegibilidade dos estudos (H.A.F.N, B.S.Y, T.H.L.P.C, I.E.S, e Y.M.R.B). Inicialmente, foi realizada uma primeira triagem para identificar estudos que avaliaram o uso de qualquer tipo de IA durante o exame ecocardiogr&#xE1;fico. Em uma segunda etapa, selecionamos os artigos a serem lidos na &#xED;ntegra por meio da plataforma Rayyan e avaliamos a relev&#xE2;ncia dos dados, bem como se havia mais de um artigo avaliando a mesma coorte de pacientes. Uma terceira etapa consistiu em rever as refer&#xEA;ncias dos artigos selecionados para obter outras fontes n&#xE3;o reveladas na pesquisa digital. Classificamos os estudos de acordo com seu desenho (caso-controle, randomizado, coorte retrospectiva, coorte prospectiva, relato de caso), em rela&#xE7;&#xE3;o ao objetivo principal do uso de IA e ao principal resultado encontrado no estudo. Para cada estudo, foram extra&#xED;das as seguintes informa&#xE7;&#xF5;es: primeiro autor; institui&#xE7;&#xE3;o do estudo; data de in&#xED;cio e fim; ano de publica&#xE7;&#xE3;o; centro &#xFA;nico ou multic&#xEA;ntrico; caracter&#xED;sticas da popula&#xE7;&#xE3;o estudada (idade m&#xE9;dia, distribui&#xE7;&#xE3;o por sexo); par&#xE2;metros ecocardiogr&#xE1;ficos avaliados; m&#xE9;todos e janelas utilizados para avaliar a fun&#xE7;&#xE3;o card&#xED;aca; modelo de IA; desenho do estudo; conclus&#xF5;es do estudo.</p>
</sec>
<sec sec-type="results">
<title>Resultados</title>
<p>Foram identificadas 729 refer&#xEA;ncias em pesquisas em bases de dados eletr&#xF4;nicas, descartando-se 336 repeti&#xE7;&#xF5;es e outros 359 resultados que foram exclu&#xED;dos por t&#xED;tulo e/ou resumo. Obtivemos 33 artigos para leitura completa, dos quais tr&#xEA;s foram descartados por n&#xE3;o se enquadrarem no escopo da revis&#xE3;o segundo os autores. Apenas um artigo adicional foi inclu&#xED;do ap&#xF3;s pesquisa manual das refer&#xEA;ncias (<xref ref-type="fig" rid="f3">Figura 1</xref>). Um total de 31 artigos foram inclu&#xED;dos na an&#xE1;lise qualitativa desta revis&#xE3;o, sendo classificados em quatro categorias principais: 1) estudos que utilizam IA para an&#xE1;lise de imagens e par&#xE2;metros ecocardiogr&#xE1;ficos (N = 14, <xref ref-type="table" rid="t5">Tabela 1</xref>); 2) estudos que utilizam IA no processo de aquisi&#xE7;&#xE3;o de imagens (N = 11, <xref ref-type="table" rid="t6">Tabela 2</xref>); 3) estudos que utilizam IA para integra&#xE7;&#xE3;o com dados cl&#xED;nicos (N = 3, <xref ref-type="table" rid="t7">Tabela 3</xref>); e <xref ref-type="table" rid="t8">4</xref>) estudos que treinaram software de IA para aplica&#xE7;&#xE3;o ao fluxo do exame (N = 3, <xref ref-type="table" rid="t8">Tabela 4</xref>); todos os artigos foram publicados em ingl&#xEA;s. Vinte e seis artigos s&#xE3;o provenientes de um &#xFA;nico centro, sendo 7 (%) dos Estados Unidos, 2 (%) de Israel, 2 (%) da Noruega, 2 (%) da It&#xE1;lia, 2 (%) da Inglaterra, 1 (%) da Alemanha, 1 (%) da Gr&#xE9;cia, 1 (%) da Fran&#xE7;a, 1 (%) do Uganda, 1 (%) de Portugal, 1 (%) da Irlanda, 1 (%) da &#xC1;ustria, 1 (%) dos Pa&#xED;ses Baixos, 1 (%) da China, 1 (%) do Canad&#xE1; e 1 (%) do Jap&#xE3;o. Um &#xFA;nico artigo foi desenvolvido em diversos centros (13 centros diferentes). Quatro artigos n&#xE3;o informaram o pa&#xED;s onde o estudo foi realizado.</p>
<fig id="f3">
<label>Figure 1</label>
<caption>
<title>PRISMA flow diagram of study screening and selection</title></caption>
<graphic xlink:href="2675-312X-dic-36-04-e20230085-gf01-pt.tif"/>
</fig>
<table-wrap id="t5">
<label>Tabela 1</label>
<caption>
<title>Estudos que utilizam IA para an&#xE1;lise de imagens</title></caption>
<table frame="hsides" rules="groups">
<colgroup width="14%">
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
</colgroup>
<thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
<tr style="background-color: #C58874;">
<th align="left" valign="middle">Autor/ano</th>
<th align="center" valign="middle">Pa&#xED;s</th>
<th align="center" valign="middle">Desenho e objetivo</th>
<th align="center" valign="middle">N</th>
<th align="center" valign="middle">Par&#xE2;metros avaliados</th>
<th align="center" valign="middle">Modelo de IA</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">Cotella <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B8">8</xref></sup></td>
<td align="left" valign="middle"/>
<td align="center" valign="middle">Coorte retrospectiva<break/> Avaliar o c&#xE1;lculo da FEVE e do SLG feito pela IA em pacientes com amiloidose</td>
<td align="center" valign="middle">51</td>
<td align="center" valign="middle">FEVE, SLG</td>
<td align="center" valign="middle">EchoGo Core 2.0, Ultromics</td>
<td align="center" valign="middle">A FEVE e o SLG calculados por IA totalmente automatizada s&#xE3;o compar&#xE1;veis &#xE0;s medi&#xE7;&#xF5;es manuais em pacientes com amiloidose card&#xED;aca.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Sveric <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B14">14</xref></sup></td>
<td align="center" valign="middle">Alemanha</td>
<td align="center" valign="middle">Coorte prospectiva<break/> Investigar a viabilidade da IA para o c&#xE1;lculo da FEVE no fluxo de trabalho</td>
<td align="center" valign="middle">889</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">LVivo Seamless<sup>TM</sup></td>
<td align="center" valign="middle">A aplica&#xE7;&#xE3;o baseada em IA identificou automaticamente as janelas apicais padr&#xE3;o do VE e mediu a FEVE. Produziu resultados semelhantes aos do m&#xE9;todo biplano de Simpson modificado atualmente utilizado e pode melhorar a quantifica&#xE7;&#xE3;o da FEVE.</td>
</tr>
