A metric for characterizing the arm nonuse workspacein poststroke individuals using a robot arm

Nathaniel S. Dennler, Stefanos Nikolaidis, Maja MataricScience Robotics 2025 · OpenReview

TL;DR

Stroke survivors often underutilize their affected arm despite regaining some function, a phenomenon called "arm nonuse." This paper introduces the Bimanual Arm Reaching Test with a Robot (BARTR), a novel system that uses a robot arm to objectively measure this nonuse in a 3D workspace. BARTR proves to be a valid, reliable, and user-friendly metric, offering a promising tool for assessing and tracking recovery in neurorehabilitation.

What Problem Does This Paper Address?

Following a stroke, individuals often develop a reliance on their less-affected limb, even if their affected (paretic) limb has regained some capacity. This "arm nonuse" is a significant barrier to full functional recovery, as it goes against the "use it or lose it" principle of rehabilitation. Current methods for assessing arm nonuse, like the Motor Activity Log (MAL) and Actual Amount of Use Test (AAUT), have limitations. The MAL relies on self-reporting, which can be inaccurate, and the AAUT requires participants to be unaware they are being tested for validity, making it unsuitable for repeated assessments. The challenge, therefore, is to develop an objective, reliable, and ecologically valid method to quantify arm nonuse in a way that can be used in clinical settings and track progress over time.

Methods

The core of this work is the Bimanual Arm Reaching Test with a Robot (BARTR). This system uses a general-purpose robotic arm to present reaching targets in a 3D space in front of the participant. A socially assistive robot (SAR) provides instructions and motivation. The BARTR involves two phases:

  1. Spontaneous Phase (sBARTR): Participants are instructed to use either hand to reach a lit target as quickly and accurately as possible within a time limit. This encourages a natural, spontaneous hand choice, even if they know they are being tested.
  2. Constrained Phase (cBARTR): Participants are instructed to use only their stroke-affected arm to reach the target.

The system collects data on hand choice and reaching time. Arm nonuse (nuBARTR) is then calculated by comparing the performance in the spontaneous phase to the constrained phase, and also against normative data from neurotypical individuals. The researchers modeled these interaction metrics (reaching time and arm choice) using Gaussian Processes, a probabilistic modeling approach.

Datasets

The study utilized data from two groups:

  • Post-stroke participants: 17 individuals with chronic stroke were recruited from the Los Angeles, California area.
  • Neurotypical participants: 10 adults without stroke were recruited to establish normative data for comparison.

The paper also references data from previous studies using the Actual Amount of Use Test (AAUT) for validation purposes.

Key Insights and Ideas

  • 3D Workspace is Crucial: Evaluating arm nonuse solely in a 2D plane is insufficient. The BARTR's ability to assess movement in a 3D workspace provides a more nuanced understanding of where and how individuals avoid using their affected arm.
  • Robot as an Unbiased Observer: The robot arm provides objective, repeatable measurements, overcoming the subjectivity and potential inaccuracies of self-report or observer-based assessments.
  • Social Robot Enhances Engagement: The inclusion of a socially assistive robot not only provides instructions but also offers motivation, which is critical for sustained engagement in rehabilitation exercises.
  • BARTR Meets Clinical Criteria: The study demonstrates that BARTR is valid (correlates with existing metrics), reliable (consistent results over time), and easy to use, making it a strong candidate for clinical adoption.

Key Results

  • Validity: The nuBARTR metric showed a significant correlation with the AAUT AOU score (r(13)=.693, p = .016), indicating it measures a similar construct of nonuse.
  • Reliability: The BARTR demonstrated high test-retest reliability, with an intraclass correlation coefficient (ICC) of .908 (p < .001) across multiple sessions. Reliability improved after the first session, suggesting habituation is beneficial.
  • Ease of Use: Participants rated the BARTR system highly on usability, with an average System Usability Scale (SUS) score of 8.93 ± 11.67, placing it in the 80th percentile of evaluated systems and indicating above-average user experience.
  • Qualitative Feedback: Participants found the system safe, easy to learn, and interpreted the robot's actions predictably. They also suggested improvements for personalization and the inclusion of more complex tasks.

Limitations and Future Work

  • Task Simplicity: The current BARTR task involves simple reaching and button pressing. Future work should incorporate fine manipulation tasks (e.g., grasping, object manipulation) to more comprehensively assess nonuse, similar to tasks in the AAUT.
  • Participant Homogeneity: The study primarily included right-hand dominant participants. Future research should investigate left-handed stroke survivors and explore domain adaptation techniques to leverage data from right-handed participants.
  • In-Home Deployment: While the system is designed for clinical use, its potential for in-home deployment needs further investigation, particularly regarding cost, data privacy, and the need for remote expert support.
  • Long-Term Impact: The study suggests that participants would miss the robot, highlighting the need to evaluate the long-term effects of robot removal after extended use.
  • Personalization: Further research could explore adapting SAR personality and communication styles over time to enhance participant engagement.

Summary updated

Abstract

An overreliance on the less-affected limb for functional tasks at the expense of the paretic limb and in spite ofrecovered capacity is an often-observed phenomenon in survivors of hemispheric stroke. The differencebetween capacity for use and actual spontaneous use is referred to as arm nonuse. Obtaining an ecologicallyvalid evaluation of arm nonuse is challenging because it requires the observation of spontaneous arm choice fordifferent tasks, which can easily be influenced by instructions, presumed expectations, and awareness that oneis being tested. To better quantify arm nonuse, we developed the bimanual arm reaching test with a robot(BARTR) for quantitatively assessing arm nonuse in chronic stroke survivors. The BARTR is an instrument thatuses a robot arm as a means of remote and unbiased data collection of nuanced spatial data for clinical evalu-ations of arm nonuse. This approach shows promise for determining the efficacy of interventions designed toreduce paretic arm nonuse and enhance functional recovery after stroke. We show that the BARTR satisfies thecriteria of an appropriate metric for neurorehabilitative contexts: It is valid, reliable, and simple to use.
BibTeX
@inproceedings{dennler2025metric,
  author = {Nathaniel S. Dennler and Stefanos Nikolaidis and Maja Mataric},
  title = {{A metric for characterizing the arm nonuse workspacein poststroke individuals using a robot arm}},
  booktitle = {Science Robotics},
  year = {2025},
  url = {https://openreview.net/forum?id=68KWBaNG5Q},
}
Figure