Modelling the relationships between EEG signals, movement kinematics and outcome in soccer kicking

Abstract

The contribution of cortical activity (e.g. EEG recordings) in various brain regions to motor control during goal-directed manipulative tasks using lower limbs remains unexplored. Therefore, the aim of the current study was to determine the magnitude of associations between EEG-derived brain activity and soccer kicking parameters. Twenty-four under-17 players performed an instep kicking task (18 m from the goal) aiming to hit 1 × 1 m targets allocated in the goalpost upper corners in the presence of a goalkeeper. Using a portable 64-channel EEG system, brain oscillations in delta, theta, alpha, beta and gamma frequency bands were determined at the frontal, motor, parietal and occipital regions separately for three phases of the kicks: preparatory, approach and immediately prior to ball contact. Movement kinematic measures included segmental linear and relative velocities, angular joint displacement and velocities. Mean radial error and ball velocity were assumed as outcome indicators. A significant influence of frontal theta power immediately prior to ball contact was observed in the variance of ball velocity (R² = 35%, P = 0.01) while the expression of occipital alpha component recorded during the preparatory phase contributed to the mean radial error (R² = 20%, P = 0.049). Ankle eversion angle at impact moment likely mediated the association between frontal theta power and subsequent ball velocity (β = 0.151, P = 0.06). The present analysis showed that the brain signalling at cortical level may be determinant in movement control, ball velocity and accuracy when performing kick attempts from the edge of penalty area.

Trial registration number #RBR-8prx2m—Brazilian Registry of Clinical Trials ReBec.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11571-022-09786-2.

Introduction

Skilled kicking performance in soccer is evidently associated with an increased likelihood of winning in competition, implying a need to identify the determinant factors for this skill (Hunter et al. 2018; Palucci Vieira et al. 2021b). Fast ball velocity and accurate ball placement are key characteristics of kicking performance (Hunter et al. 2018). Achieving high standards in kicking requires a skilled strategy that is dependent first upon processing and cognitive planning (Collins et al. 1991; Noël and Kamp 2012) as well as optimal movement coordination between lower limb segments (Putnam 1991; De Witt and Hinrichs 2012). While the underpinning role of kinematic components of movement [e.g., hip range-of-motion (ROM), knee flexion/extension angle and foot velocity] directly involved in the development of kicking outcomes is well established (Levanon and Dapena 1998; Nunome et al. 2002; Lees et al. 2010), evidence of the central functioning during the task is lacking (Collins et al. 1991; Palucci Vieira et al. 2021a; Slutter et al. 2021). A holistic approach beyond typical kicking measures that integrates neuropsychophysiological aspects commonly related to sports outcomes, such as signalling from the cerebral cortex (Babiloni et al. 2008; Ermutlu et al. 2015; Tharawadeepimuk and Wongsawat 2017; Pluta et al. 2018) is useful in understanding neural activity patterns which may favor performance. It can also potentially assist in obtaining insights into specific mental states associated with successful behavior (Pluta et al. 2018).

In previous studies, cortical activity, in particular the alpha spectral power in various brain regions (frontal, motor and occipital), enabled prediction of outcomes related to accuracy in a range of goal-directed sport skills (e.g. shooting, archery and golf putting) (Salazar et al. 1990; Crews and Landers 1993; Landers et al. 1994; Loze et al. 2001; Babiloni et al. 2008; Baumeister et al. 2008). Recent research has reported contrasting results compared to those previously observed in investigations of movements pertaining to team sports such as ice hockey shooting (Christie et al. 2017), baseball batting (Pluta et al. 2018) and basketball throwing (Chuang et al. 2013). Higher frontal-midline theta power [6–8 Hz; (Chuang et al. 2013)] and reduced frontal-posterior beta [13–30 Hz; (Pluta et al. 2018)] in the preparatory phase were identified as cortical rhythms that were moderate-to-strongly related respectively to throwing accuracy and batting power. Of note was a lack of association between either the task (shooting, batting or throwing) accuracy and athletes’ alpha power in the considered fronto-parieto-occipital sites [7–14 Hz; (Christie et al. 2017)]. Taken collectively, this indicates a need not to restrict analysis to short (e.g. single) waveband range and particularly when little information exists on associations between brain signalling and a given task performance (e.g. Ermutlu et al. 2015). Following on from their literature review, Cheron et al. (2016) proposed a theoretical model where all EEG frequency bands are hierarchically organised and have specific functions in motor control of sports performance. As for example, in addition to the previous cited rhythms (theta-to-beta range), the capacity to extract advanced sensory information present in open sport scenarios/decision-making (Ermutlu et al. 2015) and distal muscle activation (i.e. tibialis anterior; Petersen et al. 2012) is seemingly linked to scalp delta and gamma activities, respectively. Importantly, caution is necessary when attempting to extrapolate consistently reported correlations between sport performance and cortical activity to sports that require tasks with more complex parameters to be controlled. In contrast to sports requiring relatively simple voluntary movements performed in a (semi) static upright position (Hatfield et al. 2004; Cooke 2013; Nakata 2015), soccer players, when kicking the ball must acquire and process pertinent environmental information, perform approach running to the ball, and subsequently produce a coordinated multi-joint movement while also accounting for concomitant ball velocity output in order not to compromise accuracy. The latter represents an additional issue notably across adolescence, when velocity-accuracy trade-offs seemingly occur in kicking outputs which are probably a consequence related to the peak growth spurt (Vieira et al. 2018). Since biological maturity can partly explain performance discrepancies among age-matched players, inclusion of this factor is required in modelling brain-body interactions during soccer kick analyses.

