Search papers, labs, and topics across Lattice.
The paper introduces Concept-Guided Pruning and Representation Learning (CGRL), a novel framework for weakly supervised whole-slide image classification that enhances multiple instance learning (MIL) by integrating class-level concept prototypes. By implementing concept-relevance pruning, CGRL selectively retains the most informative patches based on their similarity to disease concepts, which significantly improves the alignment of instance features with class semantics. Experimental results on TCGA-BRCA and TCGA-NSCLC datasets reveal that CGRL consistently boosts accuracy and macro-F1 scores across various MIL methods while reducing computational costs.
Class-level semantic concepts can dramatically enhance patch selection and representation learning in weakly supervised image classification, leading to significant performance gains.
Weakly supervised whole-slide image (WSI) classification is widely used in computational pathology because slide-level labels are easier to obtain than dense region annotations. Existing multiple instance learning (MIL) methods often aggregate large bags of patch embeddings using mainly visual cues, which can retain many non-informative patches and provide weak alignment between instance features and class-level disease semantics. We propose Concept-Guided Pruning and Representation Learning (CGRL), a simple framework that introduces class-level concept prototypes derived from disease prompts into the MIL pipeline. First, concept-relevance pruning ranks patch instances by their similarity to class concepts and retains the top-K concept-relevant patches for downstream MIL aggregation. Second, concept-guided contrastive representation learning constructs class-wise positive and negative patch sets from the same similarity matrix and optimizes target-class, symmetric auxiliary, and cross-class separation objectives, thereby regularizing the projected concept space. We evaluate CGRL on TCGA-BRCA and TCGA-NSCLC using multiple representative MIL methods. Experimental results show that CGRL improves several model-dataset combinations, with gains depending on the downstream MIL model and dataset. It achieves particularly clear improvements in accuracy and macro-F1 while reducing computational cost through concept-relevance pruning. These findings demonstrate that class-level semantic concepts provide an effective and practical prior for patch selection and representation learning in weakly supervised computational pathology.
