Broader Estimates of Gastrocnemius Activity Generated a More Representative Cocontraction Index: A Study in Pediatric Population

The electromyography (EMG) cocontraction index (CCI) given by the antagonistic/agonistic Root Mean Square (RMS) amplitude ratio of the same muscle is a qualified biomarker used for spastic cocontraction quantification and management in cerebral palsy children. However, this normative EMG ratio is likely subject to a potential source of errors with biased estimates when measuring the gastrocnemius plantar flexors activity. Due to the uneven distribution of electrical activity within the muscle volume, cocontraction levels can be misestimated, if EMGs are obtained from the sole traditional bipolar sensor location recommended by SENIAM. This preliminary study, on 10 healthy children (mean age 10 yr), investigated whether surface EMG detected proximally and distally via two pairs of bipolar electrodes, within the medial gastrocnemius (MG), provides a significant difference in CCI estimates during non-dynamic (isometric dorsiflexion) and dynamic (swing phases of gait) conditions. Gait cycles were extracted from Inertial Measurement Unit sensors. Medial gastrocnemius activity was greater distally than proximally during plantar flexion when it acts as an agonist (~24±18%) and it was greater proximally during dorsiflexion (~23±9%) when it is acting as an antagonist. As a direct consequence, CCI estimates from the conventional sensor location were significantly different (~36%) from the CCIs computed by considering broader MG regions. This difference arose in all subjects during isometric efforts and in two of 10 healthy children during the swing phase of gait who presented cocontraction patterns ( $\text{p} < 0.05$ ). EMG bipolar sampling encompassing proximal and distal gastrocnemius muscle regions may reduce bias in CCI computation and provide a more representative and accurate cocontraction index that is especially important for comparisons to the diseased state.


Broader Estimates of Gastrocnemius Activity Generated a More Representative Cocontraction Index: A Study in Pediatric Population
Maria Vinti , Manob Jyoti Saikia , Member, IEEE, John Donoghue , Member, IEEE, Stephane Mandigout , Maxence Compagnat , and Karen L. Kerman Abstract-The electromyography (EMG) cocontraction index (CCI) given by the antagonistic/agonistic Root Mean Square (RMS) amplitude ratio of the same muscle is a qualified biomarker used for spastic cocontraction quantification and management in cerebral palsy children.However, this normative EMG ratio is likely subject to a potential source of errors with biased estimates when measuring the gastrocnemius plantar flexors activity.Due to the uneven distribution of electrical activity within the muscle volume, cocontraction levels can be misestimated, if EMGs are obtained from the sole traditional bipolar sensor location recommended by SENIAM.This preliminary study, on 10 healthy children (mean age 10 yr), investigated whether surface EMG detected proximally and distally via two pairs of bipolar electrodes, within the medial gastrocnemius (MG), provides a significant difference in CCI estimates during non-dynamic (isometric dorsiflexion) and dynamic (swing phases of gait) conditions.Gait cycles were extracted from Inertial Measurement Unit sensors.Medial gastrocnemius activity was greater distally than proximally during plantar flexion when it acts as an agonist (∼24±18%) and it was greater proximally during dorsiflexion (∼23±9%) when it is acting as an antagonist.As a direct consequence, CCI estimates from the conventional sensor location were significantly different (∼36%) from the CCIs computed by

