*2.3. Improvement of Clinical Outcome*

When added to the armamentarium of the preoperative workup of patients harboring brain neoplasms, the question arises whether nTMS motor mapping and derived nTMSbased tractography may be capable of improving the clinical outcome, as measured by an ideally increased EOR combined with lowered rates of functional perioperative decline. In this regard, Table 3 gives an overview of studies focused on clinical outcome.

An initial study compared 11 patients who underwent preoperative nTMS to 11 patients without nTMS motor mapping, revealing that preoperative nTMS motor mapping changed the treatment plan towards early and more extensive resection in 6 out of 11 patients [57]. Furthermore, one of four patients of the nTMS group with preoperative motor deficits improved by one year, whereas increased deficits were observed in three of the eight patients of the historical group not having surgery [57]. In two retrospective studies in considerably large cohorts of patients with different types of brain tumors, the utility of nTMS motor mapping and nTMS-based tractography of the CST may have facilitated a more extensive EOR, combined with tendencies towards better motor function after surgery [58,59].

Specifically, nTMS disproved suspected involvement of the primary motor cortex in 25.1% of the 250 enrolled patients, and it enabled expanding surgical indication in 14.8%, thus facilitating planning of more extensive resections in 35.2% of patients [58]. Furthermore, the distinct add-on value of nTMS-based tractography of the CST has been evaluated in a study including 70 adult patients with different brain tumor entities, revealing that patients having nTMS-based tractography available are characterized by an improved risk-benefit profile, showed an increased EOR, and demonstrated reduced rates of worsening in motor function in cases of already preexisting preoperative motor deficits [60].

In a follow-up study investigating the role of nTMS motor mapping and nTMS-based tractography in 70 patients presenting with high-grade glioma (HGG), residual tumor tissue and unexpected tumor residuals were less frequent in the nTMS group compared with historical control patients, with patients of the nTMS group being more frequently eligible for postoperative radiotherapy and showing prolonged 3-, 6-, and 9-month survival rates [61]. A significantly higher EOR was subsequently confirmed by another study for patients with a diagnosis of glioblastoma multiforme (GBM), in which patients of the nTMS group showed a gross total resection (GTR) rate of 61% versus 45% for the non-nTMS group [62]. Analogously, in studies focusing on patients with intracranial MET, patients of the nTMS group showed a lower rate of residuals combined with comparatively low rates of perioperative decline of motor function [63,64]. Specifically, in a retrospective comparative study pooling patients with intracranial MET from three different neurosurgical centers, surgery-related paresis was clearly less frequent in patients of the nTMS group [64]. In 47 patients with MEN located in the rolandic area, nTMS motor mapping and tractography facilitated a modification of the surgical strategy in 42.5% of cases, and a new permanent motor deficit (i.e., deficit that did not resolve to the preoperative status within the follow-up interval) occurred in 8.5% of cases, which is at the lower edge of the range for motor deficits known from the literature of the pre-nTMS era [65]. Furthermore, the combination of sodium fluorescein-guided resection (FGR) with preoperative nTMS motor mapping and tractography has been explored recently, revealing a higher GTR rate for patients operated on using nTMS and FGR as well as lower rates of new surgery-related permanent motor deficits when compared with controls [66,67].

One study in 43 patients with LGG and HGG showed that 72% of patients had motorpositive nTMS points in areas frontal of the rolandic area and, thus, outside of the expected spatial dimensions of the primary motor cortex [68]. Interestingly, 10 of the 13 patients who underwent resection of motor-positive nTMS points presented with postoperative paresis (8 patients with a new permanent surgery-related paresis), suggesting that even motorpositive nTMS points within the superior or middle frontal gyrus should be considered carefully for resection planning and guidance to avoid perioperative functional decline [68]. Hence, nTMS motor mapping and derived tractography may help to understand individual functional anatomy, allowing for optimized resection that provides a high EOR and low rates of surgically induced motor function decline.


**Table 3.** Improvement of clinical outcome. This table outlines the studies published on improvements of clinical outcome through the use of motor mapping by nTMS (using an electric-field-navigated system) with or without additional nTMS-based tractography in patients harboring brain neoplasms.


**Table3.***Cont.*

 **Author Year Cohort Group for Comparison Tumor Entities nTMS Method Outcome Parameters Main Objective Main Findings** Hendrix et al. [63] 2016 61 adult patients - LGG, HGG, MET, MEN, Other Preoperative motor mapping (110% rMT) − Influence on surgery − EOR/tumor volume − Motor function (BMRC) To assess the impact of nTMS on the treatment strategy and clinical outcome − Paresis resolved or improved in 56.7% of patients one week after surgery, and 89.5% of patients with postoperative paresis improved during the follow-up interval. − Only 4.3% of patients with a metastatic lesion, but 26.3% of patients with a non-metastatic lesion experienced deterioration of motor function after surgery. − All metastatic lesions were completely resected compared with 78.9% of non-metastatic lesions. Krieg et al. [64] 2016 120 adult patients 130 historical controls MET Preoperative motor mapping (110% rMT for uE and 130% for lE muscles) and nTMS-based tractography − Influence on surgery − EOR/tumor volume − Motor function (BMRC) To assess the impact of nTMS on the treatment strategy and clinical outcome (multi-centric with three sites) − Patients of the nTMS group showed significantly smaller craniotomies. − Patients of the nTMS group showed a lower rate of residual tumor tissue after surgery (odds ratio: 0.3025, 95% confidence interval: 0.1356–0.6749). − Surgery-related paresis was significantly less frequent in patients of the nTMS group (nTMS group: improved: 30.8%, unchanged: 65.8%, worse: 3.4%, non-nTMS group: improved: 13.1%, unchanged: 73.8%, worse: 13.1% of patients). Moser et al. [68] 2017 43 adult patients - LGG, HGG Preoperative motor mapping (110% rMT for uE and 130% for lE muscles) − Latency analyses − Motor function (BMRC) To assess the impact of resection of motor-positive prerolandic nTMS points on clinical outcome − 72% of patients showed motor-positive nTMS points in the prerolandic gyri and, thus, outside of the anatomically suspected extent of the primary motor cortex. − Out of the 13 patients who underwent resection of motor-positive nTMS points, 10 patients showed postoperative paresis (2 patients with transient and 8 patients with permanent surgery-related paresis). − Motor-positive nTMS points within the superior or middle frontal gyrus should be considered carefully and can result in motor deficits when affected during resection.

