Transcranial direct current stimulation (tDCS) is a non-invasive technique capable of modulating cortical excitability in humans (Nitsche and Paulus, 2000). During conventional tDCS, a sub-threshold current is applied to the brain via two sponge electrodes placed on the scalp. Prolonged enhancement of primary motor cortex (M1) excitability following anodal tDCS has been evidenced by an increase in motor evoked potential (MEP) amplitude within hand muscles (Nitsche and Paulus, 2000). Anodal tDCS applied simultaneously with behavioural tasks facilitates enhanced motor performance over a single session (Nitsche et al., 2003; Boggio et al., 2006; Vines et al., 2006) or across multiple sessions in healthy adults (Reis et al., 2009) with effects outlasting the stimulation period (Nitsche and Paulus, 2001; Boggio et al., 2006; Reis et al., 2009; Matsuo et al., 2011; Sohn et al., 2012; Kidgell et al., 2013). tDCS shows promise for facilitating motor recovery post stroke (for review see, Kang et al., 2016) and a range of other neurological disorders (Lefaucheur, 2016). High-definition tDCS (HD-tDCS) offers improved focal targeting of cortical areas and stronger regional electric fields by centering four electrodes around a central electrode of opposite polarity in a 4 × 1 ring-like orientation (Datta et al., 2009; Dmochowski et al., 2011; Villamar et al., 2013; Alam et al., 2016). Behavioural improvements occur in healthy adults over single (Doppelmayr et al., 2016) and multiple sessions (Pixa et al., 2017) in which patterns of cortical excitability may outlast those induced by conventional tDCS (Kuo et al., 2013).

Children have been neglected in neuromodulation research. With a well-defined safety and tolerability profile (Bikson et al., 2016) in adults and clear clinical need, early translation towards tDCS applications in pediatric populations seem favourable. However, limited evidence suggests primary principles of tDCS, such as polarity and current strength, may differ in the developing brain (Moliadze et al., 2015). Preliminary investigations in pediatric populations highlight potentially varying effects of tDCS in children (Kessler et al., 2013; Rajapakse and Kirton, 2013; Bikson et al., 2016). Our team recently demonstrated that M1-tDCS over three consecutive days enhances motor learning in healthy children with effects retained at 6-weeks (Ciechanski and Kirton, 2017). We have also demonstrated that both anodal tDCS and HD-tDCS can enhance motor learning over multiple days with retained effects (Cole et al., 2018). However, the underlying mechanisms of such interventional neuroplasticity remain unknown.

Transcranial magnetic stimulation (TMS) can produce maps of cortical representations of individual muscles within M1 (Cohen and Hallett, 1988; Wassermann et al., 1992; Wilson et al., 1993; Pascual-Leone et al., 1995b; Thickbroom et al., 1998) and is increasingly applied for clinical applications such as preoperative neurosurgical planning for brain tumor removal and epilepsy surgery (Picht et al., 2011; Lefaucheur and Picht, 2016), for review see Sollmann et al. (2021). Common characteristics, such as map volume, area, hotspot magnitude, and centre of gravity (COG), may also potentially quantify changes in cortical neurophysiology following learning or stimulation-induced changes in motor performance (Cohen et al., 1993; Pascual-Leone et al., 1995a, b). For instance, an increase in motor map size following the acquisition of new fine motor skills in healthy adults (Pascual-Leone et al., 1995a) resembles cortical muscle rearrangement following behavioural motor training in primates (Donoghue et al., 1992; Recanzone et al., 1992; Nudo et al., 1996). TMS motor maps, including robotic neuronavigated methods, have also been acquired in the developing brain of healthy and clinical populations (Maegaki et al., 1999; Garvey et al., 2003; Garvey and Gilbert, 2004; Friel et al., 2017; Grab et al., 2018; Giuffre et al., 2019b, 2021). Whether such advanced mapping methods can detect interventional changes in motor cortex neurophysiology during modulated motor learning in children has not been studied. Here, we aimed to determine the potential effects of tDCS and HD-tDCS enhanced motor learning on robotic TMS motor maps. Based on preliminary evidence from adult studies (Pascual-Leone et al., 1995a; Classen et al., 1998), we hypothesised that map volume and area of the trained first dorsal interosseous muscle (FDI) would increase following active stimulation (tDCS and HD-tDCS) and be associated with enhanced motor performance. We also investigated the effects of stimulation and tDCS-enhanced motor performance on additional motor map outcomes, hotspot magnitude (MEP amplitude), and COG of both hands.

