Adaptive unified contrastive learning with graph-based feature aggregator for imbalanced medical image classification

Cong Cong, Sidong Liu, Priyanka Rana, Maurice Pagnucco, Antonio Di Ieva, Shlomo Berkovsky, Yang Song

Abstract: Highlights•A self-supervised learning framework for imbalanced medical image classification.•A adaptive unified loss for simultaneous feature and classifier learning.•A novel ConvGNN-based aggregator for obtaining more discriminative features.•Significant performance improvements on three highly imbalanced datasets.