Unveiling Multiple Sclerosis Predictors: A Proposed Artificial Intelligence Approach Using Associative Classification

Creative Commons License

Onay İ., Kirişci M., Çolak C.

Zeki sistemler teori ve uygulamaları dergisi (Online), cilt.9, sa.1708339, ss.1-10, 2026 (TRDizin)

Özet

Objective:

Although multiple sclerosis (MS) is the leading non-traumatic cause of disability among young adults, the reasons behind its increasing prevalence remain elusive. This study explores the use of artificial intelligence, specifically associative classification, to identify key predictors for the progression of multiple sclerosis (MS).

Methods:

The study utilized a publicly available dataset to forecast whether individuals have MS based on various personal traits. The relevant dataset originated from a cohort group study conducted on Mexican mestizo individuals recently identified with Clinically Isolated Syndrome (CIS). These individuals had sought care at the National Institute of Neurology and Neurosurgery (NINN) between 2006 and 2010. Associative rule mining was applied to uncover relationships between the variables and the development of Clinical Definite Multiple Sclerosis (CDMS). The effectiveness of the model was assessed using accuracy, balanced accuracy, sensitivity, specificity, positive & negative predictive values, and F1 measure with 95% confidence intervals.

Results:

Oligoclonal bands, breastfeeding history, education level, and specific MRI results were significant predictors of MS classification. Oligoclonal bands were particularly associated with a higher likelihood of CDMS. The proposed model performed with high accuracy (87.8%), sensitivity (89.6%), and specificity (86.3%), highlighting its effectiveness in predicting MS progression.

Conclusions:

Artificial intelligence, through associative classification, provides valuable insights into MS progression by identifying significant predictors. These findings can support early diagnosis and contribute to the development of personalized treatment strategies. Future research should incorporate more diverse datasets to validate these results further.