A Neural Architecture-Based Solution to Detect Multi-Class Neoplastic Brain Lesion Using Neuroimaging

A neoplastic brain lesion is an abnormal growth of cells in the brain, which can be either cancerous or non-cancerous. These conditions are among the most serious in medicine and require accurate and timely diagnosis. This project focuses on improving how neoplastic brain lesions are detected using Neuroimaging scans. The goal is to create a reliable system that can help identify different types of tumors, such as Glial neoplasms, Meningeal masses, and pituitary adenomas, as well as distinguish them from healthy brain scans. Neuroimaging images were resized and labeled appropriately during dataset preparation. The system was trained to recognize patterns in the images that indicate the presence and type of tumor, to support early diagnosis and treatment. After training with the Adam optimizer, the model achieved 92 % accuracy. This approach shows promise for early tumor detection, monitoring progression, and providing a non-invasive diagnostic tool in clinical settings, where rapid and accurate diagnoses are essential for patient care.