Enhancing Brain Tumour Detection using an Ensemble approach of Particle Swarm Optimization and Convolution Neural Network

Detecting brain tumor is vital in medical imaging research and computer-aided diagnosis to improve timely diagnosis and treatment outcomes. The diagnostic accuracy depends on the subjective analysis of radiologists applying conventional methods for brain tumor identification, which include MRI scans and biopsies. Innovations in machine and deep learning provide promising automation to improve the precision of tumor diagnosis as compared to the current methods. In this paper, image augmentation, feature extraction, and optimization methods are used to enhance the diagnosis of brain tumors. The proposed approach employs rotation augmentation, median filtering, Grey-Level Co-occurrence Matrix (GLCM) feature extraction, Particle Swarm Optimization (PSO), and a Convolutional Neural Network (CNN) to boost the accuracy and pliability of brain tumor classification. The Harvard repository dataset was used. It has a wide range of brain images for training and validation. The convolutional neural network combines particle swarm optimization techniques to detect brain tumors in MRI images. An accuracy rate of 96.71% is obtained with this integrated approach, which surpasses the present system. An effective solution for automated tumor detection in MRI images is achieved by integrating cutting-edge image processing methods such as rotation augmentation, median filtering, and GLCM feature extraction with the optimization strengths of PSO and the effective learning capabilities of CNNs.