Quantitative Assessment of the Arm/Hand Movements in Parkinson’s Disease Using a Wireless Armband Device

Sofija Spasojević, Tihomir Ilić, Ivan Stojković, Veljko Potkonjak, Aleksandar Rodić, José Santos-VictorFrontiers in Neurology 2017 · DOI

Abstract

We present an approach for quantitative assessment of the arm/hand movements in patients with Parkinson's disease (PD), from sensor data acquired with a wearable, wireless armband device (Myo sensor). We propose new Movement Performance Indicators (MPIs) that can be adopted by practitioners for the quantitative evaluation of motor performance and support their clinical evaluations. In addition, specific MPIs can indicate the presence of the bradykinesia symptom. The study includes seventeen PD patients and sixteen age-matched controls. A set of representative arm/hand movements is defined under the supervision of a movement disorder specialist. A total set of 84 MPIs are computed from the sensor readings to assist evaluations, progress monitoring, and assessment of bradykinesia. Results indicate high accuracy in discrimination between patients and controls, significant correlation with clinical scales (UPDRS-III), and the potential for this low-cost device to support objective clinical assessments.

Introduction

Clinical assessment of Parkinson’s Disease (PD) currently relies heavily on qualitative scales like the Unified Parkinson’s Disease Rating Scale (UPDRS). While effective, these evaluations are subjective, prone to inter-rater variability, and typically require hospital visits. Previous attempts to digitize this process often involved expensive, restrictive equipment like sensor gloves or wired Electromyography (EMG) systems. There is a critical need for a low-cost, portable, and objective tool that can monitor disease progression and symptom severity—specifically bradykinesia—across both sides of the body in a non-clinical setting.

Methods and Materials

Experimental Setup

The study utilized the Myo sensor, a consumer-grade wearable armband equipped with eight EMG channels to measure muscle electrical activity and an Inertial Measurement Unit (IMU) containing a 3-axis accelerometer and gyroscope. The device was placed on the forearm to capture data from both internal and external muscle groups.

Labeled channels of the armband sensor (A) and placement on the right hand (B).

Protocol and Feature Engineering

Seventeen PD patients (Hoehn and Yahr stages I–III) and sixteen age-matched controls performed six representative tasks, including hand rotations, grasping objects, and tapping tests. From the raw sensor data, the team proposed 84 Movement Performance Indicators (MPIs). These features were calculated using a sliding window approach to capture signal amplitude properties, such as Mean Absolute Value (MAV) for EMG and Simple Square Integral (SSI) for IMU data.

Sliding window approach used to calculate Movement Performance Indicators (MPIs) from acceleration signals.

Analysis Pipeline

Linear Discriminant Analysis (LDA) and LASSO regression identified the 13 most informative MPIs. Six machine learning algorithms (including SVM and Random Forests) were then tested for two primary tasks: diagnosing PD vs. controls and monitoring disease stages. Results were validated against traditional clinical scores (UPDRS-III and tapping tests).

Key Results

  • Diagnostic Accuracy: Classifiers achieved an Area Under the Curve (AUC) of nearly 1.0 for discriminating between patients and controls, and above 0.90 for monitoring disease stages.
  • Clinical Correlation: MPIs showed strong statistically significant correlations with clinical scales. For example, a Spearman’s correlation of -0.67 was found between EMG-based MPIs and the UPDRS-III motor score.
  • Quantifying Bradykinesia: The researchers successfully quantified the "sequence effect"—the progressive decrease in movement amplitude over time—through the temporal evolution of MPIs during tapping tasks.
  • Bilateral Advantage: Unlike right-hand-only sensor gloves, the armband's ambidextrous nature allowed for "left-right side analysis," revealing that the performance gap between hands grows as the disease progresses.

Discussion and Conclusion

This study demonstrates that a low-cost, wireless armband can provide objective and precise quantitative evaluations of arm and hand movements in PD patients. The proposed MPIs offer a reliable way to assess bradykinesia and monitor disease progression. While the current study focused on patients in their "ON" (medicated) state, future work will explore "OFF" state characteristics and expand the application of this technology to balance, stability, and stroke recovery monitoring. This approach has significant potential to afford physicians objective data during clinical evaluations, bridging the gap between subjective observation and quantitative measurement.

Summary updated

Abstract

We present an approach for quantitative assessment of the arm/hand movements in patients with Parkinson's disease (PD), from sensor data acquired with a wearable, wireless armband device (Myo sensor). We propose new <i>Movement Performance Indicators</i> that can be adopted by practitioners for the quantitative evaluation of motor performance and support their clinical evaluations. In addition, specific <i>Movement Performance Indicators</i> can indicate the presence of the bradykinesia symptom. The study includes seventeen PD patients and sixteen age-matched controls. A set of representative arm/hand movements is defined under the supervision of movement disorder specialist. In order to assist the evaluations, and for progress monitoring purposes, as well as for assessing the amount of bradykinesia in PD, a total set of 84 <i>Movement Performance Indicators</i> are computed from the sensor readings. Subsequently, we investigate whether wireless armband device, with the use of the proposed <i>Movement Performance Indicators</i> can be utilized: (1) for objective and precise quantitative evaluation of the arm/hand movements of Parkinson's patients, (2) for assessment of the bradykinesia motor symptom, and (3) as an adequate low-cost alternative for the sensor glove. We conducted extensive analysis of proposed <i>Movement Performance Indicators</i> and results are indicating following clinically relevant characteristics: (i) adequate reliability as measured by ICC; (ii) high accuracy in discrimination between the patients and controls, and between the disease stages (support to disease diagnosis and progress monitoring, respectively); (iii) substantial difference in comparison between the left-hand and the right-hand movements across controls and patients, as well as between disease stage groups; (iv) statistically significant correlation with clinical scales (tapping test and UPDRS-III Motor Score); and (v) quantitative evaluation of bradykinesia symptom. Results suggest that the proposed approach has a potential to be adopted by physicians, to afford them with quantitative, objective and precise methods and data during clinical evaluations and support the assessment of bradykinesia.
BibTeX
@inproceedings{spasojevic2017quantitative,
  author = {Sofija Spasojević and Tihomir Ilić and Ivan Stojković and Veljko Potkonjak and Aleksandar Rodić and José Santos-Victor},
  title = {{Quantitative Assessment of the Arm/Hand Movements in Parkinson’s Disease Using a Wireless Armband Device}},
  booktitle = {Frontiers in Neurology},
  year = {2017},
  doi = {10.3389/fneur.2017.00388},
}
Figure