Skip to yearly menu bar Skip to main content


Poster
in
Affinity Workshop: Muslims in ML

Unraveling the Effects of Age-Based Distribution Shifts on Medical Image Classifiers

Kumail Alhamoud · Yasir Ghunaim · Motasem Alfarra · Philip Torr · Tom Hartvigsen · Bernard Ghanem · Adel Bibi · Marzyeh Ghassemi

Keywords: [ model evaluation ] [ Dataset Shift ] [ Medical Imaging ] [ Fairness ] [ robustness ]


Abstract:

Medical AI underperforms when faced with new data distributions. A notorious example of this is the Epic Sepsis Model (ESM), whose predictive capabilities diminished upon testing with unseen patient data distributions. We study the effects of subpopulation shifts on medical image classifiers using two representative datasets. Our results highlight the nuanced effects of class distribution imbalances on performance drops, the significance of comprehensive evaluation strategies, and the need to collect diverse samples to advance medical AI deployment.

Chat is not available.