Evidence gathered over the last decade recognises that the motor cortex controls the lower limbs during dynamic multiarticular tasks such as walking/running (Gwin et al. 2010; Petersen et al. 2012; Presacco et al. 2012). Although these publications investigating locomotion have already demonstrated a corticomuscular coupling in the legs, similar to that occurring in other regions like upper extremities, a functional segregation of movement-related cortical activity is generally assumed (Salmelin et al. 1995). The muscular activation and kinematic features of the lower contact limb during kicking modulates ensuing ball flight such as its placement in the goal and velocity (Palucci Vieira et al. 2021b), implying the likely participation of the supra-spinal inputs to the variance in movement parameters and performance. There is also preliminary evidence in soccer pointing to a substantial contribution of the measures derived from resting state EEG to skill-related performance fluctuations as determined by subjective ratings of coaches (Tharawadeepimuk and Wongsawat 2017). To our knowledge however, only one study including EEG metrics has investigated drills such as kicking directly involving a ball. Despite reporting novel information [i.e. higher alpha power in temporal sites during successful trials (ball passing or not cones)], the study included collegiate athletes, who were asked to perform unopposed kicks aimed at small targets on the ground, at a very short distance (7 m) and with high-accuracy low-effort demand required (Collins et al. 1991). Kicking in soccer is generally ballistic in nature and involves open circuit control. As such, it is plausible to state that the existing knowledge on motor control during ball kicking has been derived only from artificial experimental situations. Recent recommendations have been formulated to better align kicking assessment constraints with real-world soccer demands. For example, analysis of motor control is necessary during kicking actions performed over longer distances and including the presence of opponents contesting kicks, such as a goalkeeper. This is important in order to avoid unreliable ball placement and velocity during shots at goal (Palucci Vieira et al. 2021b). In addition, most evidence relating brain activity and technical performance in soccer is derived from studies in senior athletes (Collins et al. 1991; Tharawadeepimuk and Wongsawat 2017; Slutter et al. 2021). Owing to the fact that various kicking outputs (e.g. movement mechanics, ball placement and velocity) are highly age-dependent (Vieira et al. 2018; Palucci Vieira et al. 2021a), results from adult categories cannot be simply extrapolated to developing players. Furthermore, future studies need to make use of recent advances in contemporary imaging technologies enabling collection of EEG data during movement and outside a laboratory context (Park et al. 2015; Cheron et al. 2016; Perrey and Besson 2018). Consequently, opportunities to gather empirical data in field conditions will further knowledge of the behavior of underlying mechanisms that potentially play a role in the planning phase, motor adjustments and effectiveness during soccer kicking.

Therefore, the main purpose of the current study was to analyse, in youth soccer players, the possible associations between brain activity and concomitant lower limb kinematic and performance measures derived from ball kicking. A secondary goal was to evaluate whether (unknown) relationships among EEG-derived brain activity and kicking outcomes are mediated by movement characteristics and inter-individual estimated maturity. Based upon previous findings, our hypotheses were that (1) motor cortex high-frequency signals will contribute to the development of kicking velocity (Pfurtscheller and Lopes da Silva 1999; Gwin et al. 2010), (2) occipital alpha oscillations will have significant power to predict the magnitude of error in placing the ball in the goal (Haufler et al. 2000; Loze et al. 2001; di Fronso et al. 2016) and (3) associations between EEG and kicking outcomes will be mediated by movement features (e.g. ankle kinematics) (Katis and Kellis 2010; Ishii et al. 2012; Palucci Vieira et al. 2021a).

Materials and methods

Sample

Altogether, twenty-four youth U17 soccer players (15.9 ± 0.8 years-old; 60 ± 8.3 kg; 172 ± 9 cm and 1.92 ± 0.74 years from the estimated age of peak height velocity [APHV] (Moore et al. 2015)) competing in state-level (1^st place in São Paulo Interior League 2020, Brazil) representing all outfield playing positions, completed the current study from a total of 28 initially recruited. According to an a priori calculation (G*Power v.3.1.9.4; Franz Faul, Universität Kiel, Germany), the final sample is justified based on the assumption of possible strong EEG influence on kicking performance variance (R² > 0.26; effect size f² = 0.37; power = 80%; alpha = 0.05). The study protocol was pre-registered, as a part of an umbrella research project, in the Brazilian Registry of Clinical Trials (register number RBR-8prx2m) and approved by the Local University Human Research Ethics Committee (protocol CAAE85994318.3.0000.5398), in accordance with 466/2012 resolution of the National Health Council. Players’ guardians provided previous verbal and written permissions, while the players signed an assent form to agree to take part in the study as a volunteer.

Kicking task

The participants wore the same shoes and clothes they habitually used in training and competition in an attempt to minimise any potential effects of the experimental testing conditions on the players’ natural kicking patterns. Each participant wore eight spherical markers (25 mm Ø) which were fixed with the aid of small kinesiotaping strips (Tmax Medical Co., Ltd, South Korea) in the bone protuberances of anterior superior iliac spine, greater femoral trochanter (also in the non-dominant limb), lateral femoral epicondyle, lateral malleolus, calcaneus and distal phalanx of fifth metatarsal head (only in dominant limb).

A standard warm-up routine composed of 4 min of moderate-intensity running (3/4 rating of perceived exertion–Borg CR-10 scale) followed by dynamic stretching was performed prior to kick testing. The kicking protocol was conducted during the habitual training schedule for the age group assessed (between 15:00 and 17:00 h) on a natural grass pitch with and in the presence of sunlight. Two 1 m² square targets were fixed in the upper extremities of the goalpost (7.32 × 2.44 m). The participants completed 20 trials in total (three conditions: seven kicks directed to each side and six for the goal centre; randomized using the https://www.random.org interface) distributed into four blocks of five kicks each. An interval of 3 min and 40 s for passive recovery (van den Tillaar and Fuglstad 2017) was implemented between blocks and trials, respectively. Prior to each attempt, the ball (PENALTY® FIFA-approved; 5 sized, 70 cm of diameter and 430 g weight) was positioned at a distance of 18 m from the midpoint goal line. Ball pressure was controlled across the experiment (Poker digital calibrator 09,053, Cauduro Ind e Com Vest LTDA, Brazil) to be constant at 0.7 atm.

Players kicked the stationary ball using the instep foot region, after a beep (electronic whistle FOX 40®, Fox 40 International Inc., Canada) and with an approach run of 3.5 m and 45 degrees in relation to the kick mark (Kellis et al. 2004). The aim of the task was explained to participants: ‘kick the ball to hit the target centre so that the goalkeeper cannot intercept the ball’. Two youth goalkeepers of similar age, height and skill level (determined according to coaching staff) were asked to try to defend the kicks, remaining 1.2 m in front of the midpoint goal line, keeping their hands on the knee until the instant the ball was kicked and without prior knowledge of where the players would aim their kick (i.e. centre, ipsilateral or contralateral side conditions). Reliability of this kick testing protocol has recently been confirmed (Palucci Vieira et al. 2022).

