Machine learning methods evaluation for identification of cognitive phenotypes in multiple sclerosis and their MRI...

Area of Science:

• Neurology
• Neuroimaging
• Cognitive Science

Background:

• Cognitive impairment (CI) is prevalent in multiple sclerosis (MS) but not well-assessed by standard scales.
• Current methods like neuropsychological testing and MRI are used separately, lacking an integrated framework for predicting cognitive decline.
• Integrating cognitive performance with brain atrophy metrics may define clinically useful cognitive phenotypes.

Purpose of the Study:

• To develop a clinically applicable framework integrating neuropsychological testing and MRI measures.
• To identify distinct cognitive phenotypes in patients with multiple sclerosis (PwMS).
• To correlate cognitive phenotypes with specific patterns of brain atrophy.

Main Methods:

• Collected data from 79 PwMS, including comprehensive neuropsychological assessment and brain MRI.
• Applied feature selection (variance, MI) to neuropsychological data, followed by Partitioning Around Medoids (PAM) clustering.
• Evaluated differences in brain atrophy measures between identified clusters using ANOVA or Kruskal-Wallis tests.

Main Results:

• Feature selection identified key neuropsychological variables for clustering.
• PAM algorithm revealed three distinct cognitive phenotypes in PwMS.
• These phenotypes varied in clinical characteristics, brain atrophy patterns, and cognitive status, from preserved to globally impaired cognition.

Conclusions:

• Three cognitive phenotypes integrating neuropsychological and MRI data offer a clinically applicable framework for PwMS.
• This approach can bridge the gap between imaging findings and cognitive assessment, improving screening and monitoring.
• Early detection of CI and tailored rehabilitation planning can be enhanced by this integrated method.