**Table3.***Cont.*

 **Author Year Cohort Group for Comparison Tumor Entities nTMS Method Outcome Parameters Main Objective Main Findings** Raffa et al. [60] 2018 70 adult patients (50% also having nTMS-based fiber tracking) 35 historical controls LGG, HGG, MET, Other Preoperative motor mapping (120% rMT) and nTMS-based tractography − Influence on surgery − EOR/tumor volume − Motor function (BMRC) − KPS To assess the impact of nTMS with or without nTMS-based tractography on the treatment strategy and clinical outcome − Patients of the nTMS and nTMS + nTMS-based tractography groups received significantly smaller craniotomies and had better postoperative motor performance and KPS scores than patients of the non-nTMS group. − Patients of the nTMS-based tractography group exhibited an improved risk-benefit analysis, a significantly increased EOR in absence of preoperative motor deficits, and significantly less motor and KPS score worsening (in case of preoperative motor deficits when compared with the nTMS group). − Risk-benefit analysis, EOR, and outcome could be improved when nTMS-based tractography is added to nTMS motor mapping. Raffa et al. [66] 2019 79 adult patients 55 historical controls HGG Preoperative motor mapping (120% rMT) and nTMS-based tractography − EOR/tumor volume − Motor function (BMRC) To assess the impact of nTMS and nTMS-based tractography with sodium-fluorescein guidance on the treatment strategy and clinical outcome − In patients operated on considering nTMS + FGR, the GTR rate was significantly higher compared with controls (64.5% vs. 47.2%). − Surgery-related permanent motor deficits were reduced in the nTMS + FGR group compared with controls (11.4% vs. 20%). Raffa et al. [67] 2019 41 adult patients 41 historical controls HGG Preoperative motor mapping (120% rMT) and nTMS-based tractography − Influence on surgery − EOR/tumor volume − Motor function (BMRC) − KPS To assess the impact of nTMS and nTMS-based tractography with sodiumfluorescein guidance on the treatment strategy and clinical outcome − Use of nTMS motor mapping and nTMS-based tractography reliably identified the spatial tumor-to-function relationship with an accuracy of 92.7%. − Patients of the nTMS group showed an increased EOR and higher rate of GTR (73.2% vs. 51.2%). − The number of cases with new surgery-related permanent motor deficits was lower in the nTMS group compared with controls (9.8% vs. 29.3%). − The number of cases with KPS worsening was lower in the nTMS group compared with controls (12.2% vs. 31.7%).

**Table3.***Cont.*

 **Author Year Cohort Group for Comparison Tumor Entities nTMS Method Outcome Parameters Main Objective Main Findings** Raffa et al. [65] 2019 47 adult patients - MEN Preoperative motor mapping and nTMS-based tractography − Influence on surgery − EOR/tumor volume − Motor function (BMRC) − Arachnoidal cleavage plane To analyze the role of nTMS motor mapping for planning resection of rolandic meningiomas and predicting arachnoidal cleavage plane. − Use of nTMS motor mapping and nTMS-based tractography was considered useful in 89.3% of patients and changed the surgical strategy in 42.5% of patients. − A new permanent motor deficit occurred in 8.5% patients. − A higher rMT and the lack of an intraoperative arachnoidal cleavage plane were independent predictors of poor motor function outcome. − A higher rMT and perilesional edema predicted the lack of an arachnoidal cleavage plane.

Abbreviations: nTMS—navigated transcranial magnetic stimulation; LGG—low-grade glioma; HGG—high-grade glioma; MET—metastasis; MEN—meningioma; rMT—resting motor threshold; uE—upper extremity; lE—lower extremity; EOR—extent of resection; GTR—gross total resection; BMRC—British Medical Research Council; KPS—Karnofsky performance status; PFS—progression-free survival; OS—overall survival; FGR—fluorescein-guided resection; GBM—glioblastoma multiforme.

**Table3.***Cont.*

#### *2.4. Risk Stratification and Prediction*

Besides the role for preoperative planning and intraoperative resection guidance, nTMS motor mapping and tractography could also be efficiently used for risk stratification and prediction of the motor status in patients with brain neoplasms. This has already been acknowledged by a growing body of studies, which are summarized in Table 4.

An early study characterized the neurophysiological status as derived from nTMS motor mapping in 100 patients, and already suggested that interhemispheric differences for MEP latencies may be considered as potential warning signs for the motor system at risk as comparatively similar latencies are commonly observed between the two hemispheres [69]. On a similar note, a high interhemispheric rMT ratio (i.e., the ratio between the two hemispheres regarding the rMT, which is commonly higher in a tumor-affected hemisphere) could suggest immanent deterioration of the functional motor status [69]. Furthermore, two studies in patients with various tumor entities investigated the role of nTMS-based tractography of the CST for risk stratification, evaluating the cut-off value for the lesion-to-CST distance that amounted to 8 mm and 12 mm to avoid new surgery-related permanent motor deficits, respectively [70,71]. Hence, patients that showed a lesion-to-CST distance above this cut-off value based on preoperative nTMS-based tractography were unlikely to suffer from surgery-related postoperative permanent paresis [70,71]. Moreover, statistically significant negative correlations were observed between the rMT value and lesion-to-CST distances in patients with a new surgery-related paresis, emphasizing the interplay between the SI used during motor mapping and results of nTMS-based tractography [71]. Correspondingly, motor function did not improve in cases with the rMT being significantly higher in the tumor-affected hemisphere than in the contralateral hemisphere, as expressed by an interhemispheric rMT ratio of >110% [70]. In a study investigating patients harboring HGG, lower FA values within the tumor-affected CST and higher average apparent diffusion coefficient (ADC) values were significantly correlated to worsened postoperative motor function, thus further exploring the contribution of dMRI-derived metrics to risk modelling [72].

In an innovative approach investigating postoperative nTMS motor mapping—instead of standardly used presurgical mapping—compared with intraoperative neuromonitoring (IONM) for predicting recovery of motor function, it was revealed that IONM and postoperative nTMS motor mapping were equally predictive for long-term motor recovery [73]. Specifically, when postoperative motor mapping was able to elicit MEPs, motor strength recovered to a score of at least 4/5 on the British Medical Research Council (BMRC) scale within one month after surgery, whereas when postoperative nTMS motor mapping did not elicit MEPs, the patient did not recover [73]. Furthermore, when implementing presurgical nTMS motor mapping and tractography in multi-modal neuroimaging with multi-sequence MRI and dedicated positron emission tomography (PET) protocols, it has been demonstrated that PET may be superior to contrast-enhanced T1-weighted MRI for proposing a motor deficit prior to surgery, and that the highest association with clinical impairment was revealed for the T2-weighted lesion overlap with functional brain tissue (i.e., the spatial overlap between the lesion volume on T2-weighted images of MRI and the functional primary motor cortex and/or CST volumes as derived from nTMS motor mapping and nTMS-based tractography) [74]. Future research may further explore the role of nTMS in multi-modal environments, given that data from various methods are frequently available for clinical needs prior to surgery. Opportunistic use of data from adjunct modalities (e.g., PET) as well as performance of dedicated longitudinal motor mapping (e.g., during the immediate postoperative course and during long-term follow-up examinations) could pave the way for a more efficient use of nTMS motor mapping and related tractography.


