The features of Gait Exercise Assist Robot: Precise assist control and enriched feedback

1 Background

Robot has received high expectations as a usefulness tool for rehabilitation. The characteristics of robot include the ability of reproducing movements accurately without being affected by fatigue (Schwartz et al., 2009; Westlake et al., 2009; Bae et al., 2014). However, repeating the same movement alone is not adequate to provide highly efficient exercise. What is important here is the concept of motor learning. The major variables of motor learning are transferability, motivation, performance change, and retention or application. The factors that affect performance change include feedback, amount of exercise (frequency), and difficulty of exercise (Schmidt et al., 2013). Therefore, a robot that assists learning has to be developed taking into account the above variables.

Various types of robot assisted gait exercise systems have been developed, and an increasing number are introduced into clinical practice. In the meta-analysis reported by Mehrholz et al. (2013), when patients in the subacute phase after stroke underwent gait exercise using robotic equipment in addition to conventional physiotherapy, the proportion of patients walking independently increased significantly compared to physiotherapy alone, although there was no significant difference in walking velocity. In two randomized controlled trials of individual robots, Pohl et al. (2007) used the Gait Trainer and Schwartz et al. (2009) used the Lokomat for gait exercise in combination with routine physiotherapy, and reported significant improvement in the rate of independent walking compared to the group that underwent physiotherapy alone during the same period. Both of these robots are of the type that controls both lower limbs. Basically, exercise is conducted with a symmetrical gait pattern. However, the gait that post-stroke hemiplegic patients regain is not uniform. In the case that residual paralysis is severe, compensatory motion is required (Knutsson et al., 1979; Gaviria et al., 1996; Roth et al., 1997; Woolley 2001; Oken al., 2008). In these patients, the final gait may show left-right asymmetry. If left-right symmetrical gait is imposed from the start of exercise, the gait during exercise would differ greatly from the final gait, and task transferability may be limited.

We have engaged in the development of the Gait Exercise Assist Robot (GEAR), a robot designed to provide a highly efficient gait exercise environment for post-stroke hemiplegic patients, which assists only the hemiplegic lower limb and allows flexible adjustments of the motor learning variables. In this article, the design concept and characteristics of GEAR are described, the clinical course of a patient actually using the GEAR for gait exercise is reported, and the effectiveness of GEAR is discussed.

2 Issues of gait exercise using conventional orthoses

Gait exercise for patients with hemiplegia after stroke generally utilizes lower limb orthosis (Hesse et al., 1999; Teasell et al., 2001; Danielsson et al., 2004). Use of lower limb orthosis limits the degree of freedom of lower limb joints and simplifies the motions, thereby stabilizing gait even in patients with paralysis. If paralysis is severe, greater limitation of the degree of freedom is necessary (Krebs et al., 2015). For example, in a patient with severe hemiplegia, the risk of the knee giving way is high if an ankle-foot orthosis is used during the early stage of gait exercise, and use of a knee-ankle-foot orthosis is thus desirable (Yamanaka et al., 2004). However, use of a knee-ankle-foot orthosis has the following issues:

Since the patient has difficulty launching the foot off the ground by him/herself, the amount of assistance inevitably increases, and control is mainly by the therapist rather than the patient him/herself.

With the objectives of resolving the above issues and providing highly efficient gait exercise from the motor learning point of view, we collaborated with Toyota Motor Corporation to develop the GEAR.

3 Basic components of GEAR

The GEAR is composed of a knee-ankle-foot robot, a low floor treadmill, a safety suspending device (can be used as a body weight support device), a robot weight support device, a monitor for patient use, and a control panel (Fig. 1). The knee-ankle-foot robot, which is equipped with a motor attached to the knee joint, has a weight of approximately 5.7 kg. However, since the weight is canceled by the robot weight support device, the patient does not feel the weight. The plantar region of the robot is equipped with a pressure sensor. The robot judges the gait cycle from the pressure sensor and the knee joint angle, and effects flexion and extension of the knee joint at the appropriate timing. All the operations of the robot are controlled by the control panel. For most hemiplegic patients, gait exercise can be conducted by one therapist. A skillful therapist is able to help the patient wear the robot in 2–3 min and take off the robot in 1-2 min. The knee-ankle-foot robot can be adjusted according to the patient’s leg length and the eversion and inversion angles of the knee joint. One robot may be adapted to be used by multiple patients.

