Controlled Search Diversification-Based Catch Fish Optimization-Driven Feature Selection for Brain Tumor Classification

Brain tumors are defined as the abnormal growth of cells originating from various central nervous system regions, which leads to patient risk. However, existing tumor detection and classification models face difficulties in identifying precise tumor boundaries in Magnetic Resonance Imaging (MRI) data due to the presence of noise, low contrast, and complex tissue structures. To address this limitation, an automated brain tumor classification model, namely Controlled Search Diversification-based Catch Fish Optimization Algorithm and Visual Geometry Group-19 (CSD-CFOA-VGG19), is proposed in this research. Initially, the raw MRI data are acquired from benchmark datasets and preprocessed to enhance quality for accurate detection of tumor boundaries. Next, the Davies–Bouldin Index-based Fuzzy C-Means (DBI-FCM) clustering method is used to localize and segment the affected brain regions. After that, discriminative features are extracted using a CNN based model, and relevant features are selected by the proposed Controlled Search Diversification-based Catch Fish Optimization Algorithm (CSD-CFOA) for differentiating tumor features from normal tissues. Finally, the selected features are fed to a modified VGG19 approach for multi-class brain tumor classification. Experimental results show that the proposed model reported an accuracy of 98.53% on the Figshare dataset, which is a competitive performance when compared to benchmarks such as the Chronological Jaya Honey Badger Algorithm (CJHBA) under identical conditions.