Exploring interpretable echo analysis using self-supervised parcels

Sylwia Majchrowska, Anders Hildeman, Ricardo Mokhtari, Tom Diethe, Philip Teare

Abstract: Highlights•Creating human-interpretable features without the need for labels.•Robust heart sub-region identification via self-supervised parcels.•Underscore trade-off between model performance and robustness.•Consistently good performance across varied clinical and patient settings.

External IDs:doi:10.1016/j.compbiomed.2025.110322