Enhancing Brain Tumour Classification from MRI's Dataset Using CNNs with Attention Mechanism for Improved Accuracy

The human brain serves as the central processing unit of the body, and abnormal cellular proliferation within the brain can lead to tumourigenesis, often culminating in malignancies such as brain cancer. In the domain of medical imaging, Computer Vision (CV) plays a pivotal role by providing objective, data-driven insights that minimize the impact of human cognitive biases. This study focuses on utilizing advanced CV techniques for the automated detection of brain tumours from MRI scans. The methodology leverages bilateral filtering to effectively reduce image noise, followed by binary thresholding for enhanced contrast, and Convolutional Neural Network (CNN) segmentation to accurately delineate tumour regions. The model's robustness is ensured through extensive training, testing, and validation using diverse datasets. Key performance metrics, including accuracy, sensitivity, specificity, precision, recall, and F1 score, are utilized to evaluate the model's diagnostic capability, with the goal of outperforming existing methods. The proposed Hybrid CNN (HCNN) model achieves an accuracy of 96.4%, precision of 93.2%, recall of 94%, and an F1 score of 92%, demonstrating its superior efficacy in brain tumour detection.