An international research team led by a Filipino computer scientist has developed an artificial intelligence model that predicts how effectively the human heart pumps blood with 97.78 percent accuracy — using only basic readings taken from non-invasive sensors placed on the skin, according to a study published in the April 2026 issue of the peer-reviewed journal Bioengineering.
The model was developed by a team led by Patricia Angela R. Abu of the Ateneo de Manila University Department of Information Systems and Computer Science, in collaboration with researchers from Taiwan and China. Their paper, titled "Robust Non-Invasive Cardiac Index Prediction via Feature Integration and Data-Augmented Neural Networks," marks a significant step in making advanced cardiac monitoring accessible beyond major hospitals.
What the Cardiac Index Measures — and Why It Has Been Hard to Access
The cardiac index is a clinical metric that describes how much blood the heart pumps relative to a patient's body size. Physicians use it to evaluate heart function and guide treatment decisions, drawing on physiological indicators including heart rate, stroke volume index, and cardiac output.
Measuring the cardiac index through conventional methods requires a specialized hemodynamic analyzer, a controlled clinical setting, and personnel trained to operate the equipment. According to the study, this combination is largely confined to major hospitals in urban centers, placing the diagnostic tool out of reach for patients in provincial and rural communities.
Skin Sensors and a Neural Network at the Core
The researchers fed data from non-invasive Internet of Things sensing devices into an artificial neural network. The instruments used were the TERUMO ES-P2000 blood pressure monitor, the PhysioFlow PF07 Enduro cardiac hemodynamic analyzer, and the InBody 720 body composition analyzer. Physiological measurements were collected through adhesive sensor stickers placed directly on the patient's skin — no injections, catheters, or invasive procedures required.
With three physiological parameters as input, the neural network achieved a classification accuracy of 97.78 percent, which the authors report substantially outperforms traditional approaches. The model also held up under a two-parameter input condition, suggesting the number of measurements required could be reduced further without a significant loss in predictive reliability.
Institutional Oversight and Research Credentials
The study received institutional review board approval under application number 202501987B0, confirming the research met established ethical standards for clinical data use. The full author list includes 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, Jia-Ching Wang, and Patricia Angela R. Abu.
Rising Cardiovascular Disease Among Young Adults Worldwide
The World Health Organization has flagged rising cardiovascular disease among adults aged 20 to 29, driven by increases in obesity, hypertension, hyperlipidemia, and diabetes in younger populations worldwide. The study situates its findings within this broader public health concern, noting that earlier and more accessible cardiac assessment tools are increasingly needed.
In the Philippines, advanced diagnostic capacity is concentrated in Metro Manila and a limited number of regional centers, the study notes. Detailed heart monitoring remains difficult to access for patients in provincial and rural areas where resident cardiologists are scarce or absent entirely.
Portable Devices Could Shift Assessments to Rural Health Units
The researchers argue that a model operating from readings taken with portable, non-invasive devices could shift part of cardiac assessment to clinics and rural health units that currently lack specialized hemodynamic equipment or trained cardiology staff. This would not replace cardiologist oversight but could enable frontline health workers to flag patients who need urgent referral.
The study's findings suggest that the barrier to cardiac index measurement may be reducible from a multi-thousand-peso specialist consultation in a city hospital to an adhesive sticker and a portable sensor — a shift with significant implications for community-level health screening in lower-resource settings.
Validation in Broader Populations Still Required
According to the published paper, the 97.78 percent accuracy figure reflects laboratory performance on the study cohort. Clinical deployment would require validation in real-world patient populations across more diverse demographic groups. The research team has stated plans to validate the approach in more diverse populations and to explore reducing the number of measurements the model requires before any broader clinical application.
By the Numbers
- 97.78% — classification accuracy achieved by the AI model using three physiological parameters
- 3 — number of sensing instruments used (TERUMO ES-P2000, PhysioFlow PF07 Enduro, InBody 720)
- 2 — minimum number of parameters under which the model remained effective
- 13 — total number of authors on the published paper
- 202501987B0 — institutional review board application number for ethical approval
- 20 to 29 — age range flagged by the World Health Organization for rising cardiovascular disease risk
Why This Matters
The Ateneo-led study offers a credible pathway to democratizing cardiac monitoring in a country where advanced hemodynamic diagnostics are geographically and economically out of reach for millions of Filipinos outside major urban centers. With the WHO flagging cardiovascular disease as a growing crisis among young adults, early detection tools that function without specialist equipment carry direct public health weight. Clinical validation across broader populations remains the critical next step before real-world deployment.
Photo credit: Photo courtesy of Ateneo de Manila University / Bioengineering Journal