Automatic Brain Tumor Detection System using Convolutional Neural Network

Abstract

Magnetic Resonance Imaging (MRI) segmentation plays a crucial role in modern medical diagnosis by enabling accurate extraction of abnormal regions from brain images. Manual brain tumour segmentation is a time-consuming and expertise-dependent process, which motivates the need for automated and efficient solutions. This work presents a computer-aided Brain Cancer Detection and Classification System based on Convolutional Neural Networks (CNN) for accurate tumour segmentation and diagnosis from MRI images. The proposed system employs image preprocessing techniques such as histogram equalization and image enhancement to improve image quality, followed by segmentation and feature extraction to identify tumour regions. The CNN model is used to automatically detect tumour blocks and classify astrocytoma brain tumours from MRI datasets of multiple patients. The MATLAB-based implementation displays results on a computer and transmits the output to an embedded system through wired communication for real-time monitoring. By combining image processing and deep learning, the proposed approach improves diagnostic accuracy while reducing manual effort and computational complexity. The system provides a reliable and efficient tool for early detection and classification of brain tumours, which can assist medical professionals in timely treatment and decision-making.