I. INTRODUCTION
E LECTROMYOGRAPHY (EMG) and related techniques are highly useful tools for muscle tone assessment and therapeutic purposes in gait disorders [1].Spastic cocontraction is a form of muscle overactivity that alters gait in children with cerebral palsy from a very young age, producing the so called "spastic paretic gait" [2], [3], [4].In this type of pathologic gait where muscle overactivity, contracture, stiffness, and paresis co-occur [5], the information provided by EMG is recognized to be a highly informative and objective measure to overcome the subjective nature of clinical scales used at rest.However, there is a need for qualified EMG biomarkers to provide precision, objective functional measurements and to guide adaptive therapies.The Cocontraction Index, CCI, which is the antagonistic/agonistic activity ratio of the studied muscle is currently used as a normative EMG method to quantify muscle overactivity.CCI is typically scaled to activation levels recorded during isometric maximal voluntary contraction, but not during movement (dynamics) [6].In the spastic pediatric population, multiple factors make the measurement of normative EMG from maximal efforts very difficult [7], though as alternative, a more ethological approach of bipedal heel-rise has been proposed to derive the CCI for plantar flexors to assess spastic gait [4].This index is calculated using the ratio of the Root Mean Square EMG (RMS EMG) from ankle plantar flexors (PF) during the swing phase of gait, when PF serves as an antagonist, to the peak RMS EMG obtained from voluntary plantar flexion during 3 seconds of bipedal heel-rise (BHR) standing, the period when the PF is used as an agonist, CCI = [PflexSwingGait EMG /PflexBHR EMG].
Envelope-based indices may suffer from electrode location variation and non-uniform spatial muscle activation [8], [9], [10], [11], which may result in inaccurate CCI gait measurements.In particular, we previously identified [12] a potential technical error that could lead to cocontraction misestimation when EMGs are detected by the traditional bipolar montage.That is, different sections of the medial gastrocnemius (MG) have been shown to be activated during both plantar and dorsiflexion (both used for the index) [12].More precisely, in that study which aimed to compare CCI estimates derived from bipolar and grid electrodes (High-Density EMG) during isometric efforts, we showed that MG activity was greater proximally during isometric dorsiflexion (antagonist) and distally during isometric plantarflexion (agonist).We, therefore, suggested that proximal-distal sampling may be necessary to reliably estimate MG cocontraction and to provide an accurate CCI.To provide a more complete measurement of MG activity using tools more readily available in gait labs, we proposed positioning two pairs of closely spaced electrodes on the proximal and distal halves of MG as a conservative approach, which will increase inter-electrode distance and reduce crosstalk from nearby muscles (e.g., soleus) [13].
The overall goal of the present preliminary study, conducted in healthy children, was to investigate whether using broader estimates of MG activity including proximal and distal muscle regions, would generate a significant difference in CCI estimates in healthy children.We first compared the MG EMG agonist and antagonist values obtained from a single conventional electrode pair, positioned in accordance with the SENIAM recommendations [14], using two pairs of widely spaced electrodes configurations, one proximal and one distal on the MG, to take the entire muscle into account.Agonist MG values were obtained during bipedal heel rise (CCI denominator), and MG antagonist values from the isometric dorsiflexion and swing phase of gait (CCI numerators).Secondly, we compared CCI estimates computed from MG EMG of the two-electrode configuration with the conventional single one.We hypothesized that the use of two pairs of EMG sensors, as opposed to the typical single traditional sensor and site, to record respectively proximal and distal MG regions' activity, gives more representative EMG measurements and in turn, it increases the reliability and accuracy of the CCI calculation.

A. Participants
Ten healthy children with no history of motor pathology were included in the study (seven females and three males mean±SD age 10±2 years) as demographic information presented in Table I

