A Transfer Learning-Based Approach for Brain Tumor Classification with Attention Module

Unusual cell growths in the brain or surrounding tissues are termed brain tumors, which can be benign (noncancerous) or malignant (cancerous). Their prevalence, accounting for approximately 2% of all cancer diagnoses according to the World Health Organization (WHO), raises significant public health concerns. Patients often experience symptoms such as headaches, seizures, memory loss, and other neurological issues, impacting their quality of life depending on the tumor's location and size. This study presents a system for classifying brain tumors using a transfer learning model enhanced by a self-attention mechanism. This approach aims to improve feature representation and emphasize relevant regions in input images. We compare the proposed model's performance against existing classification methods, achieving an impressive accuracy of 98.14%.