Deep Learning Approaches for Multiple Sclerosis Detection in MRI Images

Multiple sclerosis (MS) is a chronic neurological disorder commonly diagnosed through magnetic resonance imaging (MRI). Manual interpretation of MRI scans is time-consuming and prone to observer variability. This study explores the use of deep learning (DL) models to automate MS lesion detection in MRI images. Four convolutional neural network (CNN) architectures-AlexNet, VGG16 pretrained on ImageNet, ResNet-10, and DenseNet-121-were utilized with slight modifications to optimize their performance for medical imaging tasks. The dataset, comprising 2831 FLAIR and T2 MRI slices from 60 patients, was extensively preprocessed, including normalization, resizing, and augmentation. The models were trained using the Adam optimizer and binary cross-entropy loss function and evaluated on metrics such as accuracy, precision, recall, and F1-score. VGG16 demonstrated the highest performance, with an area under the curve (AUC) of 0.94, followed by DenseNet-121 with an AUC of 0.91, and both AlexNet-C and ResNet-10 with an AUC of 0.90. VGG16 also achieved the highest precision (0.81) and recall (0.93), resulting in the top F1-score (0.86). ResNet-10, however, showed the best balance between efficiency and performance, making it highly suitable for deployment in resource-constrained environments, while DenseNet-121 and AlexNet-C were resource-inefficient given their comparable performance to ResNet-10. These findings underscore the potential of DL models to automate MS lesion detection, offering reliable, efficient, and scalable solutions for clinical diagnostics.