B. EMG Data Acquisition
For surface EMG recording, we used a wireless EMG recording system that was also integrated with Inertial Measurement Unit (IMU) sensors for biomechanical recording (Trigno™ from Delsys Inc., Boston, USA).In this sensor, if we look at the electrode face of the sensor [Fig 1 (b)], positioned upright (arrow on the backside pointing up), the electrodes on the left are the inputs, and the electrodes on the right are the references.The inputs on the left side use a differential/bipolar scheme.We set the sensor to sample at Fs= 1111.11Hz and an internal gain of 300.Each axis of the gyroscope (x, y and z) was sampled at 148.15 Hz with an internal gain of 16.4.Since the inter-electrode spacing was 10mm, bandwidth of 10-850 Hz, EMG input range of 11mVpp, and unit size of 27 × 37 × 13 mm, the sensor was suitable for our study.Also, the baseline noise was < 750 nV R M S and the Common Mode Rejection Ratio, CMRR, was >90dB that provided a superior signal quality.
The EMG data was recorded and analyzed by using the "EMGworks® 3.7 Software".The program is divided into two modules: the Real-Time Acquisition module (EMGworks acquisition) and the Data Analysis Module (EMGworks analysis).Bipolar electrodes positioned as suggested by currently available surface EMG recommendations [14] covered about 2.5 cm 2 of the centro-proximal region.In our study, the two sensor units, each with two pairs of bipolar electrodes, were respectively positioned proximally and distally on the MG surface, as shown in Fig. 1.The distance between the bar electrode and the sensor's frame is 1 cm as shown in Fig. 1 (b).The proximal IMU sensor was located ∼1 cm distal to the popliteal crease (total distance of the EMG bar electrode: 2 cm above the popliteal crease) with its most lateral part located ∼2 cm medial to the junction between the MG heads.We adopted this positioning following our previous work on the MG muscle with a matrix of 64 electrodes (13 × 5 electrodes with one missing electrode; 8 mm inter-electrode distance) positioned on a skin region covering exclusively the muscle superficial aponeurosis (Fig. 1 from [3]).The top row of electrodes was located ∼1 cm distal to the popliteal crease.The most lateral column was located 2 cm medial to the junction between gastrocnemius heads.The distal IMU sensor was aligned 1 cm apart from the proximal.
Surface EMG sensors were placed on skin sites previously abraded, cleaned, and dried above the MG muscle of the leg-dominant side of participants.Two sensor units, each with two pairs of bipolar electrodes, were respectively positioned proximally and distally on the MG surface, as shown in Fig. 1.The distal IMU sensor was aligned 1 cm apart from the proximal sensor (total distance of the EMG bar electrode: 3 cm above the bar electrode of the proximal sensor) and located in the area corresponding to the most prominent bulge of the MG muscle region in accordance with SENIAM guidelines [14].A third IMU was placed on the tibialis anterior (Fig. 1 (a): IMU-TA) location recommended by the SENIAM, thus was aligned with the anatomical axis of the child's legs, and positioned at 1/3 of the distance of the line between the tip of the fibula and the tip of the medial malleolus.The gyroscope's x-axis output signal from this IMU-TA was used for stride marker calculation to obtain the angular velocity for gait cycle analysis as introduced and validated by Hunza and collaborators in 2014 [15].As shown in Fig. 2 (a), for this analysis the start of each cycle was defined by the termination of the forward swing phase as determined by linear interpolation between the sample just before and just after the negative zero crossing of the gyroscope signal.The swing phase is the period of time between the termination and the initiation of the forward swing phase.The TA muscle activity during the swing phase of gait was also reported in Fig. 2 (b).This data is shown only for further information and not discussed because it is not directly relevant to the main topic of the paper.

C. Experimental Procedure
Muscle activity was recorded during three conditions: #1.Bipedal Heel-Rise (BHR): children in the standing position were asked to raise their heels with straight knees, using slight manual support from the examiner's hands.The examiner counted out loud for a 3-second period during which children were encouraged to maintain the heel-raised (and knee-extended) position as previously described by authors in 2021 [4].BHR is a commonly employed clinical test to evaluate calf muscle function and its neuromuscular activation [16].It has been considered to provide a better estimate of maximal gastrocnemius EMG activity than isometric plantar flexion efforts in patients with upper motor neuron lesions [17].
#2. Isometric Dorsiflexion (IDF): children were seated with the ankle positioned back at 90 • dorsiflexion and knee in extended position.Children were instructed to "lift up the foot as strongly as they could" against an isometric resistance which was an opposing force made by the examiner's hand lasting 3 seconds.#3.Gait: at the end of the BHR and IDF testing, subjects were asked to walk at a self-selected speed for a straight-line distance of 10 meters, with 1 minute rest between trials.The BHR, IDF, and walking were performed 3 times in the above indicated serial order.