OPEN IN VIEWER

Fig.1

Basic components of GEAR.

3.2 Characteristics of GEAR

One of the characteristics of GEAR is enriched feedback. As a feedback item for patients, the monitor at the front can display either the full-length image (mirror image) or the foot image. When foot image is selected, the target position of foot contact with the floor is displayed as overlay. As acoustic feedback, the device can be set to emit a sound of success when the weight on the hemiplegic side exceeds the set value, and a sound of failure when the knee gives way. Patient may have difficulty integrating and understanding multiple sources of information. In that case, the information with higher priority is selected and used as feedback. On the control panel for use by the therapist, more detailed information including weight bearing on the hemiplegic side and trajectory of the center of foot pressure can be displayed at real time.

Another characteristic of GEAR is precise assist control. The adjustable parameters include knee extension assist, swing assist, timing of beginning knee flexion, time of knee flexion and extension, and amount of body weight support. These parameters can be adjusted on the control panel (Fig. 2). With this feature, it is possible to always provide exercise at the optimal level of difficulty. For example, the knee extension assist is a function that assists knee joint extension during the stance phase, and can be set in 10 steps from level 10 (maximum) to level 1 (minimum). At level 10, even when the patient has absolutely no muscular strength for knee joint extension, the robot assists knee joint extension at an adequate strength (96 Nm) that does not cause knee joint flexion. As a result, the knee does not give way. At level 1, the motor provides no assistance for knee joint extension, and the patient him/herself has to execute knee joint extension voluntarily as when wearing an ankle-foot orthosis. The swing assist is a function that supports the weight exerted on the hemiplegic lower limb during the swing phase, and can be set from level 6 (maximum) to level 1 (minimum). At level 6, a weight of approximately 2 kg plus the weight of the robot is supported in the vertical direction, so that the patient can easily launch the hemiplegic lower limb. At level 1, a small weight of 0.5 kg less than the weight of the robot is supported. Therefore, the patient has to swing the hemiplegic lower limb while feeling a similar weight as if wearing a conventional ankle-foot orthosis. By decreasing the level of assist stepwise according to functional recovery of the hemiplegic lower limb, it is possible to always develop the maximum muscular strength of the patient during each exercise.

OPEN IN VIEWER

Fig.2

Screenshot of the control panel.

The merits of GEAR from the viewpoint of motor learning are summarized below.

From the beginning, it is possible to exercise a natural gait accompanied by knee flexion, similar to the final gait (transferability).

By providing a necessary and adequate level of assistance, it is possible to achieve gait at supervision level from the early stage of exercise, thus providing motivation (motivation).

Since knee flexion facilitates swinging and improves cadence, practicing a large number of steps becomes possible (amount of exercise).

By selecting the appropriate item from multiple feedback modes, success or failure of the patient to achieve a necessary task can be demonstrated clearly at real time (feedback).

Sophisticated adjustability of the level of assist by the robot allows setting of the appropriate level of difficulty for exercise (difficulty of exercise).

One case of actual gait exercise using the GEAR is reported.

A 70-year-old man presented with left thalamus hemorrhage and right hemiplegia. Exercise using the GEAR was started 17 days after symptom onset. Evaluation of the affected lower limb using the Stroke Impairment Assessment Set (SIAS) (Chino et al., 1994; Chino et al., 1996) immediately before starting GEAR showed the following results: hip-flexion test score 0, knee-extension test score 1, foot-pat test score 0. The patient lost superficial and positional sensations at this time. Although the patient showed mild aphasia and attention disturbance, the severity did not impede implementation of exercise. In regular gait exercise, a knee-ankle-foot orthosis was necessary because of marked giving way of the knee when using an ankle-foot orthosis. While wearing the knee-ankle-foot orthosis, assistance to shift the center of gravity was constantly required. The Functional Independence Measure (FIM) score for walk (FIM-walk score) (Data management service of the Uniform Data System for Medical Rehabilitation and the Center for Functional Assessment Research 1990) was evaluated as 2.

