JMIR Publications

ABSTRACT

Background:

Parkinson's Disease (PD) patients (PDp) with subjective cognitive decline (PD-SCD) are considered as an intermediate status between those with normal cognition (PD-NC) and those with mild cognitive impairment (PD-MCI). Wearable digital monitoring technologies and machine-learning models offer significant potential for assessing cognitive impairment in PD patients.

Objective:

We aimed to evaluate the utility of wearable technology and machine learning for identifying ordinal cognitive stages (OCS) in PD based on Timed up and go (TUG) tests (including single-task TUG (TUGst) and dual-task cognition-gait TUG (TUGdt)).

Methods:

PDp with SCD, MCI and NC were recruited in a movement disorder clinic. Patients performed single-task TUG (TUGst) and dual-task TUG (TUGdt) gait trials wearing a wearable motion and gait quantitative evaluation system. Two-hundred-and-nine kinematic parameters were synthesized for individual TUG to illustrate patients’ motion profiles. And, we constructed dual-task cost parameters (DTC), describing the magnitude of the effect of the cognitive challenge on motion performance. Covariates adjusted Ordered Logistic Regression were employed to compared parameter differences among three groups. Pairwise Wilcoxon Rank Sum Test was employed to multiply compare differences between PD-MCI and PD-SCD, PD-MCI and PD-NC or PD-NC and PD-SCD, respectively. Random Forest (RF) model was employed to classify the subjects into three cognitive impairment levels. The total population was randomly divided into training set and independent validation set in a 7:3 ratio. And, 50 rounds of random splitting for training and testing sets were performed to assess the model's robustness. The mean absolute SHapley Additive exPlanations value (Mean(|SHAP|)) was employed to explain our final RF model.

Results:

The study included 146 patients (PD-NC: PD-SCD: PD-MCI= 33:27:86). Forty-one kinematic parameters were statistically significant differed (P<.05) among the three groups. Compared to TUGst and DTC, 80.33% of the kinematic parameters derived from TUGdt exhibited a stronger correlation with ordinal cognitive stages. PD-MCI exhibited significantly reduced gait speed (P<.01), amplitude (P<.01) and pace (P<.01) in TUGdt, when compared to other two groups. PD-SCD and PD-NC showed similar gait profile. Our multiclass RF classification model with kinematic parameters achieved a recall rate above 0.70 in both training and validation dataset.

Conclusions:

This suggests PD-SCD patients could have early kinetic signs of cognitive impairment, positioning them between PD-NC and PD-MCI, and cognitive-related features from TUGdt included Shank - Max Sagittal Angular Velocity， Trunk - Sway Max Std, Shank - Swing Speed _MINLR, Shank - Forward_Backward Swing Max _MINLR, and Gait Speed could serve as predicators to distinguish cognitive stage of PD patients.