This prospective cohort study was conducted according to the Declaration of Helsinki and was approved by the Ethics Committee of Juntendo University Urayasu Hospital (approval number: 29-114). The purpose and procedures of this study were explained in detail, and written, informed consent was obtained from all participants. The study outline is shown in Figure 1.

A prospective cohort study was conducted. Participants were recruited from stroke patients within 3 days of stroke onset who were admitted to Juntendo University Urayasu Hospital in Japan from March 2018 to March 2023. The definition of acute rehabilitation initiation is often adopted within 3 days after stroke onset (21). Therefore, the participants in this study were patients who received acute rehabilitation. The diagnosis of stroke was based on each patient’s clinical history, neurological examination, and head computed tomography or magnetic resonance imaging. The inclusion criteria were as follows: (1) first unilateral hemispheric stroke; and (2) severe gait disturbance (AIM score 1 or 2) (11).

A patient with an AIM score of 1 is not able to walk with the physical assistance of one therapist using an AFO, cane, or crutch (11). A patient with an AIM score of 2 requires physical assistance to support body weight or maintain balance but shows severe knee buckling (knee flexion angle in the paretic stance phase ≥30°) during walking with an AFO, cane, or crutch (11).

The exclusion criteria were as follows: (1) inability to walk independently before onset; (2) inability to follow instructions due to various symptoms such as severe aphasia or loss of consciousness; (3) other complications or comorbidities that would alter the results of physical assessment (osteoarticular disease, recurrent stroke during hospitalization, etc.); and (4) refusal to participate in the study. A total of 50 eligible patients consented to participate in this study. The present study was conducted according to the Declaration of Helsinki and was approved by the Ethics Committee of Juntendo University Urayasu Hospital (approval number: 29-114). The purpose and procedures of this study were explained in detail, and written, informed consent was obtained from all participants.

All patients received conventional individualized inpatient rehabilitation based on the stroke rehabilitation guidelines. The physical therapy program consisted primarily of gait training with a KAFO and included range of motion training, muscle strengthening training, sitting balance training, and standing balance training as needed. The duration and frequency of physical therapy treatments ranged from 40 to 60 min per session, 5–6 times per week, depending on the patient’s physical condition.

Gait function was assessed daily using the AIM (score range: 1–7, best score: 7) (11). At the beginning of the gait training (baseline), paretic lower extremity motor function and trunk function were assessed. Paretic lower extremity function was assessed with the lower extremity motor scores of the Stroke Impairment Assessment Set (each score range: 0–5, best score: 5) (22). Trunk function was assessed using the Trunk Impairment Scale (score range: 0–21, best score: 21) (20).

Surface electromyography (EMG) was used to evaluate paretic lower extremity muscle activity during the walking task. EMG was measured using wireless surface electromyography (Pacific Supply, Osaka, Japan). The sampling rate was 1,000 Hz. The minimum sampling rate must be at least 800 Hz (400 × 2), as specified by Nyquist’s theorem (23). Therefore, the sampling rate in this study satisfied this criterion. The target muscles were the rectus femoris, biceps femoris, soleus, and tibialis anterior on the paretic side. The target muscles were selected from the major muscles involved in the knee and ankle joints that were fixed by a KAFO. Electrode placement followed the Surface EMG for Non-Invasive Assessment of Muscles guidelines (24). Before electrode placement, the skin was rubbed with alcohol to reduce impedance.

Electromyography analysis was initially band-pass filtered at 20–250 Hz. Next, full-wave rectification was performed. The analysis section was identified based on the video synchronized with the EMG examination. The video was taken from frontal and sagittal planes. The times of initial contact and toe-off were identified from the videos using a motion analysis application (Dartfish software). If there was no clear toe-off, toe-off was defined as the time when the foot moved forward even slightly. Based on the times of initial contact and toe-off, the first double-limb support phase (1st DS), single-limb support phase (SS), second double-limb support phase (2nd DS), and swing phase (Sw) were identified (25). EMG signals were averaged for each gait phase (26). Furthermore, the same analysis was performed for the three gait cycles, and the average values for the three gait cycles were calculated. The EMG was normalized using the average value of the entire gait cycle (27).

Electromyography recording during the walking task was performed twice during hospitalization. The baseline was the beginning of gait training. The second assessment was performed when the AIM score improved to 3 or more during hospitalization. An AIM score of 3 or more indicates that a KAFO is clinically unnecessary for the patient, and the patient is clinically able to convert from a KAFO to an AFO (11). For those who did not improve to an AIM score of 3 or more during hospitalization, the second EMG measurement was performed at discharge.

The walking task used for EMG recording was as follows. The walking condition was a comfortable speed. All patients used no canes or crutches. All patients used a KAFO. The knee joint control of the KAFO was fixed at 0° of knee extension using a ring-lock knee joint. The ankle joint control of the KAFO was fixed at the neutral position of dorsiflexion and plantarflexion. Assistance during walking tasks was provided by the same physical therapist familiar with gait training with a KAFO for stroke patients. The therapist assisted the patient by standing behind the patient. One of the therapist’s hands supported the patient’s anterior chest. The therapist’s other hand held the thigh cuff belt of the KAFO. Throughout the gait cycle, the therapist provided continuous balance assistance to keep the patient’s trunk in a vertical position using the hand supporting the patient’s anterior chest. During the swing phase, the physical therapist pulled up the thigh cuff belt with minimal force to assist with swinging. During the stance phase, no special assistance was provided except for continuous balance assistance throughout the entire gait cycle. For both the baseline and second assessments, the conditions during walking were the same, and the KAFO was used.