D. Data Analysis
All EMG data analysis was performed using EMG-works®software from Delsys Inc. Signals were high-pass filtered to remove movement and other common EMG artifacts, using a third-order Butterworth high-pass filter cutoff frequency at 30 Hz.We obtained the Root Mean Square (RMS) values from both proximal and distal medial gastrocnemius EMG samples.Through our comparative analysis in the past (100 ms vs 50 ms epoch), we found that the 50 ms epoch amplitude around the peak was greater and based on previous research [4], the RMS of the Medial Gastrocnemius (RMSMG) was calculated over the 50 ms epoch centered around the peak EMG during the agonist plantarflexion action of BHR (#1) (R M S M G B H R ) to obtain the Cocontraction Index (CCI) Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
denominator, and during the antagonist action within IDF (#2) (R M S M G I D F ) to obtain the CCI numerator.The RMS of the antagonist MG during the swing phase of gait (#3) was calculated over the entire duration of the swing phase.For the gait analysis, eight gait cycles in total et per subject were analyzed.The first and last strides were excluded due to walking initiation and termination.We computed the Cocontraction Index (CCI) as the ratio of the RMS value for the MG muscle acting as antagonist to the effort intended (IDF and swing phase of gait) to the RMSMG value during agonist plantar flexion effort within the BHR [4].We obtained CCI for the IDF and the swing phase of gait normalized first, uniquely from EMG signals of distal sensors to obtain the CCIs conventional, and secondly from EMG signals of proximal (numerator) and distal (denominator) sensors to obtain the experimental CCIs.The CCIs from these 2 electrode configurations were compared as follows:

E. Statistics
We calculated mean and standard deviations (SD) for the resulting readings of the BHR, IDF testing sessions and swing phase of gait using the mean data for each individual child to obtain the group means and SDs.The mean agonist and antagonist gastrocnemius RMSEMG for each subject is presented in Fig. 3

A. Medial Gastrocnemius RMS Agonist Values During the BHR and Antagonist RMS Values During the IDF
The calculated mean and SD of the RMSMG values (proximal and distal) for each child from the three trials are graphically presented in Fig. 3 for the BHR and Fig. 4 for the IDF.During the BHR test (#1) when the MG was acting as agonist plantar flexor, we found greater distal rather than proximal EMG in 8 out of the 10 children, but no difference for the other 2. For the entire group, we found ∼24±18% greater agonist RMSMG values from distal rather than proximal sensors in 8 out of the 10 children.By contrast, subjects 3 and 9 showed a different result (i.e., proximal slightly greater than distal activation level).During the IDF (#2), (Fig. 4), when the MG was acting as antagonist, the opposite muscle activity distribution was observed: RMSMG antagonist values were ∼23±9% (mean±SD) greater from proximal rather than distal sensors, and this result was found in all subjects.This EMG activity distribution was observed during the gait swing phase in subjects 5 and 7 who manifested MG antagonist cocontraction patterns.These two children showed ∼26±2% (mean±SD) greater RMSMG at proximal rather than distal sensors (illustrated for child 7 in Fig. 2).The RMSMG values for the 2 sensors were not different in the other 8 children.Notably, these subjects did not show cocontraction patterns during the swing phase.

B. Cocontraction Index
During the IDF condition, the comparison between experimental and conventional methods showed that Cocontraction Indices (CCIs) were significantly different in all subjects.During the swing phase of gait, this same comparison showed a significant difference in children 5 and 7 presenting antagonist MG activity during the swing phase.Results for the different cocontraction levels are summarized in Table II.In sum, the cocontraction levels were greater when derived from Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.experimental than conventional methods, Fig. 5 (a).No difference was observed between cocontraction estimates from the two methods for the swing phase of gait when we pulled all children data, Fig. 5 (a), however, a significant difference was observed in CCIs for the two children (subjects 5 and 7) who had MG antagonist cocontraction patterns, Fig. 5 (b).The CCIs experimental and conventional were 0.07± 0.006 vs 0.005±0.01,p=0.009 for the subject 5 and 0.08±0.03vs 0.05±0.01,p=0.004 for the subject 7.

