A decision support system based on rapid progression rules to enhance baseline evaluation of Parkinson's disease patients

Abstract

Parkinson's Disease (PD) can express different rates of symptoms progression which in some patients has been found to be extremely aggressive. In this work, a set of rules are extracted using the Repeated Incremental Pruning to Produce Error Reduction (RIPPER) algorithm to enable the prediction of such rapid disease progression cases from the baseline evaluation. After selecting the most effective of the initial 139 baseline features using the Wrapper algorithm, the best RIPPER rules are isolated based on the reported accuracy in assigning PD patients into rapid or normal progression groups. In total, about 120 unique rules that signal faster worsening of PD symptoms at baseline evaluation were extracted, assigning patients into the rapid progression group with a mean classification accuracy of 86.3%. The most informative of these rules are presented in this work, while all of them are used to develop a decision support system (DSS) to facilitate rapid progression risk assessment at the baseline evaluation. The rules were extracted using baseline evaluation data from the Parkinson's Progression Markers Initiative (PPMI) dataset. The DSS for prediction of rapid progression is included in the PD_Manager mHealth platform to offer better management of PD patients when rapid progression is expected.