Facial AI Shows Promise for BP Screening

Deep learning model identifies hypertension from facial images with 83% accuracy, with zygomatic and cheek regions performing nearly as well as whole-face analysis

By
Mouj Hijazi
January 14, 2026
3 min

Conexiant

Overview

A study demonstrates that deep learning analysis of facial images can identify hypertension with 83% accuracy, using specific facial regions. This non-invasive method shows comparable performance to traditional diagnostic techniques and may enhance hypertension screening adherence.

Background

Hypertension is a prevalent condition that often goes undetected due to low screening adherence and asymptomatic disease onset. Traditional blood pressure measurement methods can be invasive and may introduce biases, such as white-coat hypertension. The development of non-invasive, scalable screening tools is essential for improving hypertension detection and management.

Data Highlights

{'AUC': {'Zygomatic region': 'N/A', 'Cheek region': 'N/A', 'Forehead': 'N/A', 'Nose': 'N/A', 'Jaw': 'N/A', 'Lip': 'N/A'}, 'F1-score': {'Zygomatic region': 'N/A', 'Cheek region': 'N/A', 'Forehead': 'N/A', 'Nose': 'N/A', 'Jaw': 'N/A', 'Lip': 'N/A'}}

Key Findings

Deep learning analysis achieved 83% accuracy in identifying hypertension from facial images.
Models using only the zygomatic or cheek regions achieved 82% accuracy, comparable to the full-face model.
The proposed method may address barriers in hypertension detection, including low screening adherence.
Facial regions such as the zygomatic and buccal areas showed specificity in identifying hypertension.
The approach requires only standard cameras, making it operable in everyday environments.

Clinical Implications

This facial AI screening method could serve as a complementary tool for hypertension detection, potentially increasing screening rates in various populations. Clinicians should consider integrating this technology into routine practice while remaining cautious about its current limitations and the need for further validation.

Conclusion

The study highlights the potential of facial AI as a non-invasive screening tool for hypertension, warranting further research to validate its effectiveness across diverse populations.