The following is an abstract of “Evaluation of clinical disability classification and clinical deterioration prediction in multiple sclerosis using machine learning” published in the June 2024 issue. Neurology Noteboom et al.
Current limitations in predicting multiple sclerosis (MS) progression hinder efforts to aggressively manage the disease.
Researchers conducted a retrospective study to evaluate the ability of machine learning (ML) to classify clinical disability and predict deterioration in patients with multiple sclerosis (pwMS), comparing different combinations of clinical data, MRI features, and ML algorithms for optimal performance.
The researchers utilized baseline clinical and structural MRI data from two MS cohorts (Berlin: n = 125; Amsterdam: n = 330) to evaluate the performance of five ML models to classify baseline clinical disability and predict clinical deterioration after 2 and 5 years. Clinical deterioration was defined by increases in Expanded Disability Status Scale (EDSS), Timed 25-foot Walk Test (T25FW), 9-Hole Peg Test (9HPT), or Symbol Digit Modality Test (SDMT). Different combinations of clinical and volumetric MRI measures were systematically evaluated to predict outcome. ML models were evaluated using Monte Carlo cross-validation, area under the curve (AUC), and permutation tests of significance.
Results showed that in the Amsterdam cohort, the ML model effectively identified clinical disability at baseline but did not significantly predict clinical deterioration over 2 and 5 years. Severe disability (EDSS ≥ 4) was best identified by a support vector machine (SVM) classifier using clinical and global MRI volumes (AUC = 0.83 ± 0.07, PCognitive impairment (SDMT Z score ≤-1.5) was best identified by SVM using regional MRI volumes (thalamus, ventricles, lesions, hippocampus), with an AUC of 0.73 ± 0.04 (P=0.008).
The researchers found that ML could help identify MS patients with clinical impairment and associated biomarkers, but could not predict future worsening of the disease.
sauce: link.springer.com/article/10.1007/s00415-024-12507-w
