China tests AI tools to extend heart care as doctor shortage

by Xiong Weisheng

BEJING, June 15 (China Economic Net) - For many cardiac patients, risk does not end with surgery. It follows them home after discharge, when follow-up can be delayed and warning signs may be missed.

That gap is global. Pakistan, for example, has fewer than one physician per 1,000 people, while Sub-Saharan Africa averaged about 0.2 in 2022, according to World Bank data. Chinese hospitals, universities and technology firms are testing whether AI ECG models, specialist disease agents and smartphone follow-up tools can extend heart care beyond the ward.

What China is building

At Fuwai Hospital, China’s National Center for Cardiovascular Diseases and one of the world’s largest cardiovascular centres, a team working with cardiologist Wu Yongjian has described a system designed to keep patients visible to clinicians after discharge. In the model presented at the recent Beijing Academy of Artificial Intelligence Conference (BAAI) in Beijing, inpatient records would be combined with follow-up data from questionnaires and wearable devices, with patients interacting through mobile tools, to generate risk scores and escalation alerts between hospital visits.

Fuwai’s broader plan, as presented at the conference, is to encode specialist clinical expertise into disease models that could support long-term patient management at scale. The team said at the forum that it is conducting a multi-centre study to gather clinical evidence, but the study details have not been independently verified in public trial registries.

At Peking University, Hong Shenda is approaching the same problem from the device side. His group developed ECGFounder, an ECG foundation model reported in NEJM AI in 2025, built on more than 10.7 million ECGs from about 1.8 million subjects across 150 label categories. The model was designed to work on single-lead readings as well as standard clinical ECGs, making it relevant to mobile monitoring.

Hong’s team is also testing compact ECG devices that could make AI-assisted ECG analysis more available between hospital visits: a user presses two fingers against a small device and waits about 30 seconds to record a single-lead ECG through a mobile interface. The aim is to create a signal where currently there is often silence.

For such systems to work, alerts must be trusted by doctors. Yidu Tech, a Chinese company that builds clinical AI systems for hospitals, says its evidence-based medicine agent links conclusions to the underlying guideline or literature source, allowing clinicians to verify the original material. The approach reflects a broader theme at the forum: medical AI is moving toward traceable support systems, rather than autonomous diagnosis.

The data problem

These efforts point to the same structural wall: health data often does not flow.

Patient records, device data and insurance information often sit in different systems. No AI model can bridge them without access. The Fuwai team identified this explicitly: its long-term vision requires inpatient data, home monitoring data and insurance data to be connected. The forum presentation framed that connectivity as a goal, not a settled reality.

There is also the question of responsibility. Continuous monitoring only works if the AI knows when to stop and hand a decision back to a human, and if it is clear who bears responsibility when it does not. Regulation is another open question. Speakers noted that approval, safety evaluation and responsibility for open-ended medical AI systems remain unsettled.

Follow-on work from Hong’s group is exploring more ambitious uses of AI ECG, including cardiac biological ageing and future disease-risk prediction. Much of that remains research-stage work. The gap between a compelling demo and a validated clinical tool is where many medical AI projects face their hardest test.

Why it matters beyond China

The approach could matter in countries where doctors, including specialists, are scarce and follow-up care is hard to sustain.

In Pakistan and parts of Africa, low physician density can make regular specialist follow-up difficult. AI-assisted ECG tools and phone-based follow-up systems, if validated, could help health workers identify higher-risk patients earlier.

That does not mean replacing cardiologists. It means using specialist knowledge to triage patients, flag warning signs and support routine follow-up closer to home.

The model would still need local data, clinical validation, affordable devices and clear rules on who responds when software raises an alert. Without those, it risks remaining a hospital pilot. If it works, China’s experiments could offer a template for extending cardiac care in doctor-scarce health systems.