EEG measures

To record signals from the cerebral cortex (Fig. 1), a portable system was used (1024 Hz; eegoTM sports EEG system LE-200, ANT Neuro b.v., Enschede, Netherlands) (Christie et al. 2017). This includes a cap with 64-channel (WaveGuardTM original cap; 10/10 international electrode positioning system), connected to an amplifier (2 kHz; eego amplifier CE Class IIa medical device) and a tablet (TRAVELline T10-B5 Pro Tablet, Atom(TM) × 5-Z8350 1.44 GHz, Bluechip Computer AG, Germany), inserted into a small backpack, which recorded and displayed the signal online via a Wi-Fi connection with an off-field computer (Dell Inc., Intel(R) Core™ i7-10510U). The reliability of this system is reported elsewhere (Fiedler et al. 2015; di Fronso et al. 2019). Using conductive gel (neurgel; Spes medica Srl-Italy), the impedance of electrodes was maintained below 30 kΩ during the experiment (ground < 10 kΩ).

Fig. 1.

Experimental setup adopted for data collection/player montage (upper panel) and example of EEG raw (bottom left) and cleaned data (bottom right)

The collected data were subsequently exported to the MATLAB® (R2019a; MathWorks Natick, MA, EUA) and processing conducted using the open-source toolbox EEGLAB v2020.0 (Delorme and Makeig 2004). Raw data were treated with a Butterworth band-pass 0.3–50 Hz, in addition to a 60 Hz notch filter, and then were re-sampled to 512 Hz. The artifacts were initially rejected by visual inspection considering excessively noise periods (Petersen et al. 2012) and automatically, eliminating those channels with kurtosis > 3 standard deviations from the mean value and epochs with absolute difference > 150 mV (Pluta et al. 2018; Duru and Assem 2018). All remaining channels (57 ± 3 per participant) were then re-referenced to the common average and some typical artifacts such as coming from linear trend, eye blink/movement, muscle or cardiac activity were identified and removed with assist of data decomposition by independent component analysis [ICA; see Delorme et al. (2007) for more information].

Trials were divided into three periods, lasting -6 to -3 s before kicking (preparatory phase), -3 to -1 s (approach phase) and -1 s until approximately the foot-ball contact moment (impact phase). Ball impact was marked by the examiner using a proprietary program feature (eegoTM software, ANT Neuro b.v., Enschede, Netherlands) and confirmed by the EEG burst at such time. For each, the average power spectral density of the frontal (F3, Fz, F4), motor (C3, Cz, C4), parietal (P3, Pz, P4) and occipital (O1, Oz e O2) (Presacco et al. 2012) regions of interest (ROIs) considering delta (δ; 0.5–3 Hz), theta (θ; 4–7 Hz), alpha (α; 8–12 Hz), beta (β; 13–30 Hz) and gamma (γ; 31–50 Hz) frequency bands were extracted. Event-related synchronisation/desynchronisation (ERS/ERD) was calculated using input parameters (spectral power during time-windows of -5 to -4 s as the baseline period and -1 s to impact as the event period) and equation (ERS/ERD = [(event period – baseline period) / baseline period] × 100) following literature-based recommendations (Babiloni et al. 2008; Pfurtscheller and Lopes da Silva 1999). Inter-trials reliability of discrete EEG spectral power and ERS/ERD outcomes were determined, separately for all kick phases (exclusively concerning power measures in this case), as well as considering whole trials (see below). Furthermore, for each ROI event-related spectral perturbation (ERSP) images (Makeig 1993) and inter-trial coherence (ITC) images (Makeig et al. 2004) were computed through their respective "STUDY" functions available in the EEGLAB environment (Delorme and Makeig 2004). For this analysis, zero indicates the beginning of impact phase. Four players among those initially recruited were excluded from the additional analysis owing to identification of an unacceptable percentage of artifacts (> 30%) contaminating their EEG data (valid trials per participant included = 94 ± 6%).

Kinematic procedures

Four digital video cameras were used (GoPro® Hero 7 Black Edition, GoPro GmbH, München, Germany), adjusted at a sampling frequency of 240 frames/s (Wide field-of-view 1280 × 960 pixel of resolution; 1/480 s shutter speed and NTSC standard). These were synchronized using Wi-Fi (GoPro smart remote control) and coupled with fixed tripods and positioned laterally (~ 2.5 m) to where the kicks took place, allowing capture of the kicker’s movement and ball trajectory immediately after contact.

In the DVIDEOW interface (version 4.0; Laboratory of Instrumentation for Biomechanics & Institute of Computing UNICAMP, Brazil) (Figueroa et al. 2003), the following sequence of steps was completed: (i) calibration (4.4 × 3.5 × 1.3 m respectively in anterior–posterior, medio-lateral and vertical directions), (ii) semiautomatic tracking of markers and the ball and (iii) three-dimensional (3-D) reconstruction using the direct linear transformation method (DLT), embedded into the software. Thereafter, data matrices containing the spatial information as a function of time were exported into MATLAB R2019a environment (MathWorks Natick, MA, USA). Using the same procedures employed in a previous study (Vieira et al. 2018) radial distortion of image sequences was corrected before 3-D DLT reconstruction (uncertainties = 0.9 cm [precision], 2.1 cm [bias] and 2.3 cm [overall accuracy] in the present study). With the intention of minimizing systematic errors generally arising from the foot-ball impact, the movement data were linearly extrapolated (20%, at the beginning and end of time-series) and filtered using Butterworth (cut-off frequency = 25 Hz) followed by non-parametric locally weighted function rloess (span = 0.1) (Barbieri et al. 2015; Palucci Vieira et al. 2021a); in both cases, the smoothing parameters were selected after residual analysis.