<tr>
<td align="left" valign="middle">He <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B9">9</xref></sup></td>
<td align="center" valign="middle">EUA</td>
<td align="center" valign="middle">Estudo cl&#xED;nico randomizado<break/> Comparar a avalia&#xE7;&#xE3;o inicial da FEVE pelo ultrassonografista e pela IA</td>
<td align="center" valign="middle">3495</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">EchoNet-Dynamic</td>
<td align="center" valign="middle">Os cardiologistas n&#xE3;o conseguiram distinguir as avalia&#xE7;&#xF5;es iniciais realizadas pela IA e pelo ultrassonografista. A avalia&#xE7;&#xE3;o inicial da FEVE por IA n&#xE3;o foi inferior &#xE0; avalia&#xE7;&#xE3;o por ultrassonografistas.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Wang <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B15">15</xref></sup></td>
<td align="center" valign="middle">EUA</td>
<td align="center" valign="middle">Coorte retrospectiva<break/> Avaliar a usabilidade de um software de IA em pacientes com COVID-19</td>
<td align="center" valign="middle">50</td>
<td align="center" valign="middle">FEVE, SLG, VDF, VSF</td>
<td align="center" valign="middle">EchoGo, Ultromics</td>
<td align="center" valign="middle">Os resultados apresentados pelo software EchoGo demonstraram correla&#xE7;&#xF5;es elevadas com os volumes do ventr&#xED;culo esquerdo e correla&#xE7;&#xF5;es moderadas com a FEVE e o SLG em compara&#xE7;&#xE3;o com as medi&#xE7;&#xF5;es padr&#xE3;o de ecocardiograma transtor&#xE1;cico.</td>
</tr>
<tr>
<td align="left" valign="middle">Liel-Cohen <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B16">16</xref></sup></td>
<td align="center" valign="middle">Israel</td>
<td align="center" valign="middle">Coorte retrospectiva<break/> Avaliar a capacidade da IA de identificar proje&#xE7;&#xF5;es e comparar os resultados do SLG com os m&#xE9;todos convencionais</td>
<td align="center" valign="middle">100</td>
<td align="center" valign="middle">SLG</td>
<td align="center" valign="middle">LVivo Seamless<sup>TM</sup></td>
<td align="center" valign="middle">Os resultados fornecidos pela solu&#xE7;&#xE3;o LVivo Seamless baseada em IA demonstraram excelente capacidade para identificar automaticamente as proje&#xE7;&#xF5;es 4CH, 2CH e 3CH, juntamente com excelentes resultados referentes ao SLG em compara&#xE7;&#xE3;o com imagens funcionais automatizadas.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Papadopoulou e<italic>t al.</italic>, 2022<sup><xref ref-type="bibr" rid="B10">10</xref></sup></td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Coorte retrospectiva<break/> Avaliar a precis&#xE3;o de um novo HUD com algoritmo assistido por IA para calcular automaticamente a FEVE</td>
<td align="center" valign="middle">100</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">Auto EF, Vivid&#x2122;</td>
<td align="center" valign="middle">O algoritmo autoEF assistido por IA em um novo HUD &#xE9; capaz de calcular com precis&#xE3;o a FEVE em tempo real, em compara&#xE7;&#xE3;o com o m&#xE9;todo biplano manual de Simpson recomendado, empregado em sistemas fixos em uma popula&#xE7;&#xE3;o geral de pacientes.</td>
</tr>
<tr>
<td align="left" valign="middle">Salte <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B11">11</xref></sup></td>
<td align="center" valign="middle">Noruega</td>
<td align="center" valign="middle">Coorte retrospectiva<break/> Examinar medi&#xE7;&#xF5;es automatizadas do SLG utilizando aprendizado profundo e IA e comparar com m&#xE9;todos convencionais</td>
<td align="center" valign="middle">200</td>
<td align="center" valign="middle">SLG</td>
<td align="center" valign="middle">EchoPAC SWO v.202, GE Ultrasound</td>
<td align="center" valign="middle">Medi&#xE7;&#xF5;es totalmente automatizadas do SLG por meio de uma nova tecnologia de aprendizado profundo baseada em IA para estimativa de movimento s&#xE3;o vi&#xE1;veis e r&#xE1;pidas e produzem resultados compar&#xE1;veis &#xE0;queles obtidos com o software semiautomatizado mais utilizado.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Hanif <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B33">33</xref></sup></td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Transversal<break/> Testar se a medi&#xE7;&#xE3;o do strain por IA pode ser mais confi&#xE1;vel em compara&#xE7;&#xE3;o com a interpreta&#xE7;&#xE3;o humana</td>
<td align="center" valign="middle">52</td>
<td align="center" valign="middle">SLG</td>
<td align="center" valign="middle">EchoGo, Ultromics</td>
<td align="center" valign="middle">A an&#xE1;lise do strain longitudinal por IA foi vi&#xE1;vel na maioria dos ecocardiogramas sem qualquer interven&#xE7;&#xE3;o do operador. O vi&#xE9;s entre o strain longitudinal medido pelo EchoGo e pelo software convencional parece ser pequeno.</td>
</tr>
<tr>
<td align="left" valign="middle">Penso <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B12">12</xref></sup></td>
<td align="center" valign="middle">It&#xE1;lia</td>
<td align="center" valign="middle">Transversal<break/> Viabilidade, compara&#xE7;&#xE3;o e avalia&#xE7;&#xE3;o da medi&#xE7;&#xE3;o do strain por IA em compara&#xE7;&#xE3;o com m&#xE9;todos convencionais</td>
<td align="center" valign="middle">203</td>
<td align="center" valign="middle">FEVR, SLPLVD</td>
<td align="center" valign="middle">3D Auto RV, Philips Medical Systems</td>
<td align="center" valign="middle">A r&#xE1;pida an&#xE1;lise e a excelente reprodutibilidade da an&#xE1;lise do ventr&#xED;culo direito em ecocardiograma 3D por IA fundamentam o uso rotineiro do sistema.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Di Candia <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B17">17</xref></sup></td>
