Original scientific paper
SUSTAINED COGNITIVE DECLINE IN MULTIPLE SCLEROSIS: INVESTIGATING THE ROLE OF WHITE MATTER LESION LOAD USING AN AI-DRIVEN BRAIN IMAGING APPROACH
Vito Tota
; Department of Neurology, Centres Hospitaliers Universitaires HELORA, Mons, Belgium
Astrid Mehuys
; Department of Neuroscience, Research Institute for Health Science and Technology, University of Mons, Mons, Belgium
Tanguy Vansnick
; Computer Science, Software and Artificial Intelligence Unit (ILIA), University of Mons, Mons, Belgium
Otmane Amel
; Department of Computer Science, University of Tiaret, Tiaret, Algeria
Fatma Chahbar
; Laboratoire de Génie Energétique et Génie Informatique (L2GEGI), University of Tiaret, Tiaret, Algeria
Lamia Mahmoudi
; Laboratoire de Génie Energétique et Génie Informatique (L2GEGI), University of Tiaret, Tiaret, Algeria
Sidi Ahmed Mahmoudi
; IMT Nord Europe, Institut Mines-Télécom, Center for Digital Systems, Lille, France
Giovanni Briganti
; Department of Computational Medicine and Neuropsychiatry, Faculty of Medicine, University of Mons, Mons, Belgium
Laurence Ris
; Department of Computational Medicine and Neuropsychiatry, Faculty of Medicine, University of Mons, Mons, Belgium
Said Mahmoudi
; Department of Computational Medicine and Neuropsychiatry, Faculty of Medicine, University of Mons, Mons, Belgium
Abstract
Background: Multiple sclerosis (MS) is a chronic inflammatory and neurodegenerative disease of the central nervous system, where cognitive impairment can occur even without physical disability. The underlying mechanisms remain poorly understood. This study investigates the role of white matter lesion load (WMLL) in sustained cognitive decline (SCD) in a real-life MS cohort, using an artificial intelligence(AI)-based brain imaging approach. Methods: Patients from the CHU Helora MS database with ≥3 SDMT assessments and serial brain MRIs were included. SCD was defined as a ≥4-point or ≥10% SDMT drop, confirmed 6 months later. Patients were stratified into two groups: those with SCD (COG) and those without (N-COG). WMLL was measured using a AI-based model that provides segmentation masks. Lesion volume was calculated by multiplying segmented voxels by voxel size. Results: Of 109 eligible patients, 43 met inclusion criteria. Seven showed SCD; 36 did not. Imaging data were available for 5 COG and 21 N-COG patients. There was no significant difference in WMLL or its progression between patients with and without SCD. Fewer than half of the patients in the COG group showed an increase in WMLL over time, and those who did were older than the group average. WMLL changes were not a reliable marker of SCD. Consistent with previous findings, the COG group included more males, and disease control appeared more challenging. Vascular pathology may be misclassified by segmentation algorithms, which partially explain why the two patients with WMLL progression were older. Gray matter was not assessed, though it may play a key role in this phenomenon. Conclusion: SCD did not consistently correlate with WMLL progression. Affected patients were predominantly male, consistent with a more aggressive disease course. WMLL may also be influenced by age-related factors. Alternative imaging biomarkers are needed to explain SCD in MS.
Keywords
multiple sclerosis; white matter lesion load; cognitive impairment; artificial intelligence; neuroimaging
Hrčak ID:
344114
URI
Publication date:
20.9.2025.
Visits: 158 *