**Table 4.** Risk stratification and prediction. This table provides an overview of the studies published on risk stratification and prediction using motor mapping by nTMS (using an electric-field-navigated system) with or without additional nTMS-based tractography in patients harboring brain neoplasms.


**Table 4.***Cont.*

Abbreviations: nTMS—navigated transcranial magnetic stimulation; LGG—low-grade glioma; HGG—high-grade glioma; MET—metastasis; MEN—meningioma; CST—corticospinal tract; ROI—region of interest; FA—fractional anisotropy; FAT—fractional anisotropy threshold; FL—fiber length; uE—upper extremity; lE—lower extremity; rMT—resting motor threshold; MEP—motor-evoked potential; PET—positron emission tomography; MSO—maximum stimulator output.

#### *2.5. Plasticity and Reallocation of Motor Function*

Repeated application of nTMS motor mapping and tractography has potential to provide insights into brain plasticity that is likely to occur to a certain degree due to the presence and growth of a brain tumor. Few studies have already tried to investigate the role of nTMS motor mapping in this regard, and these studies are outlined in Table 5.

The non-invasive character of nTMS makes possible the acquisition of data from multiple time points, ideally spanning from the preoperative to the postoperative and follow-up interval. Correspondingly, an early explorative study in five patients and five controls used preoperative motor mapping by nTMS as well as mapping during follow-up examinations, revealing a shift of CoGs over a mean interval of 18 months of 6.8 ± 3.4 mm and a shift of motor hotspots of 8.7 ± 5.1 mm for the dominant hemispheres [75]. In a case report on a patient with a LGG that was situated within the frontal lobe and affected the suspected primary motor cortex, motor representation shifted from the precentral to the postcentral gyrus over an interval of 18 months according to serial nTMS motor mappings, which was confirmed by DES mapping during re-resection [76].

In general, a connection between the distinct location of the motor map as enclosed by nTMS as well as its extent and tumor location has been demonstrated in the sense of tumor location-dependent changes in the distribution of polysynaptic MEP latencies and spread of motor maps, especially along the anterior-posterior direction [77]. In the further course, it was revealed that in a majority of patients with mixed tumor entities, MEP counts, when elicited by nTMS to the precentral gyrus, were higher than average, potentially reflecting robust and less variable motor representations within the primary motor cortex [78]. Additionally, patients with tumors affecting the postcentral gyrus and other parietal areas primarily showed high MEP counts when stimulation by nTMS was delivered to the postcentral gyrus [78]. Hence, functional reorganization patterns seem to be reflected by a reorganization within anatomical constraints, such as of the postcentral gyrus [78]. Using again serial nTMS motor mappings from presurgical and follow-up sessions, the initial observation of CoG or motor hotspot shifts have been confirmed in further series including 22 and 20 patients with different tumor entities, respectively [79,80]. Additionally, motor representations appeared to shift more clearly toward the tumor mass if the lesion was anterior to the rolandic region than if it was located posterior to the rolandic region, and a preferential regrowth pattern of tumor recurrence towards the primary motor cortex and/or CST as defined by nTMS-based motor mapping and tractography has been suggested by exploratory approaches [80,81].

**Table 5.** Plasticity and reallocation of motor function. This table outlines the studies published on plasticity and reallocation of motor function as revealed by motor mapping using nTMS (using an electric-field-navigated system) with or without additional nTMS-based tractography in patients harboring brain neoplasms.




Abbreviations: nTMS—navigated transcranial magnetic stimulation; LGG—low-grade glioma; HGG—high-grade glioma; MET—metastasis; rMT—resting motor threshold; uE—upper extremity; lE—lower extremity; DES—direct electrical stimulation; fMRI—functional magnetic resonance imaging; APB—abductor pollicis brevis; CoG—center of gravity; ADM—abductor digiti minimi; MEP—motor-evoked potential; WHO—World Health Organization.

#### *2.6. Integration into the Clinical Environment*

For broad application of nTMS motor mapping and derived nTMS-based tractography in neuro-surgical oncology, seamless integration into existing hospital infrastructure and processes is key for acceptance and optimal use of generated data. In this regard, a structured workflow has already been proposed [82]. It starts with admission of the patient and when the indication for mapping is made and includes, amongst other steps, transfer of nTMS data to a hospital-intern picture archiving and communication system (PACS) as well as reporting within dedicated masks for the hospital-intern electronic patient charts [82].

An example of inter-disciplinary integration into different systems requiring nTMS data transfer is represented by the versatile use of the motor maps for planning and treatment purposes in radiosurgery and radiotherapy [83–87]. The first published approach achieved easy and reliable integration of nTMS, fMRI, and tractography data for radiosurgery treatment planning, which led to an average radiation dose reduction of 17% to functional brain areas in a cohort with mixed entities of pathologies [83]. Another study approved flawless integration of specifically nTMS data for radiosurgery, which influenced the radiosurgical planning procedure by improving risk-benefit balancing in all cases, achieved dose plan modifications in 81.9%, facilitated treatment indication in 63.7%, and reduced radiation doses in 72.7% of cases [86]. Compared with radiosurgery plans without nTMS data, treatment plans with integration of nTMS data demonstrated a significant decrease in dose to eloquent cortex volume, which was achievable without a reduction of the dose applied to intracranial MET [87]. Moreover, integration of nTMS motor maps with radiotherapy planning software for hypofractioned stereotactic treatment regimens for patients diagnosed with intracranial MET has been proposed, and by constraining the dose applied to the nTMS motor maps outside the planning target volume (PTV) to 15 Gy, the mean dose was significantly reduced from 23.0 Gy to 18.9 Gy, while the mean dose of the PTV increased [85]. Analogously, in patients with HGG, mean dose to the nTMS motor maps was significantly reduced by 14.3% when constraining the dose to nTMS motor areas, while the dose to the PTV was not compromised [84].

