Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ITC-RWKV: Interactive Tissue-Cell Modeling with Recurrent Key-Value Aggregation for Histopathological Subtyping

Yating Huang, Qijun Yang, Lintao Xiang, Hujun Yin

TL;DR

ITC-RWKV tackles fine-grained histopathology subtyping by integrating cell-level morphology with global tissue context. It introduces a dual-stream architecture with Aggr-RWKV for scalable nuclear feature aggregation and a bidirectional tissue–cell interaction module for mutual refinement across scales, achieving linear-time aggregation $O(n)$ versus traditional $O(n^2)$ self-attention. The method attains state-of-the-art performance on BRACS and demonstrates robust cross-domain generalization in prostate datasets, while providing interpretable tissue- and cell-level attributions. This work advances scalable, interpretable computational pathology by effectively bridging micro- and macro-scale representations.

Abstract

Accurate interpretation of histopathological images demands integration of information across spatial and semantic scales, from nuclear morphology and cellular textures to global tissue organization and disease-specific patterns. Although recent foundation models in pathology have shown strong capabilities in capturing global tissue context, their omission of cell-level feature modeling remains a key limitation for fine-grained tasks such as cancer subtype classification. To address this, we propose a dual-stream architecture that models the interplay between macroscale tissue features and aggregated cellular representations. To efficiently aggregate information from large cell sets, we propose a receptance-weighted key-value aggregation model, a recurrent transformer that captures inter-cell dependencies with linear complexity. Furthermore, we introduce a bidirectional tissue-cell interaction module to enable mutual attention between localized cellular cues and their surrounding tissue environment. Experiments on four histopathological subtype classification benchmarks show that the proposed method outperforms existing models, demonstrating the critical role of cell-level aggregation and tissue-cell interaction in fine-grained computational pathology.

ITC-RWKV: Interactive Tissue-Cell Modeling with Recurrent Key-Value Aggregation for Histopathological Subtyping

TL;DR

ITC-RWKV tackles fine-grained histopathology subtyping by integrating cell-level morphology with global tissue context. It introduces a dual-stream architecture with Aggr-RWKV for scalable nuclear feature aggregation and a bidirectional tissue–cell interaction module for mutual refinement across scales, achieving linear-time aggregation versus traditional self-attention. The method attains state-of-the-art performance on BRACS and demonstrates robust cross-domain generalization in prostate datasets, while providing interpretable tissue- and cell-level attributions. This work advances scalable, interpretable computational pathology by effectively bridging micro- and macro-scale representations.

Abstract

Accurate interpretation of histopathological images demands integration of information across spatial and semantic scales, from nuclear morphology and cellular textures to global tissue organization and disease-specific patterns. Although recent foundation models in pathology have shown strong capabilities in capturing global tissue context, their omission of cell-level feature modeling remains a key limitation for fine-grained tasks such as cancer subtype classification. To address this, we propose a dual-stream architecture that models the interplay between macroscale tissue features and aggregated cellular representations. To efficiently aggregate information from large cell sets, we propose a receptance-weighted key-value aggregation model, a recurrent transformer that captures inter-cell dependencies with linear complexity. Furthermore, we introduce a bidirectional tissue-cell interaction module to enable mutual attention between localized cellular cues and their surrounding tissue environment. Experiments on four histopathological subtype classification benchmarks show that the proposed method outperforms existing models, demonstrating the critical role of cell-level aggregation and tissue-cell interaction in fine-grained computational pathology.
Paper Structure (12 sections, 7 equations, 4 figures, 3 tables)

This paper contains 12 sections, 7 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Overall Pipeline
  5. Cell Pathway with Aggr-RWKV
  6. Tissue-Cell Interaction Modeling
  7. Experiments
  8. Datasets
  9. Implementation Details
  10. Performance Comparisons
  11. Interpretability
  12. Conclusion

Figures (4)

  • Figure 1: Overview of the proposed ITC-RWKV model.
  • Figure 2: UMAP visualization of feature embeddings from the ablation study: (a) without cell branch, (b) without tissue–cell interaction (simple concatenation), and (c) full model.
  • Figure 3: Ablation studies on key model components on the BRACS dataset. Left: Comparison of different fusion strategies for combining tissue and cell-level features. Right: Impact of varying the layer depth of the Aggr-RWKV module.
  • Figure 4: Tissue region importance heatmaps across different breast pathologies. For each pair, the left shows original H&E images and the right shows corresponding importance maps where yellow-green highlights indicate regions most influential for classification.