Developing clinically useful cell-level analysis tools for cancer digital pathology in H&E-stained whole slide images remains challenging due to limitations in dataset granularity, inconsistent annotations, high computational demands, and difficulties in integrating new technologies into clinical workflows. To address these issues, we propose a solution that enables finer-grained cell labels at inference, improves model performance, and enhances usability by creating a lightweight, extensible cell segmentation and classification model. First, we update data labels through cross-relabeling to refine annotations of PanNuke and MoNuSAC, producing a unified dataset with seven distinct cell types relevant to tumor pathology. Second, we leverage the H-optimus-0 foundation model as a fixed encoder to improve feature representation for simultaneous segmentation and classification tasks. Third, to address foundation models’ computational demands, we distill knowledge to reduce model size and complexity while maintaining comparable performance. Finally, we integrate the distilled model into QuPath, a widely used open-source digital pathology platform. Results demonstrate improved segmentation and classification performance using a model built on H-optimus-0 compared to a CNN-based model. Specifically, average R2 improved from 0.575 to 0.871, and average PQ score improved from 0.450 to 0.492, indicating better alignment with actual cell counts and enhanced segmentation quality. The distilled model maintains comparable performance while reducing parameter count by a factor of 48. By reducing computational complexity and integrating into workflows, this approach may significantly impact diagnostics, reduce pathologist workload, and improve outcomes. Although the method shows promise, the present evaluation is restricted to H&E-stained patches from PanNuke and MoNuSAC. Extending the approach to additional cell types and tissues will require further fine-tuning and validation.

Keywords: Digital pathology; Cancer; H&E; Whole slide images; Cell segmentation; Cell classification; Foundation models; Knowledge distillation; Cross-relabeling; QuPath integration

‍