The relationship between the EMG during walking at baseline and improvement of gait ability was examined. An AIM score of 3 or more indicates that a KAFO is clinically unnecessary for the patient, and the patient is clinically able to convert from a KAFO to an AFO (11). Those who had an AIM score of 3 or higher in gait ability at discharge were defined as the good recovery group, whereas those who remained at an AIM score of 2 or lower in gait ability at discharge were defined as the poor recovery group. Repeated-measures two-way ANOVA (28) was performed for differences between the group (good recovery group and poor recovery group) and gait phase (1stDS, SS, 2ndDS, and Sw) in muscle activity (rectus femoris, biceps femoris, soleus, and tibialis anterior on the paretic side) during walking. If the interaction between the two factors was significant, the Bonferroni method was used to analyze significant differences between the groups in each gait phase.

The relationship between temporal EMG changes during walking and improvement of gait ability was examined. Repeated-measures two-way ANOVA (28) was performed for differences between the group (good recovery group and poor recovery group) and time (baseline and second assessments) in muscle activity during walking. The statistical analysis was performed with SPSS version 24. A p value <0.05 was considered significant.

Twenty-three patients (46%) improved to an AIM score of 3 or higher and were able to walk with an AFO (good recovery group). Twenty-seven patients (54%) remained with an AIM score of 1 or 2 and were unable to walk with an AFO (poor recovery group).

Figure 2 shows typical muscle activity in a patient with complete hemiparesis (All lower extremity motor scores of the Stroke Impairment Assessment Set are 0) who was unable to move his paretic lower extremity even slightly, showing three gait cycles. Paretic rectus femoris, biceps femoris, and soleus showed periodic muscle activity. In particular, the patient’s knee joint was fixed in 0° extension by a KAFO, but paretic rectus femoris activity was observed in the early stance phase to resist knee buckling. This patient improved to an AIM score of 3 or higher at discharge (good recovery group) and was able to walk with an AFO (good recovery group).

Two-way ANOVA showed that rectus femoris muscle activity had a significant interaction between the group (good recovery group and poor recovery group) and gait phase (1stDS, SS, 2ndDS, and Sw) (F = 10.598, p < 0.001) (Figure 3). The results of the post hoc test using the Bonferroni method showed that the good recovery group had significantly higher paretic rectus femoris muscle activity in the 1stDS (adjusted p = 0.003) and significantly lower activity in the 2ndDS and Sw (both adjusted p < 0.001) than the poor recovery group.

Biceps femoris and soleus muscle activities had a significant main effect of the gait phase (F = 21.522 and 24.593, respectively, all p < 0.001), but no significant main effect of the group (F = 0.939 and 0.958, p = 0.338 and 0.387, respectively) and no significant interaction of the two factors (F = 2.262 and 0.958, p = 0.104 and 0.387, respectively) (Figure 3).

Tibialis anterior muscle activity showed no significant main effect of gait phase (F = 1.860, p = 0.155), no significant main effect of group (F = 0.320, p = 0.575), and no significant interaction of the two factors (F = 0.859, p = 0.464) (Figure 3).

For rectus femoris muscle activity in the 1stDS, which showed significant differences between groups, temporal EMG changes were compared between groups. Two-way ANOVA for paretic rectus femoris muscle activity in 1stDS showed significant main effects of time (F = 5.629, p = 0.022) and group (F = 802.219, p < 0.001), but no significant interaction of the two factors (F = 0.947, p = 0.335) (Figure 4).

A limitation of this study is that no control group was used and the direct effect of gait training using a KAFO on the improvement of muscle activity and gait function could not be examined. Another limitation is that muscle activity was assessed during assisted walking, and muscle activity was influenced by the method of assistance and gait speed, the training time and frequency were not identical, the amount of gait training could not be quantitatively assessed, the observation period was short, and the final gait ability could not be confirmed. In this study, during the stance phase, no special assistance was provided, except for continuous balance assistance throughout the entire gait cycle, but in body weight support treadmill gait training using a robotic device, quantitative evaluation was possible by adjusting levels of bodyweight support and guidance force. However, the number of facilities that can provide gait training using special equipment is limited. Therefore, we believe that this study is highly valuable because it demonstrates the effect of gait training with a KAFO, which can be easily performed in clinical settings. In the future, a randomized controlled trial should be conducted to clarify whether gait training with a KAFO contributes to the enhancement of quadriceps muscle activity during the early stance phase and to the improvement of gait ability.

Among hemiplegic stroke patients with severe gait disturbances who had difficulty walking with an AFO, 46% were able to acquire walking with an AFO and 54% were unable to acquire walking with an AFO throughout the acute rehabilitation phase. One of the differences between the two groups was the paretic quadriceps muscle activity during walking with a KAFO. The patients who were able to acquire walking with an AFO had significantly higher quadriceps muscle activity during the early stance phase than those who did not. In addition, both groups showed a significant increase in this quadriceps muscle activity throughout the acute rehabilitation phase. Thus, the induction and enhancement of paretic quadriceps muscle activity during walking with a KAFO may play an important role in the acquisition of walking with an AFO.