HYBRID ARITHMETIC PUZZLE OPTIMIZATION ALGORITHM BASED GRAPH NEURAL NETWORK FOR EARLY GASTRIC CANCER DIAGNOSIS AND CLASSIFICATION

Abstract

In the worldwide, the most popular and one of the severe health issue is Gastric cancer (GC). The infection of Helicobacter pylori (H. pylori) commonly affect the chronic atrophic gastritis. To improve the patient survival, early diagnosis of GC is the challenging one because the diagnostic theories are did not explained and concretized with utmost of the existing computer-aided-diagnosis (CAD) model. Nowadays, the clinical management full spectrum quickly reshape via artificial intelligence (AI) systems to diagnose GC. This work focuses on hybrid optimization model with deep learning (DL) techniques for the effective diagnosis and classification of GC. Prior to the classification, the input data images are pre-processed using Wiener filter is to remove the noise from the GC image pixels and followed by Equilibrium Optimized k-means Clustering (EOk-C) performs segmentation of GC images thereby enhancing the challenges of irregular edges and mucosal features diversity. Adaptive Gray Level Co-occurrence Matrix (AGLCM) feature extraction is performed and the GC detection and classification is employed via Proposed Hybrid Arithmetic Puzzle Optimization Algorithm (HAPOA) with Graph Neural Network (GNN). MATLAB software simulates the proposed work and it is analyzed with various performance measures for the identification of proposed GC classification efficiency. The experimental analytical results disclose that the proposed GC classification performance is superior to other state-of-art techniques.