AttriMIL: Revisiting attention-based multiple instance learning for whole-slide pathological image classification from a perspective of instance attributes

Linghan Cai, Shenjin Huang, Ye Zhang, Jinpeng Lu, Yongbing Zhang

Abstract: Highlights•We present AttriMIL with multi-branch attribute scoring.•Region-wise attribute constraint uses spatial patterns to boost sensitivity.•Slide-wise attribute constraint models instance correlations across WSIs.•Pathology adaptive learning exploits the two constraints for refined features.•Experiments show AttriMIL’s superiority, achieving state-of-the-art performance.