Enhancing multiple sclerosis diagnosis and prognosis through a dual Patch-Wise CNN architecture

Abstract

Multiple sclerosis (MS) is a complex neurological disorder that requires accurate early prediction and prognosis for timely intervention, which can be facilitated through the detection and tracking of lesion development from medical images. In medical image processing, traditional manual segmentation methods can be very time-consuming. To overcome these limitations, the present study introduces a novel dual 2D patch-wise parallel convolutional neural network (CNN) model designed to improve both lesion segmentation accuracy and the prediction of MS. In detail, the first CNN is dedicated to accurately segmenting lesions from MRI scans, while the second CNN reduces false positives, thereby increasing overall efficiency. Moreover, by integrating T1-w, T2-w, and FLAIR MRI sequences, the model achieves enhanced accuracy, adapting to variations across different MRI scanners. Following lesion identification, another CNN model to predict MS disease is developed, specifically tailored to MRI, and achieves an overall accuracy of 91% in the prediction of MS disease. The high accuracy, along with a precision of 87%, recall of 77%, and an F1 score of 80.5%, demonstrates the effectiveness of the proposed model as a robust tool for early diagnosis and prognosis in MS. The present dual CNN approach not only improves lesion segmentation but also provides clinicians with valuable insights into disease trajectory, offering a new dimension of predictive analysis for MS management.