A deep learning digital biomarker to detect hypertension and stratify cardiovascular risk from the electrocardiogram

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Hypertension is a significant risk factor for cardiovascular disease (CVD).
• Current blood pressure measurement methods are often intermittent and suboptimal.
• There is a need for improved methods to identify hypertension and assess CVD risk.

Purpose of the Study:

• To develop and validate a deep learning model (HTN-AI) for hypertension detection and CVD risk stratification using 12-lead electrocardiogram (ECG) waveforms.
• To assess the association between HTN-AI predicted hypertension probability and adverse cardiovascular outcomes.

Main Methods:

• Trained a deep learning model (HTN-AI) on 752,415 ECGs from 103,405 adults.
• Externally validated HTN-AI in 56,760 adults.
• Utilized Fine-Gray regression to analyze associations between predicted hypertension probability and CVD events.

Main Results:

• HTN-AI demonstrated accurate hypertension discrimination in internal (AUROC 0.803) and external (AUROC 0.771) validation.
• Model-predicted hypertension probability was significantly associated with increased risk of mortality, heart failure (HF), myocardial infarction (MI), stroke, and aortic dissection/rupture.
• Hazard ratios indicated substantial increases in risk for these adverse outcomes.

Conclusions:

• HTN-AI effectively identifies hypertension and stratifies CVD risk using ECG data.
• The model shows potential as a digital biomarker for hypertension-associated CVD.
• HTN-AI may aid in earlier diagnosis and risk assessment for cardiovascular conditions.