<td align="center" valign="middle">It&#xE1;lia</td>
<td align="center" valign="middle">Transversal<break/> Comparar os resultados de uma an&#xE1;lise totalmente automatizada e de uma an&#xE1;lise manual do tra&#xE7;ado utilizando um novo software de IA</td>
<td align="center" valign="middle">28</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">IA Simpson e IA Wall Motion Tracking (Rastreio do Movimento de Parede)</td>
<td align="center" valign="middle">A an&#xE1;lise totalmente automatizada por IA revelou-se tecnicamente simples, menos demorada e altamente reprodut&#xED;vel. Isso reduz a variabilidade das medi&#xE7;&#xF5;es manuais entre diferentes ultrassonografistas e em momentos diferentes.</td>
</tr>
<tr>
<td align="left" valign="middle">Papadopoulou <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B10">10</xref></sup></td>
<td align="center" valign="middle">Gr&#xE9;cia</td>
<td align="center" valign="middle">Transversal<break/> Avaliar a confiabilidade e a precis&#xE3;o do diagn&#xF3;stico de um HUD assistido por IA para calcular automaticamente a fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o</td>
<td align="center" valign="middle">100</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">Kosmos autoEF</td>
<td align="center" valign="middle">A utiliza&#xE7;&#xE3;o de um novo HUD assistido por IA forneceu resultados de FEVE semelhantes aos obtidos pelo m&#xE9;todo biplano manual de Simpson em sistemas fixos e demonstra potencial cl&#xED;nico.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Popoff <italic>et al.</italic>, 2020<sup><xref ref-type="bibr" rid="B18">18</xref></sup></td>
<td align="center" valign="middle">Fran&#xE7;a</td>
<td align="center" valign="middle">Transversal<break/> Avaliar a precis&#xE3;o e reprodutibilidade da ferramenta de IA para calcular volumes do VE e FE em compara&#xE7;&#xE3;o com a resson&#xE2;ncia magn&#xE9;tica card&#xED;aca como refer&#xEA;ncia</td>
<td align="center" valign="middle">114</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">Um modelo de redes neurais convolucionais profundas</td>
<td align="center" valign="middle">Em compara&#xE7;&#xE3;o com o contorno manual, o volume ventricular esquerdo e a FE por IA demonstraram precis&#xE3;o compar&#xE1;vel ou melhor e maior reprodutibilidade. Esses resultados demonstram o valor das ferramentas de IA, com potencial para automa&#xE7;&#xE3;o total, para avalia&#xE7;&#xE3;o objetiva e acompanhamento da fun&#xE7;&#xE3;o card&#xED;aca.</td>
</tr>
<tr>
<td align="left" valign="middle">Asch <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B13">13</xref></sup></td>
<td align="center" valign="middle">EUA</td>
<td align="center" valign="middle">Coorte prospectiva<break/> Determinar se as estimativas automatizadas feitas por IA podem ser utilizadas para classificar com precis&#xE3;o a fun&#xE7;&#xE3;o do VE</td>
<td align="center" valign="middle">233</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">Algoritmo de aprendizado de m&#xE1;quina, Caption Health</td>
<td align="center" valign="middle">O novo algoritmo de aprendizado de m&#xE1;quina permite uma avalia&#xE7;&#xE3;o precisa e automatizada da fun&#xE7;&#xE3;o do VE a partir de proje&#xE7;&#xF5;es ecocardiogr&#xE1;ficas comumente utilizadas &#xE0; beira leito. Esta abordagem permitir&#xE1; que mais profissionais &#xE0; beira leito avaliem com precis&#xE3;o a fun&#xE7;&#xE3;o do VE.</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<fn id="TFN5">
<p><italic>VDF: volume diast&#xF3;lico final do ventr&#xED;culo esquerdo; SLG: strain longitudinal global; FEVE: fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o do ventr&#xED;culo esquerdo; IA: intelig&#xEA;ncia artificial; EF: fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o autom&#xE1;tica; A4C: apical quatro c&#xE2;maras; A2C: apical duas c&#xE2;maras; VD: ventr&#xED;culo direito; LVOT-VTI: via de sa&#xED;da do ventr&#xED;culo esquerdo integral da velocidade-tempo; PLAX: corte paraesternal eixo longo; corte paraesternal eixo curto; IVC: veia cava inferior; 4CH: 4 c&#xE2;maras; 2CH: 2 c&#xE2;maras; 3CH: 3 c&#xE2;maras</italic></p></fn>
</table-wrap-foot>
</table-wrap>
<table-wrap id="t6">
<label>Tabela 2</label>
<caption>
<title>Estudos que utilizam IA para aquisi&#xE7;&#xE3;o de imagens.</title></caption>
<table frame="hsides" rules="groups">
<colgroup width="14%">
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
</colgroup>
<thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
<tr style="background-color: #C58874;">
<th align="left" valign="middle">Autor/ano</th>
<th align="center" valign="middle">Pa&#xED;s</th>
<th align="center" valign="middle">Desenho e objetivo do estudo</th>
<th align="center" valign="middle">N</th>
<th align="center" valign="middle">Par&#xE2;metros avaliados</th>
<th align="center" valign="middle">Modelo de IA</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">Peck <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B25">25</xref></sup></td>
<td align="center" valign="middle">Uganda</td>
<td align="center" valign="middle">Transversal. Avaliar a capacidade de n&#xE3;o especialistas em obter imagens de qualidade diagn&#xF3;stica em pacientes com cardiopatia reum&#xE1;tica empregando orienta&#xE7;&#xE3;o por IA com Doppler colorido</td>
<td align="center" valign="middle">50</td>
