MRI Brain Tumour Detection Using a Dual-Stage Image Enhancement and Deep Learning Classifier

Abstract

Accurate detection and segmentation of brain tumors in MRI scans is important for early diagnosis and effective treatment planning. The project focuses on developing a robust, deep learning framework that automatically classifies and classifies brain tumors in the most modern network architecture. The aim is to improve diagnostic accuracy and maintain computational efficiency at the same time. Dennenet, on the other hand, uses tightly connected layers that promote improved reuse and gradient flow, leading to more accurate and reliable classification. The system is structured to categorize MRI images into two main classes: tumor and non-tumor. MobileNet and DenseNet are utilized as the backbone models to strike a balance between speed and accuracy—MobileNet excels in scenarios requiring fast processing, while DenseNet offers higher classification performance due to its advanced connectivity pattern. In addition to classification, the framework can be extended to include tumor localization by integrating segmentation techniques that identify the exact regions affected within the brain. This combined approach enhances the system's diagnostic capabilities by delivering automated and consistent results that support clinical decisions. The effectiveness of the proposed methods will be assessed using publicly available medical imaging datasets. Evaluation will be based on key metrics such as accuracy, precision, recall, and segmentation performance. Ultimately, this research aims to reduce reliance on invasive diagnostic methods and enable integration into real-time clinical workflows for improved patient outcomes.