Bengaluru: Narayana Health’s Clinical Research team, in collaboration with its Advanced Analytics and AI division, Medha AI, has developed an artificial intelligence (AI) model which is indicated to be capable of predicting left ventricular ejection fraction (EF) from standard ECG images in near real-time.
The tool is designed with the aim to support early identification of heart failure and expand diagnostic access, particularly in resource-limited settings.
According to the company, the AI model leverages widely available ECG machines and cloud-based analytics to address limited access to echocardiography (ECHO) especially in rural areas.
It is indicated that the model is trained on over 100,000 ECG images paired with ECHO reports. According to the company, the model has shown strong performance in external validation across 14 tertiary centres, covering more than 57,000 patients. It identified 97% of individuals with severely reduced EF (≤35%) and flagged them an average of 58 days before ECHO was eventually ordered.
The tool has been integrated into Narayana Health’s in-house Electronic Medical Record system, Athma, and is undergoing validation at key locations including the Narayana Institute of Cardiac Sciences, Mazumdar Shaw Medical Center, select clinics in Bengaluru, and the Rabindranath Tagore International Institute of Cardiac Sciences in Kolkata.
Designed to deliver results in near real-time, the model aims to aid triage and care planning without disrupting clinical workflows. It is also indicated that future development is underway to expand its capabilities to detect other echocardiographic abnormalities and early signs of coronary artery disease.
The organization plans to deploy the model via its mobile medical units, targeting underserved regions as part of its broader digital public health strategy. Regulatory submissions and efforts to enable open-access distribution are also in progress.
The model has been presented at several international conferences including the Heart Rhythm Society Conference (Boston), APHRS (Sydney), IHRSCON (Kolkata), and HRX (Atlanta). It has also been published in the Indian Heart Journal and received the BMJ South Asia Award for Digital Innovation.
