Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Hierarchical Geometry-guided Transformer for Histological Subtyping of Primary Liver Cancer

Anwen Lu, Mingxin Liu, Yiping Jiao, Hongyi Gong, Geyang Xu, Jun Chen, Jun Xu

TL;DR

This paper tackles the challenging histological subtyping of primary liver cancer (distinguishing HCC from ICC and subtyping ICC) from gigapixel whole-slide images, where tissue heterogeneity and the tumor microenvironment complicate analysis. It introduces ARGUS, a hierarchical, geometry-guided transformer framework that fuses macro- and meso-scale tissue context with micro-level cellular geometry. The approach combines a micro-geometry feature derived from a nuclei graph, a Hierarchical FoVs Alignment module to aggregate multi-scale features, and a Geometry Prior Guided Fusion mechanism to integrate geometry with morphology, all under weak supervision. Across TCGA-Liver and in-house DTH-ICC datasets, ARGUS achieves state-of-the-art performance and yields interpretable attention maps and subtype-specific cellular patterns, offering a practical tool for clinical liver cancer subtyping.

Abstract

Primary liver malignancies are widely recognized as the most heterogeneous and prognostically diverse cancers of the digestive system. Among these, hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC) emerge as the two principal histological subtypes, demonstrating significantly greater complexity in tissue morphology and cellular architecture than other common tumors. The intricate representation of features in Whole Slide Images (WSIs) encompasses abundant crucial information for liver cancer histological subtyping, regarding hierarchical pyramid structure, tumor microenvironment (TME), and geometric representation. However, recent approaches have not adequately exploited these indispensable effective descriptors, resulting in a limited understanding of histological representation and suboptimal subtyping performance. To mitigate these limitations, ARGUS is proposed to advance histological subtyping in liver cancer by capturing the macro-meso-micro hierarchical information within the TME. Specifically, we first construct a micro-geometry feature to represent fine-grained cell-level pattern via a geometric structure across nuclei, thereby providing a more refined and precise perspective for delineating pathological images. Then, a Hierarchical Field-of-Views (FoVs) Alignment module is designed to model macro- and meso-level hierarchical interactions inherent in WSIs. Finally, the augmented micro-geometry and FoVs features are fused into a joint representation via present Geometry Prior Guided Fusion strategy for modeling holistic phenotype interactions. Extensive experiments on public and private cohorts demonstrate that our ARGUS achieves state-of-the-art (SOTA) performance in histological subtyping of liver cancer, which provide an effective diagnostic tool for primary liver malignancies in clinical practice.

A Hierarchical Geometry-guided Transformer for Histological Subtyping of Primary Liver Cancer

TL;DR

This paper tackles the challenging histological subtyping of primary liver cancer (distinguishing HCC from ICC and subtyping ICC) from gigapixel whole-slide images, where tissue heterogeneity and the tumor microenvironment complicate analysis. It introduces ARGUS, a hierarchical, geometry-guided transformer framework that fuses macro- and meso-scale tissue context with micro-level cellular geometry. The approach combines a micro-geometry feature derived from a nuclei graph, a Hierarchical FoVs Alignment module to aggregate multi-scale features, and a Geometry Prior Guided Fusion mechanism to integrate geometry with morphology, all under weak supervision. Across TCGA-Liver and in-house DTH-ICC datasets, ARGUS achieves state-of-the-art performance and yields interpretable attention maps and subtype-specific cellular patterns, offering a practical tool for clinical liver cancer subtyping.

Abstract

Primary liver malignancies are widely recognized as the most heterogeneous and prognostically diverse cancers of the digestive system. Among these, hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC) emerge as the two principal histological subtypes, demonstrating significantly greater complexity in tissue morphology and cellular architecture than other common tumors. The intricate representation of features in Whole Slide Images (WSIs) encompasses abundant crucial information for liver cancer histological subtyping, regarding hierarchical pyramid structure, tumor microenvironment (TME), and geometric representation. However, recent approaches have not adequately exploited these indispensable effective descriptors, resulting in a limited understanding of histological representation and suboptimal subtyping performance. To mitigate these limitations, ARGUS is proposed to advance histological subtyping in liver cancer by capturing the macro-meso-micro hierarchical information within the TME. Specifically, we first construct a micro-geometry feature to represent fine-grained cell-level pattern via a geometric structure across nuclei, thereby providing a more refined and precise perspective for delineating pathological images. Then, a Hierarchical Field-of-Views (FoVs) Alignment module is designed to model macro- and meso-level hierarchical interactions inherent in WSIs. Finally, the augmented micro-geometry and FoVs features are fused into a joint representation via present Geometry Prior Guided Fusion strategy for modeling holistic phenotype interactions. Extensive experiments on public and private cohorts demonstrate that our ARGUS achieves state-of-the-art (SOTA) performance in histological subtyping of liver cancer, which provide an effective diagnostic tool for primary liver malignancies in clinical practice.

Paper Structure

This paper contains 17 sections, 8 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Data Preprocessing and Feature Extraction
  4. Data Preprocessing
  5. Histological Feature Extraction
  6. Micro-Level Geometric Feature Extraction
  7. Hierarchical FoVs Alignment Module
  8. Geometry Prior Guided Fusion Module
  9. Experiments and Results
  10. Datasets
  11. Implementation Details
  12. Training settings
  13. Hyper-parameters
  14. Comparison with State-of-the-art Methods
  15. Ablation Study
  16. ...and 2 more sections

Figures (2)

  • Figure 1: Overview of the proposed ARGUS. (a) Overall workflow of our framework for histological subtyping, (b) Hierarchical FoVs Alignment (HFA) module, (c) Geometry Prior Guided Fusion (GPGF) module, (d) Legend for the symbols used.
  • Figure 2: Model visualization and interpretability analysis of the proposed ARGUS on DTH-ICC dataset. (a) the input WSI, associated corresponding attention heatmap for histological subtyping. (b) Representative high-attention patches from three ICC subtypes, overlaid with corresponding cell-type annotations. (c) Quantitative analysis of cell types in the top 10% high-attention patches.