Eight subjects with stroke and 8 healthy controls participated in this study. Healthy subjects underwent only BMCA protocol. Post-stroke patients admitted to the Department of Rehabilitation Medicine of Konyang University Hospital from May 2016 to March 2017 were enrolled according to the following inclusion criteria: (1) diagnosis of stroke confirmed by brain computed tomography or magnetic resonance imaging; (2) patients older than 18 who experienced their first stroke; and (3) patients with ankle plantarflexor spasticity of ≥1 point on the MAS. Exclusion criteria were (1) patients who could not perform motor tasks because of severe cognitive impairment; (2) recurrent stroke before assessment; (3) other neurological or muscular disorders; (4) previous botulinum toxin or neurolytic agent injection into the medial gastrocnemius muscle (GCM) in the last 6 months prior to enrollment; and (5) fixed ankle contracture.

This study was approved by the Institutional Review Board of Konyang University Hospital (No. 2016-05-015), and informed consent was obtained from all subjects.

Ankle plantarflexor spasticity was evaluated using the MAS and MTS. For convenience, the MAS scores were substituted by setting different grades: Grade 1 to 1, Grade 1+ to 2, Grade 2 to 3, Grade 3 to 4, and Grade 4 to 5. The Penn Spasm Frequency Scale (PSFS) and the Spasm Frequency Score (SFS) were also used to examine the validity of a clinical measure of spasms and clonus. The motor assessment scales used were the Medical Research Council grading system (MRC, 0–5) and Fugl–Meyer Motor Assessment Scale (FMA). The motor subscore of the FMA is a validated measure with a score ranging from 0 to 34 points for the lower extremity (FMA-LE) in which a higher score indicates better lower limb function. Before conducting the BMCA protocol, all patients underwent clinical evaluation by the same clinician.

In this study, the BMCA protocol was modified based on adapted rules from the protocol originally reported by Sherwood et al. [7]. The test was conducted with participants placed in the supine position in a quiet and warm room with minimal distractions. The sEMG of 10 muscles (five muscles from each side) were recorded with pairs of surface electrodes. These electrodes were placed in the midline of the muscle belly and oriented on a line parallel to the muscle fibers for the right and left quadriceps (Q), hip adductor (A), hamstring (H), tibialis anterior (TA), and triceps surae (TS) muscles. The skin under the electrodes was prepared to reduce electrode impedance below 5 kΩ. In this paper, we focused on the voluntary movement segment of the BMCA protocol. The maneuver included a bilateral task with two phases (ankle dorsiflexion/plantarflexion) and unilateral task with two phases that were conducted bilaterally. All voluntary tasks were repeated three times.

The originally recorded, full bandwidth sEMG data was reduced using a root mean square (RMS) algorithm over the 5-second interval to produce sEMG envelopes with an effective sampling rate [11]. We averaged three repeated activities for each muscle, corrected for the baseline by subtraction of the average activity computed over the 1 second immediately preceding the maneuver [12]. This set of values, one for each muscle, comprised the response vector (RV) for each phase of a task [11]. They were presented in units of μV_RMS for each of the scalar elements. This RV was also reduced to a single variable by averaging five muscle activities to create the average limb response (ALR), again with units of μV_RMS [13]. The RV for each task from eight neurologically intact participants was used as a normal pattern to evaluate voluntary movements in stroke patients. The average of RVs across the eight control subjects was used to generate a prototype response vector (PRV) [11]. These values were used to calculate the similarity index (SI) that compares relative distribution across the set of muscles selected for the task [11]. The SI value, a numerical expression of the relationship of the normalized RV to the PRV, was expressed as a radian between the RV and PRV vector computed as the inner product [11]. Therefore, SI ranges from -1 to +1. A value of 1.0 for the SI suggests that the test participant’s RV showed an identical pattern of sEMG activity across muscles to the neurologically intact group PRV for that task regardless of the EMG magnitude for those muscles [14]. The voluntary response index (VRI) value was computed separately for each maneuver. The VRI contains an element relatively independent of absolute amplitude that is SI. The second element is based on the magnitude of the RV that is an absolute measure of sEMG. These two elements are included in the VRI based on the significance of relative distribution and absolute level of activity [11].

The sEMG signals were digitized at a rate of samples per second (s/s) per channel with 12-bit accuracy using the WINDAQ data acquisition system version 2.91 (DATAQ Instruments Inc., Akron, OH, USA).

The ultrasound system with a 9–15 MHz linear transducer (GE LOGIQ E9; General Electrical Healthcare, China) was used for ultrasound measurements. The subjects were asked to lie prone with their feet hanging from the edge of the examination table and maintain the ankle joint in an anatomically neutral position (0°=perpendicular angle between the shank and the foot). Ultrasonography was conducted at the fixed point of the medial GCM, in the middle of two reference points. One of the reference points was at the proximal one-third of a longitudinal line from midway between the medial and lateral malleoli to midway between the medial and lateral epicondyles. The other point was at the medial end on a transverse line perpendicular to the point on the longitudinal line [15]. The muscle architecture was measured using electronic calipers in real-time B-mode ultrasonography. Muscle thickness was defined as the shortest distance between the superficial and deeper border of the muscle in the transverse view. The muscle fascicle length was defined as the length of hyperechoic inclined line between the superficial and deeper aponeurosis in the longitudinal view [16]. The fascicle pennation angle was defined as the angle between the muscle fascicle and the deeper aponeurosis in the longitudinal view [16] (Fig. 1).

