Predicting Enhancing Lesions in Multiple Sclerosis

This study evaluates whether deep learning can predict enhancing lesions in multiple sclerosis (MS) using unenhanced MRI scans. A convolutional neural network trained on 1970 MRI scans achieved a sensitivity of 78% and specificity of 73% in detecting enhancing lesions. The method demonstrated moderate to high accuracy, with area under the curve (AUC) scores of 0.82 for slice-wise predictions and 0.75 for participant-wise predictions, suggesting that deep learning can potentially replace contrast-enhanced MRI for lesion detection.