The patient underwent regular rehabilitation 6 days a week, in addition to exercise with the GEAR 5 days a week, for 40 minutes per day. Rehabilitation for one day, which included physiotherapy, occupational therapy, speech therapy and GEAR exercise, did not exceed 180 min. On the first day of GEAR exercise, both knee extension assist and swing assist were set at the maximum levels, and the gait exercise speed was also set at a low level of 0.2 km/h. As the patient was unfamiliar with robot exercise, maximum assistance (equivalent to FIM 2) was required. Because of difficulty in perceiving weight-bearing on the affected side, audio feedback was provided by a sound of success when the weight loaded on the anterior part of the foot exceeded the threshold value. Visual feedback was also given by displaying the foot image on the front monitor to reduce variability of foot contact position due to severe sensory deficit. As a result, swing and stance were both improved. After 3 days of exercise, gait at supervision level was possible when the GEAR was used. From this time onward, stepwise decreases of knee extension assist and swing assist were started. After one week of GEAR exercise, knee extension assist was decreased to level 1, but swing assist remained at level 5 and a high level of assistance for launching was still required. The FIM-walk score remained at 2. In the next week, while the level of swing assist continued to decrease gradually, high-intensity gait with GEAR was conducted at a speed exceeding the comfortable walking speed on level ground. As a result, swing assist was decreased to level 2 after 2 weeks of exercise. At this time point, although knee joint instability in the stance phase was observed, level ground gait using an ankle-foot orthosis and a four-legged cane was at supervision level. Because of loss of deep and superficial sensory function, launching of the affected lower limb was probably executed depending on visual feedback displayed on the front monitor. From the third week, the frequency of displaying information on the front monitor was reduced gradually to utilize intrinsic feedback. When the visual feedback was discontinued, marked variation in stride length was observed in the beginning, but thereafter improved gradually. In addition, GEAR exercise continued at a speed exceeding the comfortable speed of level ground gait, paying attention to train at an appropriate level of difficulty. After 4 weeks of GEAR exercise, two-point forward gait using a 4-legged cane and an ankle-foot orthosis was stabilized, and the level ground gait speed was improved to 0.84 km/h. Since gait using a T-shaped cane and ankle-foot orthosis also reached supervision level, GEAR exercise was terminated after 4 weeks. A summary of the course of exercise is shown in Table 1.

The course of exercise of the present case was compared with the average course of patients who underwent gait exercise using conventional orthosis. The control group consisted of 15 patients aged 20–75 years with hemiplegia after primary stroke, who were admitted to the convalescent rehabilitation ward of another hospital of our university and fulfilled the following criteria: initial evaluation after admission was done within 60 days after onset; FIM scores for cognitive comprehension, memory and social interaction were 3 or above; FIM-walk score and SIAS total lower limb score were identical to those of the present patient. The control patients underwent rehabilitation 7 days a week, with the total time of physiotherapy, occupational therapy and speech therapy not exceeding 180 minutes per day. The primary outcome measure was improvement efficiency of FIM-walk as defined by the following equations, and compared between the patient and the control group:

(In the case that FIM-walk score reached 5 within 4 weeks from the start of GEAR exercise or admission) Improvement efficiency of FIM - walk = 5 - ( FIM - walk score at start ) No . of weeks until achieving FIM - walk score 5

(In the case that FIM-walk score did not reach 5 at 4 weeks from the start of GEAR exercise or admission) Improvement efficiency of FIM - walk = ( FIM - walk score after 4 weeks ) - ( FIM - walk score at start ) 4 weeks

The characteristics and Improvement efficiency of FIM-walk of the present case and the control group are shown in Table 2. In addition, the changes in FIM-walk score over time are shown in Fig. 3. Since the data of the control group were available every two weeks, the Improvement efficiency of FIM-walk of the patient who underwent GEAR exercise was also calculated every two weeks for comparison. As only one patient underwent GEAR exercise, statistical comparison was difficult. However, the Improvement efficiency of FIM-walk was markedly higher in the patient using GEAR than in the control group, suggesting that gait exercise using GEAR is effective for improving gait capability in post-stroke hemiplegic patients.

OPEN IN VIEWER

Fig.3

Changes in FIM-walk score over time.

