Diagnosing Multiple Sclerosis from Magnetic Resonance Imaging Images: Highlights from the Second Isfahan Artificial...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Multiple sclerosis (MS) is a debilitating autoimmune disease affecting the central nervous system, causing disability in young adults.
• MS involves immune system attacks on myelin sheaths, leading to lesions in the brain.
• Manual lesion detection in MRI scans is time-consuming and prone to inaccuracies.

Purpose of the Study:

• To organize a challenge focused on developing AI-based methods for diagnosing MS from MRI images.
• To advance the segmentation and localization of MS lesions using artificial intelligence.
• To foster innovation in medical image analysis for neurological disorders.

Main Methods:

• A challenge titled "Diagnosing MS from magnetic resonance imaging (MRI) Images" was conducted.
• Participants utilized AI-based methods, primarily deep learning, for lesion segmentation and localization.
• The challenge involved distinct training, testing, and final in-person evaluation phases with MRI datasets.

Main Results:

• The best Dice score achieved was 0.33, with a sensitivity of 0.349 and precision of 0.3.
• The lowest centroid distance for lesion localization was 53.025.
• Lesion detection accuracy varied by brain region, with the highest at 80.28% in periventricular areas.

Conclusions:

• AI methods show capability in segmenting and localizing MS lesions from MRI data.
• Current AI performance in MS lesion detection requires significant improvement to reach desired accuracy levels.
• Further research and development are essential to enhance AI-driven diagnostic tools for multiple sclerosis.