Current review and next steps for artificial intelligence in multiple sclerosis risk research

Area of Science:

• Neurology
• Computational Science

Background:

• Multiple sclerosis (MS) prevalence is increasing globally.
• Artificial intelligence (AI) and machine learning (ML) are gaining traction in MS research.
• Existing AI/ML studies in MS primarily focus on lesion detection, neglecting risk factor assessment.

Purpose of the Study:

• To review AI/ML methods applied to multiple sclerosis (MS) research.
• To identify gaps in the current literature regarding MS risk factor assessment.
• To propose a causal Bayesian network approach for improved MS risk factor analysis.

Main Methods:

• Systematic review of AI/ML research using keywords related to MS, AI, ML, Bayes, and Bayesian.
• Analysis of 216 identified papers, focusing on 90 relevant and recently published studies.
• Proposal of a causal Bayesian network model for risk factor assessment.

Main Results:

• Over half of reviewed AI/ML studies in MS focus on lesion detection and segmentation.
• Clinical and lifestyle risk factor assessment and prediction in MS research are largely overlooked.
• Many existing risk factor studies lack consideration of confounding factors and bias.

Conclusions:

• There is a significant gap in AI/ML research concerning the assessment and prediction of multiple sclerosis (MS) risk factors.
• Causal Bayesian networks offer a robust approach to address deficiencies in current epidemiological methods for MS risk measurement.
• The proposed causal Bayesian network approach can improve the utilization of observational data for understanding MS etiology.