TABLE II COCONTRACTION INDICES (CCIS). EXPERIMENTAL AND CONVENTIONAL CCIS DURING (A) ISOMETRIC DORSIFLEXION (IDF), AND (B) SWING PHASE OF GAIT (SW) IN
IV. DISCUSSION Abnormal co-contraction patterns are important features of spasticity to measure and clinically assess.This study investigated whether a more broadly bipolar surface EMG differentiating proximal and distal gastrocnemius muscle samples would show differences in cocontraction estimates and may improve the accuracy of this quantitative clinical assessment tool.To establish this measure, MG antagonist cocontraction was estimated in healthy children during isometric dorsiflexion efforts and the swing phase of gait, this last being affected by cocontraction in paretic children.
The main findings showed that MG activity is greater distally than proximally during plantar flexion when it acts as an agonist and it is greater proximally during dorsiflexion, when it is acting as an antagonist.Since the MG cocontraction index is defined as the ratio between RMS values obtained during dorsi and plantar flexions (antagonists/agonists ratio), exclusively sampling a single aggregate site as from the MG centro-proximal region (SENIAM recommendation) might result in lower muscle cocontraction representativity.The estimation of Cocontraction Indices (CCIs) from the conventional sensor location was significantly different from the CCIs computed by considering broader MG regions.This difference arose in all subjects during isometric efforts and in two of 10 healthy children during the swing phase of gait who presented cocontraction patterns.

A. Agonist MG Responses at the Distal Sensor Location
Two pairs of EMG electrode sensors were used to record the proximal and distal regions of the MG muscle, in contrast to the typical single location used for clinical studies of tone abnormalities.Our results from the present study showed that the traditional SENIAM distal position provided greater amplitudes than the proximal one during the MG agonist actions in most children (8 out of 10).Similar results have previously been reported during unilateral plantarflexion in standing with the most activity at the distal site for the medial [18] and lateral gastrocnemius [19].A possible biomechanical explanation could be that since the studied MG acts as an agonist muscle among the other plantar flexors involved, greater activation of the distal portion is more efficient in transmitting force transmission to the ankle joint.The heel-rise task used in this study for the measure of maximal MG EMG requires shortening at the calcaneal insertion under a full body loading condition while the knee joint is stiffened, and the body weight is thus lifted as the leg is stabilized on the foot as well as for the femur on the tibia [18], [19].Although results were not presented in this paper, a similar predominantly distal activity was also shown during the stance phase of the gait cycle where the MG also acts as an agonist during plantar flexion to mainly maximize the body weightlifting force, to stabilize and control ankle dorsiflexion, and to prepare for foot off during gait [20].

B. Antagonist MG Responses at the Proximal Sensor Location
In accordance with our previous findings in healthy adults [12], EMG during the MG antagonist action yielded the largest values from the proximal site.Antagonist MG activity was on average 25% greater proximally in all participants tested during isometric dorsiflexion (Fig. 4).A similar response was also observed in this study during the gait swing phase in only 2 out of 10 children, which would increase CCI MG as shown in Fig. 2 for subject 7. If levels Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
of gastrocnemius cocontraction during dorsiflexion action are known to be increased during isometric efforts [6], [21], during walking activity the main task of gastrocnemius is rather to prevent the tibialis anterior action from causing sagittal plane ankle movement [22].Generally, most of the muscle activity is centered in the gastrocnemius muscle during stance and in the tibialis anterior muscle during the swing phase, when there is no gastrocnemius cocontraction [23].Our findings are consistent with this view: in this study, 8 out of 10 children showed no difference in the antagonist activity level from the medial gastrocnemius during the gait swing phase.A similar level of EMG values was observed in the proximal and distal muscle regions.While requiring further study, these results seem to suggest that during gait profiles with no antagonist-increased activity, EMG samplings from different MG volumes do not seem to matter.However, when cocontraction is manifested during the swing phase of gait, as in subjects 5 and 7, the same MG non-uniform activation that occurred during isometric efforts was observed.In these cases, the amplitudes from proximal sites were on average 25% larger than distal.Why were cocontraction patterns manifested during the swing phase in these 2 children?A possible factor could be the child's age, although children in this age range can be considered to have a mature walk, variability about the mean performance continues to develop into adulthood [24], [25].An alternation of the presence/absence of a swing phase activity for gastrocnemius has been described as a factor of EMG variability in children [26].