We investigated the effects of natural and modulated (tDCS and HD-tDCS) motor learning on robotic TMS motor maps. There were no significant effects of time or group on FDI motor map volume, area, or hotspot magnitude of the left-hand (trained) or right-hand (untrained), or within any of the secondary muscles. We did observe interactions between time and group on COG-x in muscles of the left-hand over time. Map metrics did not correlate with behavioural changes. Our results suggest robotic TMS motor mapping is a potentially valuable tool for studying plasticity in the developing motor system, but multiple factors must be considered, most notably the large variability in maps across individuals and time (Ngomo et al., 2012).

Primary motor cortex plays a pivotal role during skilled voluntary movements and learning of motor skills (Cohen et al., 1993; Rossi et al., 1999; Sanes and Donoghue, 2000; Muellbacher et al., 2001; Rioult-Pedotti and Donoghue, 2003). TMS has advanced our knowledge of such M1 plasticity, revealing changes in corticospinal excitability and cortical reorganization in response to upper-limb behavioural training (Pascual-Leone et al., 1995b, a; Classen et al., 1998; Butefisch et al., 2000; Ziemann et al., 2001; Gallasch et al., 2009). However, how motor mapping can further inform these mechanisms remains poorly understood. Robotic TMS may offer additional potential in the study of M1 neurophysiology and plasticity and may be particularly applicable in pediatric populations for its improved accuracy, movement-compensation, and increasingly well-established safety and tolerability profile (Zewdie et al., 2020). We recently demonstrated that robotic TMS can safely quantify M1 neurophysiology in healthy children (Grab et al., 2018) and investigated motor map outcomes and their associations with motor performance (Giuffre et al., 2021).

The present study is the first to demonstrate the application of such robotic TMS motor map outcomes to explore the underlying M1-plasticity effects of multiple days of modulated learning in healthy children (Cole et al., 2018). Our findings suggest COG may be a valuable neurophysiological motor mapping outcome. Following five consecutive days of intervention, the weighted cortical motor map of the trained LFDI appeared to shift differently between those participants receiving active tDCS compared to Sham. The significance of this finding is supported by our reliability study of robotic TMS motor mapping measures over both short- and long-term timeframes in young adults (Giuffre et al., 2020). Our reported standard error of COG-x (0.365) was below our previously reported standard error of measurement for COG-x (0.40). Such observations of COG-x being more reliable than other map outcomes is consistent with other reports (Wassermann et al., 1992; Mortifee et al., 1994; Jones–Lush et al., 2010; Ngomo et al., 2012; Littmann et al., 2013). COG may reflect changes in cortical physiology of upper-limb muscles following behavioural training or tDCS-modulated motor skill learning. Medial-lateral shifts in COG have been previously reported to shift medially with age in preadolescents (Säisänen et al., 2021) following one-hour of motor training (Liepert et al., 1999; Tegenthoff et al., 2004) and suggested to reflect shifts in motor cortex excitability (Wassermann, 2002) and reorganization following motor learning (Liepert et al., 2000). As M1 is distributed in a medial-lateral plane, shifts along COG-x potentially identify scalp coordinates corresponding to a larger number of the most excitable corticospinal neurons influencing recorded muscle activity (Röricht et al., 2001). Other reports suggest COG may be a more precise measure of anatomical reorganization assessable by TMS, although additional mapping outcomes such as map volume, area, and hotspot magnitude are more commonly investigated.

