Heart Disease Clues Found in an Opera Singer's Voice

A research team at Florida International University (FIU) in the United States has developed technology that uses artificial intelligence (AI) to predict myocardial infarction from heart sounds alone. The team, led by Professor Joshua Hutcheson, successfully detected cardiac abnormalities with 95% accuracy in experiments with rats and is currently preparing for clinical application in humans.

The starting point for this research was unexpectedly opera. Professor Hutcheson, who is married to an opera singer, realized that just as vocal nodules change a singer's voice, calcium deposits forming on the aortic valve would also alter heart sounds. He subsequently developed an AI-based heart sound analysis system with Research Professor Valentina Dargam.

AI Captures Subtle Abnormalities That Doctors Miss

Professor Dargam explained, "While doctors excel at distinguishing between normal and abnormal heart sounds, it's nearly impossible to immediately diagnose specific diseases." In current medical practice, even when abnormalities are detected through a stethoscope, additional tests are required to determine the exact condition.

The AI algorithm developed by the research team fills this gap. The system learns from a pre-collected database of heart sounds and then analyzes incoming heart sounds in real-time through comparison. Experimental results showed a 95% detection rate for cardiac abnormalities in rats, with accurate identification of specific diseases reaching 86%.

Heart Disease: The Leading Cause of Death Worldwide

The World Health Organization (WHO) identifies cardiovascular disease as the leading cause of death globally. Heart diseases, including myocardial infarction, are difficult to detect early due to unclear initial symptoms, and once they occur, the treatment window is extremely short. This is why prevention and early diagnosis are paramount.

Currently, diagnosing heart disease requires expensive equipment such as electrocardiograms (ECG), echocardiography, and CT scans, along with specialized personnel. In contrast, the newly developed AI heart sound analysis system can be used with simple tools like a stethoscope, potentially serving as a screening tool for heart disease even in areas with limited medical access.

Human Clinical Trials: The Next Step

The research team is currently preparing for human clinical trials, moving beyond the animal testing phase. Since the heart structure and beating patterns of rats and humans differ, the AI model must be retrained and more human heart sound data must be collected.

Professor Hutcheson predicted, "If we can predict diseases simply by listening to heart sounds, the patient screening process in medical settings will be revolutionarily simplified." Particularly if used as a rapid screening tool in primary care facilities or emergency rooms, it could enable early detection of serious heart conditions and advance treatment timing.

[AI Analysis] Medical AI: From Auscultation to Diagnosis

This research demonstrates that AI is evolving from an 'assistive tool' to a 'diagnostic partner' in medical settings. In recent years, AI has proven human expert-level performance in various fields including medical imaging analysis, pathological diagnosis, and drug development.

Heart sound analysis AI is particularly significant as a non-invasive and cost-effective diagnostic method. If combined with smartphone-based stethoscopes in the future, an era could open where people can easily monitor their heart health at home.

However, to improve the accuracy of AI diagnosis, sufficient data from patients of diverse races, age groups, and underlying conditions must be secured. Additionally, regulatory approval and gaining the trust of medical professionals are essential for the clinical adoption of medical AI. This is why the research team's next steps are drawing attention.