Validation of a Novel Instrument Task for Assessing Upper Limb

In movement disorders, such as multiple sclerosis, impairments in sensorimotor control of goal-directed upper limb movements are commonly present, leading to a decreased ability to perform daily life activities and increased dependence on caregivers . Due to the disabling nature of upper limb impairments, they are often a focus during the neurorehabilitation process of persons with movement disorders. This process relies on a precise description of the presence and severity of sensorimotor impairments through so-called assessments. Specifically, these assessments are essential to individualize therapeutic interventions, to provide documentation for insurances justifying further therapy, and to shed light on the often unknown mechanisms underlying the impairments and their temporal evolution. Technology-aided assessments can provide detailed movement (kinematic) and grip force (kinetic) data that allows an objective and traceable description of upper limb behaviour. This offers the potential to sensitively characterize sensorimotor impairments and could help significantly reducing sample sizes required in resource-demanding clinical trials. Such technology-aided assessments typically consist of a robotic interface (e.g., exoskeleton or end-effector) that functions as a joystick and a virtual reality environment with a goal-directed manipulation task that is rendered on a 2D computer screen located next to the robotic interface. While these approaches are widely used for research purposes and showed promising results in detecting sensorimotor impairments, certain subjects were observed to have atypical movement patterns and high variability when repeatedly performing the assessment task (high intra-subject variability). These artifacts can results from unintuitive perception of depth on the 2D screen and missing visuomotor collocation, which can induce a bias in the outcome measures of the assessment and reduce the ability of the assessment to capture longitudinal changes (reliability and sensitivity). To overcome these challenges, assessment tasks relying on virtual collocated visuomotor feedback or immersive head mounted displays have been proposed. However, the virtual collocated visuomotor feedback still has its limitations, especially when subjects are supposed to interact with the virtual objects, such as in 3D object manipulation tasks that are typically used in clinical settings. In addition, head mounted display are expected to have limited clinical feasibility in persons that are naïve to technologies or have cognitive impairments. The aim of this collaborative project between ETH Zurich (Prof. Dr. Gassert, Dr. Lambercy, Dr. Kanzler) and the Kliniken Valens (Dr. med. Dr. sc. nat. Roman Gonzenbach, Ramona Sylvester) is to validate the concept of assessing upper limb sensorimotor impairments with a robotic interface embedded in a novel physical 3D object manipulation task, the Physical Peg Insertion Test (PPIT), and compare it to an existing virtual assessment approach. In more details, the aim is to perform a longitudinal study with the PPIT and its previously validated virtual pendent, the Virtual Peg Insertion Test (VPIT), in persons with movement disorders. This will allow to compare the feasibility, validity, reliability, and sensitivity of both physical and virtual instrumented assessments. Both approaches rely on a CE-certified haptic end-effector device (Geomagic Touch, 3D Systems) and an instrumented handle. For the PPIT, an electromagnet that can be controlled through the applied grip forces is attached to the end-effector and is required to pick up physical magnetic pegs in a physical pegboard task. For the VPIT, a personal computer renders a virtual reality environment with virtual pegs that need to be transported into virtual holes. These approaches have been developed at the Rehabilitation Engineering Laboratory of ETH Zurich and have been shown to be feasible, safe, and time-efficient (5 repetitions of the task, approximately 15 min duration per body side) in able-bodied controls, and persons with stroke, multiple sclerosis, or cerebellar ataxia that have mild to moderate disability levels. While the PPIT has only been developed rather recently, the VPIT has already been extensively used in clinical studies world-wide. Most importantly, the VPIT has already been successfully established as a clinical routine assessment for all persons with multiple sclerosis that are admitted to the Kliniken Valens and is therefore well known to the clinic, which provides the basis for the smooth implementation of the planned project. Through these and other previous tests, sensor data from 120 able-bodied subjects, that serve as a normative database, and over 200 persons with movement disorders could already be recorded. The collected raw kinematic and kinetic data is typically transformed through a signal processing framework into 10 sensor-based metrics that provide information on different aspects of task performance, including for example movement smoothness, accuracy, and speed. Given that the PPIT provides the same sensor data as the VPIT, the signal processing framework of the VPIT can also be applied to the data collected with the PPIT, allowing to extract the 10 sensor-based metrics in both scenarios and serves as a basis for their comparison.