AI Detects Hypertension and Diabetes From Facial Video Scans

What Happened

Scientists have demonstrated that artificial intelligence systems can detect hypertension and diabetes by analyzing short videos of a patient's face, with separate research showing that retinal scans processed by AI may identify individuals at elevated risk of dementia. The findings, reported this week, add to a growing body of evidence that non-invasive imaging combined with machine learning can flag serious chronic conditions before traditional diagnostic pathways are engaged.

How the Technology Works

The facial-video approach works by detecting subtle physiological signals embedded in ordinary camera footage. AI models analyze micro-variations in skin tone caused by blood flow beneath the surface, a technique sometimes called remote photoplethysmography. By processing these signals across thousands of frames, the system extracts cardiovascular markers that correlate with elevated blood pressure and abnormal blood glucose levels, two of the defining features of hypertension and type 2 diabetes respectively.

In the retinal scanning application, AI models examine the fine network of blood vessels visible at the back of the eye. Changes in vessel structure, diameter, and pattern have been associated in clinical literature with neurological and vascular conditions. The AI systems in these studies were trained to identify configurations linked to dementia risk, conditions that may otherwise go undetected until symptoms appear years later.

Why Hypertension and Diabetes Are the Focus

Hypertension affects an estimated 1.28 billion adults worldwide, according to the World Health Organization, and a significant proportion of those individuals remain undiagnosed. Type 2 diabetes affects more than 500 million people globally, with the International Diabetes Federation estimating that roughly half of all cases go undetected. Both conditions are leading risk factors for cardiovascular disease, kidney failure, and stroke. Early detection is widely regarded in clinical medicine as one of the most effective tools for reducing long-term complications and healthcare costs.

Dementia, including Alzheimer's disease, currently has no cure and limited early-stage treatment options. Identifying at-risk populations before cognitive decline becomes apparent is a central goal of current neurological research.

The Role of Retinal Imaging

The eye has long been described by clinicians as a window into systemic health because its blood vessels are directly visible without invasive procedures. Retinal photography is already used in ophthalmology and diabetic screening programs. Applying AI to that existing imaging infrastructure would not require new hardware in clinical settings where retinal cameras are already deployed, potentially lowering the barrier to broader screening programs.

Limitations and Current Status

The wire report does not specify the size of the datasets used to train or validate these models, the clinical settings in which they were tested, or whether the research has been published in peer-reviewed journals. Sensitivity and specificity figures, which measure how accurately a diagnostic tool identifies true positives and true negatives, were not included in the available details. These metrics are standard requirements for regulatory review of medical diagnostic software.

AI-based diagnostic tools in the United States require clearance from the Food and Drug Administration before they can be used clinically. The FDA has established a dedicated pathway for software as a medical device, and a growing number of AI imaging tools have received clearance in recent years across cardiology, radiology, and ophthalmology.

What It Means in Practice

If validated at scale, a facial-video screening tool could be deployed through standard webcams or smartphone cameras, requiring no specialized medical equipment. This would make population-level screening theoretically accessible in low-resource settings and in primary care environments where access to laboratory testing is limited. Retinal-based dementia screening could similarly expand the utility of equipment already present in optometry and ophthalmology clinics.

Neither application described in the reports has received regulatory clearance, and neither has been announced for commercial deployment. Further peer-reviewed publication and independent validation studies would represent the standard next steps before clinical adoption.