A Random Forest–Driven Clinical Decision Model for Early Detection of Clinically Definite Multiple Sclerosis Using Multimodal Diagnostic Data

Early and precise diagnosis of Clinically Definite Multiple Sclerosis (CDMS) continues to be a significant issue in neurology, especially during the progression from initial symptoms to a conclusive diagnosis. This study introduces a machine learning-driven predictive model that utilizes multimodal diagnostic data—comprising MRI results, somatosensory evoked potentials (LLSSEP and ULSSEP), visual and auditory evoked potentials (VEP, BAEP), and clinical-demographic variables—to categorize patients as CDMS or non-CDMS. A Random Forest classifier was developed and verified using a dataset of 273 actual patients. The model had a very high predictive performance, with an area under the ROC curve (AUC) of more than 0.90 and very good precision-recall metrics. Feature importance analysis indicated that spinal cord MRI, oligoclonal bands, and lower limb SSEPs are the most significant predictors of CDMS. These results correspond with established biomarkers of MS development, illustrating the model's clinical significance. The suggested framework provides understandable, highly accurate decision assistance that could help neurologists diagnose MS early and figure out how likely someone is to have it. This research advances the emerging domain of AI-driven neurodiagnostics and underscores the efficacy of integrating electrophysiological, imaging, and historical patient data for precise classification of neuroinflammatory illnesses. Future validation in multi-center cohorts may facilitate its implementation in clinical decision systems.