BiasPruner: Mitigating bias transfer in continual learning for fair medical image analysis

Nourhan Bayasi, Jamil Fayyad, Alceu Bissoto, Ghassan Hamarneh, Rafeef Garbi

Abstract: Highlights•We analyze bias transfer in continual learning for medical image classification.•Empirical study quantifies bias transfer effects in CL across multiple benchmarks.•We introduce BiasPruner, a framework that mitigates forward and backward bias.•BiasPruner prunes biased units, preserving fairness across sequential tasks.•Our empirical study quantifies bias transfer effects in CL across multiple benchmarks.

External IDs:dblp:journals/mia/BayasiFBHG25