Online compensation detecting for real-time reduction of compensatory motions during reaching: a pilot study with stroke survivors

So you obviously missed the point that you learn faster by having errors and correcting them. Good thing you aren't training toddlers to walk, they would never get to perfection and actually walk.  

Online compensation detecting for real-time reduction of compensatory motions during reaching: a pilot study with stroke survivors

• Siqi Cai,
• Xuyang Wei,
• Enze Su,
• Weifeng Wu,
• Haiqing Zheng &
• Longhan Xie 

Journal of NeuroEngineering and Rehabilitation volume 17, Article number: 58 (2020) Cite this article

Abstract

Background

Compensations are commonly observed in patients with stroke when they engage in reaching without supervision; these behaviors may be detrimental to long-term functional improvement. Automatic detection and reduction of compensation cab help patients perform tasks correctly and promote better upper extremity recovery.

Objective

Our first objective is to verify the feasibility of detecting compensation online using machine learning methods and pressure distribution data. Second objective was to investigate whether compensations of stroke survivors can be reduced by audiovisual or force feedback. The third objective was to compare the effectiveness of audiovisual and force feedback in reducing compensation.

Methods

Eight patients with stroke performed reaching tasks while pressure distribution data were recorded. Both the offline and online recognition accuracy were investigated to assess the feasibility of applying a support vector machine (SVM) based compensation detection system. During reduction of compensation, audiovisual feedback was delivered using virtual reality technology, and force feedback was delivered through a rehabilitation robot.

Results

Good classification performance was obtained in online compensation recognition, with an average F1-score of over 0.95. Based on accurate online detection, real-time feedback significantly decreased compensations of patients with stroke in comparison with no-feedback condition (p < 0.001). Meanwhile, the difference between audiovisual and force feedback was also significant (p < 0.001) and force feedback was more effective in reducing compensation in patients with stroke.

Conclusions

Accurate online recognition validated the feasibility of monitoring compensations using machine learning algorithms and pressure distribution data. Reliable online detection also paved the way for reducing compensations by providing feedback to patients with stroke. Our findings suggested that real-time feedback could be an effective approach to reducing compensatory patterns and force feedback demonstrated a more enviable potential compared with audiovisual feedback.

Background

Stroke is the leading cause of long-term disability in adults worldwide [1], and many post stroke patients suffer from varying degrees of upper extremity motor dysfunction [2]. Skilled reaching is an important aspect of upper limb motor ability but is impaired after stroke [3, 4]. Patients with stroke usually develop adaptive compensatory patterns, particularly by recruiting excessive trunk or shoulder movements during reaching [5, 6]. The use of compensatory motions could be beneficial for an immediate improvement in function; however, such a functional improvement occurs because of a reinforcement of compensation instead of true motor recovery [7]. Patients with stroke who commonly use compensatory strategies may form nonoptimal motion patterns, hindering long-term recovery of their impaired arms [8, 9]. Previous studies have demonstrated that reducing compensatory patterns has the potential to improve the final functional outcome. Improvements were accompanied by lager active joint range [8], higher FMA-UE score [10] and recovery of lost motor patterns [7]. Therefore, therapists correct undesired compensatory motions when they supervise therapeutic exercises. However, stroke patients perform many exercises without supervision, such as home therapy, which highlights the need to detect and reduce compensation in unsupervised rehabilitation [10].
Automatic detection of compensation can ensure subsequent intervention to prompt the patient into the correct pose. Previous studies have evaluated the feasibility of sensor-based and camera-based systems to detect compensation without the supervision of a therapist [11,12,13,14,15]. However, camera systems are not appropriate for application in clinical or home settings, which face challenges such as object obstruction, complex setups and privacy [13, 16]. Sensor-based systems suffer from inducing unnatural motions due to the attachment of sensors. Moreover, the reliability of the outcome estimates from these sensors is still a challenge for researchers [14, 16]. While the detection of compensatory patterns still lacks a simple, unobtrusive and practical method, we have proposed a pressure distribution-based compensation detection system [17, 18]. With a pressure mattress mounted on the chair, participants performed seated reaching tasks, and the pressure distribution data were recorded. Several features were extracted from the pressure distribution data that reflected the information for different kinds of compensatory motions. Different models were applied to recognize compensatory patterns and achieved excellent offline recognition accuracy. Our previous studies pave the path toward detecting compensation based on pressure distribution data and machine learning methods. However, there is still a gap between online and offline detection performance, and few studies on the real-time detection of compensation have been reported. To our knowledge, no previous study has evaluated the feasibility and validity of detecting compensatory motions based on pressure distribution data and machine learning methods in real time. Therefore, the purpose of this study is to investigate whetherthe pressure distribution-based method can be implemented in the real-time monitoring of compensatory motions in patients with stroke.
Based on the real-time detection of compensation, various feedback strategies, in the form of visual [19, 20], auditory [21, 22], or force feedback [23], were provided to patients with stroke to modify their motion patterns. However, there is still no consensus on the kind of feedback modalities that would be effective in reducing compensation. In this study, virtual reality (VR) technology was employed to provide audiovisual feedback, while a rehabilitation robot was employed to provide force feedback. This pilot study aimed to investigate whether the compensation of stroke survivors during reaching can be reduced by audiovisual and force feedback and to examine whether one feedback method is superior to the other.
Therefore,the main contributions of this paper are as follows:

1. 1)The implementation and validation of the presented compensation-detecting method using pressure distribution data and machine learning algorithms in real time;
2. 2)The use of virtual reality and a rehabilitation robot to reduce compensatory motions in patients with stroke during reaching tasks; and
3. 3)The comparison of audiovisual and force feedback for reducing compensation.

Methods

Participants

Eight poststroke subjects were recruited from the Third Affiliated Hospital, SUN Yat-sen University. A summary of the participants’ demographics was provided in Table 1. Ethical approval was obtained from the Guangzhou First People’s Hospital Department of Ethics Committee. All participants reviewed and signed an informed consent form prior to entering the study.

Table 1 Demographic and clinical data for participants with stroke

Full size table

Participants were included if they had experienced their first-ever stroke, if they were either in the subacute (between 1 to 6 months poststroke) or chronic (over 6 months poststroke) stage of recovery, if they were able to remain in a sitting posture (without back or arm rests) for at least 10 min and if they were able to perform the required motions. Participants were excluded from the study if they had the following: upper limb pain > 4/10 on a visual analogue scale (VAS) [24]; upper limb spasticity > 2 on the Modified Ashworth Scale (MAS) [25]; Mini-Mental State Examination (MMSE) score ≤ 23 [26]; or visual spatial neglect based on clinical judgment.