TMS motor mapping was performed while the children were seated on their parent's lap (see Figure 1A for example). The children were allowed to play with toys or to watch TV during the study. The motor evoked potentials (MEPs) elicited by TMS were recorded by surface electromyography (EMG) from bilateral adductor pollicis brevis (APB), brachioradialis, and tibialis anterior (TA) muscles using disposable electrodes (Neuroline 720, Ambu Inc., Maryland, USA) and sampled at 3 kHz and band-pass filtered from 10 to 500 Hz. In each hemisphere, the mapping procedure began with application of TMS at an intensity set at 100% of the stimulator output (SO) around the middle part of the precentral gyrus (i.e., the hand knob area). The starting TMS intensity was set at 100% SO to compensate for the decreased efficacy of the standard figure-of-eight TMS coil in infants and children due to their immature motor system, brain size, and brain tissue conductivity (36, 37) as well as to reduce testing time due to the limited cooperation expected in this cohort. In children with difficulty tolerating this stimulation intensity, SO was decreased gradually until adequate tolerability was achieved. At this point, if no MEPs or CSPs were elicited, the mapping session was ended. TMS stimulation was applied as one pulse at each location with stimulation repeated as needed to cover surrounding cortex, including the precentral and postcentral sulci. The leg motor cortex was mapped by applying TMS along the paracentral lobule and posterior medial frontal gyrus. As the patients could not maintain a true baseline or the incidence of MEP was variable, motor threshold determination (resting or active) was not attempted. The TMS time-locked EMG epochs were analyzed offline to determine the presence of MEP, and, when applicable, to calculate its latency and peak-to-peak amplitude. Since the patients could not maintain relaxed muscles and had ongoing muscle contractions during TMS stimulation, we also examined the EMG recordings for any interruption of this voluntary activity following TMS, i.e., the cortical silent period (CSP) (38). CSP has proven to be a useful diagnostic biomarker in many neurological disorders including epilepsy in adults (39–42). We and others have previously shown that CSP can also be used to localize the motor cortex in individuals including young children in whom the SO required to elicit an MEP is at or near 100% (12, 20, 43).

TMS was used to localize the language-specific cortex by employing the “virtual lesion” paradigm (44) in 13 pre-school children between the ages of 5 and 6 years. Twelve of the 13 children also completed motor mapping prior to language mapping. The color-naming task was used in eight children (see Figure 1B for example) and an object-naming task was used in five children. We have previously shown that young and developmentally delayed children who could not consistently name objects could still name colors accurately and have successfully used the color-naming task to successfully map speech and language in this cohort (32, 45). In other children, pictures of objects included in the NexSpeech module was used (46). Since children were able to name common objects or colors used in this study, we did not individualize the stimuli based on their linguistic abilities.

The participants were seated in a chair and viewed the stimuli on a monitor. Colors (or objects) were displayed for 1,000 ms with an interstimulus interval adjusted according to the individual participant's ability ranging from 3.5 to 5 s. Patients were asked to correctly name the colors or the drawings of common objects as quickly as possible. Colors/pictures erroneously named were removed from the stimulus pool, so that all stimuli presented during TMS had a corresponding correct baseline recording. The TMS SO was adjusted to deliver an E-field of 80–100 V/m at the cortical surface based on previous reports (47, 48) and our own experience in older children and adults (45, 49). Patients were continuously monitored visually and by electromyography for signs of the intracortical spread of excitation or seizures (50). Discomfort during TMS was evaluated using the visual analog scale (VAS) for pain, recorded at baseline, four times throughout mapping, and whenever a patient spontaneously expressed discomfort. The intensity was decreased if participants rated pain ≥3 on the VAS but not below an E-field of 50 V/m. The mapping was discontinued if the pain persisted, and E-field fell below 50 V/m. Stimulus presentation and TMS onset were simultaneous with no delay to include early cortical activity (51) and decrease false negative results (52). The TMS train frequency was set at 5 Hz (five pulses). For mapping language cortices, TMS was applied from the supramarginal and angular gyri and extending to the superior, middle, and inferior temporal gyri, as anteriorly as the patient could tolerate. Then stimulation of the middle and inferior frontal gyri, including the pars opercularis and pars triangularis, as well as premotor regions was attempted. Baseline performance, stimuli presented, participant's response, and cortical locations of TMS were recorded for post-hoc analysis. Across compliance levels, the entire language mapping procedure was completed in ~1 h. In each participant, the hemisphere to be stimulated first was decided by prioritizing clinical need and therefore, the lesional hemisphere was mapped first.

