Page 113 - AIH-1-4

Artificial Intelligence in Health

ORIGINAL RESEARCH ARTICLE
Discovering predictive features of multiple

sclerosis from clinically isolated syndrome with
machine learning

Minh Sao Khue Luu * , Bair N. Tuchinov 1,2† , Anna I. Prokaeva 2,3† ,
1†
Denis S. Korobko 2,3 , Nadezhda A. Malkova 2,3 , and Andrey A. Tulupov 1,2
1 Stream Data Analytics and Machine Learning Laboratory, Novosibirsk State University, Novosibirsk,
Russia
2 The Institute International Tomography Center of the Russian Academy of Sciences, Novosibirsk, Russia
3 State Novosibirsk Regional Clinical Hospital, Novosibirsk, Russia

Abstract

Accurately predicting the progression of clinically isolated syndrome (CIS) to multiple
sclerosis (MS) is crucial for early intervention and management. This study employs
a range of machine learning models, including categorical boosting, extreme
gradient boosting, light gradient boosting machine, random forest, support vector
machine, and logistic regression, to classify CIS patients based on their likelihood
† These authors contributed equally of developing MS. Our best model achieves and demonstrates superior predictive
to this work. accuracy of 0.9312, measured using the area under the curve metric. In addition, we
*Corresponding author: apply explainability techniques to determine the most influential features driving
Minh Sao Khue Luu the predictions, identifying which CISs are most indicative of MS progression.
(khue.luu@g.nsu.ru) Furthermore, we explore feature interactions to detect relationships between
Citation: Luu MSK, Tuchinov BN, features, providing a deeper understanding of the underlying mechanisms. The
Prokaeva AI, Korobko DS, study utilizes public data from 273 CISs patients, offering significant contributions to
Malkova NA, Tulupov AA. the clinical management and early diagnosis of MS.
Discovering predictive features of
multiple sclerosis from clinically
isolated syndrome with machine
learning. Artif Intell Health. Keywords: Clinically isolated syndromes; Multiple sclerosis; Machine learning; Binary
2024;1(4):107-122. classification; Predictive features; Model explainability
doi: 10.36922/aih.4255
Received: July 16, 2024
Accepted: August 26, 2024
1. Introduction
Published Online: September 24,
2024 Multiple sclerosis (MS) is a chronic, inflammatory, demyelinating disease of the central
nervous system that affects approximately 2.9 million people worldwide. MS primarily
3
1,2
Copyright: © 2024 Author(s).
This is an Open-Access article affects young adults, with 70 – 80% of patients having an age of onset between 20 and
distributed under the terms of the 40 years, and represents the leading cause of non-traumatic disability in young adults.
3
Creative Commons Attribution The consequences of MS can be profound, affecting physical abilities, cognitive functions,
License, permitting distribution,
4-8
and reproduction in any medium, emotional well-being, and overall quality of life. As such, MS exerts a large personal
provided the original work is and societal socioeconomic burden due to the high cost of treatment and additional
properly cited. care related to permanent neurologic disability. MS is often first suspected when an
9,10
Publisher’s Note: AccScience individual experiences a clinically isolated syndrome (CIS), which is a sudden onset of
Publishing remains neutral with neurological symptoms lasting at least 24 h due to inflammation or demyelination in
regard to jurisdictional claims in 11
published maps and institutional the central nervous system. As these initial episodes recur and additional neurological
affiliations. symptoms appear over time, the condition may develop into clinically definite MS
Volume 1 Issue 4 (2024) 107 doi: 10.36922/aih.4255