The sonoelastography technique, described by Kesikburun et al. [17], was used. In the transverse plane, at least three compression-relaxation cycles were used to validate reproducible findings. Light repetitive compression was applied manually with the hand-held transducer by providing visual feedback from the quality bar indicator. The images were stored as cine loops in the memory of the ultrasound system. The recorded motion images were replayed to select the best representative image. The elastogram of the region of interest (ROI) was obtained according to the color-coded image of the ROI. The ROI was set to cover the entire medial GCM muscle, excluding its hyperechoic epimysium. The elasticity index is defined as a value ranging from 0.0 to 6.0, created by the reflected color distribution within the ROI. A higher value indicates ‘higher stiffness’ and ‘blue color’ is dominant (Fig. 2). If muscle movement or contraction was detected during measurement, scanning was discontinued. The ultrasonographic evaluation was conducted by a radiologist without knowledge of the results of clinical spasticity scores.

We used Wilcoxon signed-rank test to compare the BMCA protocols between the post-stroke spastic group and the normal control group; the ultrasonography parameters between the hemiplegic leg and the unaffected leg. The correlation of the MAS and MTS category with ultrasonography parameters and sEMG activities was assessed using the Spearman correlation coefficient (r). We used multiple regression analysis to validate the accuracy of BMCA and sonoelastography in evaluating spasticity. The significance level was set at p<0.05. All statistical tests were conducted using SPSS version 18.0 (SPSS Inc., Chicago, IL, USA).

The sEMG activity for voluntary ankle dorsiflexion of an stroke subjects was compared with a healthy subject (Fig. 3). Voluntary ankle movements in healthy subjects result in efficient reponses characterized by well-organized, low-level activation of a minimal number of muscles required for maneuvers, involving only movement against gravity (Fig. 3A). By contrast, the hemiplegic side of stroke subjects revealed generally increased muscle activity, that spread into many more channels, and with less movemet (Fig. 3B, 3C). The degree of coactivation greatly affected the movement observed during clinical examination.

The VRIs for ankle dorsiflexion/plantarflexion were obtained from the control and stroke groups (Fig. 4). All SI values for these tasks were approximately 1.0 in the control group (SI=0.99 for ankle dorsiflexion; SI=0.94 for ankle plantarflexion), which suggests that all participants revealed a similar pattern of muscle activation during these tasks. In the stroke patient group, the mean value of SI was 0.14 for ankle dorsiflexion and the mean value of SI was 0.43 for ankle plantarflexion. The SI values were significantly lower in the stroke group compared with the control group for hemiplegic ankle dorsiflexion and plantarflexion (p=0.012 for ankle dorsiflexion; p=0.012 for ankle plantarflexion). The magnitudes of RV in sEMG were also lower in the stroke group when compared with the control group (63.07 μV vs. 105.43 μV for dorsiflexion; 48.24 μV vs. 58.21 μV for plantarflexion), without any significant difference (p=0.263 for ankle dorsiflexion; p=0.674 for ankle plantarflexion).

The averaged magnitudes of RV in sEMG revealed significant increases in the stroke group compared with the control group, with mean ALR values of 4.58 μV and 3.44 μV, respectively (p=0.036). Fig. 5 shows the relationship between RV elements averaged across the voluntary ankle movements and clinical spasticity measurement tools. These average sEMG activities of RV significantly increased corresponding to the MAS and MTS spasticity grade (X), and the difference between R1 and R², defined as the dynamic component (R²–R1) (r=0.913 and p=0.001 for MAS; r=0.869 and p=0.005 for X-value; r=0.926 and p=0.002 for R²–R1 value) [18] (Fig. 5). PSFS and SFS showed no significant relationship with the sEMG activity.

Table 2 shows the muscle architecture parameters measured using conventional B-mode ultrasonography and sonoelastographic data. The fascicle length and pennation angle were significantly decreased in the medial GCM on the hemiplegic side compared with the unaffected side (p=0.012 for fascicle length; p=0.012 for fascicle pennation angle). Muscle thickness on the hemiplegic side was lesser than on the unaffected side, without significant difference (p=0.173). The MAS and MTS showed no significant relationship with the difference in muscle architecture values between hemiplegic and unaffected sides.

The elasticity index of the spastic medial GCM was significantly increased compared with that of the muscle on the unaffected side (p=0.01). The MAS score was positively correlated with the elasticity index of the spastic GCM (r=0.371), without significant correlation (p= 0.228). The X and R²–R1 values were significantly correlated with the elasticity index in the hemiplegic GCM (r=0.913 and p=0.002 for X value; r=0.737 and p=0.037 for R²–R1 value) (Fig. 6). PSFS and SFS had no significant relationship with the elasticity index.

In the multiple regression analysis, the relationship between clinical evaluation tools and both BMCA and sonoelastography was linear (R²=0.905 for MAS; R²=0.847 for X value; R²=0.809 for R²–R1 value), without statistical significance.