C. Potential Factors of Uneven Distribution of Activity Within MG
In this study higher MG RMS values were observed at the proximal region during ankle dorsiflexion (antagonist role) in comparison to those at distal regions during ankle plantarflexion (agonist role).If in muscles whose fibers are parallel to the skin, surface electrodes detect the propagation of motor units, [27] in the gastrocnemius muscle, characterized by oblique fiber orientations, the amplitude of surface EMGs is influenced by the number of active fibers beneath and depends on the muscle's pinnation angle [13].Although, the relative amount of active MG fibers, cannot be exclusively deduced from a given RMS amplitude measured on a small skin region, the findings from this study, may suggest the likelihood of a variation in the distribution of active fibers within the MG muscle across different functional roles or specificities.When the MG acts as an agonist (serving as primary mover), there appears to be recruitment of a greater number of motor units and muscles fibers at the distal region (Fig. 3) while when it functions as an antagonist (acting as an opposing mover) the active fibers appear to be distributed more widely within the MG muscle, spanning a larger proximal region (Fig. 4).Most of bi-articular muscles in animals including humans are innervated by multiple motor nerves or nerve branches creating neuro-muscular compartment within a muscle [28], [29].Significant differential activation patterns between proximal and distal gastrocnemius regions are also reported by other authors during standing [19], [30] or isometric plantar flexion [31], [32].The concept of muscle compartments independently controlled by the CNS is similarly supported by these studies.The gastrocnemius has highly heterogeneous tissues characterized by regional variations in muscle architecture and histochemical fiber type composition [33].A marked nonuniformity exists within these muscle regions, with fascicle length being shortest in the most proximal regions and longest in the most distal regions which are thought to be determinants of muscle function, controlled independently and with distinct biomechanical functions [34], [35].One possible reason for the proximal and distal distribution of MG activity during plantar/dorsal flexion contractions could be related to the muscle mechanical action on the ankle and knee joints which is currently an open issue.For example, for the trans-joint rectus femoris muscle, activity seems to distribute predominantly proximal and distal regions of the muscle during hip flexion and knee extension respectively, both in isometric contractions and during the swing phase of gait [10], [36].Isometric plantar flexion contractions performed with knee flexed lead to smaller values of EMG amplitude rather than extended [37].However, during knee flexion, despite the smaller RMS amplitude in relation to knee-extended position, surface EMGs with relatively greater RMS amplitude were previously observed over a more proximal skin region during isometric efforts [12], [32].

D. Impact on Cocontraction Estimates: Underestimation Risk
The uneven muscle activity distribution within MG has a direct impact on CCIs estimation since the MG cocontraction index used in this study is defined as the agonist/antagonist RMS ratio of the same muscle [6].As shown in Fig. 5, during isometric dorsiflexion efforts, when we take into account samples from proximal and distal MG, experimental CCIs were significantly different from those issued from the sole bipolar distal sampling site (conventional method).Similarly, for the swing phase of gait, CCIs were significantly affected in the 2 children who presented an increased CCI.Using the proximal electrode placement could potentially lead to a biased estimation of the gastrocnemius antagonistic activity if we had only considered values from distal regions.Exclusively sampling from the MG location recommended by SENIAM, might therefore result in cocontraction index underestimation.In the present study, this error amounted to 22% on average for the isometric condition (Fig. 5 (a)) and 25% for the swing phase of gait in the 2 children with significant differences in CCIs (Fig. 5(b)).

