A Novel Approach for the Automatic Detection of Brain Plaques in MRI Images of MS Patients using Transformers

Articles in Press

Document Type : Original Article

Authors

Department of Computer Engineering, Faculty of Engineering, Yazd University, Yazd, Iran

Abstract

In recent years, medical imaging has made significant advancements with the use of deep learning techniques. Multiple sclerosis (MS) is a chronic disease caused by the demyelination of the central nervous system. This disease is characterized by plaques visible in MRI scans. Accurate detection of these plaques is crucial for prognosis and treatment monitoring. Manual segmentation of MS plaques by experts is a time-consuming process and prone to human error. This study presents a transformer-based approach for MS plaque segmentation. The proposed model consists of a swin transformer-based encoder for feature extraction and a specialized decoder for segmentation map reconstruction. The output is a labeled mask of the segmented plaques. The model was evaluated on ISBI2015 dataset, and its performance was compared with the classical U-Net model. The results show that the proposed model improved IoU and Dice scores to 0.71 and 0.83, respectively, outperforming U-Net model, which achieves IoU and Dice scores of 0.64 and 0.78.

Keywords

Tabriz Journal of Electrical Engineering

Articles in Press, Corrected Proof
Available Online from 22 September 2025

Files

Share

How to cite

Statistics