Using explainable AI to investigate electrocardiogram changes during healthy aging -- from expert features to raw signals

Explainable AI revealed hidden patterns in aging hearts: By using interpretable machine learning methods on electrocardiogram data from healthy individuals, researchers identified specific ECG changes associated with aging that traditional analysis might have missed—particularly changes in breathing rates and heart rhythm variability (SDANN values) in elderly individuals.

The P-wave emerged as a key aging indicator: Deep learning models highlighted that the P-wave (a specific component of the ECG signal) plays a central role in predicting age across all age groups, suggesting this feature changes systematically as people age and warrants closer clinical attention.

Raw signals provided insights beyond traditional measurements: The study compared analysis of raw ECG signals directly against conventional ECG features, demonstrating that examining the underlying data can reveal age-related relationships that feature-based approaches alone might overlook—important for developing more comprehensive diagnostic tools.

Responsible AI methods enabled trustworthy medical insights: By employing explainable AI techniques, the researchers made their models' decision-making processes transparent and interpretable, allowing clinicians and researchers to understand why the models identified certain patterns as age-related rather than treating them as "black boxes."

Distinguishing normal aging from disease is clinically crucial: This work specifically focused on healthy individuals to isolate age-related ECG changes from disease-related changes, helping establish what constitutes normal cardiovascular aging versus pathological conditions—essential for accurate diagnosis and appropriate clinical care.