Motor mapping recordings of the EMG were reviewed, and cortical locations where MEPs were observed on the EMG were identified by comparing it with the patient's baseline EMG data, as well as with previously established criteria for identification of cortical motor cortex responses (50). For each MEP, peak-to-peak amplitude and corticomotor latency (time from TMS stimulation to MEP onset) were initially estimated using the automated algorithm included in Nexstim software and verified by visual inspection. The MEP amplitude was used as a threshold to determine whether TMS elicited a response or not. Since the EMG electrode positions were variable between children and the baseline EMG also varied greatly, from being quiet (asleep or paresis) to vigorously active (holding a toy, being agitated, etc.), the MEP amplitudes were not further analyzed. When present, the onset and offset of CSP were calculated by visual inspection. When CSP was observed following an MEP, the MEP onset was also considered to be the CSP onset. When isolated CSP was observed, its onset was determined to be the point when EMG became quiet, and its offset as when the EMG activity returned to baseline pre-TMS levels. Similar to MEPs, the presence of CSP was used to localize response to TMS, and no further analysis was carried out. Examples of TMS-elicited MEPs and CSPs in APB, brachioradialis, and TA muscles are shown in Figure 2. Cortical locations of identified MEPs and CSPs were then marked on the patient's MRI.

Videos of patient performance during TMS language mapping were reviewed and potential speech and language errors were compared with the corresponding baseline response for the same item. Observed errors were independently categorized using the recommended criteria (9, 53) by two authors blinded to the site of stimulation. Any disagreement in categorization was resolved by consensus. The performance was coded as follows: speech arrest errors: when TMS resulted in an inability to produce any response; semantic errors: when a semantically related or associated word is substituted for the target word; and performance errors: form-based distortions, slurring, stuttering, imprecise articulation, or delayed response when compared to baseline recordings of the patient naming the same color or object. Speech errors attributed to discomfort or distraction were excluded from analysis. Further, speech errors resulting from the stimulation of primary mouth and laryngeal motor cortices or lips, jaw, and tongue muscles were also removed. The cortical location, type of error, and TMS intensity (%SO and E-field) for each TMS train were recorded. Similar to language mapping by fMRI (54, 55), we have previously shown that TMS-induced speech and language errors are also amenable to calculating a laterality index (LI) (45, 49). The LI was calculated as E_left-E_right/E_left+E_right, where E_left was the total number of speech and language errors in the left hemisphere, and E_right was the total number of speech and language errors in the right hemisphere. We also calculated LI by weighing the speech arrests (3x) and semantic errors (2x) more than the performance errors. Similar to LI thresholds used in fMRI (54, 55), TMS LI values 1.0–0.1 indicated left hemisphere dominance (HD), −1.0 to −0.1 indicated right HD, and values between −0.1 and +0.1 were considered a balanced bilateral representation of language.

The motor mapping cohort (n = 36 children, 47 total maps) consisted of 19 males and 17 females with a mean age (± SD) of 1.68 (± 0.8) years. The youngest child studied was a 2-month-old female, and 13 children were aged younger than 1 year. Nine children were reported to be right-handed, eight were left-handed, one was ambidextrous, two were not reported, and 16 were indeterminate due to their young age. Each child's handedness was determined via clinical evaluation by their attending neurologist. Patients were evaluated for refractory epilepsy (n = 32) or a brain tumor (n = 4). The most common etiologies for seizures were tuberous sclerosis complex type 2 (TSC-2, 22%), cortical dysplasia (22%), and perinatal stroke (17%). All children had an identifiable lesion on the MRI, with the lesion found in the left hemisphere in 33%, in the right hemisphere in 42%, and bilateral in 25% of the children. Thirty-two children were on AEDs (average 2.7 ± 1; range 1–5). The most common AEDs prescribed were levetiracetam (51% of children), clobazam (30%), oxcarbazepine (30%), and lacosamide (26%). The details of the demographic, clinical, and AED information of the motor mapping cohort are listed in Table 1.

