AI-powered cardiometabolic assessment from a single facial video
Key result
Facial video machine learning detects hypertension with ~95% accuracy and accurately screens for diabetes.
- ROC-AUC 0.97 (30s diabetes); r=0.968 (BP estimation)
- n=285
Why the study?
Many individuals with hypertension and diabetes remain undiagnosed because current screening relies on clinical encounters or wearable devices that limit population-level coverage. A passive, contactless monitoring system using short-duration spectroscopic facial video could overcome these barriers to enable early detection without physical contact or blood sampling.
Does a machine-learning algorithm applied to short-duration spectroscopic facial and hand video accurately detect hypertension and diabetes and estimate blood pressure compared to cuff-based blood pressure and HbA1c/clinical diagnosis in patients and healthy volunteers?
Study Design
Type
Cohort (n=285)
Multicenter
No
Structured PICO
Does a machine-learning algorithm applied to short-duration spectroscopic facial and hand video accurately detect hypertension and diabetes and estimate blood pressure compared to cuff-based blood pressure and HbA1c/clinical diagnosis in patients and healthy volunteers?
P
Population
285 consenting patients and healthy volunteers in a single-center prospective cohort evaluated for hypertension and diabetes using 5- to 30-second spectroscopic facial and hand videos.
E
Exposure
Machine-learning algorithm applied to short-duration (30-second and 5-second) high-speed spectroscopic video recordings of the face and hands extracting pulse wave dynamics, skin blood-flow patterns, and spectral characteristics without physical contact or blood sampling.
C
Comparator
Concurrent cuff-based blood pressure measurements (reference standard for hypertension: systolic ≥130 mmHg or diastolic ≥80 mmHg per AHA 2017 criteria) and HbA1c values (≥6.5%) or physician-diagnosed diabetes (reference standard for diabetes).
O
Outcome
Diagnostic accuracy for detecting hypertension and diabetes, and blood pressure estimation accuracy (Mean Absolute Percentage Error and model correlation coefficient) using 30-second and 5-second spectroscopic video recordings.surrogate
Main Result
Effect estimate: ROC-AUC 0.97 (30s diabetes); r=0.968 (BP estimation)
Limitations
- Single-center study warranting further validation in larger, multi-center cohorts
- Single-center study requiring further validation in larger, multi-center cohorts
Key expert perspectives
Captured external expert commentary on this paper, strongest first. Original sources are linked where available.
“Our machine-learning algorithm accurately detected hypertension and diabetes from facial spectroscopic video recordings as short as 5 seconds. We intend to validate these findings in larger cohorts across more diverse populations to support real-world application. If validated, this contactless approach could allow people to be screened in everyday settings – without cuffs, blood sampling or a dedicated clinic visit – helping to identify at-risk individuals who would otherwise remain undiagnosed and therefore untreated.”
“It is remarkable that AI-supported technologies are enabling the development of such powerful tools for early disease prevention. Because this approach is quick, easy and contactless, it could be used in many settings beyond hospitals, giving it the potential to reach far more people than traditional screening methods. Detecting these conditions early means treatment and lifestyle changes can start sooner, helping to prevent heart attacks, strokes and other cardiovascular diseases.”
Implication
A contactless machine-learning algorithm applied to spectroscopic facial video can accurately screen for hypertension and diabetes from recordings as short as 5 seconds and estimate blood pressure without a cuff.
Abstract
Introduction Hypertension and diabetes are leading modifiable risk factors for cardiovascular disease, yet a substantial proportion of affected individuals remain undiagnosed. Current screening relies on dedicated clinical encounters or wearable devices, both of which limit population-level coverage. A passive, contactless monitoring system that can be embedded in everyday environments could overcome these barriers and enable early detection at scale. Purpose We aimed to develop and evaluate a machine-learning algorithm that detects hypertension and diabetes and estimates blood pressure from short-duration facial video captured by a spectroscopic camera, without any physical contact or blood sampling. Methods & Results In this single-center prospective cohort study, 285 consenting participants (patients and healthy volunteers) underwent high-speed spectroscopic video recording of the face and hands. Concurrent cuff-based blood pressure measurements served as the reference standard for hypertension (systolic ≥130 mmHg or diastolic ≥80 mmHg, per AHA 2017 criteria), while HbA1c values and physician-diagnosed diabetes (≥6.5% or clinical diagnosis) were used as the reference for diabetes. The algorithm extracted three categories of features from the video data: pulse wave dynamics, skin blood-flow patterns, and spectral characteristics. For hypertension detection, the pulse wave-based analysis achieved 95.0% accuracy using a 30-second recording in an ensemble model (normal blood pressure sensitivity 100.0 ± 0.00%, hypertension sensitivity 89.2 ± 2.30%), and 90.3% accuracy using 5 seconds of data (normal blood pressure sensitivity 98.3 ± 6.22%, hypertension sensitivity 84.5 ± 7.2). For diabetes detection, a model using 30-second videos based on blood flow patterns achieved 88.2% accuracy (Fine Gaussian SVM, ROC-AUC 0.97), and a model using 5-second videos achieved 81.2% accuracy (SVM kernel, ROC-AUC 0.98). Five-fold cross-validation was used for these binary classifications. In addition, for blood pressure estimation using 30-second videos of the face only, a model using three parallel models—Random Forest, XGBoost, and LightGBM—achieved a Mean Absolute Percentage Error (MAPE) of 9.7% and a model correlation coefficient of 0.968. Conclusion A machine-learning algorithm applied to spectroscopic facial video can detect hypertension and diabetes with high accuracy from recordings as short as 5 seconds, and can estimate blood pressure without a cuff. These findings suggest that contactless biometric sensing may serve as a scalable screening tool for cardiovascular risk factors, warranting further validation in larger, multi-center cohorts.ESC2026
Discussion
No takes yet. Share an insight, caveat, or question.
Cite This Study
Uchida et al. (2026) conducted a cohort in Hypertension and diabetes (n=285). Machine-learning algorithm applied to short-duration spectroscopic facial and hand video vs. Concurrent cuff-based blood pressure (AHA 2017 criteria) and HbA1c (≥6.5%) or physician-diagnosed diabetes was evaluated on Detection accuracy for hypertension and diabetes, and blood pressure estimation (Mean Absolute Percentage Error) (ROC-AUC 0.97 (30s diabetes); r=0.968 (BP estimation)). Machine-learning analysis of 30-second spectroscopic facial videos detected hypertension with 95.0% accuracy and diabetes with 88.2% accuracy (ROC-AUC 0.97), and estimated BP with 9.7% MAPE.