Portfolio Details

Description

Brain tumors pose significant health challenges, making accurate diagnosis and classification vital for effective treatment. MRI scans play a key role in detecting and characterizing brain tumors, but their complex and heterogeneous nature makes classification challenging. This project leverages deep learning to classify brain tumors into glioma, meningioma, no tumor, and pituitary categories using 7,023 MRI images from datasets like Figshare, SARTAJ, and Br35H. After data augmentation, we implemented a self-built convolutional neural network and state-of-the-art transfer learning models (VGG, ResNet, and GoogleNet). Among these, VGG16 achieved the best performance with 97% accuracy, correctly classifying 95% of gliomas, 94% of meningiomas, and 99% of both no tumor and pituitary cases. Our custom model also performed well, achieving 94% test accuracy.