<td align="center" valign="middle">Presen&#xE7;a / aus&#xEA;ncia de cardiopatia reum&#xE1;tica, morfologia anormal da v&#xE1;lvula mitral e regurgita&#xE7;&#xE3;o mitral</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">A orienta&#xE7;&#xE3;o por IA com Doppler colorido &#xE9; vi&#xE1;vel para permitir a triagem de cardiopatia reum&#xE1;tica por n&#xE3;o especialistas, com desempenho significativamente melhor para avalia&#xE7;&#xE3;o da v&#xE1;lvula mitral do que da a&#xF3;rtica.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Klempfner <italic>et al.</italic>, 2023<sup><xref ref-type="bibr" rid="B5">5</xref></sup></td>
<td align="center" valign="middle">EUA</td>
<td align="center" valign="middle">Coorte prospectiva. Determinar a precis&#xE3;o diagn&#xF3;stica de estudos ecocardiogr&#xE1;ficos obtidos por usu&#xE1;rios inexperientes guiados por um algoritmo de IA</td>
<td align="center" valign="middle">240</td>
<td align="center" valign="middle">Dimens&#xE3;o e fun&#xE7;&#xE3;o do ventr&#xED;culo esquerdo e direito, derrame peric&#xE1;rdico, morfologia valvar, dimens&#xE3;o da atrial esquerda e da veia cava inferior</td>
<td align="center" valign="middle">Algoritmo UltraSight</td>
<td align="center" valign="middle">Ap&#xF3;s treinamento m&#xED;nimo com um novo software de orienta&#xE7;&#xE3;o por IA, usu&#xE1;rios inexperientes conseguiram adquirir imagens em 10 proje&#xE7;&#xF5;es ecocardiogr&#xE1;ficas card&#xED;acas padr&#xE3;o com qualidade diagn&#xF3;stica que se aproxima daquela obtida por ultrassonografistas especialistas.</td>
</tr>
<tr>
<td align="left" valign="middle">Dadon <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B27">27</xref></sup></td>
<td align="center" valign="middle">Israel</td>
<td align="center" valign="middle">Ensaio cl&#xED;nico randomizado avaliar o uso da IA como uma ferramenta did&#xE1;tica para melhorar a avalia&#xE7;&#xE3;o da FEVE por cl&#xED;nicos-gerais n&#xE3;o cardiologistas a partir da proje&#xE7;&#xE3;o A4C</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">LVivo EF</td>
<td align="center" valign="middle">O uso de uma ferramenta de IA por cl&#xED;nicos-gerais que trabalham no servi&#xE7;o de emerg&#xEA;ncia melhorou a precis&#xE3;o da avalia&#xE7;&#xE3;o da FEVE em compara&#xE7;&#xE3;o com o grupo de controle.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Varudo <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B19">19</xref></sup></td>
<td align="center" valign="middle">Portugal</td>
<td align="center" valign="middle">Coorte prospectiva. Comparar as medi&#xE7;&#xF5;es em tempo real realizadas por profissionais inexperientes e por especialistas com as medi&#xE7;&#xF5;es de refer&#xEA;ncia da FEVE realizadas manualmente por ecografistas</td>
<td align="center" valign="middle">95</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">Algoritmo EF em Tempo Real</td>
<td align="center" valign="middle">As medi&#xE7;&#xF5;es em tempo real da FEVE por aprendizado de m&#xE1;quina apresentaram forte correla&#xE7;&#xE3;o com as medi&#xE7;&#xF5;es manuais realizadas por especialistas.</td>
</tr>
<tr>
<td align="left" valign="middle">Burke <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B20">20</xref></sup></td>
<td align="center" valign="middle">Irlanda</td>
<td align="center" valign="middle">Transversal.<break/> Avaliar a qualidade do estudo por profissionais inexperientes com orienta&#xE7;&#xE3;o por IA</td>
<td align="center" valign="middle">120</td>
<td align="center" valign="middle">Qualidade da imagem para determinar a FEVE, derrame peric&#xE1;rdico</td>
<td align="center" valign="middle">Sistema Caption AI</td>
<td align="center" valign="middle">A tecnologia Caption AI permite, com seguran&#xE7;a, que profissionais inexperientes realizem ecocardiogramas eficientes e precisos &#xE0; beira leito em diferentes contextos cl&#xED;nicos que sejam compar&#xE1;veis aos realizados por profissionais especializados.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Narang <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B21">21</xref></sup></td>
<td align="center" valign="middle">EUA</td>
<td align="center" valign="middle">Coorte prospectiva.<break/> Testar se profissionais inexperientes conseguem obter estudos ecocardiogr&#xE1;ficos transtor&#xE1;cicos de 10 proje&#xE7;&#xF5;es de qualidade diagn&#xF3;stica utilizando um software de aprendizado profundo</td>
<td align="center" valign="middle">240</td>
<td align="center" valign="middle">Tamanho do ventr&#xED;culo esquerdo, fun&#xE7;&#xE3;o, derrame peric&#xE1;rdico e tamanho do ventr&#xED;culo direito</td>
<td align="center" valign="middle">Orienta&#xE7;&#xE3;o Software Caption</td>
<td align="center" valign="middle">O algoritmo de aprendizado profundo permite que iniciantes sem experi&#xEA;ncia em ultrassonografia obtenham estudos ecocardiogr&#xE1;ficos transtor&#xE1;cicos para fins de diagn&#xF3;stico, para avalia&#xE7;&#xE3;o do tamanho e fun&#xE7;&#xE3;o do ventr&#xED;culo esquerdo, tamanho do ventr&#xED;culo direito e presen&#xE7;a de derrame peric&#xE1;rdico n&#xE3;o trivial.</td>
</tr>
<tr>
<td align="left" valign="middle">Surette <italic>et al.</italic>, 2020<sup><xref ref-type="bibr" rid="B22">22</xref></sup></td>
<td align="center" valign="middle">EUA</td>
<td align="center" valign="middle">Ensaio cl&#xED;nico multic&#xEA;ntrico prospectivo. Avaliar se as imagens adquiridas por iniciantes com orienta&#xE7;&#xE3;o do software foram de qualidade diagn&#xF3;stica em pacientes com e sem dispositivos eletrofisiol&#xF3;gicos implantados</td>
<td align="center" valign="middle">240</td>
