Improving the calibration of machine-learning models for predicting disease progression of Multiple Sclerosis patients

Abstract

Background: Evoked potentials (EPs) are electrical signals that are produced by the nervous system in response to an external stimulus. They are used to monitor disease progression of Multiple Sclerosis (MS) patients. Previous studies have used several machine learning algorithms to prove this relationship, but until now the calibration properties of these models have not been sufficiently investigated. This research performs a machine learning analysis on latencies of motor EP and investigates how well the probabilistic outputs of the model are calibrated.