Abstract

The central organ of the human nervous system consists of brain that sends and receive stimuli to the different organs of the body in the engagement of daily routine activities. Irregular growth of the cells near the brain that disturb the normal function of the healthy brain cells that result in brain tumor. Timely diagnosis of a brain tumor is essential for effective treatment, improved results, and minimizing complications. Machine learning techniques are reliable for brain tumor detection that enables accurate analysis of medical data for identification of tumors with early and improve diagnostic efficiency. Transfer learning which is a machine learning algorithms can be able to achieve high accuracy with minimal training data by leveraging pre-trained models which can be able to adapt new tasks efficiently. Our study aims in brain tumor segmentation using MRI through Xception model which is a deep convolutional neural network. The aim of using xception model for its efficiency and strong feature extraction capabilities. Pre processed MRI images are given to caption based segmentation framework that recognizes tumor regions with high accuracy. Explainable AI (XAI) techniques such as Grad-CAM is integrated to enhance transparency and trust that highlights the critical regions influencing the model’s prediction. This approach also ensures that model focuses on relevant regions rather than irrelevant surroundings. Xception attained the high accuracy of 97.35% whereas MobileOne S2 had given the strongest training performance indicating healthy learning capabilities. The combination of AI and image processing in MRI based tumor segmentation improves diagnostic accuracy that enables early detection and personalized treatment with speedy more reliable results than manual methods.