CGAT: Cell Graph ATtention Network for Grading of Pancreatic Disease Histology Images

Abstract

Early detection of Pancreatic Ductal Adenocarcinoma (PDAC), one of the most aggressive malignancies of the pancreas, is crucial to avoid metastatic spread to other body regions. Detection of pancreatic cancer is typically carried out by assessing the distribution and arrangement of tumor and immune cells in histology images. This is further complicated due to morphological similarities with chronic pancreatitis (CP), and the co-occurrence of precursor lesions in the same tissue. Most of the current automated methods for grading pancreatic cancers rely on extensive feature engineering involving accurate identification of cell features or utilising single number spatially informed indices for grading purposes. Moreover, sophisticated methods involving black-box approaches, such as neural networks, do not offer insights into the model’s ability to accurately identify the correct disease grade. In this paper, we develop a novel cell-graph based Cell-Graph Attention (CGAT) network for the precise classification of pancreatic cancer and its precursors from multiplexed immunofluorescence histology images into the six different types of pancreatic diseases. The issue of class imbalance is addressed through bootstrapping multiple CGAT-nets, while the self-attention mechanism facilitates visualization of cell-cell features that are likely responsible for the predictive capabilities of the model. It is also shown that the model significantly outperforms the decision tree classifiers built using spatially informed metric, such as the Morisita-Horn (MH) indices.

Prediction of Tuberculosis From Lung Tissue Images of Diversity Outbred Mice Using Jump Knowledge Based Cell Graph Neural Network

“…It is assumed that nearby cells are more likely to interact with one another. Researchers frequently construct graphs using Delaunay triangulation [ 38 ], [ 39 ] or the K-nearest-neighbor (KNN) approach [ 18 ], [ 19 ], [ 40 ], [ 41 ] to depict these interconnections. The Waxman model [ 42 ] is another alternative strategy that uses exponential decay based on Euclidean distance to represent cell interactions.…”

“…Additionally, a Transformer module was incorporated to capture long-distance dependencies. The authors in [ 19 ] introduced the CGAT network for precisely classifying pancreatic cancer and its precursors from immunofluorescence histology images. It integrated a unique self-attention mechanism at its output, enhancing interactions among graph nodes.…”

“…An emerging solution to fully leverage the rich information within WSIs is the use of cell graph representations that map the granuloma into a graph. However, existing methodologies [18], [19] for constructing cell graphs often employ edge thresholding techniques, which can inadvertently discard vital biological information. This oversight may result in the loss of subtle yet crucial insights by producing overly sparse or dense graphs.…”

Explainable AI (XAI) In Biomedical Signal and Image Processing: Promises and Challenges

“…Specifically, the spatial patterns of cell-cell organization permit the characterization and inference of spatial biomarkers [22] that might be prognostic of disease state or prognosis. [24] proposed a framework, Cell Graph ATtention Network (CGAT) that uses deep learning graph architectures to analyze and interpret mIF images of tissue. While the GCN framework is used for classification of disease grade, the network architecture includes an attention layer that identifies particular nodes that are most relevant to the prediction task.…”