Accordingly, our primary map outcome measures were area, volume, and hotspot magnitude. We originally hypothesised that map volume and area would increase following learning and stimulation, based in part on evidence of increased cortical excitability (MEP amplitude) following anodal tDCS in healthy adults and children (Nitsche and Paulus, 2000; Nitsche et al., 2003; Ciechanski and Kirton, 2017; Cole et al., 2018) and enlarged motor map representations following motor performance alone in both animal (Nudo et al., 1996) and human models (Pascual-Leone et al., 1995a). Since map volume quantifies MEP amplitude at responsive grid-points, taking map area into account, we predicted the effects of tDCS stimulation might unveil different patterns of change within these M1 representations of hand muscles. Both measures have the potential to be complementary metrics capable of quantifying M1-plasticity effects such as strengthened synaptic connections and unmasking of latent horizontal connections (Huntley, 1997; Sanes and Donoghue, 2000). Contrary to our hypothesis, however, map volume, area, and cortical excitability showed no significant effect of time or group. Since executing the current study, we have learned that inter-session reliability of these outcomes is moderate with large minimal detectable differences estimated at more than 40–50% (Giuffre et al., 2020). Accordingly, with our modest sample divided into three treatment groups, we were likely underpowered to show significant changes in these standard map metrics. Additional reasons behind the absence of findings may stem from diverse individual variability in tDCS-induced changes in excitability, individual differences in motor skill learning, or differences in tDCS current density among participants (Ciechanski et al., 2018).

Increased map volume and area have been associated with repeated motor task learning (Pearce et al., 2000) and acquiring motor skills (Pascual-Leone et al., 1995b; Reis and Fritsch, 2011). In healthy subjects, MEP amplitude and cortical representations of fingers (map size) increased following repetitive practice of a piano sequence (Pascual-Leone et al., 1995a), while showing no change following repeated PPT performance (Garry et al., 2004) or during non-specific motor training (Ngomo et al., 2012). It is postulated that while behaviour improves, cortical output maps to the muscles involved become progressively larger until explicit knowledge is achieved, after which they may reduce in size (Pascual-Leone et al., 1994). For instance, while many TMS studies reported behavioural training increases cortical excitability (MEP amplitude), suggesting a correlation with functional outcome (Pascual-Leone et al., 1995a; Muellbacher et al., 2001; Ziemann et al., 2001; Garry et al., 2004), others have reported no change (Carroll et al., 2002; Ngomo et al., 2012). Inconsistent findings are also found in motor map outcomes, such as volume and area, following behavioural training though these have been studied almost exclusively in adults (Brasil-Neto et al., 1992; Wilson et al., 1993; Mortifee et al., 1994; Thickbroom et al., 1998; Malcolm et al., 2006). These may represent potential reasons for our results showing no definitive changes in motor map volume and area following motor training. That stimulation with tDCS or HD-tDCS might also alter the natural direction of effects of motor learning on motor maps must also be considered.

More expansive representations of a trained sequence of movements in M1 have been associated with performance, suggesting specific representations of movement sequences may be implemented at the cortical level as a new functional unit and distributions of cortical sites may be functionally linked during coordinated movements (Tyč and Boyadjian, 2011). As we demonstrated in the behavioural arm of the interventional trial on which the current study is based, days 4 and 5 showed the most significant increases of tDCS-induced effects on motor performance (Cole et al., 2018). We observed similar results in an earlier trial of tDCS-enhanced motor learning in children (Ciechanski and Kirton, 2017). The effect of time (day of training) appears to be dynamic and possibly altered by stimulation, adding yet another source of variance between individuals to our measurements which were limited to 3 timepoints. Acquiring TMS mapping measures on each day of such interventional trials could shed light on such shorter-term, day-to-day changes in cortical representations (map volume or area) during the course of stimulation-induced motor learning but may be limited by resource utilization and participant compliance.

