Artificial Intelligence

IJCS

Editorial

AI-Enhanced ECGs for the Diagnosis of HFpEF: A Promising Frontier, But Not Yet a Standalone Solution

Heart Failure with preserved Ejection Fraction (HFpEF) is among the most challenging syndromes in contemporary cardiology. Despite accounting for nearly half of all heart failure cases, HFpEF remains chronically underdiagnosed and is associated with significant morbidity and mortality.^, The diagnosis of HFpEF is challenging and relies on clinical history, natriuretic peptide measurements, and echocardiography. Diagnostic scores have been developed that integrate these three key elements.^, Echocardiography is indispensable for identifying structural and functional criteria, but its availability and cost make population-level screening impractical in many healthcare systems. In addition, in many places, a comprehensive echocardiogram is not performed. Likewise, natriuretic peptides are not always available. Therefore, new solutions to facilitate diagnosis are welcome.

In this context, the systematic review and meta-analysis presented in this issue of International Journal of Cardiovascular Sciences, provide important insights into an increasingly discussed alternative: Artificial Intelligence-enhanced Electrocardiography (AI-ECG). Drawing from more than 170,000 patients across three observational studies, the authors report a pooled sensitivity of 0.89, specificity of 0.80, and an AUC of 0.89 for AI-ECG in detecting HFpEF. These findings suggest that deep learning models can extract subtle electrical signatures associated with diastolic dysfunction – signals that may be visually imperceptible on standard ECGs. The appeal is clear: the ECG is universal, inexpensive, rapid, and deeply embedded in clinical workflows. If Artificial Intelligence could reliably identify patients requiring echocardiographic confirmation, screening strategies might be revolutionized.

[…]

Ler artigo completo

IJCS

Editorial

Ethical Use of Artificial Intelligence: Guidelines for Peer Review

Cristiano Marciano Duarte et al. Claudio Tinoco Mesquita

Introduction

Peer review is a key component to the success of scientific articles. High-quality peer review ensures reliable information, improves the efficiency of communication, and facilitates the interpretation of findings for better applicability of the data presented in studies. Recently, with the increasing availability of artificial intelligence (AI) tools, especially large language models (LLMs), several scientific journals have published policies and guidelines regarding their use in the peer review process. LLMs are generative AI tools capable of producing text, figures, and other forms of content based on the datasets on which they have been trained. Their use in scientific publications has gained prominence in improving the quality of scientific writing, translating texts, reformatting references, and supporting peer review. There are still no clear guidelines on the use of LLMs in peer review. In 2024, Li et al. reviewed the editorial practices of 100 high-impact medical journals and found that 78% had clear guidelines on the use of AI in peer review. Among these, 59% (46 journals) explicitly prohibited the use of AI in generating reports, whereas 32 journals (41%) allow its use in a limited manner, provided that confidentiality was maintained and copyright protected. The objective of this article was to present a review on this topic and to propose a rationale for implementing AI in accordance with best practices in ethics and scientific integrity.

[…]

Ler artigo completo

IJCS

Original Article

Diagnostic Performance of Artificial Intelligence-Driven Electrocardiogram Algorithms in Detecting Heart Failure with Preserved Ejection Fraction: A Systematic Review and Meta-Analysis

Gustavo Barbosa David et al. Nicolas Ramos Ramon Huntermann Juan Peres de Oliveira … Maria Eduarda Molinari Caroline de Oliveira Fischer Bacca

Abstract

Background:

Heart failure with preserved ejection fraction (HFpEF) is a highly prevalent syndrome associated with substantial morbidity and mortality. Accurate diagnosis is often challenging and costly. Recently, artificial intelligence-based electrocardiogram (AI-ECG) algorithms have emerged to make the diagnosis of HFpEF less expensive and more accessible. However, there is a lack of data regarding their accuracy.

Objective:

We aimed to conduct a systematic review and meta-analysis to evaluate the diagnostic performance of AI-ECG algorithms in detecting HFpEF.

