The following is a summary of “Evaluation of machine learning-based classification of clinical impairment and prediction of clinical worsening in multiple sclerosis,” published in the June 2024 issue of Neurology by Noteboom, et al.

Current limitations in predicting multiple sclerosis (MS) progression hamper efforts to manage the disease proactively.

Researchers conducted a retrospective study to assess the ability of machine learning (ML) to classify clinical impairment and predict worsening in people with multiple sclerosis (pwMS), comparing various combinations of clinical data, MRI features, and ML algorithms for optimal performance.

They utilized baseline clinical and structural MRI data from two MS cohorts (Berlin: n = 125, Amsterdam: n = 330) to evaluate five ML models’ performance in classifying baseline clinical impairment and predicting clinical worsening over 2 and 5 years. Clinical worsening was defined by increases in the Expanded Disability Status Scale (EDSS), Timed 25-Foot Walk Test (T25FW), 9-Hole Peg Test (9HPT), or Symbol Digit Modalities Test (SDMT). Various combinations of clinical and volumetric MRI measures were systematically assessed to predict outcomes. ML models were evaluated using Monte Carlo cross-validation, the area under the curve (AUC), and permutation testing for significance.

The results showed that ML models effectively identified clinical impairment at baseline in the Amsterdam cohort but did not significantly predict clinical worsening over 2 and 5 years. High disability (EDSS ≥ 4) was best identified by a support vector machine (SVM) classifier using clinical and global MRI volumes (AUC = 0.83 ± 0.07, P=0.015). Impaired cognition (SDMT Z-score ≤ -1.5) was best identified by an SVM using regional MRI volumes (thalamus, ventricles, lesions, and hippocampus), achieving an AUC of 0.73 ± 0.04 (P=0.008).

Investigators found that ML could help identify patients with MS who have clinical impairment and relevant biomarkers, but it could not predict the future worsening of the disease.

Source: link.springer.com/article/10.1007/s00415-024-12507-w