To date, there is no gold standard for TMS motor mapping that provides the highest accuracy of measurements. Mäki and Ilmoniemi (2010) showed variability of MEP amplitudes at the same cortical grid-point, one-third of the largest responses were on average 10 times higher than one-third of the lowest responses (Mäki and Ilmoniemi, 2010). Estimating the size of motor representations based on MEP amplitudes obtained within a finite number of stimulation points is challenging (Julkunen, 2014) and depends on the accuracy of stimulation, density of stimulus locations, and MEP variability at the borders of the maps (Brasil-Neto et al., 1992; Classen et al., 1998; Chernyavskiy et al., 2019). Other groups have used a greater number of stimuli (6–10 pulses) per grid-point to accommodate (Cirillo et al., 2010; Ngomo et al., 2012) and reduce variability (Bastani and Jaberzadeh, 2012; van de Ruit et al., 2015), however, the feasibility of this approach in the developing brain is questionable. Additional recent investigations using neuronavigated single biphasic TMS pulses suggest increased efficacy to induce motor responses (Pitkänen et al., 2018), which may accommodate motor mapping challenges in young children, such as high motor thresholds (Julkunen, 2014; van de Ruit et al., 2015), though investigations in pediatric populations are lacking. Robotic neuronavigated TMS may be a useful tool to overcome these and other challenges in children. In addition to reducing human error of accurate coil placement, neuronavigated robotic TMS reduces acquisition time, provides consistent coil positioning, and near real-time motion correction, accommodating subject movement (Ginhoux et al., 2013). Robotic motor mapping appears safe, feasible in children (Grab et al., 2018; Giuffre et al., 2021), and is well-supported by growing evidence of TMS safety in children (Zewdie et al., 2020).

An ability to measure motor map plasticity has translational implications for children with early brain injury and cerebral palsy, the leading cause of lifelong neurological disability. The sensorimotor network is one of the earliest developing networks, well-established very early in life but highly refined throughout of development where motor maps may undergo dramatic developmental plasticity (Khazipov and Milh, 2018). How these natural processes are altered after early unilateral injury such as perinatal stroke are increasingly understood from preclinical models (Martin et al., 2007; Wen et al., 2018) and human imaging and brain mapping studies (Kirton et al., 2021). A recent mouse model of perinatal stroke evaluated motor map size and movement latency following cortical stimulation and motor training (Zhang et al., 2021). In line with our findings, motor map area (size) showed no overall expansion following 10 min a day over three-weeks of pellet training. Interestingly, mice with smaller map areas at baseline showed the greatest improvement in skilled forelimb training. TMS motor maps appear to be altered in children with unilateral cerebral palsy (CP) undergoing therapeutic interventions. For example, Friel et al. (2016) has shown increased motor map area (size) and MEP amplitudes of the affected hand in children with unilateral spastic CP following 3-weeks (6 h/day, 5 days/week) of bimanual therapy (Friel et al., 2016). Additional investigations by their group have shown medial-lateral shifts in FDI COG, in addition to investigating size and excitability of cortical motor maps, may lead toward further defining the relationship between changes in motor maps and motor function in children with CP (Kesar et al., 2012; Marneweck et al., 2018). Such models are informing large-scale randomised neuromodulation clinical trials (Kirton et al., 2016, 2017) where the addition of robotic TMS mapping with comprehensive outcomes such as COG before and after modulation may shed further light on mechanisms of interventional plasticity.