Methods:

We searched PubMed, Embase, and Cochrane databases for studies evaluating AI-ECG algorithms in diagnosing HFpEF. We extracted true-positive, true-negative, false-positive, and false-negative events to estimate pooled sensitivity, specificity with 95% confidence intervals (CI), and area under the curve. Statistical analysis was performed using R under a random-effects model. Heterogeneity was assessed with I² statistics.

Results:

We included three studies comprising 171,073 patients (66.86% with HFpEF), in which an AI-ECG was evaluated alongside echocardiography, with a left ventricular ejection fraction of ≥ 50%. The patients’ mean age was 65.86 ± 4.5 years, and 48.12% were male. AI-ECG algorithms demonstrated a sensitivity of 0.89 (95% CI: 0.64 to 0.97), specificity of 0.80 (95% CI: 0.65 to 0.89), and an area under the receiver operating characteristic curve of 0.89 (95% CI: 0.85 to 0.92).

Conclusion:

AI-ECG algorithms demonstrate moderate to high sensitivity, specificity, and accuracy in diagnosing HFpEF, supporting their potential for routine screening and early detection.

Ler artigo completo

IJCS

Review Article

How Artificial Intelligence Can Transform Women's Healthcare Across All Life Stages

Larissa Neto Espíndola et al. Elise Sant’Ana Isaías Gláucia Maria Moraes de Oliveira Alexandra Monteiro

Abstract

Artificial intelligence (AI) has been driving significant transformations in healthcare by enhancing diagnostics, personalizing treatments, and optimizing clinical workflows. In women’s health, it offers innovative approaches for prevention, early detection, and management of various conditions throughout life — from adolescence to old age.

Over the past decade, advances in machine learning algorithms, neural networks, and natural language processing have enabled the analysis of large volumes of clinical data with previously unimaginable precision and speed. However, the integration of AI into women’s healthcare remains underutilized, representing a valuable opportunity to reduce social inequalities that influence adverse health outcomes.

However, the use of AI in women’s healthcare raises important ethical concerns, particularly the need to protect patients’ sensitive data, avoid algorithmic biases that may perpetuate stigma or inequality, and ensure safe, equitable, and responsible implementation. Ethical governance — with national and international guidelines and regulations — is essential to ensure that AI benefits all women, including vulnerable and underrepresented populations, without exacerbating existing disparities.

This manuscript aims to synthesize the main applications of AI in women’s health across different life stages, identifying advances, opportunities, and challenges in this context. Given the complexity and multidisciplinary nature of the topic, we conducted a scoping review – an appropriate methodology for exploring emerging and still underdeveloped fields. In addition to scientific databases, we included a complementary search in digital health app libraries to identify technological solutions already implemented or under development for women’s health.

Ler artigo completo

IJCS

Editorial

Forty Years of Silent Revolutions in ICUs: From Clinical Records to Intelligent and Predictive Surveillance

Rubens Carmo Costa-Filho et al. Hugo Caire Castro Faria-Neto

Introduction

In the mid-1980s, as modern intensive care medicine was still consolidating as a specialty, a conceptual challenge emerged that would redefine critical care: how to objectively and comparably quantify the severity of illness across different patients and units. In 1985, William A. Knaus et al. published the Acute Physiology and Chronic Health Evaluation II (APACHE II). This landmark model translated clinical observations into a mathematical framework capable of estimating hospital mortality based on 12 physiological and chronic health variables collected during the first 24 hours of ICU admission.

At that time, computing power was limited, and data were collected manually, the technological ceiling of that era.