In accordance with definitions available in the scientific literature regarding the reference frame of the joint and segment centers, it was possible to compute the angular joint displacement of the hip (internal/external rotation, abduction/adduction and flexion/extension), knee (flexion/extension) and ankle (plantarflexion/dorsiflexion, abduction/adduction and eversion/inversion), as well as the associated angular joint velocity in each case (Palucci Vieira et al. 2021a). Relative linear velocities between joints were also obtained (De Witt and Hinrichs 2012). Range-of-motion (ROM) was defined as the difference between minimal and maximal values for each movement component observed during the duration of the trial. To measure foot velocity, the barycentre of the ankle, calcaneus and toe coordinates was used (Palucci Vieira et al. 2021a). Ball velocity was determined using its centroid trajectory and calculated as an average of 10 frames following ball contact [for more details of calculation see (Levanon and Dapena 1998)].

To capture ball placement after kicking, one camera operating at 60 frames/s (Linear field-of-view 1920 × 1080 pixel of resolution), of the same model and brand of the previous, was fixed frontally 23 m away from the midpoint between goalposts. An auxiliary camera (configured at the same parameters of frontal one) was positioned in the intersection between the penalty area and goal lines, with its focus parallel to the ball trajectory. For each trial, using the same software and kinematic procedures described above, 2-D ball position was digitized in the instant it crossed the goal line. The Euclidean distance between the ball centroid and the target centre was then measured and the mean radial error calculated, separately for each of the experimental conditions (Vieira et al. 2018). Kick attempts were classified as successful (i.e. kicks on-target/goal) or unsuccessful (i.e. missed kicks).

Statistical analysis

Initially, outliers were automatically identified and removed (1.47% of all eligible data cells) using the z-score principle (i.e. − 3.29 > z >  + 3.29; (Tabachnick and Fidell 2019)). As many variables did not fulfil the normality assumption evaluated by the Shapiro–Wilk test, kurtosis and visual inspection, we applied a Box-Cox transformation to the dataset, aiming to reduce problems arising from non-additivity, non-normality and heteroscedasticity. Intraclass correlation coefficient (ICC) were taken as reliability estimates of EEG spectral power and ERS/ERD measurements across trials (using an Excel® spreadsheet freely available in sportsci.org/2015/ValidRely.htm domain), and were interpreted as being poor (< 0.50), moderate (0.50–0.75), good (0.75–0.90) or excellent (> 0.90). One-way analysis of variance tests (ANOVAs) were used to compare lower limb kinematics, ball velocity and EEG between experimental conditions while Student t test for paired samples was used in mean radial error comparisons. Repeated measures ANOVAs (RM-ANOVAs) designed in 2 (performance [kicks on-target, kicks off-target]) × 3 (phase [preparatory, approach, impact]) × 4 (ROI [frontal, motor, parietal, occipital]) were performed, separately for each EEG power frequency band. Additional RM-ANOVAS of 2 (performance) × 4 (ROI) were performed, separately for each EEG ERS/ERD frequency band. Resulting grand average ERSP and ITC images were obtained across all subjects considering kicks on-target/goal and missed kicks and then compared using paired t-test; its non-significant outcomes are masked in plots. In an attempt to control for inflation in type I error rate, Pearson’s product-moment correlation coefficients were computed between mean radial error in the kicks and lower limb kinematic parameters/ball velocity with EEG spectral power measures only considering the preparatory and impact phases, respectively. In the case of ERS/ERD, such restriction was not applied. Linear regressions (stepwise method) were subsequently run to identify the relative contribution of the EEG to the variance in kicking performance indicators. Structural equation modelling (SEM) was used to analyse if lower limb movement kinematics or estimated biological maturity (APHV) were significant moderators in the significant associations between EEG and kicking performance (mean radial error and ball velocity), as well as whether there was any trade-off between the two last mentioned dependent factors. Software used included MATLAB® (outliers exclusion, ERSP and ITC comparisons; R2019a, MathWorks Natick, MA, USA), IBM SPSS Statistics (data transformation, ANOVAs, correlations and regressions; v.25, IBM Corp.©, USA) and Stata (SEM; v.13 for Windows, StataCorp LP, Texas, USA) and for all cases, a statistical significance level of P ≤ 0.05 was assumed unless otherwise stated.

Results

Comparisons between kicking conditions

The ERSP and associated ITC results are shown separately for each ROI in Figs. 2–5. In general, regarding ERSPs it was possible to observe that successful kick attempts, as compared to missed kicks, showed a significant higher decline in frontal theta power around -200 ms to zero and increase in alpha power starting at 200 ms until the end of the window (Fig. 2). In the motor region, beta power raised (50 to 200 ms) in successful kicks while it decreased in the same period for missed kicks. Also, an increase in motor alpha activity (300 ms to the end) was significantly greater in successful compared to missed kicks (Fig. 3). No change or slightly decrease were identified in parietal theta (-100 ms to zero) during successful kicks whilst missed kicks presented augmented power is this same frequency band/period. There was a decrease in occipital theta (-300 to -200 ms) and increase in alpha and beta from 300 ms to the end of successful kicks when compared to missed kick attempts (Fig. 5). The ITCs were significantly greater in motor theta (-300 to -100 ms), frontal (-200 to ~ 0) and parietal (100 to 200 ms) alpha and lower in occipital theta (300 ms to end) in successful versus unsuccessful kicks.

Fig. 4.

Grand average ERSP and ITC outcomes, respectively in the first and second lines, representative of parietal brain region according to kick conditions

Fig. 2.

Grand average event-related spectral perturbation (ERSP) and inter-trial coherence (ITC) outcomes, respectively in the first and second lines, representative of frontal brain region according to kick conditions. The green area represents absence of statistically significant values in both group level or resulting plots for the comparison made; the same is valid for remainder similar figures

Fig. 5.

Grand average ERSP and ITC outcomes, respectively in the first and second lines, representative of occipital brain region according to kick conditions

Fig. 3.