Furthermore, integrating nTMS motor mapping in clinical workflows has provided initial evidence for the usefulness of the method for planning of a stereotactic tumor biopsy, performing endoscopic cystoventriculostomy, or facilitating a transparietal approach to the trigone of the lateral ventricle in patients with brain neoplasms [88–90]. In a special environment such as the intensive care unit with critically ill patients, an approach for safe and reliable use of nTMS motor mapping has been described recently, yet preliminarily in patients suffering from other diseases than brain tumors (e.g., central cord syndrome after trauma, ischemic or hemorrhagic stroke) [91]. In particular, the use of computed tomography (CT) instead of MRI data may help to establish nTMS motor mapping also in special environments with patients who may only be eligible to undergo CT due to specific infrastructural constraints (e.g., non-availability of timely imaging by MRI) or medical conditions (e.g., specific implanted devices as contraindications for MRI) [91,92]. While this underlines the broad applicability of nTMS motor mapping, which requires little patient interaction while creating valuable data on the motor system in a non-invasive way, high accuracy has to be ensured and other imaging sources than MRI have to be regarded as second-line alternatives in selected cases.

#### **3. Methodological Considerations on Application of nTMS Motor Mapping**

#### *3.1. Current Practices and Protocols*

Previous large-scale studies have converged to feasible routines for nTMS motor mapping in clinical practice [93]. As an example, the usual SI is normalized to the rMT, and 110% rMT has become the standard, though also 105% rMT is often used. Alternative methods, such as using Mills–Nithi upper threshold (UT), exist as well [93].

There is large variance in the motor mapping protocols since the studies intend to answer specific research questions. On a similar note, the terminology used is not always clear and uniform. New methods and analysis tools have emerged. In addition to formerly used metrics, new and parallel map measures (e.g., area and volume in defining the extent of motor maps) are increasingly used and reported, which brings variance to the studies and makes them more difficult to compare with each other. For instance, motor map topography based on counting the number of discrete peaks, which in turn was based on MEP amplitudes, was introduced in 2017 [94]. Another measure that seems to be interesting and reliable and may also have clinical relevance is the overlap of the motor representation of muscles, though its potential meaning is not yet fully understood [95–97].

When it comes to quantitative mapping, a lot of research on quantitative parameters derived from nTMS motor mapping is ongoing. Despite many motor mapping results in patients (such as greater or smaller area of motor maps, closeness of CoGs, or location and shift of CoGs) have already been published, there is clear need of comparison data on healthy volunteers as a basis against which to evaluate cortical reorganization in clinical populations [98,99]. In addition, the normativity of hemispheric side-to-side differences needs to be ensured before comparison between potentially affected and non-affected hemispheres takes place. Fine-scale topography seems to be complex and variable between subjects. To understand it better, multi-modal approaches would be important to better track and understand nTMS-induced effects across the brain [100,101]. Furthermore, instead of mapping single muscles, the importance of groups of muscles, their synergy, and the role of movements and their relation to posture and biomechanics have been pointed out [102].

It should not be forgotten that the steps during initial mapping to locate a motor hotspot, sometimes called coarse mapping or technical mapping, are an important part of the examination to define the precise and correct hotspot. Regarding motor areas of lE muscle representations, a double-cone coil is recommended to reach deeper [103]. Different coil types hinder direct comparison between studies. The coil orientation may also have impact, which partly depends on the area of interest (uE, lE, or face muscle representations) [104–106]. The need for preactivation of muscles is a special issue that needs to be taken into account, particularly for lE and face muscle representations [106]. Another important issue is that when mapping the extent of several muscle representations within a limb, the rMT is usually only determined for one specific small hand muscle (abductor pollicis brevis (APB) or first dorsal interosseous (FDI) muscle), and this is used as a reference for motor mapping of representations of other muscles as well. This could perhaps be tackled by targeted post-hoc analysis [95]. For the lE, there is much more variance when choosing the muscle of primary interest, which should preferably also have the lowest rMT. The clinical importance is, however, mostly unknown.

In the analysis of motor mapping data, large variability in MEP amplitudes is a challenge. Another challenge is that the amplitudes are often small. The usual response criterion for an accepted MEP in rest is often 50 µV [26], but lower amplitude criteria have been successfully applied [107]. Some of the measures need to be normalized to the maximum recorded MEP amplitude. Mapping-related biomarkers of sensorimotor plasticity could be used in the study of pathophysiology of different diseases and an important application is rehabilitation. Based on these reflections on the current status and routine procedures for nTMS motor mapping, in the following we outline the most important quantitative parameters and methods for ensuring feasible accuracy of nTMS motor mapping.

#### *3.2. Selecting Stimulation Intensity*

The proper SI used in quantitative mapping is conventionally and fundamentally dependent on the rMT, which can be determined in several ways [47,108–112]. These days, the most convenient and most widespread method is adaptive threshold hunting (ATH), which estimates the threshold SI in an iterative fashion with excellent confidence [109,110,113–115].

Selection of the SI used in motor mapping overall is a crucial part of the experimental design, as the SI defines the amplitude and spread of the responses, and the spatial accuracy of individual stimuli [25]. In general, the map size increases with the SI [24,25]. A common practice is to use a SI that is related to the individual rMT by percentage increase, i.e., 110% of the rMT. A workshop report including recommendations for nTMS motor mapping in patients harboring brain tumors suggest the use of 105% rMT when mapping the uE muscle representations and 110% (with additional 20 V/m) to map the representations of lE muscle representations [93]. While these are practical solutions easily applied for clinical practice, they are likely to cause protocol-induced variation to the results. The individual input–output characteristics vary with factors such as age [116,117]. Furthermore, they are then also affecting the outcome of the mapping, as the used supra-threshold SI is dependent on those characteristics [24,25,107,118,119].

From a risk-benefit perspective, the risks in selection of the SI are the following: (1) too low SI will not activate the cortex that contains motor functions in mapped areas (and, as a result, false-negative responses are gained), and (2) too high SI that excites neurons outside the stimulated target region (resulting in a response that is falsely positive, i.e., positive responses not associated with the stimulated target). Provided that the fluctuation in cortical excitability is normal, and muscles are maintained in rest, the benefits corresponding with the above-mentioned risks are that with (1) low SI if the stimulation of a cortical target is producing a response, it can be assumed with high confidence as a true-positive response, and with (2) high SI the probability of inducing a response is greater (Figure 5), and it is likely that if there is no response the stimulated target is a truenegative response. Therefore, in selecting the SI, the risks and benefits should be weighted regarding the information that is needed to be acquired. In preoperative motor mapping in patients with brain tumors, a motor map that only shows a minimum number of falsepositive stimulation points is warranted in order to avoid overestimation of the extent of the motor map. Such overestimation could lead to incomplete tumor resection given that false-positive spots are unnecessarily spared from resection due to anticipated, but faulty "true" motor function representations. However, high fractions of false-negative points would put the patient at a theoretically higher risk of functional deterioration in cases in which such points are included in the surgical resection. Correspondingly, Thordstein et al. speculated that using a low SI could include an additional risk since the activation area of a muscle could be distributed non-continuously along the motor cortex [107,120].

