Selecting and Evaluating Key MDS-UPDRS Activities Using Wearable Devices for Parkinson's Disease Self-Assessment

Abstract

Parkinson's disease (PD) is a complex neurodegenerative disease in the elderly. This disease has no cure, but assessing these motor symptoms will help slow down that progression. Inertial sensing-based wearable devices, such as mobile phones and smartwatches have been widely employed to analyze the condition of PD patients. However, most studies purely focused on a single activity or symptom, which may ignore the correlation between activities and complementary characteristics. In this article, a novel technical pipeline is proposed for fine-grained classification of PD severity grades, which identify the most representative activities. We also propose a multiactivities combination scheme based on MDS-UPDRS. Utilizing this scheme, symptom-related and complementary activities are captured. We collected 85 PD subjects of different severity grades using a single wrist sensor. Our best results demonstrate F1 scores of 95.75 $\%$ for PD diagnosis and the fine-grained classification accuracy of PD disease grade is 82.41 $\%$ when combing four activities which improved by 11.02 $\%$ over a single activity. The experiments and theoretical analyses can serve as a useful foundation for future investigations into the effect of proposed solutions for PD diagnosis in uncontrolled environment setup, ultimately leading to self-PD assessment in the home environment.