AI detects hypertension and diabetes via short facial videos

Researchers from the University of Tokyo and the Institute of Science Tokyo have developed an AI algorithm capable of detecting hypertension and diabetes through short video clips of a person's face and palms. The technology, presented at the European Society of Cardiology (ESC) Congress 2026 in Munich, offers a non-invasive alternative to traditional screening methods.

The system utilizes a specialized spectroscopic camera to analyze subtle physiological signals invisible to the naked eye. By monitoring pulse-wave dynamics, skin blood flow patterns, and the spectral characteristics of skin coloring, the AI screens for chronic conditions without blood draws or blood pressure cuffs. In a study involving 215 participants—comprising both healthy volunteers and diagnosed patients—the AI demonstrated significant precision. For hypertension, the system achieved 90.3% accuracy using a video clip as short as five seconds, rising to 95% accuracy with a 30-second clip. For diabetes, the AI identified the condition with 81.2% accuracy in five seconds and 88.2% accuracy in 30 seconds.

The challenge of silent killers

Hypertension and diabetes are frequently described as "silent killers" because they often progress for years without obvious early symptoms. This lack of visibility leads to millions of undiagnosed cases globally. Because traditional screening requires clinical equipment or invasive tests, many high-risk individuals avoid check-ups until they experience severe symptoms or acute health crises.

Implications for mass screening

This shift toward optical detection could fundamentally change how preventative healthcare is delivered. By integrating this technology into mundane environments—such as pharmacies or via smartphone interfaces—health providers could implement mass screening on a scale previously impossible. The ability to identify high-risk individuals through a brief camera session allows for earlier medical intervention, potentially preventing the catastrophic cardiovascular or organ damage associated with these conditions. Early detection is critical in reducing the global incidence of strokes and heart attacks.

Future outlook

While the screening accuracy is high, the researchers are continuing to refine the system's capabilities. The current focus remains on the binary detection of these conditions rather than precise measurement. Future developments will likely focus on expanding the participant pool to ensure the algorithm remains accurate across diverse skin tones and demographics. For now, the technology serves as a powerful triage tool to prompt necessary clinical tests for those flagged by the AI.