Fist‐Palm Test May Identify Mild Cognitive Impairment and Freezing of Gait in Parkinson's Disease: A Machine Learning Approach

Background: Mild cognitive impairment (MCI) and freezing of gait (FOG) are two common symptoms in Parkinson's disease (PD). Objectives: The objectives were to test the strength of association of fist-palm test (FiPaT), a nonverbal motor test, with both MCI and FOG in PD and investigate the predictive ability of FiPaT in the identification of PD patients with MCI or FOG. Methods: We enrolled 74 PD patients: 47 of 74 patients had MCI (PD + MCI), 27 of 74 were cognitively unimpaired (PD-NC), 29 of 74 presented FOG (PD + FOG), and 45 of 74 were without FOG (PD-FOG). We performed univariate statistical analysis, including binary logistic regressions, to determine the variables that distinguish PD + MCI from PD-NC and PD + FOG from PD-FOG. We implemented machine learning algorithms with feature selection, using clinical-demographic features and FiPaT scores as input variables, and the presence of MCI or FOG as features to predict. Results: FiPaT total error and topography errors (P = 0.049 and P = 0.026) significantly discriminated PD + MCI and PD-NC, whereas disease duration (P = 0.039), FiPaT total error (P = 0.026), and attention errors (P = 0.046) significantly discriminated PD + FOG and PD-FOG. Applying machine learning analysis, the models reached an accuracy of 87.4% for MCI and 78.2% for FOG. Conclusions: A worse performance on FiPaT and its subscores is closely related to MCI and FOG in PD. The early identification of MCI and FOG is of particular interest as both are important risk factors for PD dementia. Moreover, the association between FiPaT and FOG strengthens the close relationship between motor system, namely gait, and higher-order cognitive functions.