<td align="center" valign="middle">Tamanho e fun&#xE7;&#xE3;o do ventr&#xED;culo direito</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">O novo software de aprendizado profundo pode orientar profissionais inexperientes a realizar ecocardiogramas transtor&#xE1;cicos que permitam a avalia&#xE7;&#xE3;o qualitativa do tamanho do VD, mesmo na presen&#xE7;a de dispositivos eletrofisiol&#xF3;gicos implantados.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Schneider <italic>et al.</italic>, 2020<sup><xref ref-type="bibr" rid="B23">23</xref></sup></td>
<td align="center" valign="middle">&#xC1;ustria</td>
<td align="center" valign="middle">Transversal.<break/> Avaliar se o novo algoritmo de aprendizado de m&#xE1;quina pode ajudar operadores inexperientes a adquirir loops de imagem de diagn&#xF3;stico</td>
<td align="center" valign="middle">14</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">O algoritmo de aprendizado de m&#xE1;quina pode orientar ultrassonografistas inexperientes a adquirir loops de eco de diagn&#xF3;stico e fornecer um c&#xE1;lculo automatizado da FEVE que seja compar&#xE1;vel ao realizado por um especialista humano.</td>
</tr>
<tr>
<td align="left" valign="middle">Motazedian <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B24">24</xref></sup></td>
<td align="center" valign="middle">Holanda</td>
<td align="center" valign="middle">Coorte prospectiva.<break/> Determinar a precis&#xE3;o da FEVE em exame US &#xE0; beira leito guiado por IA em compara&#xE7;&#xE3;o com o ecocardiograma transtor&#xE1;cico formal</td>
<td align="center" valign="middle">449</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">EchoNous Kosmos</td>
<td align="center" valign="middle">A US &#xE0; beira leito assistida por IA realizada por profissionais inexperientes e experientes pode determinar com precis&#xE3;o a FEVE em compara&#xE7;&#xE3;o com um ecocardiograma transtor&#xE1;cico abrangente.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Zhai <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B26">26</xref></sup></td>
<td align="center" valign="middle">China</td>
<td align="center" valign="middle">Transversal.<break/> Identificar o grau de correla&#xE7;&#xE3;o entre LVOT-VTI automatizada e LVOT-VTI manual realizado por m&#xE9;dicos especialistas em US &#xE0; beira leito</td>
<td align="center" valign="middle">46</td>
<td align="center" valign="middle">LVOT&#x2010;VTI</td>
<td align="center" valign="middle">Software de IA</td>
<td align="center" valign="middle">A IA pode fornecer orienta&#xE7;&#xE3;o em tempo real a operadores iniciantes que n&#xE3;o possuem experi&#xEA;ncia para adquirir a proje&#xE7;&#xE3;o padr&#xE3;o ideal.</td>
</tr>
<tr>
<td align="left" valign="middle">Fung <italic>et al.</italic>, 2019<sup><xref ref-type="bibr" rid="B34">34</xref></sup></td>
<td align="center" valign="middle">Canad&#xE1;</td>
<td align="center" valign="middle">Coorte prospectiva. Avaliar a precis&#xE3;o do sistema de IA para avaliar a qualidade da imagem em compara&#xE7;&#xE3;o com a an&#xE1;lise subjetiva realizada por especialistas</td>
<td align="center" valign="middle">53</td>
<td align="center" valign="middle">PLAX, PSAX, A2C, A4C e IVC subcostal</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">O sistema de IA &#xE9; capaz de avaliar a qualidade da imagem com precis&#xE3;o modesta.</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<fn id="TFN6">
<p><italic>A2C: apical duas c&#xE2;maras; A4C: apical quatro c&#xE2;maras; IVC: veia cava inferior; FEVE: fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o do ventr&#xED;culo esquerdo; LVOT-VTI: via de sa&#xED;da do ventr&#xED;culo esquerdo integral da velocidade-tempo; PLAX: corte paraesternal eixo longo; PSAX: corte paraesternal eixo curto; VD: ventr&#xED;culo direito; SLG: strain longitudinal global; VDF: volume diast&#xF3;lico final do ventr&#xED;culo esquerdo; VSF: volume sist&#xF3;lico final do ventr&#xED;culo esquerdo; HUD: dispositivo de ultrassom port&#xE1;til; AutoEF: fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o autom&#xE1;tica; FEVR: fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o do ventr&#xED;culo direito; SLPLVD: strain longitudinal da parede livre do ventr&#xED;culo direito; US: ultrassonografia.</italic></p>
</fn>
</table-wrap-foot>
</table-wrap>
<table-wrap id="t7">
<label>Tabela 3</label>
<caption>
<title>Estudos que utilizam IA para integra&#xE7;&#xE3;o com dados cl&#xED;nicos</title></caption>
<table frame="hsides" rules="groups">
<colgroup width="14%">
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
</colgroup>
<thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
<tr style="background-color: #C58874;">
<th align="left" valign="middle">Autor/ ano</th>
<th align="center" valign="middle">Pa&#xED;s</th>
<th align="center" valign="middle">Desenho e objetivo do estudo</th>
<th align="center" valign="middle">N</th>
<th align="center" valign="middle">Par&#xE2;metros avaliados</th>
<th align="center" valign="middle">Modelo de IA</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">Kusunose <italic>et al.</italic>, 2019<sup><xref ref-type="bibr" rid="B35">35</xref></sup></td>
<td align="center" valign="middle">Jap&#xE3;o</td>
<td align="center" valign="middle">Coorte prospectiva. Investigar se uma DCNN poderia detectar RWMAs com mais efici&#xEA;ncia</td>
<td align="center" valign="middle">300</td>
<td align="center" valign="middle">Presen&#xE7;a de anomalias do movimento da parede</td>
<td align="center" valign="middle">ResNet, DenseNet, Inception-ResNet, Inception e Xception</td>
<td align="center" valign="middle">Os resultados apoiam a possibilidade de utilizar DCNN para o diagn&#xF3;stico automatizado de RWMAs no campo da ecocardiografia.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">O&#x27;Driscoll <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B28">28</xref></sup></td>