Upper extremity (UE) primary motor cortex localization was attempted in both hemispheres in 40 sessions, in only the hemisphere of clinical interest in four, and in the only intact hemisphere in three. Six children were mapped twice (range 2–13 months apart), one child mapped three times (at ages 0.4, 1, and 2 y), and another child mapped four times (at ages 0.5, 0.6, 0.9, and 2.4 y). Motor mapping was considered successful if MEP or CSP was observed even for one stimulation. Three children had only one clear response in each hemisphere, but most often, five or more clear responses were observed (average number of responses = 23; 14% of stimulations). On average, 167.2 stimulations were delivered per session with a range of 26–535 single pulses. Using these criteria, the motor representation for APB and brachioradialis muscles were successfully localized in at least one hemisphere in 44 sessions (33 children) and in both hemispheres in 33 sessions (27 children). In two children, a 5-month-old with cortical dysplasia and a 7-month-old with a tumor, TMS did not elicit any MEPs or CSPs in APB or brachioradialis muscles despite extensive stimulation of both hemispheres at 100% SO. In the 5-month-old infant, the repeat motor mapping at 11 months of age was successful. Finally, the hand/forearm motor cortex could not be identified in one 10-month-old infant with non-lesional refractory epilepsy who could not tolerate increasing TMS intensity > 70% SO.

The primary leg motor cortex localization was attempted in both hemispheres in 18 children (11 males, seven females; average age ± SD: 2.1 ± 0.5 y). Two children were mapped twice. The youngest child studied was a 1-year-old male, and nine children were younger than 2 years. The motor representation of TA was successfully localized in both hemispheres in eight children, with only one hemisphere successfully mapped in seven children. In five children (four males; average age ± SD: 2.0 ± 0.8 y), despite extensive stimulation of both hemispheres at 100% SO, TMS did not elicit any MEPs or CSPs in the lower extremity muscles.

All mapping sessions were completed under nursing supervision. We were able to accurately apply TMS in all children as they were seated in their parent's lap. In all children, except for the one child noted above, the loud TMS clicks and the sensation of tapping at an intensity of 100% SO was well-tolerated without any serious adverse effects. We did not use earplugs in an attempt to minimize exposure to TMS clicks as the children could not keep them in place. The entire motor mapping session, including placing EMG electrodes, registering to the MRI, and surveying brain areas, was completed most often in 30–45 min. Five children (four males, age range 0.4–2.9 y) had seizures during the motor mapping procedure, and five (three males, age range 1.7–2.6 y) had seizures after the mapping procedure was completed, while being transferred to the wheelchair or transported to their inpatient room. In all children, the seizure semiology and duration during/following TMS were consistent with their clinical seizures observed at home and/or in the epilepsy monitoring unit, and all had a history of refractory epilepsy with frequent seizures. The seizures were either focal motor involving face and/or extremity twitching (n = 5) or generalized atonic seizures characterized by drop attacks (n = 5). In all children, seizures lasted <1 min (range 5 s−50 s) and did not require administration of rescue medication. In children who had seizures during mapping, the procedure was successfully completed, and meaningful data were derived.

Over 47 motor mapping sessions, an average of 167 ± 89 stimulations (range 26–532) with an intensity of 98 ± 8% SO was applied. The equivalent E-field in the UE motor cortex was 243 ± 97 V/m and in the leg motor cortex was 209 ± 73 V/m, both measured at a peeling depth of 17 ± 3 mm. There was no significant difference in the TMS E-field between the left and right hemispheres. The corticomotor latency from motor cortex to APB was 17.8 ± 3.1 ms; for brachioradialis, 16.6 ± 4.6 ms; for TA, 26 ± 4.5 ms. When isolated CSPs were observed, the latency for UE muscles was 34.1 ± 4.7 ms and for LE muscles 37.7 ± 6.7 ms. The details of the TMS parameters for upper and lower extremity mapping are listed in Table 2.