[…]

Ler artigo completo

IJCS

Original Article

Performance Benchmarking of Open-Source Large Language Models on the Brazilian Society of Cardiology's Certification Exam

João Victor Bruneti Severino et al. Matheus Nespolo Berger Pedro Angelo Basei de Paula Filipe Silveira Loures … Solano Amadori Todeschini Eduardo Augusto Roeder Maria Han Veiga José Knopfholz Gustavo Lenci Marques

Abstract

Background:

Large language models (LLMs) have made a significant impact in medicine and demonstrate substantial promise for further development. However, most of the existing research has predominantly centered on English-language tasks with lower medical complexity. This underscores the importance of investigating the performance of state-of-the-art LLMs in more complex specialties, such as cardiology, and in languages beyond English, such as Portuguese.

Objective:

This study aimed to evaluate and compare leading LLMs based on their performance on the validated cardiology knowledge assessed by the Brazilian Society of Cardiology’s (SBC) Certification Exam.

Methods:

This study conducted a comparative analysis of 23 LLMs in the context of the SBC’s Certification Exam. The exam consists of 100 multiple-choice questions, 20 of which include images that cannot be processed by all LLMs. Therefore, these image-based questions were excluded from the analysis.

Results:

Proprietary LLMs showed a varying performance, with GPT-4o achieving the highest success rate at 62.25%, followed by Claude Opus at 60.25%. In the medium-sized model category (up to 100 billion parameters), Claude Haiku reached 47.25%. Among open-source models, Llama3 70B Instruct scored 53.50% in the large model category (over 100 billion parameters), while Llama3 8B achieved 36.25% in the small model category (under 20 billion parameters).

Conclusions:

Both proprietary and open-source LLMs underperformed on the test, failing to meet the exam’s cutoff score. Although larger models generally achieved better results, some medium-sized models — such as Llama3 70B Instruct and Claude Haiku—showed noteworthy results. The LLMs lacked specialized knowledge in cardiology and faced challenges in understanding Portuguese, revealing a significant gap in current AI capabilities and emphasizing the need for improvements.

Ler artigo completo

IJCS

Editorial

Dr. Strange, Artisanisation, and the Percolative Artificial Intelligence: Where Should We Go?

Erito Marques de Souza Filho et al. Leonardo Silva de Lima Raphael de Moraes Dutenkerfer Célia Martins Cortez

Percolation appears as a ubiquitous phenomenon in nature, with significant practical relevance in the study of networks. Its origins connect to the pioneering work of Paul Flory in 1940 and the foundational research of Broadbent and Hammersley in 1957, which established its intricate mathematical principles,^, incorporating concepts from probability, physics, and graph theory. In a network composed of nodes and edges, understanding how percolation operates enables researchers to evaluate the effects of adding or removing an edge or a node on network connectivity. A simple example emerges in coffee brewing: an excessive amount of ground coffee can obstruct the water, preventing full percolation. This theory applies to porous materials development, including membranes, filters, and absorbent substances. The extracorporeal circulation membrane exemplifies this application, serving as a crucial tool for sustaining cardiac and respiratory functions when conventional methods prove ineffective. Additionally, a pilot study proposed these ideas as a conceptual framework for explaining atrial fibrillation termination.

Percolation Theory also plays a crucial role in distribution networks, including energy and water systems. It contributes to disease propagation assessments, such as in epidemiology,^, the study of geometric characteristics of viral capsids to understand specific geometries favored in natural evolution, the dissemination of information in social networks, and forest fire analysis. Kotlarz et al. applied the theory to examine the network effects of beta-amyloid protein deposition in Alzheimer’s disease. Rabinovitch et al. used percolation theory to investigate fibrosis in cardiac cells. Their findings suggest that fibrosis propagation in these cells mirrors the behavior observed in forest fires.