<td align="center" valign="middle">Inglaterra</td>
<td align="center" valign="middle">Coorte retrospectiva. Avaliar se a FEVE e o SLG, calculados automaticamente pela IA, aumentam o desempenho diagn&#xF3;stico da ecocardiografia sob estresse para a detec&#xE7;&#xE3;o de DAC</td>
<td align="center" valign="middle">512</td>
<td align="center" valign="middle">FEVE e SLG em repouso e stress</td>
<td align="center" valign="middle">EchoGo Core 1.0, Ultromics</td>
<td align="center" valign="middle">O c&#xE1;lculo da FEVE e do SLG por IA por meio do contorno em ecocardiografias sob estresse com e sem contraste, em repouso e sob estresse, &#xE9; vi&#xE1;vel e melhora independentemente a identifica&#xE7;&#xE3;o da DAC obstrutiva para al&#xE9;m dos m&#xE9;todos convencionais.</td>
</tr>
<tr>
<td align="left" valign="middle">Asch <italic>et al.</italic>, 2022<sup><xref ref-type="bibr" rid="B36">36</xref></sup></td>
<td align="center" valign="middle">Multic&#xEA;ntrico</td>
<td align="center" valign="middle">Coorte prospectiva. Testar a hip&#xF3;tese de que a an&#xE1;lise por IA de imagens ecocardiogr&#xE1;ficas poderia prever a mortalidade com mais precis&#xE3;o do que a an&#xE1;lise convencional por um especialista humano</td>
<td align="center" valign="middle">870</td>
<td align="center" valign="middle">FEVE, SLVE</td>
<td align="center" valign="middle">EchoGo Core, Ultromics</td>
<td align="center" valign="middle">A an&#xE1;lise por IA da FEVE e do SLVE apresentou viabilidade semelhante &#xE0; an&#xE1;lise manual e as an&#xE1;lises por IA, mas n&#xE3;o as manuais, foram um fator significativo de predi&#xE7;&#xE3;o de mortalidade hospitalar e durante per&#xED;odo de acompanhamento.</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<fn id="TFN7">
<p><italic>DAC: doen&#xE7;a arterial coronariana; DCNN: rede neural convolucional profunda; SLG: strain longitudinal global; FEVE: fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o do ventr&#xED;culo esquerdo; SLVE: strain longitudinal do ventr&#xED;culo esquerdo; RWMAs: anormalidades no movimento regional da parede; SE: ecocardiograma sob estresse; IA: intelig&#xEA;ncia artificial.</italic></p>
</fn>
</table-wrap-foot>
</table-wrap>
<table-wrap id="t8">
<label>Tabela 4</label>
<caption>
<title>Estudos que desenvolveram uma IA</title></caption>
<table frame="hsides" rules="groups">
<colgroup width="14%">
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
<col/>
</colgroup>
<thead style="border-top: thin solid; border-bottom: thin solid; border-color: #000000">
<tr style="background-color: #C58874;">
<th align="left" valign="middle">Autor e ano</th>
<th align="center" valign="middle">Pa&#xED;s</th>
<th align="center" valign="middle">Desenho e objetivo do estudo</th>
<th align="center" valign="middle">N&#xFA;mero de pacientes</th>
<th align="center" valign="middle">Par&#xE2;metros avaliados</th>
<th align="center" valign="middle">Modelo de IA</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">Asch <italic>et al.</italic>, 2019<sup><xref ref-type="bibr" rid="B32">32</xref></sup></td>
<td align="center" valign="middle">EUA</td>
<td align="center" valign="middle">Transversal<break/> Desenvolver um algoritmo e testar sua viabilidade e precis&#xE3;o para quantifica&#xE7;&#xE3;o automatizada da FEVE por um computador especialista bem treinado em compara&#xE7;&#xE3;o &#xE0; refer&#xEA;ncia</td>
<td align="center" valign="middle">99</td>
<td align="center" valign="middle">FEVE</td>
<td align="center" valign="middle">AutoEF (BayLabs)</td>
<td align="center" valign="middle">O algoritmo de aprendizado de m&#xE1;quina para a estimativa da FEVE que evita a segmenta&#xE7;&#xE3;o de imagens e as medi&#xE7;&#xF5;es de volume &#xE9; altamente vi&#xE1;vel e pode produzir resultados estritamente compar&#xE1;veis aos obtidos por profissionais experientes utilizando a metodologia convencional.</td>
</tr>
<tr style="background-color: #E8CCBF;">
<td align="left" valign="middle">Stowell <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B30">30</xref></sup></td>
<td align="center" valign="middle">Inglaterra</td>
<td align="center" valign="middle">Transversal<break/> Desenvolver um m&#xE9;todo baseado em aprendizado de m&#xE1;quina, treinado e validado por especialistas credenciados da AI Echocardiography Collaborative</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Strain longitudinal do ventr&#xED;culo esquerdo</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">A medi&#xE7;&#xE3;o de strain por IA de c&#xF3;digo aberto e independente do fornecedor automaticamente produz valores compat&#xED;veis com o consenso de especialistas e com os obtidos por especialistas individualmente.</td>
</tr>
<tr>
<td align="left" valign="middle">Tromp <italic>et al.</italic>, 2021<sup><xref ref-type="bibr" rid="B31">31</xref></sup></td>
<td align="center" valign="middle">Cingapura</td>
<td align="center" valign="middle">Transversal<break/> Desenvolver um fluxo de trabalho de aprendizado profundo totalmente automatizado para classificar, segmentar e anotar v&#xED;deos bidimensionais e modalidades Doppler em ecocardiogramas</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">-</td>
<td align="center" valign="middle">Os algoritmos de aprendizado profundo podem anotar automaticamente v&#xED;deos 2D e modalidades Doppler com precis&#xE3;o semelhante &#xE0;s medi&#xE7;&#xF5;es manuais por ultrassonografistas especializados.</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<fn id="TFN8">
