Multi-instance curriculum learning for histopathology image classification with bias reduction

Zihao Mi, Jianan Zhang, Xueyu Liu, Guanghui Yue, Junhong Yue, Mingqiang Wei, Yidi Li, Yongfei Wu

Abstract: Highlights•A MIL curriculum learning framework addresses false positives and class imbalance.•Designed novel HNM and instance augmentation strategies to improve performance.•Post-discrimination mechanism enables effective instance augmentation.•The effectiveness of the proposed method is validated on four datasets.