Age and Sex Estimation Using Artificial Intelligence From Standard 12-Lead ECGs

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Sex and age influence electrocardiogram (ECG) readings.
• Biologic and anatomic factors contribute to sex and age-related ECG variations.
• Deep learning models can analyze ECG signals for demographic prediction.

Purpose of the Study:

• To train a convolutional neural network (CNN) to predict sex and age from 12-lead ECG signals.
• To investigate if the difference between predicted and chronological age can serve as a physiological health indicator.

Main Methods:

• Trained CNNs on 10-second ECG samples from nearly 500,000 patients for sex and age prediction.
• Tested models on an independent cohort of over 275,000 patients.
• Assessed within-individual age estimation accuracy using longitudinal ECG data from 100 patients.

Main Results:

• The model achieved 90.4% accuracy in sex classification and estimated age with an average error of 6.9 years.
• Most patients (51%) showed less than a 7-year age prediction error over decades.
• A predicted age significantly exceeding chronological age was linked to conditions like hypertension and coronary disease.

Conclusions:

• Artificial intelligence applied to ECGs can predict patient sex and estimate age.
• Validated AI-driven physiological age estimation may offer a novel measure of overall health.