Grand average ERSP and ITC outcomes, respectively in the first and second lines, representative of motor brain region according to kick conditions

Non-time-series EEG extracted measures (spectral power: F_2,53 ≤ 2.11; P ≥ 0.13; η² ≤ 0.07; ERS/ERD: F_2,52 ≤ 2.79; P ≥ 0.07; η² ≤ 0.07), lower limb kinematics (F_2,67 ≤ 1.48; P ≥ 0.24; η² ≤ 0.04) and kicking performance (ball velocity: F_2,67 = 0.591; P = 0.56; η² = 0.02 and mean radial error: t_1,23 = − 1.688; P = 0.11; d = 0.49) did not exhibit significant differences between experimental conditions, allowing us to group all trials for further statistical analysis (i.e. 20 representative trials of the ball velocity/lower limb kinematic results and 14 used to compute the variable of ball placement in the goal– this was not determined for kicks directed to the goal centre). ICC values (ranged from poor to excellent) are presented as supplementary online materials 1 and 2. The results of RM-ANOVAs also pointed to the absence of a statistically significant interactive effect between performance × period × ROI (Fig. 6), for all EEG power frequency bands included (delta: F_6,54 = 0.769; P = 0.60; η² = 0.08; theta: F_6,60 = 0.677; P = 0.67; η² = 0.06; alpha: F_6,72 = 1.420; P = 0.22; η² = 0.11; beta: F_6,72 = 1.026; P = 0.42; η² = 0.08 and gamma: F_6,54 = 0.508; P = 0.80; η² = 0.05) and no significant performance × ROI interaction (Table 1) concerning ERS/ERDs (delta: F_3,27 = 0.510; P = 0.68; η² = 0.05; theta: F_3,27 = 0.778; P = 0.52; η² = 0.08; alpha: F_3,27 = 0.596; P = 0.62; η² = 0.06; beta: F_3,27 = 0.504; P = 0.68; η² = 0.05 and gamma: F_3,27 = 0.106; P = 0.96; η² = 0.01) as well as separate performance main effects (P ≥ 0.43; η² ≤ 0.07).

Fig. 6.

EEG spectral power verified in the kicking trials on- and off-target

Table 1.

EEG ERS/ERD verified in the successful and unsuccessful kicking trials

		ERS/ERD [Box-Cox(log)µV2/Hz]
		Kicks on-target/goal	Missed kicks	p-value	Effect size (d)
Frontal	Delta	3.52 ± 1.03	3.43 ± 1.21	0.88	0.08
	Theta	1.14 ± 0.69	1.33 ± 1.04	0.53	− 0.22
	Alpha	0.06 ± 0.82	0.53 ± 1.08	0.25	− 0.50
	Beta	− 0.44 ± 0.70	− 0.04 ± 0.75	0.29	− 0.55
	Gamma	− 0.90 ± 0.74	− 0.61 ± 0.97	0.53	− 0.34
Parietal	Delta	3.27 ± 1.23	2.84 ± 1.33	0.54	0.34
	Theta	0.74 ± 0.79	0.35 ± 1.10	0.38	0.41
	Alpha	− 0.27 ± 0.78	− 0.04 ± 1.19	0.52	− 0.24
	Beta	− 0.81 ± 0.73	− 0.63 ± 1.03	0.69	− 0.19
	Gamma	− 1.34 ± 0.91	− 1.03 ± 1.16	0.51	− 0.30
Motor	Delta	3.15 ± 1.11	3.15 ± 1.85	1.00	0.00
	Theta	0.61 ± 0.71	0.55 ± 0.94	0.86	0.08
	Alpha	− 0.05 ± 0.81	− 0.02 ± 1.11	0.96	− 0.03
	Beta	− 0.83 ± 0.75	− 0.37 ± 0.99	0.35	− 0.53
	Gamma	− 1.25 ± 0.83	− 0.87 ± 1.16	0.49	− 0.38
Occipital	Delta	3.20 ± 0.98	2.80 ± 1.50	0.58	0.32
	Theta	1.01 ± 0.78	0.87 ± 1.15	0.81	0.13
	Alpha	0.35 ± 1.04	0.35 ± 0.73	0.98	0.01
	Beta	− 0.33 ± 0.57	− 0.23 ± 0.94	0.80	− 0.12
	Gamma	− 0.86 ± 0.94	− 0.69 ± 0.75	0.65	− 0.20

Correlations between EEG, kinematics and outcomes

Significant correlation coefficients (with their respective confidence limits) were obtained from the associations between selected kicking components with the EEG spectral power (Table 2) and ERS/ERD values (Table 3). Specifically, the knee ROM (1.15 ± 0.16 rad) was moderately correlated with frontal, parietal gamma and frontal delta powers (2.75 ± 3.05, 1.74 ± 2.06, 518.43 ± 638.12 µV²/Hz, respectively) and largely correlated with motor, parietal and occipital gamma ERS/ERDs (-22.98 ± 166.25%, -49.77 ± 154.94%, -98.65 ± 182.31%, respectively). Frontal theta power (30.19 ± 32.37 µV²/Hz) was moderately and largely correlated with ankle eversion (0.05 ± 0.32 rad) and hip flexion angles (0.57 ± 0.12 rad) at impact, respectively. Hip flexion angle at impact was also moderately and largely correlated with parietal and motor delta, theta and alpha powers (340.03 ± 328.66, 19.55 ± 15.29 and 4.96 ± 4.16 µV²/Hz, respectively for the latter), while the ankle eversion angle at impact showed moderate inverse relationships with frontal alpha power (7.71 ± 7.88 µV²/Hz) and occipital theta ERS/ERD (+ 86.73 ± 212.53%). Significant large and moderate relationships occurred between hip flexion angle at impact and motor theta ERS/ERDs (+ 100.76 ± 215.35%), parietal gamma and theta (-9.70 ± 302.02%) ERS/ERDs, respectively. Furthermore, the relative velocity between foot centre of mass and knee linear velocities at impact instant (50.15 ± 5.35 km/h) exhibited moderate correlations with occipital beta (3.59 ± 2.39 µV²/Hz) and gamma (2.61 ± 1.93 µV²/Hz) powers.

Table 2.

Correlations (± 90% confidence limits) between selected kicking performance parameters and lower limb kinematics with EEG spectral power in each frequency band/brain region

Table 3.