**Figure 5.** Example of the cumulative distribution function (*black line*) fitted to the experimental data of one subject. The *red dots* indicate the probability of response based on repeated MEP trials. The number of repetitions at each SI (given as %-MSO) is reflected in the marker size, which has been used to weight the fitting of the cumulative distribution function.

When using a SI that is at supra-threshold level normalized to the rMT (e.g., 110% rMT), it is unclear what the individual likelihood for induction of a response is (Figure 6).

However, it may largely avoid false-negative stimulation points within the motor map by arbitrarily creating some sort of a "safety margin" around unequivocally motor-positive stimulation spots. Certainly, the likelihood for producing a response at the motor hotspot where the rMT was determined is greater than 50%—but is it 60% or 95%? This dilemma makes it difficult to estimate the real accuracy of individual motor mapping if the input– output characteristics are unknown. Alternative techniques for determining the mapping intensity based on a threshold value that relies on greater probability of responses than 50% has been suggested as an alternative way to determine the SI [25]. Specifically, Kallioniemi and Julkunen proposed that the use of the so-called Mills–Nithi UT could be used directly as the SI for mapping, and demonstrated that it indeed reduces the inter-individual variation in the quantified motor map size [25,108]. The core principle in using the UT instead of a SI normalized to the rMT seems justified as with UT the probability of a response is ~90%, hence reducing the dependence of mapping outcome from the individual input–output characteristics. However, there are drawbacks in that methodology: (1) the confidence in the estimated UT is likely lower than for the rMT that is determined using ATH [109,110], and (2) it may take a few minutes more time to determine the UT [25].

**Figure 6.** Demonstration of the cumulative distribution function representing the probability of inducing a response at different SIs. As the slope, represented as relative spread [110], can differ between subjects, the use of the SI related to the rMT can induce differences to the absolute size of the motor map [25]. In the plot, mean relative spread (*solid black line*) and the minimum (*dotted black line*) and maximum (*dashed black line*) found in the study population were used to compare probabilities of inducing a response at 110% rMT (*blue dots*) and 120% rMT (*red dots*). At a low relative spread, 110% and 120% rMT produce an MEP at the stimulation target with a probability close to 100%, but with the high relative spread, the probabilities of MEPs at the target at 110% and 120% rMT are 68% and 83%, respectively.

To demonstrate the variability of motor maps due to uncertainty in response occurrence probability, we simulated motor maps based on experimental motor mapping data by assuming that 10% of the responses were falsely negative to estimate the general confidence of motor mapping based on uncertainties related to response occurrence (Figure 7). The simulation demonstrated that despite the uncertainty, the SI-dependent differences in the motor map area were still apparent, and the shape and extent of the motor maps were maintained from simulation to simulation. Unlike the area of the motor map, the location of CoGs does not appear to depend on the SI [24]. It was demonstrated by Thickbroom et al. that when moving the coil from one cortical location to another, the shape of the

input–output curve does not change significantly, only the offset that is the crucial part being represented as the rMT [121].

**Figure 7.** Visualization of the motor map resulting from mapping with three SIs (*blue*: 110% of the rMT, *green*: Mills–Nithi UT method [108], *red*: 120% rMT). Data from each of the experiments were bootstrapped 1000 times, assuming that 10% of the responses observed were false negative. The image is visualized in neurological projection. The stimulations were placed on average 0.4 mm apart. The 95% confidence limits are indicated for quantified areas in the images.

The SI is commonly represented as a percentage of the maximum stimulator output (%-MSO), by definition making it dependent on the maximal stimulator performance that is highly dependent on the used instrumentation including the characteristics of stimulation coils and the stimulating pulse [122–128]. This means that when comparing the used SIs between individuals, one has to consider the characteristics of the instrumentation. In addition, the SI, when represented as %-MSO, is not considering the individual distance of the stimulated cortex from the stimulating coil that is placed on the scalp [129–133]. To account for individual coil-to-cortex distance by estimating the cortically-induced EF by stimulation, an EF estimate could be used [134–136]. However, the EF estimate may not account for differences in pulse characteristics [137]; yet it could reduce the difference in representing SIs as EFs between stimulator manufactures differ [126]. Nevertheless, when applying EFs in different individuals, there exists a challenge to determine the anatomic location where the EF should be estimated, as the exact location of response induction has not been unambiguously determined since the activation by stimulation is not limited to the crown of specific layers of the cortex, but, instead, the coil distance-dependent EF affects a large part of the cortex [29,105,138].

Recently, Nazarova et al. mapped the representations of multiple muscles simultaneously to distinguish between muscle representations of the individual muscles [98]. As previously suggested, they observed that the use of a single SI may be a possible limitation as the different muscles could potentially have different rMTs [98]. Previously, it has been observed that the different somatotopically adjacent muscle representations could have different excitability profiles [95,107,118]. Albeit, at the group level the effect size may be minor or acceptable, at the individual level the clinical significance for such different profiles may be crucial [95,119,139–141]. This means that if a muscle has a lower rMT for activation and a steeper rise for the input–output curve than the other mapped muscles,

the motor map will be biased due to the responses of that muscle, and will mostly present the representation area of that specific muscle over the other mapped muscles. Therefore, when determining quantitative characteristics for a group of muscles, the mapped outcome may be biased with certain muscles due to the differences in the individual muscle rMTs, and perhaps also due to individual motor hotspots.

Furthermore, the coil-to-cortex distance varies with stimulated cortical regions, which may require adjustment of the applied SI [130,131,135,136,142–144]. Because of the differences between target sites, the SI needs to be adjusted by taking into account the differences in coil-to-cortex distances, the secondary field caused by charge accumulation at conductivity discontinuities, and the coil orientation, and adjustment based only on the SI or primary EF is not sufficient [135].

