Deep transfer learning for Multiple Sclerosis Detection and classification using pre-trained model

Multiple sclerosis (MS) is a neurological condition that damages the central nervous system and is characterized by demyelination, inflammation , and axonal destruction. Early and correct MS diagnosis is essential for prompt treatment and better patient outcomes. With the help of total 3427 MRI images in the form of four different types of classes — Control Axial, Control Sagittal, MS-Axial, and MS-Sagittal. We Proposed a unique method for MS detection in this study. The pre-processing of the MRI data is the initial step of the proposed methodology. This research paper presents a comprehensive study aimed at enhancing MS classification through the augmentation of MRI data. By expanding the initial dataset of 3427 images to a robust collection of 10000 images using advanced techniques, the study offers an invaluable resource for training and evaluating MS detection algorithms, thereby fostering improved diagnostic outcomes. The VGG-16, VGG-19 and xception model are trained on the hybrid dataset, consisting of both axial and sagittal images, allowing it to capture the intricate patterns associated with MS lesions. We have obtained the highest accuracy, out of the 3 models used here, of 97%, by using the VGG-16 model.