Optimized brain tumor identification via graph sample and aggregate-attention network with Artificial Lizard Search Algorithm

Abstract

A brain tumour is an abnormal growth of brain nerves that interferes with normal brain function.It causes a great deal of deaths. Timely detection and treatment are essential for saving the lives from this illness. Locating tumor-affected brain cells is a tedious, time-consuming process. In this manuscript, a Graph Sample and Aggregate-Attention Network with Artificial Lizard Search Optimization Algorithm for identification of brain tumour (GSAAN-ALSOA-BTI) is proposed. Initially, the input imageries are obtained from Figshare Brain Tumor Dataset. The images are preprocessed using Data-Adaptive Gaussian Average Filtering (DAGAF) to eliminate the unnecessary noise from image frames. The pre-processed image is given into the Multi kernel k-Means clustering (MKKMC) for image segmentation to identify boundaries in an image. Finally, the extracted feature features are given to Graph Sample with Aggregate-Attention Network (GSAAN) for categorizing the brain tumour identification as benign and malignant tumors. In general, GSAAN does not express any adaption of optimization techniques for determining the ideal parameters to assure exact classification of brain tumour identification. Thus, the Artificial Lizard Search Optimization Algorithm (ALSOA) is proposed to improve the weight parameter of GSAAN, which accurately categorizes the brain tumour identification of brain tumor. The proposed model is executed in Python and its efficacy is evaluated utilizing performance metrics like accuracy, precision, recall, FI-score, error rate, computation time and ROC. The GSAAN-ALSOA-BTI method provides higher accuracy, higher precision, higher recall compared with existing methods.