Wearable sensors for Parkinson's disease: which data are worth collecting for training symptom detection models

Abstract

Machine learning algorithms that use data streams captured from soft wearable sensors have the potential to automatically detect PD symptoms and inform clinicians about the progression of disease. However, these algorithms must be trained with annotated data from clinical experts who can recognize symptoms, and collecting such data are costly. Understanding how many sensors and how much labeled data are required is key to successfully deploying these models outside of the clinic. Here we recorded movement data using 6 flexible wearable sensors in 20 individuals with PD over the course of multiple clinical assessments conducted on 1 day and repeated 2 weeks later. Participants performed 13 common tasks, such as walking or typing, and a clinician rated the severity of symptoms (bradykinesia and tremor). We then trained convolutional neural networks and statistical ensembles to detect whether a segment of movement showed signs of bradykinesia or tremor based on data from tasks performed by other individuals. Our results show that a single wearable sensor on the back of the hand is sufficient for detecting bradykinesia and tremor in the upper extremities, whereas using sensors on both sides does not improve performance. Increasing the amount of training data by adding other individuals can lead to improved performance, but repeating assessments with the same individuals-even at different medication states-does not substantially improve detection across days. Our results suggest that PD symptoms can be detected during a variety of activities and are best modeled by a dataset incorporating many individuals.

Keywords: Diagnostic markers; Parkinson's disease; Rehabilitation.

Fig. 1

Data collection and sensor setup. a Individuals with PD underwent multiple clinical assessments spaced by 30 min during a first visit (day 1); assessments were done before and after each participant took their PD medication. A single follow-up assessment was performed about 2 weeks later during a second visit (day 2). During each assessment, participants performed a series of daily activities and standardized clinical tasks. b Overview of the MC10 BioStampRC senor; c Position of sensors on the body and sensor modalities; sensors were placed on both sides, although only one side is shown for clarity. Data were recorded from the accelerometer (acc) and gyroscope (gyro) sensors, or from the accelerometer and electromyography (EMG) sensor. Data from the EMG sensor was not used in the current study

Fig. 2

The effect of sensor location and number of sensors on model performance. AUROC curves for detection of bradykinesia and tremor using random forest population models a when trained on data from either a single hand (with dominant symptoms), both hands (Hand_Bi), or a combination of hand, forearm and thigh sensors unilaterally (Combo). Corresponding AUROC curves from CNN population models b for detection of bradykinesia and tremor. Using a combination of sensors did not yield any advantage to only using hand sensors. Solid lines indicate mean AUROC and shaded areas represent 95% confidence intervals

Fig. 3

Symptom detection across activities. AUROC curves for bradykinesia and tremor detection using population models (random forest), split by groups of activities. Symptoms were detected equally well during both clinical structured tasks (e.g. finger to nose movements) and most of daily functional activities (walking and fine motor tasks, e.g. typing on a keyboard). The lowest AUROC was during the execution of gross motor tasks

Fig. 4

Comparing performance of population models with different number of individuals in the training pool. Plots showing the trend in expected AUROC for bradykinesia and tremor detection when testing on a new individual, when using population models with varying numbers of individuals in the training pool. Shaded areas mark 95% confidence interval on the mean

Fig. 5

The effect of using training data from multiple assessments (session) on model performance. Boxplots show the distribution of AUROC across participants for population models trained on data from day 1 and tested on either day 1 or day 2. Adding data from multiple sessions improves the AUROC for bradykinesia detection on day 1. However, no significant changes in AUROC occur when models are tested on data from a different day (day 2). A similar pattern is observed for detection of tremor, although a modest improvement was observed on day 2