<p><italic>AutoEF: fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o autom&#xE1;tica; FEVE: fra&#xE7;&#xE3;o de eje&#xE7;&#xE3;o do ventr&#xED;culo esquerdo; IA: intelig&#xEA;ncia artificial.</italic></p>
</fn>
</table-wrap-foot>
</table-wrap>
</sec>
<sec sec-type="discussion">
<title>Discuss&#xE3;o</title>
<p>A implementa&#xE7;&#xE3;o de ferramentas de IA no estudo ecocardiogr&#xE1;fico pode melhorar o atendimento ao paciente, bem como otimizar o diagn&#xF3;stico e o tratamento. Os estudos inclu&#xED;dos nesta revis&#xE3;o sistem&#xE1;tica s&#xE3;o favor&#xE1;veis &#xE0; aplica&#xE7;&#xE3;o da tecnologia automatizada na rotina do exame devido &#xE0; sua similaridade, viabilidade, utilidade, precis&#xE3;o, exatid&#xE3;o e reprodutibilidade compar&#xE1;veis &#xE0; avalia&#xE7;&#xE3;o manual. Embora o uso de IA no exame ainda seja limitado e, portanto, propenso a modifica&#xE7;&#xF5;es que visem avan&#xE7;ar e aperfei&#xE7;oar o m&#xE9;todo, o benef&#xED;cio que a tecnologia representa para a ecocardiografia parece superar quaisquer receios ou desvantagens.</p>
<sec>
<title>An&#xE1;lise de imagem e par&#xE2;metros ecocardiogr&#xE1;ficos</title>
<p>A medi&#xE7;&#xE3;o automatizada realizada por IA de par&#xE2;metros como FEVE e strain longitudinal global (SLG) pode ser compar&#xE1;vel &#xE0; medi&#xE7;&#xE3;o manual por ultrassonografistas experientes e o uso dessas ferramentas preservou a precis&#xE3;o da medi&#xE7;&#xE3;o de forma confi&#xE1;vel.<sup><xref ref-type="bibr" rid="B8">8</xref>&#x2013;<xref ref-type="bibr" rid="B13">13</xref></sup> Isso &#xE9; relevante devido aos benef&#xED;cios adicionais oferecidos: maior efic&#xE1;cia operacional do exame e qualidade da consist&#xEA;ncia das an&#xE1;lises;<sup><xref ref-type="bibr" rid="B6">6</xref></sup> menos variabilidade entre diferentes examinadores,<sup><xref ref-type="bibr" rid="B14">14</xref>&#x2013;<xref ref-type="bibr" rid="B18">18</xref></sup> superando limita&#xE7;&#xF5;es humanas como cansa&#xE7;o ou distra&#xE7;&#xE3;o; e maior reprodutibilidade do exame.<sup><xref ref-type="bibr" rid="B1">1</xref></sup></p>
</sec>
<sec>
<title>Aquisi&#xE7;&#xE3;o de imagem</title>
<p>Grande parte da literatura se concentra na interpreta&#xE7;&#xE3;o ou an&#xE1;lise pela IA de imagens j&#xE1; adquiridas, mas h&#xE1; evid&#xEA;ncias crescentes sobre o uso da tecnologia de IA para aquisi&#xE7;&#xE3;o de imagens por n&#xE3;o especialistas.<sup><xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B19">19</xref>&#x2013;<xref ref-type="bibr" rid="B24">24</xref></sup> A tecnologia automatizada pode orientar um examinador n&#xE3;o especialista, reconhecendo imagens incorretas e a necessidade de visualiza&#xE7;&#xF5;es ou melhorias adicionais. Tamb&#xE9;m pode ajudar a padronizar a aquisi&#xE7;&#xE3;o de imagem e as medi&#xE7;&#xF5;es.<sup><xref ref-type="bibr" rid="B1">1</xref>,<xref ref-type="bibr" rid="B6">6</xref></sup> Dessa forma, seria uma ferramenta utilizada para otimizar o treinamento e os estudos ecocardiogr&#xE1;ficos,<sup><xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B26">26</xref></sup> sendo poss&#xED;vel adquirir imagens de qualidade diagn&#xF3;stica compar&#xE1;veis &#xE0;s adquiridas manualmente por especialistas.<sup><xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B19">19</xref>,<xref ref-type="bibr" rid="B20">20</xref>,<xref ref-type="bibr" rid="B22">22</xref>&#x2013;<xref ref-type="bibr" rid="B24">24</xref></sup> A tecnologia &#xE9; aplic&#xE1;vel na pr&#xE1;tica cl&#xED;nica, em emerg&#xEA;ncias e em &#xE1;reas remotas com recursos escassos e poucos profissionais experientes.<sup><xref ref-type="bibr" rid="B5">5</xref>,<xref ref-type="bibr" rid="B21">21</xref>,<xref ref-type="bibr" rid="B25">25</xref>,<xref ref-type="bibr" rid="B27">27</xref></sup></p>
</sec>
<sec>
<title>Integra&#xE7;&#xE3;o com dados cl&#xED;nicos</title>
<p>No campo da ecocardiografia, a interpreta&#xE7;&#xE3;o dos achados dos exames &#xE9; necess&#xE1;ria para melhorar o cuidado e o monitoramento dos pacientes. Assim, o treinamento de softwares de IA para associar dados de imagem a dados cl&#xED;nicos do paciente permite obter informa&#xE7;&#xF5;es de diagn&#xF3;stico relevantes sem a necessidade de habilidades avan&#xE7;adas de an&#xE1;lise por um especialista e sem despender tempo e recursos. A correla&#xE7;&#xE3;o com anomalias funcionais, estados patol&#xF3;gicos, previs&#xE3;o de desfechos e estratifica&#xE7;&#xE3;o de risco s&#xE3;o exemplos dos benef&#xED;cios que a integra&#xE7;&#xE3;o entre algoritmos e dados cl&#xED;nicos pode oferecer. Os resultados apresentados pelo estudo de Kusunose <italic>et al.</italic> apoiam a possibilidade de utiliza&#xE7;&#xE3;o de software de IA para o diagn&#xF3;stico automatizado de anomalias funcionais, e o estudo de O&#x27;Driscoll <italic>et al.</italic> para o c&#xE1;lculo da FEVE e do SLG.<sup><xref ref-type="bibr" rid="B28">28</xref>,<xref ref-type="bibr" rid="B29">29</xref></sup> No futuro, a integra&#xE7;&#xE3;o de softwares de IA com dados cl&#xED;nicos pode possibilitar a formula&#xE7;&#xE3;o de terapias personalizadas para cada paciente.</p>
</sec>
<sec>
<title>Desenvolvimento da IA</title>