In the 33 children in whom motor mapping was successful, the UE primary motor cortex was localized to the central part of the precentral gyrus around the hand knob area (See Figure 3 for examples). In five of these children (one with TSC-2, one with infection, and three with cortical dysplasia), stimulation of both hemispheres resulted in occasional MEPs in both UE muscles, representing intact uncrossed pyramidal neurons, a normal variant in this age group (see Figure 3A). In four out of six children with perinatal stroke, an interhemispheric pattern of motor reorganization was demonstrated with bilateral UE motor representation in the intact hemisphere (see Figure 4B). In two out of eight children with cortical dysplasia, an intrahemispheric pattern of motor reorganization was demonstrated with the UE representation displaced toward the lower extremity motor cortex (see Figure 4A) or the premotor cortex. In 15 children in whom LE mapping was successful, the primary leg motor cortex was localized to the paracentral lobule in the medial frontal gyrus in one or both hemispheres.

Of the 36 patients who underwent TMS motor mapping, 20 children underwent somatosensory mapping with MEG, all under sedation. MEG was successful in only five children (four bilateral, and one in one hemisphere only). Twenty-five children underwent fMRI during passive hand movement, also under sedation. The sensorimotor cortex was successfully mapped in 20 children (18 bilateral, and two in one hemisphere only). These findings are consistent with our previous report in a smaller cohort of children under 3 years of age (20). When compared to TMS, which assesses motor cortices directly, MEG and fMRI under sedation primarily mapped the somatosensory cortex. Although the primary motor and sensory cortices are closely linked and findings of one modality can, to some extent, be generalized to the other, the comparison is not a direct one. Therefore, we did not use fMRI or MEG somatosensory maps as controls for TMS motor results. One child (2.2 y/o female with TSC-2, see Figure 5) underwent subdural grid placement to confirm the epileptogenic focus. In this child, the CSM- and TMS-localized motor cortices showed excellent overlap (Figure 5B). None of the other children underwent invasive mapping. We therefore attempt to demonstrate the utility and accuracy of TMS presurgical motor mapping in this cohort through its use in surgical planning and the post-operative results.

Of the children who underwent motor mapping only, many did not proceed to surgery. TMS mapping in these children provided the parents and physicians with information regarding the location of the child's motor cortices and also provided a baseline status of motor development. In children who returned for repeat motor mappings, TMS results were used to track motor function and development over time. Motor mapping was also used to assess the risks and benefits of surgical treatment in 11 children who had lesions in the vicinity of motor cortex. They included 2 children with brain tumor, 6 children with cortical dysplasia, 2 children with TSC-2, and one child with perinatal stroke. In all children who underwent surgery, the MRI with the TMS locations marked was transferred to the surgical navigation system, and the proximity of the anatomical lesion and/or the epileptiform focus to the primary motor cortex identified by TMS was estimated. The children were evaluated clinically at follow up and 7 children were found to have good motor function with no new motor deficits. The child with perinatal stroke, in whom motor representation in the lesioned hemisphere was shown to be reorganized to the intact hemisphere, demonstrated an improvement in preexisting hemiplegia following surgery. One child had slightly decreased movement in the contralateral UE with good muscle bulk, strength, and tone. Two children were found to have mild monoparesis at follow up; in one, the resected cortical dysplasia was within the primary leg motor cortex (Figure 4A), and in the other, the frontal lobectomy extended up to the precentral sulcus (Figure 2A).

The TMS language mapping cohort consisted of eight males and five females with a mean age (± SD) of 5.6 (± 0.3) years, with all children being between 5 and 6 years of age. Eight patients were right-handed, three left-handed, and two ambidextrous. The patients were being evaluated for tumor (n = 7) or refractory epilepsy (n = 6). The causes of refractory seizures were cortical dysplasia (n = 2), tuberous sclerosis complex type 2 (n = 1), perinatal stroke (n = 1), and hippocampal sclerosis (n = 1). The lesion was in the left hemisphere in seven children, in the right hemisphere in four children, bilateral in one child, and no detectable lesion on MRI in one child. Ten children were on AEDs (average 1.4 ± 0.5; range 1–2). The most common AED prescribed was levetiracetam (50%). The details of demographic, clinical, and AED information of the language mapping cohort is listed in Table 1.