Although the patient presented in this article had severe paralysis and sensory deficit, his gait capability was improved smoothly by gait exercise using the GEAR. Because of the severe sensory disturbance, the patient had difficulties controlling the foot position at launching in the beginning of exercise. Patients with severe sensory disturbance tend to look at the feet while walking, in order to confirm the foot contact position. This often causes the hip joint to flex, leading to pelvis retraction. In the present case, by displaying the foot image together with the target position of ground contact on the front monitor, the patient was able to receive feedback on the ground contact position while maintaining a good posture. In addition, in the case of severe sensory disturbance, weight bearing on the affected side is often inadequate, resulting in unstable stance. In the present case, the patient received feedback in the form of a success sound when the weight exerted on the affected side exceeded the threshold, thus encouraging weight bearing on the hemiplegic side, which probably contributed to early stabilization of stance. Real-time evaluation of weight bearing on the hemiplegic side and feedback cannot be achieved in conventional exercise using orthosis. This exercise strategy is realized in GEAR exercise through making full use of the characteristics of robot.

In the present case, we were able to provide gait exercise at supervision level after 1 week of GEAR exercise, at a time when the patient still required maximum assistance while using conventional orthosis (FIM score 2). In gait exercise under maximum assistance, the therapist shifts the center of gravity for the patient. Therefore, even if the patient him/herself moves the trunk, in the absence of feedback on whether the trunk movement contributes to shifting the center of gravity, motor learning cannot be achieved efficiently. On the other hand, in gait exercise at supervision level, the patient obtains feedback at each step on whether shift of center of gravity is effectively achieved by controlling the trunk. In the present case, through providing appropriate levels of assist for swing and stance using the GEAR, gait exercise under supervision was accomplished at an early stage, providing the patient with many opportunities to control the trunk by himself. This feature probably contributed to early recovery of gait capability. During 2–4 weeks of GEAR exercise, while assistance provided by the robot was reduced gradually, gait exercise at a speed faster than walking using orthosis was possible. Exercise at faster gait speed implies an increase in exercise amount per unit time, which is considered to be advantageous for motor learning. Throughout the GEAR exercise period, gait exercise was conducted with the knee extension assist and swing assist always set at the lowest level possible to an extent that gait was not excessively worsened. An animal study suggested that the functional improvement effect of gait exercise for motor paralysis is higher when using an assist-as-needed strategy than using a fixed assist strategy (Cai et al., 2006). In addition, Krishnan et al. (2012) reported a case of robot exercise in a post-stroke hemiplegic patient, in which the robot provided minimum assist according to the patient’s capability, creating opportunities for the patient to actively control the four limbs. They suggested that this exercise strategy contributed to improving functional improvement compared to passive repetitive exercise of gait pattern. In exercise using GEAR, minimizing the assist level also may have increased the opportunities for active motions by the patient, which may have led to improvement in gait capability.

The present case illustrated that by conducting assist according to the ability of the patient, increasing amount of exercise maximally matching the patient’s active participation, and giving feedback based on objective information, we provided a gait exercise environment conducive to motor learning which contributed to enhance efficient motor learning.

Despite being a single case study, the fact that the patient who underwent gait exercise using GEAR showed markedly higher Improvement efficiency of FIM-walk than that of the control group suggests that gait exercise using GEAR may be effective for improving walking ability. Further larger scale study is required to validate the usefulness of this gait exercise modality. In Japan, clinical use of GEAR has started in multiple facilities, and future reports of clinical outcome are anticipated.

6 Conclusion

GEAR was developed as a robot for assisting gait exercise in patients with hemiplegia following stroke. The unique features of GEAR; enhanced feedback and precise assist control, allow flexible control of the variables of motor learning, with the goal to enhance highly efficient motor learning. In a patient with severe hemiplegia and sensory deficit in the subacute stage after stroke, conducting gait exercise using GEAR in additional to conventional physiotherapy achieved markedly better improvement of gait ability compared to control patients who underwent physiotherapy alone. This case suggests that GEAR is potentially useful for gait exercise in hemiplegic patients.

Conflict of interest

The authors declared a potential conflict of interest as follows: Dr. Saitoh got collaborative research fund from Toyota Motor Corporation which developed the robot discussed in this article. Mr. Konosu is a researcher in Toyota Motor Corporation.