University of British Columbia
Model weights not public. Contact creators for more information.
Prototype-based neural network for interpretable, uncertainty-aware classification of aortic stenosis (AS) severity from B-mode echocardiography videos. Rather than a black-box prediction, ProtoASNet bases its output on similarity scores between the input video and a set of learned spatio-temporal prototypes (typically highlighting valve calcification and restricted leaflet motion), and uses an abstention loss to flag ambiguous/uncertain cases for expert review. Evaluated on a private clinical dataset and the public TMED-2 dataset, it achieved balanced accuracy of 80.0% (private) and 79.7% (TMED-2), improving to 82.4% when uncertain cases are excluded.
Architecture
Hybrid
Prototype-based network: extracts spatio-temporal feature vectors from echo video, compares them against learned class prototypes via similarity scores, and aggregates these for both AS severity classification and aleatoric uncertainty estimation
Added to catalog
2026-08-14
TMED-2 (Tufts Medical Echocardiogram Dataset)
Public aortic-stenosis/view-classification benchmark; exact N not confirmed from paywalled sources, left blank rather than guessed
UBC/Vancouver General Hospital Private Echocardiography Dataset (ProtoASNet)
Aortic stenosis severity classification from echocardiography video with per-case aleatoric uncertainty estimate