## *3.3. Stimulation Grid*

To enable quantitative mapping and to set the spatial density for stimulation targets and, thus, spatial accuracy of the quantitative mapping, stimulation grids are frequently used [27,145–147]. The grids are especially crucial for non-navigated estimation of motor maps [145,146]. However, they are also used in nTMS approaches where the underlying anatomy is visible [27,147]. The use of the grids enables straightforward calculation of the motor map size, i.e., by calculation of the number of active squares (producing acceptable responses when that square is stimulated) within the grid [118]. The definition of the active grid square varies in terms of interpreting a response (e.g., response/no response, maximum MEP amplitude, mean MEP amplitude, MEP count, etc.). In nTMS, the size of the grid squares affects the accuracy of the motor map that can be related to anatomical structures. The accuracy is limited also by the resolution of the underlying structural MRI and the accuracy of the neuronavigation system [1]. If using a grid as aid for enabling homogeneous spacing between the stimuli, the selection of the grid size should consider the required spatial resolution of the motor map. Figure 8 demonstrates how the size of a stimulation grid square could affect the appearance of a motor map. The larger grid squares result in bulk-shaped motor maps with the grid potentially overestimating the true map size, while reduction of the grid square size converges towards the true motor map size. However, the smallest grid squares produce lower map size than the reference value because the original data were gathered with inferior density and there might be no data for all grid squares (Figure 9).

It may soon be obsolete to consider motor mapping in terms of grids and grid targets, as modern nTMS motor mapping could potentially be performed in a quantitative fashion without the use of grids as long as quality criteria are set [148]. This means that the placement of stimuli is anatomically guided with denser placement of stimuli at specific regions, e.g., in the vicinity of anatomically interesting landmarks or at the edges of a motor map. This also means that the spatial and regional accuracy within the motor map may vary, while the extent of the motor map could be more accurately defined. The effects could be inverse for the accuracy of the CoGs, and for calculation of CoGs the coverage of each individual stimulus target in the motor map needs to be accounted for and be weighted in the calculation [27].

#### *3.4. Number of Stimuli Required (per Target Location)*

The accuracy of the motor mapping has been shown to relate to the number of stimuli used [26]. The number of stimulated responses within each grid square varies, as does the size of the grid squares [26,118,140,145,147,149–153]. Previous studies have investigated the required number of stimuli in a stimulation grid, having observed that two or more stimuli should be used to improve the confidence in the resulting motor map parameters [26,154]. Specifically, Cavaleri et al. reported that at least two responses induced by targeting each grid square were required for reliable calculation of the CoG and motor map volume in non-navigated TMS with a grid square size of 10 mm [154]. The used stimulus grid, or the density of the stimulus target spacing plays an essential role when defining the parameters

of the motor map. In essence, the CoG location is dependent on the grid square size, as it is the case with the cortical area (Figure 9). The larger the grid square is, the less accurate is the CoG location or the motor area (Figure 9).

**Figure 8.** Demonstration of the use of a grid in calculating the area of the motor map. The original individual responses are presented as *red-filled dots*, with the dot size reflecting the MEP amplitude. The *black dots* are 0-amplitude responses. Average MEP amplitudes were calculated within grid squares in different size grids. The resulting average MEP amplitude size is reflected in the *yellow color* of the grid squares. The area of the motor map was evaluated based on the sum of the grid square areas with average MEP amplitudes of at least 50 µV, and by using spline interpolation (*yellow line*). The corresponding motor map areas are displayed above the grids with the grid sizes. For comparison, the motor map area is displayed for the original responses with spline interpolation, indicated by the *red line* in each plot.

When using nTMS, the conventional grid squares (e.g., 10 × 10 mm or 15 × 15 mm in size) are likely to include more stimuli. This is demonstrated in Figure 10, placing a stimulus grid of typical size over the mapped region and demonstrating that several stimuli are placed within the grid squares [146,149]. In fact, due to the spatial averaging caused by the large number of MEPs recorded during the mapping, the inherent variability effect in MEPs may be reduced, and placing multiple stimuli per location is compensated by closer spacing of the stimulus locations [95,155]. Chernyavskiy et al. showed that with an increasing number of stimuli included in the motor map, the accuracy is improved in nTMS mapping without application of a grid, as the coverage of a single response in a motor map and, hence, the contribution a response for the total motor map is decreased [155].

**Figure 9.** (**A**) The effect of grid size on the resulting motor areas with different methods of calculation. As a comparison, there is the area calculated with the spline interpolation (*red*), which is independent of the grid size, but is instead dependent on the local spacing of the stimuli. The area calculated from active grid squares is presented in *green* and, for ease of comparison, the spline interpolation area calculated from the active grid sites is shown in *blue*. The shaded areas indicate the 95% confidence interval within the study population, which was 24 mapping experiments at 110% of rMT. (**B**) The average number of stimuli falling within each grid square is shown as a function of grid size, with the shaded area indicating the 95% confidence interval within the study population. (**C**) The effect of grid size on the CoG location is shown, with the shaded area indicating the 95% confidence interval within the study population.

Often, when utilizing stimulation grids, a few stimuli may be repeated per grid to reduce the effect of MEP amplitude variation. However, then the number of stimuli is in a different scale than the number of required stimuli needed to obtain the average MEP amplitude confidently. To reach for a reliable and stable value of the MEP amplitude, previous studies have found that at least 20 repeated trials should be averaged [156,157]. However, these confidences are not fully comparable, since the overall data on MEP amplitudes within the motor maps are more extensive than in studies assessing single-target MEP amplitudes. Thus, the effect of individual grid square MEP amplitudes variability affects less the binarized map parameters such as the area, and, hence, the "spatial filtering" reduces the apparent MEP variability. Obviously, parameters utilizing absolute MEP amplitudes such as the map volume could be more affected.

**Figure 10.** The visualized grid square size used was 10 mm × 10 mm, which is a quite commonly used size [146,149]. The stimulations were placed on average 2.6 mm apart. The individual stimulus locations are shown with *red dots*, the size of which is indicative of the associated MEP amplitude (the larger the amplitude, the larger the dot). The *yellow colors* in the grid squares indicate the size of the mean MEP amplitude for the MEPs induced by stimulating points within the grid square (the brighter the color, the higher the mean MEP amplitude).

#### *3.5. Coil Orientation with Respect to Anatomy*

The degrees of freedom (DOF) of the stimulation coil include the coil location, orientation and tilt, while additionally, one system-dependent DOF is also the previously discussed SI. The coil tilt affects the efficiency of dose delivery on the cortex [158,159]. With respect to the underlying cortical anatomy, the coil orientation affects the observed response (i.e., suboptimal coil orientation may result in MEPs with low amplitude or a non-response) [105,160–162]. Balslev et al. showed with non-navigated TMS that a 45◦ angle with respect to the interhemispheric midline is generally the optimal coil orientation [163]. Further, the optimal coil orientation has been observed from experiments and simulations to be perpendicular to the gyral wall [104,105,135,138,160,164]. While there may not be group-level differences in the optimal coil orientation, the individually quantified parameters, such as the motor area, may depend on it [106].

