A study conducted in collaboration with the University of Calabria, Humanitas University, the University Federico II of Naples, the University and the University Health Service of Trieste, the University of Zurich, Imperial College London, King’s College London, and Seoul National University, has led researchers to “train” a deep neural network by analyzing 84,895 electrocardiograms and NT-proBNP tests. The algorithm has learned to capture micro-electrical signals that are imperceptible in traditional readings but linked to the presence of biomarkers. When tested on 679 patients in centers in Cosenza and Trieste, the algorithm demonstrated an accuracy rate near 90% in identifying at-risk values.
“Heart failure is one of the major health emergencies in our country,” explains Ciro Indolfi, the lead author of the study and president of the Italian Foundation for Heart and Circulation of the SIC. “In Italy, it affects over one million people, with a prevalence estimated at 1-2% in the general population, and about 80,000 new cases each year. The frequency increases rapidly with age, and the disease is one of the leading causes of hospitalization among the elderly, with a significant impact on mortality, quality of life, families, and the sustainability of the National Health Service.”
Currently, the diagnosis of heart failure often begins with general symptoms such as fatigue, shortness of breath, or swelling in the legs. “To confirm suspicion, a blood sample with laboratory analysis (the NT-proBNP dosage) and a heart ultrasound are necessary,” states Gianfranco Sinagra, President of the Italian Society of Cardiology and co-author of the study. “However, the blood test requires time, costs, and proper facilities, which are not always easily accessible in the territory or smaller centers. The electrocardiogram, on the other hand, is a cost-effective, non-invasive, immediate examination available everywhere: from family doctors’ offices to pharmacies, and even ambulances. Applying AI to the ECG allows us to transform it into a true ‘digital filter’: a tool capable of notifying the doctor immediately when it is essential to request further laboratory and specialized investigations.”
The algorithm does not replace the laboratory biomarker nor does it autonomously diagnose heart failure. Its potential clinical value is to promote earlier recognition of risk and a more efficient selection of follow-ups, especially when blood tests are not immediately available or routinely requested. Its use for opportunistic screening or, in the future, in population programs is a prospect to consider with dedicated prospective studies before widespread implementation.
“These results demonstrate that the ECG contains far more information than is visible to the human eye,” emphasizes Indolfi. “With artificial intelligence, we can for the first time derive an alert signal from a simple and universal test regarding a potential increase in NT-proBNP and guide the patient towards the appropriate assessments.”
“This approach” – continues Sinagra – “could be particularly useful for diagnosing and treating heart failure, especially in areas where access to laboratory and specialized diagnostics is more challenging, contributing to reducing territorial inequalities. Prospective studies are now needed to demonstrate the actual benefits in care pathways and clinical outcomes.”
Cardiovascular diseases remain the leading cause of death in Italy, accounting for about 30% of deaths. The country has a national strategy for oncological diseases, but has not yet adopted a similar comprehensive institutional plan for the prevention and treatment of cardiovascular diseases.
“The Italian Federation of Cardiology, with the support of the European Society of Cardiology, has developed a proposal for a National Strategic Plan for Cardiovascular Health 2024-2027, which highlights the need to make access to prevention, early diagnosis, and innovative therapies more equitable,” recalls Indolfi. “In this context, preventing heart failure, recognizing it earlier, and treating it according to evidence represent a strategic objective to reduce mortality, hospitalizations, disability, and costs,” highlights Sinagra. “The application of Artificial Intelligence to simple and already widespread tools, such as the ECG, can become one of the means for a proximity and quality cardiology for citizens, regardless of where they live. Naturally, for clinical enhancement, the data will need to be placed in the overall context, and therein human intelligence and the expertise of the physician will come into play.”
– Photo Ipa Agency –
