Eighteen individuals with hemiplegia who were diagnosed with either CP or stroke were recruited in this study. The procedure was approved by the Institutional Review Board (4-2012-0483). Informed consent was obtained from all participants after the experimental procedures were sufficiently explained and before the study began. The inclusion criteria for individuals with hemiplegia were as follows: each participant 1) had no discernible hearing deficit, 2) was able to walk independently for a distance of at least 10 m without the use of a walking aid or supporter, and 3) was able to understand the command to walk following RAS.22

Demographical and clinical characteristics including age, gender, height, weight, body mass index, diagnosis, and ambulatory status are shown in Table 1. Participants were classified as either community ambulators or household ambulators according to their ambulatory status. Community ambulators were able to independently walk around on level ground, curbs, and uneven terrain outdoors as well as indoors for a minimum of 150 feet; they were also able to manage stairs and challenging community activities, such as making visits in their neighborhood. Conversely, household ambulators were only able to walk indoors for short distances (a maximum of 50 feet) either independently or with orthopedic devices; they also encountered difficulty with stairs and uneven terrain and required assistance or walking aids when leaving the house.2137

All participants underwent the 4-week gait training with RAS on the carpet of the same clinic. The RAS treatment was performed for 30 minutes per each session, three sessions per week. The intervention procedure consisted of the following established protocol:2122 1) A participant independently walked barefoot along a 10-m flat walkway three times without RAS at the individual's preferred walking speed. 2) Walking cadence (steps/min) was calculated based on the gait parameters outlined in step 1. 3) The identified initial tempo signaled by metronome beats (beats per minute) was set to the participant's cadence obtained in step 2 in each session. 4) RAS was then provided by the music therapists, who played a live rhythmic pattern using a composed four-chord progression with metronome beats on a keyboard (PSR-E213, Yamaha Electronics Co., Hamamatsu, Japan). To ensure accuracy and consistency of the rhythmic stimulus, the same therapists performed the procedure for each patient. These specialized therapists provided a regular and stable rhythm. To avoid rhythmic monotonousness, several chord progressions were applied. 5) The same chord pattern was repeated to provide a continuous timing cue and a period of 2 minutes to help each participant quickly adapt to the RAS. In this step, the music therapist instructed each participant to finger-tap the rhythm for 1 minute to ensure that they could hear the sounds and also confirmed that each participant could walk comfortably by allowing them to adapt to the RAS for 1 minute. 6) Each participant then walked the length of 10 m three to six times with RAS and rested for 1-3 minutes between walks, depending on the endurance level of the patient. This step was repeated 5-8 times in a session. To monitor compliance of the participants to the RAS, three music therapists evaluated the change in walking speed.2122 7) The final 1-2 minutes was spent by fading out the rhythmic stimulation to monitor the independent carryover effect, which was qu-antified by calculating cadence. These seven steps were applied in each training session during a period of 4 weeks.

All participants were pre- and post-tested 2 days before starting and 2 days after conducting the 4-week gait training with RAS in the gait analysis laboratory. Kinematic and temporospatial data from the pelvis, hip, knee, ankle, and foot were collected and analyzed for the gait trials without RAS. To ensure that reflective foot markers could be recognized from the infrared signal of the motion analysis system, participants walked barefoot along a 10-m flat walkway three times at the individual's preferred walking speed without RAS. For kinematic analysis, 15 passively reflective markers were adhered with specialized tape to the sacrum, both sides of the anterior superior iliac spine, middle thigh, lateral knee, middle shank of the tibia, lateral malleolus, heel, and forefoot.22 A three-dimensional motion analysis system with six cameras (Vicon Nexus ver. 1.8.5, Vicon Motion System Ltd., Oxford, UK) was used to record kinematics by measuring the degree of joint motion during gait performance. This system comprised six infrared-sensitive cam-eras for locating and tracking the fixed retro-reflective markers through space.