The temporal and frontal lobes in both hemispheres were successfully mapped in six of 13 children. In five of the remaining children, bilateral temporal lobes were evaluated, but stimulation of frontal lobe language areas could not be tolerated. In one child, only the left hemisphere frontal lobe around the tumor was mapped. One child could not tolerate TMS stimulation and language mapping was discontinued. Twelve children also successfully completed motor mapping of the upper and lower extremities.

We were able to map at least part of the language areas in the two hemispheres in 92% of this cohort, with comprehensive language maps derived in 50% of children. The most common complaint was pain during stimulation. The TMS intensity was reduced when the pain reported on VAS was ≥3. Throughout the mapping procedure, we ensured that the pain score was below 3. No child had a seizure during or after completion of the TMS language study.

Eight children were mapped using the color-naming task and five using the object-naming task. The TMS parameters of rate and intensity used in this study were within the guidelines for safety (50). An average of 130 (± 52) trains of 5 Hz stimulation were applied. The TMS intensity was 34 ± 3 % SO (range 25–40% SO), equivalent to an E-field of 91 ± 14 V/m (range 58–107 V/m), measured at a peeling depth of 17 ± 3 mm. The hemisphere with the lesion was stimulated first (LH: 9; RH: 3). Mapping was considered successful if at least one convincing speech error was noted; however, the lowest number of errors noted in any child was six (both hemispheres included). On average, 13% of stimulations resulted in speech errors. Performance error (10 ± 7) was the most common type of speech error noted, followed by speech arrest (4 ± 5), and then semantic errors (2 ± 2). The average total number of errors in the left and right hemispheres were similar (LH: 8.1 ± 6.7; RH: 8.2 ± 7.8). The TMS intensities recorded were not significantly different for the three error types, indicating that the type of speech errors elicited by TMS were independent of TMS intensity. The details of the language mapping TMS parameters are listed in Table 2.

TMS-elicited speech disruptions were noted following stimulation of the middle and posterior parts of the superior and middle temporal gyri and the supramarginal gyrus in both hemispheres (see Figures 6, 7 for examples). Critical language areas identified in the frontal lobe included the ventral premotor cortex, pars opercularis, and pars triangularis (see Figures 6, 7). Across both hemispheres, language areas were primarily localized in the superior temporal lobe (83% of patients), the middle temporal gyrus (68% of patients), and the supramarginal gyrus (64% of patients). Language representation in the inferior frontal gyrus was identified in 94% of children in whom the frontal lobe was stimulated. Based on the localization of language cortices in the frontal and temporal lobes in the two hemispheres by TMS, the LI was estimated in 11 children. Two were deemed LH dominant, seven RH dominant, and two bilaterally dominant for language.

Of the 13 patients who had TMS language mapping, 10 underwent MEG receptive language mapping under sedation, of which only three were successful. Of the eight children who underwent fMRI during passive listening under sedation, language cortices in the temporal lobes were successfully localized in only three patients. Because the sedation required for MEG (27) and fMRI (56) in this age group often precludes successful and reliable mapping, and even when successful, mapping primarily consists of receptive language, and the comparison against expressive language maps derived by TMS is not a viable option. None of the children in this group underwent invasive mapping. However, we feel that the utility and accuracy of TMS presurgical mapping in this cohort is best shown by its use in surgical planning and the post-operative results, presented below.

Of the 13 children who underwent speech mapping with TMS, seven children underwent surgery to remove the tumor (n = 6) or tuber (n = 1). Due to the availability of TMS mapping, none underwent CSM. One child underwent placement of VNS, and the remaining five did not undergo surgery. Of those five, two declined surgery, two were experiencing adequate seizure control with medication management at that time, and one was not a surgical candidate due to non-localizable seizures. TMS mapping allowed neurosurgeons to demonstrate patient-specific functional areas and the proposed surgical approaches to patients' families, allowing both parties to accurately weigh the risks and benefits of proceeding with surgery at that time. Finally, for all children who underwent surgery, the TMS results provided a presurgical baseline of expressive language function which was used for comparison with post-operative language. In these children, the MRI with the TMS locations marked was transferred to the surgical navigation system, and the proximity of the anatomical lesion and/or the epileptiform focus to the language cortices identified by TMS was estimated. Post-operatively, the children were evaluated clinically, and none of the children were found to have speech or language deficits.