TABLE 2.

Related works.

Model	Dataset	Performance	Ref
GraphSAGE-mean	Placenta Histology Data	Accuracy: 88.94 ±0.38	[56]
MLP and Transformer	CRC-MSI, STAD-MSI, and GIST-PDL1	AUC improvement of more than 5% on various network backbones	[37]
Adaptive GraphSAGE with Graph Clustering module	Colorectal Cancer Data	Patch Accuracy : 91.60 ± 1.26, Image Accuracy: 97.00 ± 1.10 %	[40]
Hierarchical Transformer Graph Neural Network	Colorectal Cancer Dataset (CRC) and Extended Colorectal cancer dataset (Extended CRC)	Accuracy on CRC:98.55±1.26 %, Accuracy on Extended CRC : 95.33±0.58 %	[41]
GINConv+TopKPool	Gastric Cancer	AUC of Binary classification: 0.960±0.01, AUC of Ternary classification: 0.904±0.012	[45]
Cell-Graph Attention (CGAT) network	Pancreatic Diseases and Cancer	Precision: 0.73, Recall:0.65 and F1-score:0.62	[19]
Augmented Cell Graph and MLP	Brain Cancer	Sensitivity: 97.53% and Specificities of inflamed and healthy: 93.33% and 98.15%	[46]
Extracellular matrix (ECM)-aware cell-graph with Support Vector Machine (SVM)	Bone Cancer	Accuracy: 90%	[43]
Hierarchical Cell-to-Tissue (HACT) network	Breast Carcinoma Subtyping Set	Weighted F1 score : 61.53±0.87	[49]
Feature Driven Local Cell Graph with linear discriminant classifier	Lung Cancer	AUC = 0.68	[48]
Cell Cluster Graph with SVM	Prostrate Cancer	Accuracy : 83.1 ±1.2%	[47]
Hierarchical Cell Graphs and SVM	Breast Cancer	Accuracy :81.8%	[18]