Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

C2P-GCN: Cell-to-Patch Graph Convolutional Network for Colorectal Cancer Grading

Sudipta Paul, Bulent Yener, Amanda W. Lund

TL;DR

C2P-GCN tackles colorectal cancer grading by unifying local cell organization within patches and global relationships across patches into a single WSI-level graph, processed by a multi-layer GCN. The method combines nuclei detection via gLoG, patch-level graphs enriched with Voronoi/Delaunay/MST patterns, and a cosine-similarity-based image-level graph to connect patches, enabling data-efficient training. Across two CRC datasets, C2P-GCN yields competitive or superior accuracy while using orders of magnitude less training data than patch-centric CNN or GCN approaches, emphasizing improved data efficiency and interpretability. This dual-stage graph representation offers practical gains for histology grading and potentially broader tissue-structure analysis.

Abstract

Graph-based learning approaches, due to their ability to encode tissue/organ structure information, are increasingly favored for grading colorectal cancer histology images. Recent graph-based techniques involve dividing whole slide images (WSIs) into smaller or medium-sized patches, and then building graphs on each patch for direct use in training. This method, however, fails to capture the tissue structure information present in an entire WSI and relies on training from a significantly large dataset of image patches. In this paper, we propose a novel cell-to-patch graph convolutional network (C2P-GCN), which is a two-stage graph formation-based approach. In the first stage, it forms a patch-level graph based on the cell organization on each patch of a WSI. In the second stage, it forms an image-level graph based on a similarity measure between patches of a WSI considering each patch as a node of a graph. This graph representation is then fed into a multi-layer GCN-based classification network. Our approach, through its dual-phase graph construction, effectively gathers local structural details from individual patches and establishes a meaningful connection among all patches across a WSI. As C2P-GCN integrates the structural data of an entire WSI into a single graph, it allows our model to work with significantly fewer training data compared to the latest models for colorectal cancer. Experimental validation of C2P-GCN on two distinct colorectal cancer datasets demonstrates the effectiveness of our method.

C2P-GCN: Cell-to-Patch Graph Convolutional Network for Colorectal Cancer Grading

TL;DR

C2P-GCN tackles colorectal cancer grading by unifying local cell organization within patches and global relationships across patches into a single WSI-level graph, processed by a multi-layer GCN. The method combines nuclei detection via gLoG, patch-level graphs enriched with Voronoi/Delaunay/MST patterns, and a cosine-similarity-based image-level graph to connect patches, enabling data-efficient training. Across two CRC datasets, C2P-GCN yields competitive or superior accuracy while using orders of magnitude less training data than patch-centric CNN or GCN approaches, emphasizing improved data efficiency and interpretability. This dual-stage graph representation offers practical gains for histology grading and potentially broader tissue-structure analysis.

Abstract

Graph-based learning approaches, due to their ability to encode tissue/organ structure information, are increasingly favored for grading colorectal cancer histology images. Recent graph-based techniques involve dividing whole slide images (WSIs) into smaller or medium-sized patches, and then building graphs on each patch for direct use in training. This method, however, fails to capture the tissue structure information present in an entire WSI and relies on training from a significantly large dataset of image patches. In this paper, we propose a novel cell-to-patch graph convolutional network (C2P-GCN), which is a two-stage graph formation-based approach. In the first stage, it forms a patch-level graph based on the cell organization on each patch of a WSI. In the second stage, it forms an image-level graph based on a similarity measure between patches of a WSI considering each patch as a node of a graph. This graph representation is then fed into a multi-layer GCN-based classification network. Our approach, through its dual-phase graph construction, effectively gathers local structural details from individual patches and establishes a meaningful connection among all patches across a WSI. As C2P-GCN integrates the structural data of an entire WSI into a single graph, it allows our model to work with significantly fewer training data compared to the latest models for colorectal cancer. Experimental validation of C2P-GCN on two distinct colorectal cancer datasets demonstrates the effectiveness of our method.
Paper Structure (13 sections, 8 equations, 1 figure, 5 tables)

This paper contains 13 sections, 8 equations, 1 figure, 5 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Cell identification
  4. Patch graph formation
  5. Cell graph
  6. Voronoi graphs of cells for global graph
  7. Image level Graph
  8. Network architecture
  9. Experiments
  10. Dataset
  11. Implementation
  12. Experimental Results
  13. Conclusion

Figures (1)

  • Figure 1: (a) C2P-GCN overall pipeline. C2P-GCN initially breaks a WSI into multiple patches and then constructs a patch-level graph for capturing structural features within a patch. Next, it forms an image-level graph of collective patches based on similar measures. The image-level graph containing whole image structural information is fed into a multi-layer GCN structure for classification. (b) Nuclei detection with gLoG filter. (c) Cell-graph construction; the magnified section shows how a sample node (nuclei) is connected with its neighborhood. (d) Voronoi diagram. (e) Delaunay triangulation. (f) Minimum spanning tree. (g) This figure highlights the red patch, which is a randomly chosen node of interest, and its top 15 most similar patches (blue) in terms of similarity scores.