Automated Detection and Segmentation of Brain Tumors Using Efficient Predefined Time Adaptive Neural Network for Multi-Class Tumor Classification and Pixel-Level Segmentation in MRI Images

Accurate detection and segmentation of brain tumors are essential for effective diagnosis and treatment planning. However, challenges such as Magnetic Resonance Imaging (MRI) noise and low contrast between tumor and surrounding tissues often reduce the performance of automated models. To address these issues, this work proposes an Efficient Predefined Time-Adaptive Neural Network for Automated Multi-Class Brain Tumor Detection and MRI Segmentation (DSBT-MRI-EPTANN). Initially, MRI images are collected from the Figshare brain tumor dataset and pre-processed utilizing Fast Resampled Iterative Filtering (FRIF), which enhances image quality while preserving structural details. The refined images are then segmented at the pixel level using Dual Information Enhanced Multi-view Attributed Graph Clustering (DIEMAGC), accurately delineating tumor regions. Segmented tumors are classified into meningioma, glioma, and pituitary tumor categories using the Efficient Predefined Time Adaptive Neural Network (EPTANN). To further improve detection and classification performance, the Multiplayer Battle Game-Inspired Optimizer (MBGIO) is applied to optimize the EPTANN weight parameters. The proposed DSBT-MRI-EPTANN framework achieves 99.55% accuracy, 99.13% sensitivity and 99.11% specificity, demonstrating robust tumor segmentation, precise multi-class classification, and enhanced reliability in brain tumor detection utilizing MRI images.