The motion analysis system was calibrated before each gait analysis. Participants were simultaneously videotaped from the front and side, and measurements were recorded in the sa-gittal, coronal, and transverse planes. Kinematic data included the angle of pelvic tilt, pelvic obliquity, pelvic rotation, hip flexion and extension, hip adduction and abduction, hip internal and external rotation, knee flexion and extension, ankle plantar and dorsiflexion, and foot internal and external rotation. All kinematic and temporospatial data were processed and plotted, and the graphs were visualized using Polygon software ver. 3.5.1 (Oxford Metrics Inc., Oxford, UK), which was interworked with a three-dimensional motion analysis system. To perform the statistical analysis, three points from each joint range of motion from continuous raw data in the lower extremity were used: initial contact (the moment when the heel struck the floor), the minimal joint angle, and the maximal joint angle during the whole gait cycle.22

In order to evaluate the overall gait pathology, a global index of three-dimensional kinematic changes, the gait deviation index (GDI),38 was utilized in the pre- and post-RAS treatment. The GDI is a scaled distance between nine individual kinematic variables of pathological gait (pelvic tilt, pelvic obliquity, pelvic rotation, hip flexion, hip adduction and abduction, hip internal and external rotation, sagittal angles of the knee and ankle, and foot progression in the transverse plane) and the average of kinematic variables of the reference normal gait. One stride obtained from each pre- and post-treatment gait analysis was selected from the hemiplegic side of every participant. All kine-matic variables were extracted and sampled at every 2% inter-val in the whole gait cycle. Each set of 459 data points (9 joint angles×51 points) was computed as gait vectors using the gait feature of a normal gait referred to as orthonormal f-basis. The computing method is explained in detail in another referenced paper.38 The GDI score can be interpreted as follows: a score of 100 indicates a GDI equal to the average of the normal control. In our study, the GDI was utilized to ensure that individual pathological gait improved after gait training with RAS compared to the normal reference. The GDI score was calculated to take into account the overall effect of kinematic changes of the pelvis, hip, knee, ankle, and foot throughout the gait cycle. In other words, the GDI incorporated the pelvic tilt, pelvic obliquity, pelvic rotation, hip flexion, hip adduction and abduction, hip internal and external rotation, the sagittal angles of the knee and ankle, and the foot progression in the transverse plane.

Temporospatial parameters such as cadence (steps/min), walking velocity (m/sec), stride length (m), step length (m), stride time (sec), step time (sec), single limb support (%), double limb support (%), stance phase (%), and swing phase (%) were calculated using Polygon software ver. 3.5.1 (Oxford Metrics Inc., Oxford, UK). To evaluate the side-to-side asymmetry between the lower limbs, the absolute difference between the unaffected and affected sides was analyzed using temporospatial data.22

All statistical analyses were performed with SPSS Statistics 20 (IBM Corp., Armonk, NY, USA). Wilcoxon signed-rank tests were used to evaluate intra-subject pairwise comparisons between gait parameters: kinematics and temporospatial data and the GDI score in the pre- and post-RAS treatment. Mann-Whitney U tests were also used to compare the inter-group designs of general characteristics and gait parameters. All statistical significance levels were set at p<0.05.

Among total 18 hemiplegic patients, seven subjects diagnosed with CP and eleven subjects diagnosed with stroke participated in this study. Nine subjects with community ambulatory function and nine subjects with household ambulatory function participated. All participants showed mild spasticity of the lower extremities with a grade of 1 to 1⁺ on the Modified Ashworth Scale. The onset duration of stroke was 3.58±2.22 years. Detailed information on gender, age, height, weight, and body mass index is shown in Table 1. Considering that CP is a developmental disorder that affects the initial acquisition of gait while stroke patients have previous walking experience, these two neurologic disorders represented in the study population were separately analyzed. There were no significant differences in the parameters of general characteristics between CP and stroke (Table 1). In addition, there were no significant differences in the general parameters between community ambulators and household ambulators (Table 1).