[…]

Ler artigo completo

IJCS

Editorial

Rescuing Lives: Reflecting on the Future of Cardiopulmonary Arrest Care

Sergio Timerman

Current Overview

Between 2018 and 2022, data from 2,154 cases attended by a mobile emergency service were analyzed. This analysis not only reinforced the severity of CPA, with mortality rates still alarmingly high, but also emphasized the impact of early interventions. Bystander-initiated cardiopulmonary resuscitation (CPR), the use of automated external defibrillators (AEDs), and ongoing training of emergency teams are critical pillars for changing this scenario.^,

However, the effectiveness of care is still heavily reliant on one critical factor: time. Every minute of delay drastically reduces survival chances, reinforcing the importance of robust and quick response systems, both in urban and rural settings.^,

[…]

Ler artigo completo

IJCS

Review Article

Artificial intelligence for nuclear cardiology: Perspectives and challenges

Caitlyn Gilbert et al. Alec Chunta Robert Miller

Abstract

Recent advancements in artificial intelligence (AI) have significantly enhanced the efficiency and quality of nuclear cardiology imaging. AI algorithms, such as generative adversarial networks (GAN) and deep learning (DL) models, are improving image acquisition by reducing noise and correcting motion to minimize radiation exposure while maintaining diagnostic accuracy. AI has also demonstrated the potential to create synthetic attenuation correction (AC) imaging, accounting for soft-tissue attenuation without added computed tomography (CT) imaging. In patients with CT attenuation imaging, AI can automate the segmentation of coronary artery calcification (CAC), epicardial adipose tissue (EAT), and other structures to maximize the clinical information obtained from these added studies. Furthermore, machine learning (ML) models that combine clinical, imaging, and stress test data are advancing risk prediction for coronary artery disease (CAD) and ischemia, assisting clinicians in decision-making and improving patient outcomes. While many of these applications have been developed using myocardial perfusion imaging (MPI), they are increasingly applied to other nuclear cardiology studies. This review explores these AI applications and their impact on diagnostic accuracy, workflow efficiency, and risk stratification while also highlighting potential future directions and challenges.

Ler artigo completo

IJCS

Editorial

Use of Artificial Intelligence in Scientific Publications

Marcella dos Santos Lopes da Silva et al. Jayme Portasio Lopes Thaís Amadeu Pol Claudio Tinoco Mesquita

Social explosion

The creation of various types of AI has led to the generation of a large cluster of data, called “Big Data”. This data storage brings countless possibilities for use, such as in health and education, data interpretation, recognition of writing patterns, and analysis and construction of texts. The social explosion of AI occurred in November 2022, with the launch of the chatbot ChaGTP, developed by OpenAi, which was the first use of generative AI to be successful and reach millions of users quickly. Generative AI consists of algorithms that, once trained with extensive databases, can provide original texts, images, and audios, based on learned patterns.

Currently, the use of high-performance software, chatbots and virtual assistants permeates society and impacts all spheres of human thought thanks to the constant development of NLP, computer vision and robotics. However, along with these advances come ethical and social issues that need to be addressed, such as privacy, security, the use of original data created by third parties, plagiarism and the impact on the job market.

[…]

Ler artigo completo

IJCS

Original Article

Artificial Intelligence as a Tool to Support Decision-Making in the Management of Intraoperative Hypotension

Giorgio Pretto et al. Diogenes de Oliveira Silva Victor Hugo de Freitas Morales

Abstract

Introduction

Intraoperative hypotension (IOH) is a frequent complication associated with adverse cardiovascular, cerebral, and renal outcomes, with increased mortality. Recent evidence indicates that cumulative exposure time to hypotension is a significant factor, related to both the duration and severity of hypotensive episodes. The Hypotension Prediction Index (HPI) algorithm, with active alerts, showed that hypotension periods can be significantly reduced.

Objectives

To describe an artificial intelligence (AI) algorithm for hypotension prediction integrated into real-time anesthesia record-keeping software and Clinical Decision Support Systems (CDSS) that can warn anesthesiologists about the possibility of the onset of hypotension up to 20 minutes in advance.

Methods

This prediction tool incorporates four machine learning classifier models developed using the XGBoost, a supervised learning library that works with decision trees with gradient boosting. These models were trained, validated, and tested based on a database of approximately 0.5 million anesthesia records, using real world data.

