AdaptMIL: Domain Adaptation for Multiple Instance Learning in Digital Pathology

Key points

• Pathology foundation models (PFMs) support a wide range of histopathology tasks on whole-slide images through multiple instance learning (MIL).1
• AdaptMIL is motivated by our finding that directly applying domain adaptation objectives during MIL training can interfere with MIL aggregation learning, often resulting in performance degradation over standard training.1
• AdaptMIL resolves this by decoupling representation-level domain alignment learning from MIL aggregation learning through a two-stage training procedure.1
• We show that AdaptMIL improves cross-hospital performance and outperforms alternative adaptation methods across experiments spanning multiple WSI datasets, tasks, PFMs, MIL architectures, and domain adaptation objectives.1

Sources (1)

• [1]AdaptMIL: Domain Adaptation for Multiple Instance Learning in Digital Pathology
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 10:52 AM

  To address this challenge, we propose AdaptMIL, a training strategy for integrating domain adaptation into MIL training.

  Pathology foundation models (PFMs) support a wide range of histopathology tasks on whole-slide images through multiple instance learning (MIL).

Extractive summary: sentences quoted from the sources.