When kinematic characteristics of the hemiplegic side were evaluated before and after RAS treatment, pelvic anterior tilt in the sagittal plane largely tended to decrease after gait training with RAS. Hip adduction significantly increased in the mid-stance phase after the RAS treatment (p=0.039 by Wilcoxon signed-rank test) (Table 2). Maximal knee flexion increased in the mid-swing phase (p=0.022), and ankle plantar flexion decreased at initial contact (p=0.025) and push off (p=0.006), while ankle dorsiflexion increased in the terminal stance (p=0.031). When a comprehensive measure of overall gait pathology, GDI, was calculated, the GDI score was found to significantly improve in patients with hemiplegia after RAS treatment (83.89±2.72 to 87.29±2.65, p=0.043) (Table 2).

When temporospatial parameters on the hemiplegic side were evaluated both before and after RAS treatment, there were no significant changes in the statistical analysis after gait training with RAS (Table 3). However, when the side-to-side differences between the unaffected and the hemiplegic side of the temporospatial parameters were evaluated, the side-to-side difference of the stance phase (6.95±0.99% to 4.62±1.15%; p=0.006 by Wilcoxon signed-rank test) and the swing phase (12.12±5.06% to 4.62±1.15%; p=0.006) significantly decreased after the RAS treatment (Table 3), suggesting that gait training with RAS could improve the side-to-side symmetry in patients with hemiplegia.

We additionally analyzed the kinematic data of CP and stroke separately. Stroke patients showed significant kinematic improvement in maximal knee flexion in mid-swing phase (48.88±4.31 to 55.31±3.90; p=0.021 by Wilcoxon signed-rank test) and in maximal ankle dorsiflexion in terminal stance (13.79±1.27 to 16.15±1.42; p=0.026) (Table 4). Patients with CP, on the other hand, did not show kinematic improvement after gait training with RAS (Table 4). This result suggests that the previous walking experience in stroke patients who have a history of typical gait may have an effect on their response to RAS.

Based on longitudinal studies suggesting that motor recovery reaches a plateau until the first 6 months post-stroke,39 we additionally analyzed the effects of RAS on kinematic patterns in both subacute and chronic stroke patients separately. Although chronic stroke patients did not show an increase in GDI score, these patients did show improvement in hip external rotation (0.26±3.57 to -3.98±3.13; p=0.028), maximal ankle dorsiflexion at terminal stance (14.08±1.98 to 16.85±1.72; p=0.028), and foot external rotation (-6.23±2.47 to -9.69±1.44; p=0.028) at initial contact after gait training with RAS (Table 5). On the other hand, subacute stroke patients showed a significant increase in GDI score (80.88±3.82 to 88.99±5.00; p=0.043 by Wilcoxon signed-rank test) with improvement in maximal knee flexion in mid-swing phase (45.15±3.59 to 56.42±4.74; p=0.043) (Table 5). This result suggests that subacute stroke patients are more likely to respond to RAS than chronic patients.

In the analysis of kinematic characteristics, household ambulators showed that pelvic anterior tilt at initial contact (10.74±2.28 to 7.93±1.97; p=0.038 by Wilcoxon signed-rank test) and the minimal angle of anterior tilt (10.19±2.40 to 7.43±1.99; p=0.038) were ameliorated after gait training with RAS. On the other hand, community ambulators were found to have significant increases in maximal knee flexion in the mid-swing phase (60.79±4.67 to 63.82±4.40; p=0.021) and ankle dorsiflexion in the terminal stance (13.86±2.42 to 16.97±2.12; p=0.008), whereas ankle plantar flexion at push-off decreased (-9.39±2.20 to -6.92±2.06; p=0.021) (Table 6). In the analysis of GDI score, household ambulators, and not community ambulators, show-ed significant gait improvement (80.07±3.36 to 86.32±3.82; p=0.008 by Wilcoxon singed-rank test) (Table 6). The kinematic changes of the proximal pelvic joint and distal ankle joint at pre- and post-RAS treatment showed differential patterns according to the ambulatory status (Fig. 1).