Document Type : Research Article
Authors
1
Department of Electrical and Electronics Engineering, Faculty of Engineering, Shomal University, Amol, Iran
2
Department of Biomedical Engineering, Faculty of Engineering, Shahed University, Tehran, Iran
10.22080/frai.2025.29274.1015
Abstract
Multiple Sclerosis (MS) is a chronic autoimmune disease that affects the central nervous system and can lead to neurological disabilities. Early and accurate diagnosis plays a key role in managing its long-term effects. This study proposes a novel model based on convolutional neural networks (CNN) for identifying MS lesions in MRI images.
This study used an MRI dataset from 60 individuals divided into training, validation, and test sets. The preprocessing included removing initial slices and applying data augmentation (random rotations) to increase the number of training images to 1080. A customized CNN architecture was designed to learn the features related to MS lesions. The model's performance was evaluated using accuracy, sensitivity, and specificity metrics on validation and test data. The CNN performance was also compared with two machine learning algorithms: decision tree and support vector machine.
The proposed CNN model showed promising performance in detecting MS lesions. It achieved an accuracy of 99% during training and 96.44% during validation, demonstrating its ability to generalize to new data. The test accuracy was 92.6%, with sensitivity and specificity reported as 84% and 95%, respectively. Compared to other methods, the CNN outperformed the support vector machine (accuracy 85%, sensitivity 82.61%, specificity 98%) and the decision tree (accuracy 98%, sensitivity 95%, specificity 83.72%), highlighting its high capability in detecting MS lesions.
This research successfully demonstrates the capability of convolutional neural networks (CNN) in the accurate and automated detection of MS lesions in MRI images, achieving a test accuracy of 92.6%. The superior performance of CNN compared to traditional machine learning methods offers a promising approach for improving diagnostic accuracy, reducing reliance on human factors, and accelerating therapeutic interventions. The development of such tools can assist clinical specialists, enhance diagnostic efficiency, and facilitate better patient management. In the future, it is recommended to focus on improving CNN architecture, utilizing broader datasets, and exploring its application in different types of MS and disease progression monitoring.
Keywords
)