Application of Machine Learning Technique to Distinguish Parkinson’s Disease Dementia and Alzheimer’s Dementia: Predictive Power of Parkinson’s Disease-Related Non-Motor Symptoms and Neuropsychological Profile

Abstract

In order to develop a predictive model that can distinguish Parkinson’s disease dementia (PDD) from other dementia types, such as Alzheimer’s dementia (AD), it is necessary to evaluate and identify the predictive accuracy of the cognitive profile while considering the non-motor symptoms, such as depression and rapid eye movement (REM) sleep behavior disorders. This study compared Parkinson’s disease (PD)’s non-motor symptoms and the diagnostic predictive power of cognitive profiles that distinguish AD and PD using machine learning. This study analyzed 118 patients with AD and 110 patients with PDD, and all subjects were 60 years or older. In order to develop the PDD prediction model, the dataset was divided into training data (70%) and test data (30%). The prediction accuracy of the model was calculated by the recognition rate. The results of this study show that Parkinson-related non-motor symptoms, such as REM sleep behavior disorders, and cognitive screening tests, such as Korean version of Montreal Cognitive Assessment, were highly accurate factors for predicting PDD. It is required to develop customized screening tests that can detect PDD in the early stage based on these results. Furthermore, it is believed that including biomarkers such as brain images or cerebrospinal fluid as input variables will be more useful for developing PDD prediction models in the future.