Detection of Multiple Sclerosis Using Deep Learning

Multiple sclerosis is a chronic neurological disorder characterized by its adverse effects on the central nervous system. Hence, early and correct diagnosis are needed for effective management. The goal of this study is to find a deep learning-based approach of MS detection involving a combination of U-Net for segmentation purposes and EfficientNet for classification, using MRI images. The U-Net model is used to segment MS lesions from MRI scans accurately by capturing fine-grained spatial details using its encoderdecoder architecture. For classification, EfficientNet is used to classify MS cases based on two dimensions of MRI images: MS Axial and MS Sagittal. EfficientNet's optimized architecture ensures high classification accuracy with reduced computational complexity. U-Net for lesion segmentation enhances the model's interpretability and gives it valuable insights into clinical analysis. Our approach is trained and tested on benchmark MRI datasets with strong performance both in segmentation and classification tasks. Our system implemented in TensorFlow gives an efficient and scalable solution to automate MS detection for radiologists for early diagnosis and treatment planning. This project provides advanced deep learning techniques that develop tools for AI-assisted medical imaging, further enhancing patient care and diagnostic accuracy.