Background: Stroke patients usually have arm weakness, which affects trunks and arms. Objective: To investigate the effects of paretic side and non-paretic side arm training on trunk control and upper limb functions. Design: Randomized Controlled Trial (single blind). Methods: Twenty patients with stroke in hospital were enrolled in the study. Twenty subjects were randomly assigned to paretic side arm training group (PATG, n = 10) or non-paretic side arm training group (NATG, n = 10). Trunk impairment scale (TIS) was used for trunk control, and box and block test (BBT) was used for upper limb function. Training was conducted for 4 weeks. Results: PATG showed significant difference in TIS (static balance, dynamic balance, coordination, total score) and BBT. NATG showed significant differences in static balance, and dynamic balance and total score except for coordination and BBT. PATG also showed a more significant difference in BBT and coordination and total score than NATG. Conclusions: The arm training performed on the paretic side are more effective than those performed on the non-paretic side in improving both upper limb function and trunk control in stroke patients.

양손 협응을 통한 상지의 정상적인 운동은 일상생활동작 수행의 질을 결정하는 필수조건이 된다. 그러나 뇌졸중으로 인한 신경학적 손상은 뇌손상이 일어난 반대측 신체에 감각운동적 기능장애를 일으키게 된다. 마비측 상지 훈련뿐만 아니라 양측 상지의 동시적 운동 수행이 마비된 상지의 기능회복에 미치는 영향에 대한 연구가 활발히 진행되고 있다. 그러나 양측 상지 훈련을 다양한 기능 수준의 환자군에게 적용한 연구가 없다. 다양한 양측 상지 훈련 개발을 위해 양측 상지를 이용한 과제 에 대해 정확한 분석이 이루어져야 할 것이며 치료 효과를 이어갈 수 있도록 양측 상지 훈련에 대한 홈 프로그램 운동이나 프로토콜 개발이 필요할 것이라고 생각된다.

The aim of this study was to determine the effect of action-observation training on arm function in people with stroke. Fourteen chronic stroke patients participated in action-observation training. Initially, they were asked to watch video that illustrated arm actions used in daily activities; this was followed by repetitive practice of the observed actions for 3 times a week for 3 weeks. Each training session lasted 30 min. All subject participated 12 training session on 9 consecutive training days. For the evaluation of the clinical status of standard functional scales, Wolf motor function test was carried out at before and after the training and at 2 weeks after the training. Friedman test and Wilcoxon signed rank test was used to analyze the results of the clinical test. There was a significant improvement in the upper arm functions after the 3-week action-observation training, as compared to that before training. The improvement was sustained even at two weeks after the training. This result suggest that action observation training has a positive additional impact on recovery of stroke-induced motor dysfunctions through the action observation-action execution matching system, which includes in the mirror neuron system.

Reinforcement learning has been applied to various problems in robotics. However, it was still hard to train complex robotic manipulation tasks since there is a few models which can be applicable to general tasks. Such general models require a lot of training episodes. In these reasons, deep neural networks which have shown to be good function approximators have not been actively used for robot manipulation task. Recently, some of these challenges are solved by a set of methods, such as Guided Policy Search, which guide or limit search directions while training of a deep neural network based policy model. These frameworks are already applied to a humanoid robot, PR2. However, in robotics, it is not trivial to adjust existing algorithms designed for one robot to another robot. In this paper, we present our implementation of Guided Policy Search to the robotic arms of the Baxter Research Robot. To meet the goals and needs of the project, we build on an existing implementation of Baxter Agent class for the Guided Policy Search algorithm code using the built-in Python interface. This work is expected to play an important role in popularizing robot manipulation reinforcement learning methods on cost-effective robot platforms.