Accurate detection of severe aortic stenosis based on self-supervised and ensemble learning of single-view echocardiograms.
A large-scale benchmark for long-tailed learning of chest X-rays.
Self-supervised learning method for echocardiogram videos drives label-efficient fine-tuning on aortic stenosis classification. Presented at IMLH 2022, an ICML workshop.
A Transformer for weakly supervised disease localization that uses a novel feedback loop between (local) radiomics features and (global) chest X-ray features.