Correlations (± 90% confidence limits) between selected kicking performance parameters and lower limb kinematics with EEG ERS/ERD in each frequency band/brain region

Moderate to large significant correlations were observed between ball velocity (95.13 ± 7.45 km/h) and EEG spectral power during the impact phase (-1 s to ~ kick), in reference to the frontal theta (30.19 ± 32.37 µV²/Hz) and fronto-parieto-occipital delta levels (Table 2). ERS/ERD indices were non-significantly related to ball velocity (Table 3). Finally, the occipital alpha (2.40 ± 1.17 µV²/Hz) during the preparatory phase (-6 to -3 s) was the sole EEG spectral power measure (Table 2) moderately correlated with the mean radial error (2.08 ± 0.35 m). Occipital beta ERS/ERD (-87.59 ± 276.24%) also showed a moderate inverse relationship with mean radial error (Table 3).

Linear regressions and mediational models

The linear regression models (Fig. 7 and Supplementary online material 3) revealed a 35% contribution of the frontal theta EEG spectral power (-1 s to ~ kick) to the variance in ball velocity (Z_(1,19) = 9.641; standardised β coefficient = 0.591; P = 0.01) while 18 to 20% of the variance in mean radial error was explained respectively by the occipital beta ERS/ERD (Z_(1,19) = 4.911; standardised β coefficient = -0.463; P = 0.04) and EEG power expression in alpha band during the preparatory phase (-6 to -3 s) obtained in the occipital region (Z_(1,19) = 4.453; standardised β coefficient = -0.445; P = 0.049). Other EEG parameters did not demonstrate significant power to predict the kick outcome measures (i.e. ball velocity or mean radial error). The SEM models pointed to a mediation with borderline significance (Table 4) of the ankle eversion angle in the association between frontal theta power (1 s to ~ kick) and subsequent ball velocity [relative χ² (χ²/gl) = 4.208 (satisfactory value < 5.00); P = 0.06; Fig. 8(A)]. No significant mediators (i.e. indirect effects) were detected for the association between EEG and mean radial error (e.g. Figure 8(B)).

Fig. 7.

General overview of the plots resulting from the associations between EEG spectral power and soccer kicking performance. Note: The shaded area represents the confidence interval of regression line. Hip angle is in reference of extension (−) and flexion (+) and ankle angle is in reference of inversion (+) and eversion (−) movements at impact instant. ROM range of motion

Table 4.

Direct, indirect and total effects obtained from the mediational models including EEG, lower limb kinematics and soccer kicking performance parameters

	β coefficient	Standard error	Z	p-value	Confidence interval (90%)
Model A
Direct	0.3726	0.16	2.29	0.02	0.11–0.64
Indirect	0.1513	0.08	1.87	0.06	0.18–0.28
Total	0.1605	0.09	1.70	0.09	− 0.01–0.32
Model B
Direct	− 0.3857	0.19	− 2.01	0.04	− 0.76 to − 0.01
Indirect	− 0.0537	0.10	− 0.55	0.58	− 0.24–0.14
Total	− 0.4395	0.22	− 2.02	0.04	− 0.87 to − 0.01
Model C
Direct	− 0.1801	0.30	− 0.59	0.55	− 0.78–0.41
Indirect	− 0.1512	0.22	− 0.70	0.48	− 0.57–0.27
Total	− 0.3313	0.17	− 1.89	0.06	− 0.67–0.01
Model D
Direct	− 0.4056	0.13	− 3.08	0.01	− 0.66 to − 0.15
Indirect	− 0.0065	0.11	− 0.06	0.95	− 0.22–0.20
Total	− 0.4121	0.13	− 3.18	0.01	− 0.67 to − 0.16

Each model corresponds to those graphically illustrated in Fig. 8

Fig. 8.

Mediational models of the associations between EEG and soccer kicking performance. Note: MRE mean radial error; PHV estimated biological maturity (age of peak height velocity); V_BALL ball velocity; ROM range of motion. A model with final outcome being the velocity, where ankle kinematics are in reference of eversion movement; b, c and d models with final outcome being the ball placement in the goal, where knee kinematics are in reference of range-of-motion and ankle kinematics are in reference to the plantarflexion movement; *EEG spectral power; **ERS/ERD. The solid continuous lines represent significant associations between two given factors while the dashed lines indicate the absence of statistical significant at the level of P ≤ 0.05

Discussion

The goal of the current study was to determine whether an association exists between brain activity measured via mobile EEG and concomitant movement kinematics and instep kicking performance in youth soccer players. Most of our initial hypotheses have been confirmed. In sum, both the ball velocity and its placement on the goal target were correlated, generally at a moderate magnitude, with the inter-individual EEG oscillations. In contrast, no differences were observed, for any spectral power measurements, following a simple classification of kicking attempts that successfully attained or not the target. When considering successful (pooled kicks on-target and goals) versus unsuccessful attempts (missed kicks) as well as time-series (ERSP images) rather than only extracted single power values representative of a given epoch, some differences were evidenced. These included: (i) frequent (frontal, motor, parietal and occipital) increased alpha activity (~ 200 ms to the cycle end); (ii) declines in theta power (frontal, parietal and occipital) mainly around zero time-moment and (iii) augmented beta (motor and occipital) roughly from 50 ms to the end, that collectively were greater during successful compared to missed kicks. High-frequency signals were not always determinant parameters in kicking velocity. Nonetheless, different cortex regions and frequency bands of the EEG signal seemed to play a role in controlling distinct aspects required in the kicking skill (i.e. velocity and accuracy). The occipital alpha power during the preparatory phase prior to the approach run and occipital beta ERS/ERD notably influenced the mean radial error of kicks while ball velocity was more dependent upon the frontal theta power component immediately before foot-ball impact. In addition, these EEG factors generally showed moderate reliability. The results also revealed likely mediation of the ankle kinematics in the association between EEG (frontal theta power) and ball velocity. Among the mediation factors computed here, these only had non-significant influences on the associations between EEG and mean radial error. When significant, ITCs were generally greater in kicks on-target/goal compared to missed kicks over various brain regions only with the exception of in the occipital lobe. In the following paragraphs, interpretations of the observed correlations regarding EEG and performance measures are proposed focusing on the possible role of cerebral inputs to planning and motor responses when kicking.

