<ul><li>Machine learning models can have significant performance disparities between patient groups.</li><li>Traditional subgroup analysis based on metadata may not capture all performance variations.</li><li>Subgroup discovery techniques using learned feature representations can provide more granular subgroup performance reports.</li><li>Subgroup discovery can be an important tool for comprehensive performance validation and monitoring of AI in medicine.</li></ul>

Subgroup Performance Analysis in Hidden Stratifications

Discover more