Classification and Segmentation of Brain MRI images using Deep Learning

Medical images play a critical part in the doctor's ability to make the correct diagnosis and in the patient's treatment. Intelligent algorithms make it possible to swiftly recognize lesions in medical imaging, and extracting features from images is very significant. Various algorithms have been integrated into medical imaging in a number of research. The basic architecture of CNN is constructed by focusing on picture feature extraction using a convolutional neural network (CNN). The research is expanded to multi-channel input CNN for visual feature extraction in order to overcome the constraints of machine vision and human vision. Glioma tumor, meningioma tumor, pituitary tumor, and no tumor are the four classifications investigated in this study, which includes roughly 3300 MRI samples gathered from kaggel. The BrainNet that has been implemented has a 98.31 percent of training accuracy and an 87.80 percent of validation accuracy. Deep architectures such as InceptionNet, ResNet, and XceptionNet were also tested with and without transfer learning to see which strategy performed better.