The microanatomy of the gyrus may also affect the efficacy of nTMS, i.e., how aligned are the stimulations and the activated neuronal structures, and how organized or anisotropic are the activated individual neurons in their population [138,165]. While these factors cannot be directly visualized during application of nTMS, they may take relevant effect on the mapping outcome.

#### *3.6. Other Quantitative Parameters in Motor Maps*

Other quantitative parameters commonly used to characterize motor maps based on MEP amplitudes include the motor hotspot, CoG, motor area, and motor map volume. The motor hotspot location typically is defined by the location of the maximum MEP amplitude response (*xmax*; *ymax*) or the "optimal location" for stimulation [28,29,166–168]. The motor hotspot is not only used to characterize the location of the motor representation, but is also used as the location where the rMT is determined and around which the motor

mapping is extended [169]. Considering the definition of the motor hotspot, Reijonen et al. characterized a hotspot as a region instead of a unique target and found that if the definition is based on MEP amplitudes within individualized motor maps, the hotspot is on average 13 mm<sup>2</sup> in size, and if the hotspot is defined on the basis of the stimulation-induced EF, the size is on average 26 mm<sup>2</sup> [29]. These hotspot definitions consider the accuracy of the definition of the hotspot, including within-session neuronavigation system accuracy. Specifically, with nTMS, it has been shown that intra- and inter-observer variability for motor hotspot determination are on average ≤1 cm, with values ranging within the calculated precision of the used system [170].

The CoG is defined as the amplitude-weighted location in coordinates representing the location in the motor representation area where the center of motor activation lies, and is represented by the following equation:

$$\mathbf{x}\_{\text{CoG}} = \sum \mathbf{M}\_{\text{i}} \mathbf{x}\_{\text{i}} / \sum \mathbf{M}\_{\text{i}\prime} \mathbf{y}\_{\text{CoG}} = \sum \mathbf{M}\_{\text{i}} \mathbf{y}\_{\text{i}} / \sum \mathbf{M}\_{\text{i}} \tag{1}$$

where CoG location is defined in two dimensions with the Cartesian coordinates *x*CoG and *y*CoG, individual MEP amplitudes, and *M<sup>i</sup>* corresponding to single stimulation targets (*x<sup>i</sup>* ; *yi* ) [171]. The CoG locations are dependent on the muscle representation mapped [95,118,172]. The intersession repeatability of the CoG has been demonstrated to be good to excellent [28,31,98,118]. In our simulations, we found that the SI has a minor effect on the CoG location, as does the chance of false-negative responses (Figure 11).

**Figure 11.** Visualization of CoGs based on motor maps with three SIs (blue: 110% of the rMT, green: Mills–Nithi UT method [108], red: 120% rMT). Data from each of the experiments were bootstrapped 1000 times, assuming that 10% of the responses observed were false negative. The dots indicate a product of 1000 CoG locations each. The image is visualized in neurological projection.

If the spacing of the stimulus targets is not homogeneous within the motor map (i.e., when a stimulus grid is not used), then the response amplitudes *M<sup>i</sup>* need to be weighted by their stimulus targets coverage in the motor map *A<sup>i</sup>* [27]:

$$\mathbf{x}\_{\rm CoG} = \sum M\_i A\_i \mathbf{x}\_i / \sum M\_i A\_{i\uparrow} \mathbf{y}\_{\rm CoG} = \sum M\_i A\_i \mathbf{y}\_{\rm i} / \sum M\_i A\_i \tag{2}$$

The volume of a motor map is usually defined in a grid by summing the MEP amplitudes associated with the grid squares that exceed a given response threshold [24,30,31]:

$$Volume = \sum M\_i \tag{3}$$

where the index *i* refers to each grid element. Alternatively, the volume of the motor map has been determined as the volume of an interpolated amplitude surface on the cortex [96]. The repeatability of the motor map volumes has been demonstrated to be between poor and good [30,31].

The area has been defined in different ways. When using a simulation grid, previous studies have multiplied the grid square area with the number of active stimulation sites within the grid [118,147,173]. With nTMS, recent studies have utilized different means for calculating the area, such as spline interpolation or Voronoi tessellation [25,27,95,153,155,173–176]. These analysis techniques of the representation are not directly comparable, as they include systematic differences [173]. Previous studies have shown that the motor map area may suffer from poor to excellent session-to-session/within-session repeatability [27,30,31,98,118]. Here, we simulated the session-to-session repeatability of a motor map by assuming the potential of false-negative motor responses to find that the quantitative size is well-preserved within the sessions as are the shapes and locations, while minor session-to-session differences were observed in the motor map area (Figure 12). The set amplitude criterion also affects the motor area, and commonly the amplitude criterion of 50 µV is used (Figure 13).

**Figure 12.** Visualization of motor map outlines of the FDI muscle representation in repeated measurements. Data from each of the experiments were bootstrapped 1000 times, assuming that 10% of the responses observed were false negative. The image is visualized in neurological projection. The *blue maps* were mapped a bit more densely (mean distance between stimulus locations was 2.6 mm on both hemispheres, as a 3.0 mm grid was used as aid) than the *red maps* (mean distance between stimulus locations was 2.7 mm and 3.0 mm on the left and right hemisphere, respectively, without a stimulation grid). The 95% confidence limits are indicated for quantified areas in the images.

Sinitsyn et al. found, based on their experiments comparing multiple techniques and multiple stimuli/grid squares in calculating the motor map areas, that area weighted by the probability of an MEP within a grid square appeared overall best in terms of accuracy [26]. As the SI applied with respect to the rMT determines the probability of an MEP in each grid square, the reliable determination of the probability would likely require multiple repetitions per grid square, or for one stimulus per grid square the probability would be 0 or 1.

**Figure 13.** Effect of amplitude criterion for accepted MEPs measured on the resulting motor map size as determined using spline interpolation [27]. In the figure, 24 cortical mapping experiments for locating and outlining the FDI muscle representation were used to calculate the mean (*green line*) and 95% confidence interval (*green area*) for the motor area at different amplitude threshold criteria for accepted MEPs.