In addition to the above limitations, additional challenges are acknowledged. Our study could not control for other factors that may influence map size, such as the influence of hand use throughout the intervention (Ngomo et al., 2012). For instance, musicians and racket ballplayers (Pearce et al., 2000) who utilise their hand muscles more than the average person display different neurophysiology and organization of cortical motor networks. Our study was based on a secondary aim of the primary interventional tDCS trial, powered to determine the behavioural effects of tDCS on motor learning in children. Combined with new knowledge of TMS mapping reliability (Giuffre et al., 2020), it is clear that we had very modest power to detect true changes in map parameters. This emphasises the need for careful power calculations in future motor mapping studies. Additional limitations are due to the potential differences in electric field strength and distribution in our pediatric sample. Advances in computer finite element modeling (FEM) and realistic 3D head models have enabled tDCS-induced current to be predicted and modeled through the cortex (Miranda et al., 2006; Wagner et al., 2006; Bikson et al., 2012). The development of high-resolution derived head models allows for more accurate and precise modeling of current, though special considerations are required when applying tDCS in children. An early current modeling case study suggested tDCS-induced electric fields may be stronger in children (Kessler et al., 2013). Supplementing these findings in the largest pediatric sample to date, our group recently demonstrated children incur increased current densities and distributions compared to adolescents and adults (Ciechanski et al., 2018). These differences may be attributed to age-related differences in developmental changes within grey and white matter and skull thickness. As children have thinner skulls, stronger electric fields may be induced, resulting in more expansive current in underlying tissue (Opitz et al., 2015; Ciechanski et al., 2018). Recently, HD-tDCS was proposed to offer more focal current delivery and optimised targeting (Datta et al., 2009), however, investigations of modeling HD-tDCS current distributions in pediatrics are lacking. In a preliminary investigation, we studied potential differences in electric field modeling of conventional tDCS and HD-tDCS in the pediatric sample described in this study. We observed lower peak electric fields in HD-tDCS compared to tDCS and found no significant associations between electric fields and motor performance of the two stimulation groups (Giuffre et al., 2019a). As a larger sample of participants is needed to determine differences in electric field distribution of tDCS and HD-tDCS, we cannot account for possible differential effects of tDCS on motor maps. Additional investigations of current modeling various tDCS montages, especially in pediatric and clinical populations, are needed to optimise tDCS enhancement of motor function and advance therapies for clinical populations.

Patterns of motor map volume and area within participants revealed large variability that likely limited our ability to detect differences. Variability can be attributed to both subject factors (age, genetics, and sleep) (Antal et al., 2010; Li Voti et al., 2011; Ngomo et al., 2012), as well as measurement error. Variable subthreshold activation of corticospinal outputs at rest, background EMG activity, and focality of TMS-induced electric currents are all considered and can only partially be controlled. These factors may have contributed to our findings in secondary hand muscles. As different muscle representations display unique pyramidal and interneuronal orientations, the TMS coil may differentially active unique muscle representations. Although many horizontally oriented neuronal elements are perpendicular to the central sulcus, the depth from the scalp at which these elements initiate MEP response to TMS are difficult to evaluate and likely differs in children. An additional limitation may relate to our use of the LFDI RMT to determine the mapping threshold across all muscles. Although this may provide relative cortical representations of additional hand muscles and is highly practical, it cannot account for differences in thresholds, orientation, and other specific factors unique to each muscle representation that might only be determined with each mapped individually.

In summary, robotic TMS motor mapping is feasible and well-tolerated in children. The maps generated can estimate neurophysiological measures relevant to M1 physiology and plasticity. Large effects on traditional mapping outcomes such as area and volume were not observed but shifts in COG may represent an informative measure. Fully powered TMS motor mapping investigations are needed in the developing brain to determine utility in understanding mechanisms of motor developmental and interventional plasticity.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by University of Calgary Research Ethics Board (REB16-2474). Written informed consent and assent when applicable to participate in this study was provided by the participants’ legal guardian/next of kin.

AG: conceptualization, methodology, formal analysis, ethics, recruiting, investigation, data collection, supervision, and writing—original draft. EZ, JW, and LC: conceptualization, methodology, formal analysis, ethics, recruiting, investigation, data collection, and writing—manuscript and editing. HC: conceptualization, methodology, formal analysis, investigation, data collection, and writing—manuscript and editing. H-CK: conceptualization, methodology, formal analysis, recruiting, investigation, data collection, and writing—manuscript and editing. AB: formal analysis, investigation, and writing—manuscript and editing. AK: conceptualization, methodology, ethics, investigation, writing—original draft, and supervision. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.