Automatic Intelligent System Using Medical of Things for Multiple Sclerosis Detection.

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Multiple sclerosis (MS) is an immune system disorder causing nerve damage and disrupting brain-body communication.
• Early diagnosis of MS is crucial for mitigating disease severity and improving patient outcomes.
• Magnetic resonance imaging (MRI) is a standard clinical tool for assessing MS progression.

Purpose of the Study:

• To implement a convolutional neural network (CNN) framework for detecting MS lesions in brain MRI slices.
• To evaluate the effectiveness of deep and hand-crafted features, optimized by a firefly algorithm, for MS detection.
• To compare classification performance with and without the skull in MRI scans.

Main Methods:

• A CNN-based scheme involving image preprocessing, deep and hand-crafted feature extraction, firefly algorithm optimization, and serial feature integration.
• Utilizing the VGG16 architecture for feature extraction.
• Employing Random Forest (RF) and K-nearest neighbor (KNN) classifiers for lesion classification.
• Implementing five-fold cross-validation for robust performance assessment.

Main Results:

• The proposed CNN framework achieved high classification accuracy for MS detection.
• VGG16 combined with RF classifier yielded >98% accuracy for MRI scans with the skull.
• VGG16 combined with KNN classifier achieved >98% accuracy for MRI scans without the skull.

Conclusions:

• The developed CNN-supported scheme demonstrates significant potential for accurate and early detection of MS lesions in brain MRI.
• The approach offers a promising tool for clinical diagnosis, potentially leading to better management of multiple sclerosis.
• The study highlights the efficacy of deep learning models, specifically VGG16, in analyzing medical images for neurological disease identification.