In the present young soccer players, faster ball velocity during instep kicking was shown to be related to greater cortical activity in the theta band, recorded at the frontal cortex region. This main finding reinforced evidence collated in a review of EEG markers as potential indicators of sports performance. In the review (Cheron et al. 2016), the emerging role of theta oscillations to motor control of sport skills was highlighted and preliminary data presented by the authors in a pilot study (partly reported in that same text–Introduction, pp. 2) also pointed to an occurrence of this specific rhythm, in a frontal channel. Indeed, increased frontal-midline theta spectral power is sensitive in discriminating distinct levels of performance in a variety of goal-directed sports tasks, although the current literature supports this behavior only for upper limb tasks (Baumeister et al. 2008; Doppelmayr et al. 2008; Chuang et al. 2013; di Fronso et al. 2016). Additionally, the quickness of movement in ballistic tasks was previously correlated with the magnitude of brain oscillations in theta domain (Ofori et al. 2015). There is evidence of a substantial participation of frontal cortical activity in movement control during running (Suzuki et al. 2004). Traditionally, central executive processes, which integrate long-term and working memory, are recognised to rely on frontal functioning (Collette and Van der Linden 2002). In our experiment, the cognitive demand can be exemplified by the necessity to direct the ball to a far target while producing adequate velocity levels to avoid placing it outside the goal or a goalkeeper block. The frontal region may also be acting in coupling such complex task parameters. Conversely, excessive frontal theta spectral power may be indicative of high attentional control during a given movement (Haufler et al. 2000; Baumeister et al. 2008, 2010; Chuang et al. 2013). This is occasionally deemed as a possible risky strategy since errors can be generated in the task in relation to accuracy demands (Kao et al. 2013). However, in our study, there was no statistical confirmation of a velocity-accuracy trade-off in the experiment carried out (e.g. non-significant correlations between these components – see Fig. 7). Hence, some sustained concentration in the final phase of soccer kicking may help produce faster ball velocity.

To the extent of our knowledge, most studies relevant to current work have attempted to model the relationships between EEG and the outcomes of sports skills, using either indirect (Tharawadeepimuk and Wongsawat 2017; Pluta et al. 2018) or direct performance measures (Collins et al. 1991; Chuang et al. 2013; Christie et al. 2017). However, these have not considered the characteristics of the movement technique that can mediate this process, as done in the current study. Here, it is noteworthy that the frontal theta power, in addition to its influence on the development of high ball velocity, also showed important correlations with the hip flexion and ankle eversion angular joint displacements at the foot-ball impact instant. Previous studies generally confirm the benefits of movement patterns characterized as high hip flexion and ankle eversion in reaching maximal ball velocity when performing instep kicks (Katis and Kellis 2010; Ishii et al. 2012; Palucci Vieira et al. 2021a). In particular, greater ankle eversion can permit an increase in the foot-ball contact area during instep kicking. Consequently, the programming of motor action to position the foot properly and produce high ball velocity is supported by mutual relationships between frontal theta power, ankle movement and ball velocity. Differently and contrary to our hypotheses, gamma power was associated only with the knee ROM and the motor region of the cortex showed influence more frequently in hip movements as compared to knee (single significant correlation) and ankle/foot movements (no associations). These results suggest that there is likely a greater role of the motor cortex in the proximal joint involved in kicking than in the distal endpoint; and high-frequency brain signals may not be necessarily responsible for the movement velocity in this skill. On the other hand, motor cortex-hip relationships may signify that foundation signals to control kicking movement are sent to this specific joint. In fact, considering the traditional somatotopic organisation of the sensorimotor cortex, despite lower limb segments appears to recruit almost the same region, hip is located closer to the scalp as compared to ankle and foot specially in output region (motor cortex) (de Klerk et al. 2015) and this could help explain the above-mentioned result. The correlations reported between delta power in various ROIs and kick outcomes (i.e. ball velocity) require careful interpretation due to the fact that oscillations in this band have been removed from the regression models because they do not meet the established assumptions; their real contribution to lower limb kinematics is also uncertain (Castermans et al. 2014), can be highly coupled with high-frequency cortical activity (Händel and Haarmeier 2009) and disappear when considering ERS/ERD values.

Here, a single EEG spectral power measure directly associated with the ball placement error in the designed target on the goal was identified. The greater the occipital alpha power in the preparatory phase of kick immediately before the approach run to the ball, the lower the observed mean radial error. Conversely, occipital beta ERS/ERD was negatively related to the ball placement error; the higher the occipital beta ERD expression, the farther from target the kicks occurred. Augmented expression of occipital alpha EEG prior to execution of goal-directed skills also induced enhancements in subsequent accuracy in completing other sport tasks that are visuomotor in nature (Haufler et al. 2000; Loze et al. 2001; di Fronso et al. 2016). There are some possible putative mechanisms by which increased alpha power levels generally benefit athletic performance. As for example, alpha oscillations may reflect a more relaxed behavioral state where unnecessary/conflicting processes are inhibited to a greater degree (Klimesch et al. 2000; Goldman et al. 2002; Budnik-Przybylska et al. 2021) thus inducing a greater focus on a given desired task (Cheron et al. 2016). The visual behavior associated with higher observed levels of this specific EEG marker also deserves scrutiny. During the aiming phase of kick, players are collecting information about the environment (e.g. goalkeeper, ball and target positions). At this stage the system begins to make the calculations aiming to perform the most precise movements as possible. In this sense, it is recognized that a longer time fixing the gaze on the target during the immediately pre-kick phase (e.g. quiet eye; see Nagano et al. (2006)) and, inversely, reduced time looking at environmental distractors such as the goalkeeper (Noël and Kamp 2012), both assist in obtaining good kicking accuracy. This is further confirmed in a previous experiment which tested among ‘target-focused’ or ‘keeper-focused’ strategies in soccer kicking, where the former resulted in better shot performance (Wood and Wilson 2010). Despite information on eye movements not being captured in the present study, eye quietness has been positively correlated with concomitant determination of occipital alpha EEG (Janelle et al. 2000; Gallicchio and Ring 2020). Notwithstanding, occipital beta ERD linked with poor ball placement may be indicative of a potentially excessive attempt to capture visual information later in the kick cycle (Cordones et al. 2013) that was not beneficial to targeting the goal. Therefore, the effectiveness of placing the ball in the goal–upper corners–during soccer kick attempts at goal from the edge of the penalty area, apparently requires a mental state of focus and restricted selection of environmental information presented during the task while simultaneous control over the player’s state of excitability also seems essential while kicking. However, unlike ball velocity, it was not possible to completely clarify the possible path through which cortical activity may influence the kinematics of movement and subsequent ability to place the ball based only on the measurements considered in our work. Future studies using advanced EEG calculations (connectivity, chaotic and source estimation metrics) and statistics (e.g. multilayer neural network) may provide additional pertinent data.

