A CASE STUDY OF HUMAN-MACHINE INTERFACE TO CONTROL AN EMBODIED VIRTUAL BODY: AUGMENTED REALITY REHABILITATION SYSTEM FOR PATIENTS WITH STROKE

F. Kaneko¹, M. Okawada^1,2, S. Sasaki^1,2, M. Yoneta^1,2, M. Kawakami¹

Background: We previously reported the clinical effect of combining kinesthetic illusion induced by visual stimulation (KINVIS) application with therapeutic exercise. In our paradigm, KINVIS therapy is to be executed with neuromuscular electrical stimulation (NMES) to induce enhancement of primary motor cortex excitability as the after-effect and to reproduce an impression of movement in the brain. Furthermore, we have developed a human-machine computer interface (HCI) system for patients to learn the appropriate agonist muscular activation selectively involved in controlling the embodied virtual body. KINVIS-HCI is expected to be applied to patients whose joints stiffen due to spasticity or abnormal co-contraction during actual exercise.

Purpose: To provide our KINVIS-HCI system’s information and explore the different variables involved in the proposed system.

Methods: The patient was a 58-year-old woman with left paresis due to right putamen hemorrhage 3 years ago. Interventions A and B were conducted 10 times each in succession, and examinations were performed before and after each intervention (Before A, After A, After B). In intervention A, two sets of KINVIS therapy were performed for 10 minutes, using a specialized product (KiNvisTM-HMD, Inter Reha Co., Ltd., Japan). For the computer-graphics image, a hand opening and closing repeatedly was presented as the virtual body. targeted the extensor digitorum communis (EDC) and synchronized with the finger extension phase of the virtual hand. In intervention B, KINVIS therapy was performed for 10 minutes, and then KINVIS-HCI was performed for 30-50 minutes. Surface electromyogram (sEMG) was recorded to detect the activity of EDC and flexor carpi radialis (FCR) or flexor digitorum profundus (FDP), and thresholds were set individually. The patient tried to open the hand voluntarily, and when only the sEMG of the extensor muscle exceeded the threshold, the virtual hand was opened. Modified Ashworth Scale (MAS) was used to assess spasticity; Fugl-Meyer Assessment (FMA), Stroke Impairment Assessment Set (SIAS), and Action Research Arm Test (ARAT) for upper limb motor function; and Motor Activity Log (MAL) to digitize the amount of upper limb use in activities of daily living.

Results: MAS values for elbow flexors, wrist flexors, and finger flexors after each intervention stage improved over the minimal clinically important difference. The total FMA upper limb motor scores were 9/9/12 and ARAT scores were 4/6/4 (Before A/After A/After B). MAL gradually increased depending on therapy progression. In intervention B, the threshold (EDC/FCR) was set to 6.3 µV/9.2 µV on the first day, and the number of trials was 30, number of successes was 18, and success rate was 60%. On the final day, the threshold (EDC/FDP) was set to 4.4µV/4.4µV, the number of trials was 68, number of successes was 47, and success rate was 69%.

Conclusion(s): In order to use KINVIS-HCI therapy as a treatment for patients with severe motor paralysis after stroke, continued research is necessary.

Implications: Established treatments to improve upper limb motor function in patients with severe impairments due to stroke are few. Therefore, exploring clinical evidence of this therapeutic approach, based on neuroscience, is necessary to translate the system into clinical practice.

Funding, acknowledgements: This work was supported by AMED (Grant Number JP19192664), and JSPS KAKENHI (Grant Number JP19H01088).

Keywords: stroke, human-computer interface, virtual reality

Topic: Neurology: stroke

Did this work require ethics approval? No

Institution: Keio University School of Medicine

Committee: Keio University School of Medicine Ethics Committee

Reason: A case study is not required to be review by the ethics committee in Japan, however, written informed consent was obtained.