Abstract
Continuous in-home monitoring of Parkinson’s Disease (PD) symptoms might allow improvements in assessment of disease progression and treatment effects. As a first step towards this goal, we evaluate the feasibility of a wrist-worn wearable accelerometer system to detect PD tremor in the wild (uncontrolled scenarios). We evaluate the performance of several feature sets and classification algorithms for robust PD tremor detection in laboratory and wild settings. We report results for both laboratory data with accurate labels and wild data with weak labels. The best performance was obtained using a combination of a pre-processing module to extract information from the tremor spectrum (based on non-negative factorization) and a deep neural network for learning relevant features and detecting tremor segments. We show how the proposed method is able to predict patient self-report measures, and we propose a new metric for monitoring PD tremor (i.e., percentage of tremor over long periods of time), which may be easier to estimate the start and end time points of each tremor event while still providing clinically useful information.
Highlights
Parkinson’s Disease (PD) is a chronic neurodegenerative disorder that can cause a variety of motor symptoms [1]
We evaluate the performance of a full Convolutional Neural Networks (CNNs) architecture for tremor detection, as well as that of a Multi-Layer Perceptron (MLP) when hand-crafted features are used
The results show the importance of tremor spectrum extraction (-T/NT): do Mel Frequency Cepstral Coefficients (MFCCs) and CNNs show significant improvement when computed after this preprocessing step compared with raw data, but CNN-extracted features computed on raw data (CNN) perform worse than our handcrafted MFCCs computed after tremor spectrum extraction (MFCCs-T/NT)
Summary
Parkinson’s Disease (PD) is a chronic neurodegenerative disorder that can cause a variety of motor symptoms [1]. Weak labels (presence, absence, or the approximate amount of a symptom within a time segment) can be provided by participants of the data collections. Supervised algorithms, such as multiple-instance learning, explicitly account for weak labels, and tend to perform better than standard, fully supervised learning algorithms in these scenarios [4]. We propose a new metric for monitoring PD tremor (percentage of tremor time measured over longer periods), which is easier to detect than specific tremor events This new metric provides information about the symptom prevalence, and it could be used by the physician to adjust the dosage of the medication. PD motor symptom [8]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have