RESEARCHERS led by scientists from Ateneo de Manila University have developed an artificial intelligence (AI) model that can accurately predict heart function using non-invasive physiological measurements, offering a potential alternative to more complex cardiovascular tests, this was reported by Danika Geronimo of the Ateneo Research Communications Department.
According to the World Health Organization, heart disease among people ages 20 to 29 has been rising alongside metabolic conditions such as obesity, hypertension, high cholesterol and diabetes.
Advanced assessments of heart function often require specialized equipment and trained personnel typically found only in major hospitals, limiting access to early diagnosis and monitoring.
The international research team, led by Patricia Angela Abu of Ateneo’s Department of Information Systems and Computer Science, developed a neural network that estimates cardiac index, a key measure of how effectively the heart pumps blood. Cardiac index incorporates indicators such as heart rate, stroke volume index and cardiac output, which physicians use to evaluate cardiovascular performance and guide treatment.
The AI model analyzes physiological data collected through non-invasive adhesive sensors placed on a patient’s skin. According to the researchers, the system achieved a classification accuracy of 97.78 percent in predicting cardiac index.
The approach could reduce reliance on specialized hemodynamic analyzers and invasive procedures while making advanced heart monitoring more accessible in clinics that lack sophisticated diagnostic equipment.
The study used data gathered from non-invasive devices, including body composition, blood pressure and blood flow analyzers. Physiological measurements collected from adhesive sensor patches were processed by the AI model to estimate cardiac function.
The researchers said the results demonstrate that advanced cardiovascular assessment can be performed using fewer and less complex measurements while maintaining a high level of accuracy. They said the technology could expand access to heart monitoring, particularly in underserved communities and healthcare facilities with limited resources.
The team plans to validate the model using more diverse patient populations and explore whether the number of required physiological measurements can be reduced further without compromising performance.
The findings were published in the April 2026 issue of the journal Bioengineering in a paper titled “Robust Non-Invasive Cardiac Index Prediction via Feature Integration and Data-Augmented Neural Networks.” Aside from Abu, the study was authored by Chih-Hao Chang, Mei-Ling Chan, Yu-Hung Fang, Po-Lin Huang, Tsung-Yi Chen, Tsun-Kuang Chi, I. Elizabeth Cha, Tzong-Rong Ger, Kuo-Chen Li, Shih-Lun Chen, Liang-Hung Wang and Jia-Ching Wang.
