Brain Tumor Classification from MRI Images Using Xception Model

Abstract

Brain tumors, characterized by abnormal cell growth within the brain or central spinal canal, present significant challenges in diagnosis and treatment. These tumors can arise from various cell types within the brain, leading to a diverse range of tumor subtypes with distinct clinical. Existing research has achieved brain tumor detection accuracy of up to 97% using pre-trained CNNs like Yolo V7. Our proposed method, fine-tuning hyperparameters of the Xception model, significantly enhances classification accuracy to 99%, marking a notable improvement in brain tumor diagnosis. Additionally, the model's validation yielded a perfect accuracy of 1.0, an F1 score of 0.54, and a precision and recall 0.54. Our model then obtained an accuracy of 0.99, precision of 0.684, recall of 0.61, and F1 score of 0.60 on the testing dataset. These findings highlight the significance of hyperparameter tweaking in customizing deep learning architectures to the complexities of medical imaging data, ultimately resulting in better patient outcomes and treatment.