E. Implication for Assessment of Spastic Muscles
Present results may suggest that a broader muscle volume sampling of MG doesn't matter during the physiological swing phase of gait because of the poor MG antagonist cocontraction, this could have severe implications for cocontraction assessment in pathological contexts.In spastic gait high levels of cocontraction are known to be deleterious for function [3], [5], [6].Proximal and distal MG sampling may therefore be important to obtain a representative indication of antagonist and agonist activity to better overcome their loss of motor selectivity during active efforts typical of spastic muscles [5].Moreover, the gastrocnemius muscle is one of the most common targets of botulinum toxin type A injection for the treatment of plantarflexion spasticity [38].These results may suggest that better clinical outcomes could be obtained in focal muscle weakening therapy with botulinum toxin if directed at the proximal than distant sites within this muscle.Recent cadaveric studies have shown that the intramuscular endings for the gastrocnemius are located above the proximal third portion of the lower leg [39] rather than at the known mid-belly site reported by ancient studies [40].Implementing a more focused EMG surface approach quantification of the proximal gastrocnemius with ongoing monitoring throughout the botulinum toxin therapeutic regimen, may allow significant advancements in selective functional treatment for equinus foot dynamic deformation.This approach could offer the possibility of reducing the required dosage, optimizing the induction of selective muscular weakness, and mitigating the immediate risk of local diffusion.

V. CONCLUSION
Results from this study suggest a characteristic proximal and distal activity distribution within the gastrocnemius muscle during isometric effort and the swing phase of gait.A bipolar EMG sensor configuration encompassing these muscle regions results in a significant difference in CCI estimation.EMG bipolar sampling information from the proximal medial gastrocnemius (MG) region may be considered as a methodological step in EMG studies which aim to assess antagonist muscle activity, especially in rheologically modified muscles as seen in spastic paresis, to guarantee the accuracy of results and related interpretations.Beyond the gained cocontraction index accuracy, EMG evidence on the agonist and antagonist role of MG muscle could provide informed guidance in the chemodenervation procedure to ensure maximal efficacy.
Possible limitations might affect the outcomes of the present study and must be considered as a limit in finding interpretation.Since the soleus muscle is underneath the gastrocnemius muscle, there are challenges in estimating MG muscle activity using surface EMG.The soleus muscle activity could induce signal disturbance that can interfere with the EMG signal from MG muscle activity and influence our results.Performing experiments on a pediatric population is challenging, however, outcomes reported in this study must be explored and confirmed on a larger sample size to reduce variability and increase the power of the statistical analysis.The IDF condition will be very challenging to apply in spastic paretic cerebral palsy because of ankle dorsiflexor paresis, however some preliminary results show that despite the low TA activation, a similar topographic distribution on the gastrocnemius antagonist activity can be identified.

VI. FUNDING
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Fig. 1 .
Fig. 1.Sensor placement.(a) IMU-TA on the front leg (tibialis anterior SENIAM recommended location) for stride detection, (b) front and back of sensor, and (c) EMG electrodes configuration on children leg: proximal and distal MG sensor locations.

Fig. 2 .
Fig. 2. Gait cycle event detection.Stance and swing phases: (a) angular rate (degree/s) for three strides marker calculation from the gyroscope x-axis output (IMU placed at the shank); (b) RMS from Tibialis Anterior; (c) RMSMG (µV) from proximal and distal sensors for subject 7 presenting a muscle antagonist cocontraction during the swing of gait.
and 4 which were obtained by averaging values across the three trials.The average Cocontraction Index (CCI) was obtained by averaging CCI values across eight gait cycles for the swing phase of gait and across the three trials for the BHR and IDF.The averaged values of CCIs between these measurements were used for the statistical analyses.Wilcoxon's paired sample non-parametric test was used for differences between CCIs conventional and experimental during IDF and swing phase of gait trials.The level of significance was set at 0.05.Statistical analysis was done using Stata v. 15 (StataCorp, College Station, TX).

Fig. 5 .
Fig. 5. CCIs.Mean and SD of conventional and experimental CCIs: (a) during swing phase of gait and IDF for all children; (b) during swing of gait for the subject 5 and 7.

TABLE I TABLE SHOWING PARTICIPANTS
CHARACTERISTICS: AGE, SEX, SIDE TESTED, AND THE AVERAGE GAIT VELOCITY EXPRESSED IN TIME IN SECOND NEEDED TO WALK A 10-METER DISTANCE