[Paper Review] Wearable-based Mediation State Detection in Individuals with Parkinson's Disease

Murtadha D. Hssayeni, Michelle A. Burack+5 more|arXiv (Cornell University)|Sep 19, 2018

Parkinson's Disease Mechanisms and Treatments24 references3 citations

TL;DR

This study proposes a wearable sensor-based system using two tri-axial gyroscopes on the wrist and ankle to detect medication ON and OFF states in Parkinson’s disease (PD) patients. Employing a support vector machine with fuzzy labeling on handcrafted features, the method achieves 90.5% accuracy, 94.2% sensitivity, and 85.4% specificity, offering a personalized, continuous, and objective alternative to self-reported medication state tracking.

ABSTRACT

One of the most prevalent complaints of individuals with mid-stage and advanced Parkinson's disease (PD) is the fluctuating response to their medication (i.e., ON state with maximum benefit from medication and OFF state with no benefit from medication). In order to address these motor fluctuations, the patients go through periodic clinical examination where the treating physician reviews the patients' self-report about duration in different medication states and optimize therapy accordingly. Unfortunately, the patients' self-report can be unreliable and suffer from recall bias. There is a need to a technology-based system that can provide objective measures about the duration in different medication states that can be used by the treating physician to successfully adjust the therapy. In this paper, we developed a medication state detection algorithm to detect medication states using two wearable motion sensors. A series of significant features are extracted from the motion data and used in a classifier that is based on a support vector machine with fuzzy labeling. The developed algorithm is evaluated using a dataset with 19 PD subjects and a total duration of 1,052.24 minutes (17.54 hours). The algorithm resulted in an average classification accuracy of 90.5%, sensitivity of 94.2%, and specificity of 85.4%.

Motivation & Objective

• To develop a personalized, wearable-based system that objectively detects medication ON and OFF states in PD patients to reduce reliance on unreliable self-reports.
• To enable continuous, passive monitoring of medication response during daily activities using minimal sensor hardware.
• To address inter-subject variability by training individualized classifiers per patient rather than using a one-size-fits-all model.
• To improve therapy adjustment by providing accurate, real-time data on duration in ON and OFF states from free-living conditions.
• To validate the system in a clinical lab setting with diverse routine activities before future deployment in home environments.

Proposed method

• Two tri-axial gyroscopes are placed on the wrist and ankle to capture motion data during daily activities such as walking, dressing, and resting.
• A comprehensive set of time-domain, frequency-domain, and time-frequency features are extracted from the motion signals to represent motor behavior patterns.
• A feature selection process is applied to identify the most discriminative features for medication state classification.
• A support vector machine (SVM) classifier with fuzzy labeling is used to handle uncertainty in state boundaries and improve robustness.
• Each subject’s classifier is trained using labeled data from their own short-duration clinical session, ensuring personalization and improved generalization.
• The system is designed to operate passively in real-world settings after initial calibration, requiring no ongoing patient or clinician input.

Experimental results

Research questions

• RQ1 Can a two-sensor wearable system accurately detect medication ON and OFF states in PD patients during diverse daily activities?
• RQ2 How does individualized classifier training per patient improve detection performance compared to a shared model across subjects?
• RQ3 To what extent can a system based on handcrafted features and fuzzy SVM outperform existing methods in terms of accuracy and continuity of monitoring?
• RQ4 How does the system perform across different motor behaviors (e.g., walking, resting, dressing) without requiring activity-specific models?
• RQ5 What is the impact of sensor placement and minimal hardware on system feasibility and clinical utility?

Key findings

• The proposed algorithm achieved an average classification accuracy of 90.5% across 19 PD subjects during 1,052.24 minutes of data collection.
• Sensitivity of 94.2% indicates strong detection of ON states, crucial for identifying optimal medication response.
• Specificity of 85.4% demonstrates reliable identification of OFF states, reducing false positives.
• The system successfully detected medication states across multiple activities (ambulation, drinking, resting, dressing), enabling continuous monitoring.
• The use of individualized classifiers significantly improved performance by accounting for inter-subject variability in motor symptoms and disease progression.
• The system requires only two wearable sensors, making it practical for long-term, real-world use compared to multi-sensor systems.