<p>Bases de dados de estudos ecocardiogr&#xE1;ficos foram utilizadas para treinar protocolos e algoritmos de aprendizado profundo para detectar disfun&#xE7;&#xE3;o card&#xED;aca, identificar anomalias anat&#xF4;micas e definir fluxos de trabalho.<sup><xref ref-type="bibr" rid="B30">30</xref>&#x2013;<xref ref-type="bibr" rid="B32">32</xref></sup> Ap&#xF3;s o treinamento, a precis&#xE3;o observada foi semelhante &#xE0; precis&#xE3;o obtida com metodologias convencionais. Esta valida&#xE7;&#xE3;o &#xE9; indispens&#xE1;vel, pois no desenvolvimento de softwares de IA, tanto o software quanto as imagens prov&#xEA;m de fornecedores externos. A qualidade das imagens utilizadas &#xE9; um ponto relevante, uma vez que, caso os dados inseridos sejam de baixa qualidade, tendenciosos ou restritos, a interpreta&#xE7;&#xE3;o pelo algoritmo tamb&#xE9;m ser&#xE1;. Um desafio futuro ser&#xE1; desenvolver algoritmos de aprendizado profundo que possam ser empregados em diferentes equipamentos, possuam nomenclaturas padr&#xE3;o e armazenem e recuperem dados de forma independente do fornecedor.<sup><xref ref-type="bibr" rid="B1">1</xref></sup></p>
<p>A incorpora&#xE7;&#xE3;o da IA na pr&#xE1;tica cl&#xED;nica levanta preocupa&#xE7;&#xF5;es &#xE9;ticas principalmente relacionadas &#xE0; responsabilidade, &#xE0; medida que o cuidado se torna mais indireto, e &#xE0; transpar&#xEA;ncia, conforme se questiona a extens&#xE3;o daquilo que deve ser informado ao paciente e por ele consentido. &#xC9; importante lembrar que o uso da IA j&#xE1; faz parte da rotina das pessoas e esse tipo de tecnologia est&#xE1; cada vez mais dispon&#xED;vel para a sociedade. Esta intera&#xE7;&#xE3;o tende a ser irrevers&#xED;vel e, certamente, n&#xE3;o ser&#xE1; diferente na pr&#xE1;tica m&#xE9;dica e na ecocardiografia. Mesmo assim, n&#xE3;o se trata de uma competi&#xE7;&#xE3;o entre humanos e m&#xE1;quinas. N&#xE3;o conseguimos imaginar um cen&#xE1;rio no qual um substitua o outro. Pelo contr&#xE1;rio, a IA est&#xE1; a favor do m&#xE9;dico e contra a obscuridade das limita&#xE7;&#xF5;es sist&#xEA;micas e humanas, tais como: alta demanda, falta de especialistas, ergonomia e esgotamento dos servi&#xE7;os. Especialmente porque, ao aumentar a efic&#xE1;cia operacional do exame, as habilidades do m&#xE9;dico especialista podem ser empregadas de forma mais eficiente. Espera-se que o uso da IA ajude a democratizar o acesso ao exame, representando um avan&#xE7;o na assist&#xEA;ncia &#xE0; sa&#xFA;de.</p>
<p>A presente revis&#xE3;o sistem&#xE1;tica possui limita&#xE7;&#xF5;es. Em primeiro lugar, a grande maioria dos estudos tem um desenho transversal. No entanto, como nosso objetivo era a valida&#xE7;&#xE3;o diagn&#xF3;stica, estudos com esse desenho s&#xE3;o suficientes para demonstrar a n&#xE3;o inferioridade dos m&#xE9;todos. Os desenhos que estudam medidas de associa&#xE7;&#xE3;o em longo prazo podem esclarecer melhor os benef&#xED;cios da IA para a qualidade do acompanhamento longitudinal dos pacientes. Em segundo lugar, os softwares de IA geralmente necessitam de uma qualidade m&#xED;nima de imagem. Por vezes, pacientes com janelas ruins foram exclu&#xED;dos.<sup><xref ref-type="bibr" rid="B10">10</xref>,<xref ref-type="bibr" rid="B14">14</xref>,<xref ref-type="bibr" rid="B15">15</xref></sup> Em diversos estudos, as imagens analisadas foram coletadas em cen&#xE1;rios ideais por especialistas experientes ou faziam parte de pacotes provenientes de centros de fornecedores. &#xC9; necess&#xE1;rio realizar estudos maiores com imagens de baixa qualidade, especialmente em pacientes com janelas ruins, hemodinamicamente inst&#xE1;veis ou em ambientes n&#xE3;o ideais (UTI, emerg&#xEA;ncia). Em terceiro lugar, a maioria dos softwares consegue medir um n&#xFA;mero limitado de par&#xE2;metros. Assim, estudos adicionais podem validar os benef&#xED;cios do uso da IA al&#xE9;m dos par&#xE2;metros estudados nesta revis&#xE3;o.</p>
</sec>
</sec>
<sec sec-type="conclusions">
<title>Conclus&#xE3;o</title>
<p>O uso de IA na &#xE1;rea da ecocardiografia provou ser um m&#xE9;todo altamente eficaz e preciso em quatro dom&#xED;nios principais (Figura Central). A IA proporciona a automatiza&#xE7;&#xE3;o e padroniza&#xE7;&#xE3;o das imagens e respectivas medi&#xE7;&#xF5;es, o que proporciona diagn&#xF3;sticos mais precisos, superando as limita&#xE7;&#xF5;es encontradas na manipula&#xE7;&#xE3;o por ultrassonografistas humanos. Al&#xE9;m disso, a IA tamb&#xE9;m provou ser um instrumento facilitador para a obten&#xE7;&#xE3;o de resultados satisfat&#xF3;rios, mesmo quando utilizada por n&#xE3;o especialistas. As vantagens oferecidas pela IA tornam o exame muito mais r&#xE1;pido e acess&#xED;vel, uma vez que a dura&#xE7;&#xE3;o da forma&#xE7;&#xE3;o de um profissional especializado &#xE9; um dos desafios da disponibilidade de exames na sa&#xFA;de p&#xFA;blica. Portanto, tendo em vista os fatos mencionados, a IA &#xE9; um recurso favor&#xE1;vel que pode ser altamente implementado nos servi&#xE7;os de ecocardiografia, pois as vantagens apresentadas pelo m&#xE9;todo abrangem diferentes dimens&#xF5;es, desde a precis&#xE3;o dos resultados e integra&#xE7;&#xE3;o de dados para correla&#xE7;&#xE3;o cl&#xED;nica at&#xE9; a preven&#xE7;&#xE3;o de doen&#xE7;as ergon&#xF4;micas relacionadas ao trabalho do ultrassonografista.</p>
</sec>
</body>
<back>
<fn-group>
<fn fn-type="other" id="fn5">
<p>Editor respons&#xE1;vel pela revis&#xE3;o: Daniela do Carmo Rassi Frota</p></fn>
<fn fn-type="supported-by" id="fn6">
<p><bold>Fontes de Financiamento</bold></p>
<p>O presente estudo n&#xE3;o teve fontes de financiamento externas.</p></fn>
<fn fn-type="other" id="fn7">
<p><bold>Vincula&#xE7;&#xE3;o Acad&#xEA;mica</bold></p>
<p>N&#xE3;o h&#xE1; vincula&#xE7;&#xE3;o deste estudo a programas de p&#xF3;s-gradua&#xE7;&#xE3;o.</p></fn>
<fn fn-type="other" id="fn8">
<p><bold>Aprova&#xE7;&#xE3;o &#xC9;tica e Consentimento Informado</bold></p>
<p>Este artigo n&#xE3;o cont&#xE9;m estudos com humanos ou animais realizados por nenhum dos autores.</p></fn>
</fn-group>
</back>
</sub-article>
</article>