The analysis of missed kicks showed these were generally linked to less consistent neural responses. This is illustrated by lower ITC values observed over various regions (frontal, motor and parietal) during missed kicks as compared to kick attempts correctly delivered to the goal or target in its upper corners. The only exception was the occipital theta manifestation later in the analysed window that was more repeatable in kicks outside the goal. Of note, suppression but not augmentation of occipital theta band activity was demonstrated previously to benefit visuomotor performance (Beatty et al. 1974) and this was the case here earlier in the kick cycle. Additionally, from the ERSP analysis it was possible to observe again that alpha activity was raised to a greater extent in successful as compared to missed kicks, over all brain regions considered. Since alpha oscillations is often related to a decrease in cortical activation (Hatfield et al. 2004), this can provide further evidence favouring the “neural efficiency hypothesis” that associates proficient sports performance to low demands placed upon cortical resources at some point; indeed this theory postulates that successful behaviour in athletes occurs as a function of experiencing a default mode or “automatic” network functioning, in which less energy is spent in producing the more skilled actions (Haufler et al. 2000; Hatfield et al. 2004; Del Percio et al. 2009; Cheron et al. 2016; di Fronso et al. 2016; Duru and Assem 2018; Budnik-Przybylska et al. 2021). Owing to the fact that a rapid deceleration of the body centre-of-mass during the last stride is related to subsequent performance in kicking (Augustus et al. 2021) and a direct association exists between theta waves with voluntary movement velocity (Cheron et al. 2016; Ofori et al. 2015), occurrences of more pronounced reductions in theta power in successful kicks–observed near the transition between approach run to the ball and kick movement itself–suggest that such EEG pattern can be important to players regulate their velocity in last step before ball contact in order to obtain effective intersegmental control and/or allow adjustments during the impact phase. Finally, concerning the increase identified in beta (e.g. motor) power expression at ~ 50 ms, this could characterise successful kicks as having greater levels of the so called beta rebound, a mechanism that aids re-calibration of the motor system after a forcibly interrupted movement (i.e. approach run cessation and commencement of impact phase of kicking) (Mustile et al. 2021).

The present study adds to the current knowledge about the motor control of kicking ability in soccer although several limitations must be acknowledged, implying caution in the interpretations and extrapolations made. Firstly, we chose to group electrodes of both hemispheres together with midline sites to obtain representative EEG measures for each region. Despite a previous study showing that EEG responses during low effort-high accuracy demands may be similar (Collins et al. 1991), future investigations should determine separately the contribution of each cerebral hemisphere to the soccer kicking performance in relation to players’ lateral preference. A lack of concomitant collection/analysis of eye electrooculographic (EOG) and muscle electromyographic (EMG) activities related to kicking action prevents direct identifying EOG-EMG contamination of EEG signals [e.g. respectively in theta (Gasser et al. 1985) and gamma (Goncharova et al. 2003) waves] limiting consequent application of advanced data filtering procedures (Jiang et al. 2019; Mucarquer et al. 2020). Also, given our resources, the kinematics of the kick cycle were not electrically synchronized with the EEG equipment. Even reporting inter-trial reliability, it may have added some undesirable variance to the dataset. Bearing in mind that the experiment was conducted on-field, factors difficult to control across testing days (e.g., lighting) could have led to natural variations in EEG spectral power, specially the parieto-occipital alpha band (Baumeister et al. 2010). While only two of the subjects declared that the EEG apparatus used might have negatively interfered with their usual movement pattern, continual developments in technology will help improve future study designs. Regarding potential applied directions for youth athletes, obtaining accurate ball placement requires players entering into a relatively relaxed mental state which may result from: (i) anticipated programming (e.g. desired ball location defined in advance) coupled to target-focused eye behavior; (ii) avoidance of giving much weight to possible environmental distractors during preparatory phase and (iii) caution when collecting or attempts to gather visual information (feedback) available later in the kick cycle. Finally, when looking to attain fast ball velocity in kick attempts made from the edge of penalty area, (iv) volitional attentional control of the lower limb movement (i.e. ability to engage in a sustained concentration process) seems necessary during the impact phase of soccer kicking action.

Conclusion

The current study expands the notion that EEG has potential for use in forecasting sports performance, specifically in a complex manipulative task using the lower limbs in soccer. Here, we demonstrated that increased frontal theta power in the impact phase and occipital alpha in the preparatory period predict better instep kicking performance in youth players, inducing respectively faster ball velocity and lower radial error. Thus, the signalling at cortical level that may be determinant for the velocity and accuracy components of kicking varies notably in relation to the brain region and frequency band. These kicking components demands prominent and paradoxical central control mechanisms during the movement planning and execution (i.e. respectively more automaticity and sustained top-down attention), highlighting the complex nature of ball kicking aiming at a far target. While insights on central inputs that are likely acting in controlling kick velocity, can be derived from the mutual relationships observed between EEG, ankle kinematics and subsequent ball velocity, the central mechanisms responsible for motor responses determining the effective ball placement of the ball in the goal when kicking remain unclear.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

This study was funded by the São Paulo Research Foundation (FAPESP) under fellowship and grant numbers #2018/02965-7 (doctorate - Luiz Henrique Palucci Vieira), #2020/04282-4 (scientific initiation - João Pedro da Silva) and #2017/19516-8 (regular - Fabio Augusto Barbieri) and in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance code [001].

Data availability

The results reported are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The raw data/spreadsheets used in current paper has been uploaded and made publicly available at Open Science Framework research workflow (https://osf.io/4jdqg/).

Declarations

Conflict of interest

The authors state no potential conflicts of interest in relation to the content of the present study.