Results

Accuracy ranged from 84.92% to 89.07%, and sensitivity ranged from 82.15% to 90.86%, with the best results found in five-minute predictions and the worst in twenty minute predictions. The algorithm was tested in all types of surgical procedures but only using adult data. Separate analyses were performed on small, medium, and large procedures of different surgical specialties with similar results.

Conclusions

The present study demonstrated the use of an AI algorithm integrated into anesthesia record-keeping software, showing a high accuracy for hypotension prediction. This algorithm is available for all types of anesthesia care in adult procedures.

Ler artigo completo

IJCS

Review Article

Computer-Interpreted Electrocardiograms: Impact on Cardiology Practice

Shyla Gupta et al. Anthony H. Kashou Robert Herman Stephen Smith … Adam May Ana G. Múnera Echeverri Mildren Del Sueldo Ana C. Berni Juan Farina Sebastian Garcia-Zamora Adrian Baranchuk

Abstract

In the realm of modern cardiology, the integration of computer-interpreted electrocardiograms (CI-ECGs) has marked the beginning of a new era of diagnostic precision and efficiency. Contemporary electrocardiogram (ECG) integration systems, applying algorithms and artificial intelligence, have modernized the interpretation of heart rhythms and cardiac morphology. Due to their ability to rapidly analyze and interpret ECG recordings CI-ECGs have already profoundly impacted clinical practice. This review explores the evolution of computer interpreted ECG technology, evaluates the pros and cons of current automatic reporting systems, analyzes the growing role of artificial intelligence on ECG interpretation technologies, and discusses emerging applications that may have transformative effects on patient outcomes. Emphasis is placed on the role of ECGs in the automatic diagnosis of occlusion myocardial infarctions (OMI). AI models enhance accuracy and efficiency in ECG interpretation, offering insights into cardiac function and aiding timely detection of concerning patterns for accurate clinical diagnoses. The shift to AI-driven diagnostics has emphasized the importance of data in the realm of cardiology by improving patient care. The integration of novel AI models in ECG analysis has created a promising future for ECG diagnostics through a synergistic fusion of feature-based machine learning models, deep learning approaches, and clinical acumen. Overall, CI-ECGs have transformed cardiology practice, offering rapid, accurate, and standardized analyses. These systems reduce interpretation time significantly, allowing for quick identification of abnormalities. However, sole reliance on automated interpretations may overlook nuanced findings, risking diagnostic errors. Therefore, a balanced approach in integrating automated analysis with clinical judgment is necessary.

Ler artigo completo

IJCS

Review Article

Phenotype mapping of heart failure with preserved ejection fraction

Evandro Tinoco Mesquita et al. Debora Carvalho Grion Miguel Camargo Kubrusly Bernardo Barcelos Fernandes Fumagalli Silva Érico Araújo Reis Santos

Abstract

Heart failure with preserved ejection fraction (HFPEF) has become the main phenotypic model of heart failure (HF) in community and referral patients in Brazil and in the world. Despite advances in the development of new drugs for HF treatment, there has been no significant improvement in mortality of this condition.

According to many studies, this can be explained by the heterogeneous nature of HF physiopathology, whose basic mechanisms may result in different clinical presentations, culminating in the emerging of different phenogroups in this syndrome. In this context, phenotype mapping of HFPEF has emerged as a possible solution, since it enables the development of clinical trials that establish specific therapeutic strategies for each phenotypic profile.

New technologies in the field of artificial intelligence have enabled the assessment of a large volume of data and infer intrinsic patterns and different outcomes. Thereby, it is possible to obtain mutually exclusive categories of HFPEF, with a phenotype mapping of the syndrome and grouping of patients according to their phenotypic features. Besides, other diseases can have the same clinical phenotype but different pathophysiological basis, the so called “phenocopies”.

These tools enable the analysis and categorization of the wide spectrum of heart failure, contributing to solve the dilemmas of the treatment of this syndrome.

Ler artigo completo