#### **4. Perspectives and Future Directions**

Over the past decade, nTMS has found its way into clinical routine, particularly for motor mapping among patients with motor-eloquent brain neoplasms since the method combines spatially resolved identification of brain function with high accuracy even in cases with deranged anatomo-functional architecture. The good agreement between preoperative nTMS motor mapping and intraoperative DES mapping as the reference method seems one of the key factors contributing to the current role of nTMS motor mapping in neurooncological surgery [35–42]. Furthermore, motor maps derived from nTMS can be used for seeding to achieve function-based tractography, which enables the identification of the spatial course of critical subcortical structures such as the CST [51–56]. However, the initial application for preoperative planning and intraoperative resection guidance in patients harboring functionally eloquent brain neoplasms has been greatly expanded over the years, thus enabling researchers to also address basic research questions in the context of a brain tumor as the use case, spanning from risk stratification for motor function to plasticity and reshaping of functional anatomy [69–73,77–81].

The key to successful integration of nTMS motor mapping and derived tractography into the clinical workflow is closely related to feasibility aspects and potential difficulties when embedding these approaches in existing environments. In this context, seamless integration into pre-existing infrastructure (e.g., hospital information system or PACS) can be achieved by standardized data export and transfer from the nTMS system [82]. However, it has to be noted that performing the mapping procedure in the most accurate fashion requires some training and time (~60 to 90 min per patient, excluding preparation and depending on factors such as the extent of motor mapping and patient cooperation) [6]. Thus, trained personnel and dedicated time slots for mapping purposes may be required in the clinical setting, which are not always granted under economic and time constraints. However, alternatives to nTMS motor mapping for the preoperative workup in patients harboring brain neoplasms (e.g., fMRI or MEG) also come with expenses and may only be available in specialized centers. Given that an overall better agreement between nTMS motor mapping and DES mapping has been observed in comparative studies with fMRI or MEG, efforts during the preoperative setup with nTMS seem justified [35–42].

Further potential of nTMS motor mapping lies in the interdisciplinary use of derived data. Particularly, radiotherapists may take advantage of such maps to modify treatment plans with the aim of limiting dose exposure to motor-eloquent cortex, as shown in first

studies on the matter [83–87]. Integration of functional data derived from nTMS motor mapping could also be achieved for case discussions of interdisciplinary tumor boards and forwarded to follow-up treatment, which may make use of such information for rehabilitation strategies. In this context, a first example in patients suffering from acute surgery-related paresis of uE muscles after glioma resection provides evidence for the beneficial use of nTMS as a therapeutic tool in neuro-oncology, with the exact site of stimulation being determined based on nTMS motor maps [177]. Specifically, the combination of low-frequency nTMS with physical therapy for seven consecutive days after surgery improved motor function outcome according to the Fugl–Meyer assessment performed postoperatively and until the 3-months follow-up examinations [177].

Additionally, nTMS data could also be efficiently used in multi-modal scenarios. For instance, the combination of motor mapping with mapping of other functions such as language and derived tractography of motor- and language-related subcortical pathways has already been achieved in a few studies, which may help to gain a broader picture of functional anatomy in patients harboring large or critically situated neoplasms that most likely do not solely affect the motor system [178,179]. Further, to understand better the distinct effects of nTMS on the brain's connectivity profile, combinations of nTMS application with pre- and post-stimulation fMRI acquisitions and functional connectivity analyses are possible and have shown promising results in healthy volunteers [100,101]. Most notably, it has been revealed that modulation by nTMS critically depends on the connectivity profile of the target region, with imaging biomarkers derived from fMRI possibly playing a role to improve sensitivity of nTMS for research and clinical applications [100]. Based on such data, it seems likely that in patients with brain neoplasms, the impact and effects of nTMS also depend on a connectivity profile that may fluctuate over time or due to yet unidentified parameters, possibly interfering with the mapping outcome. Yet, multi-modal scenarios specifically in patients with motor-eloquent brain tumors may also exert difficulties on data acquisition, processing, and interpretation of data, which need to be addressed prior to routine application. For fMRI, presence of particularly high-grade tumors with increased cerebral blood flow characteristics can negatively interfere with signal interpretation [180,181]. Thus, combinations of fMRI with nTMS in such patients need to be considered carefully to avoid errors in calculations of connectivity characteristics.

Regarding dMRI, images can suffer from geometric image distortions compared with anatomical MRI, which may introduce spatial inaccuracies when dMRI data is linearly projected on conventional T1- or T2-weighted sequences and used for tractography. This may be retrospectively corrected for by non-linear, semi-elastic image fusion, thus potentially enabling tractography with improved accuracy and clinical feasibility [182]. Expanding on this, intraoperative MRI-based elastic image fusion for anatomically accurate tractography of the CST using nTMS motor maps has been achieved, correlating well with IONM and disproving the severity of brain shift in selected cases [183,184]. Furthermore, most work on nTMS-based fiber tracking of the CST has used standard deterministic algorithms implemented in commercially available packages for clinical use (e.g., fiber assignment by continuous tracking (FACT) algorithms) [51,52,54,55,70–72]. Using a FACT algorithm, fiber bundles are reconstructed in a voxel-by-voxel fashion with respect to the direction of the main eigenvector, which works purely data-based (no interpolation function) and needs only comparatively low computation time [185,186]. However, on the other hand, FACT algorithms create some predictable inherent errors, which limit the accuracy of the method and could lead to error-prone or incomplete tractography results [186–189]. Probabilistic approaches disperse trajectories more than deterministic methods and may delineate a greater proportion of white matter tracts, particularly when combined with more advanced dMRI sequences [189]. Additionally, the potential value of diffusion measures besides mere delineation of the spatial course of the CST by nTMS-based tractography may be of merit. One study has shown that the extent of impairment of diffusion metrics (such as FA and ADC) correlates with motor function deficits according to segmental analyses within the CST [72]. Hence, supplementing nTMS-based tractography of the CST with diffusion

metrics may improve the predictive power for postoperative motor impairment, but other parameters such as mean diffusivity (MD) have not yet been routinely considered. As such, MD is a quantitative measure of the mean motion of water and reflects the rotationally invariant magnitude of water diffusion, which could also be representative of structural integrity of white matter [190]. Future work may explore the potential benefits of further, even more elaborate quantitative markers of white matter structure, composition, and integrity, which can be derived from advanced dMRI techniques such as high angular resolution diffusion imaging, multi-shell imaging, diffusion kurtosis imaging, neurite orientation dispersion, and density imaging [191–193].

The continuous optimization of the nTMS system technique has recently enabled a novel paired-pulse nTMS (pp-nTMS) paradigm for biphasic pulse wave application to induce short-interval intracortical facilitation [194–196]. Use of pp-nTMS may increase efficacy of motor mapping in patients with brain tumors as accurate motor maps are achieved even in cases where conventional single-pulse mapping fails (e.g., due to tumoraffected motor structures or edema) [174,175]. Further, pp-nTMS would result in a lower rMT, thus allowing motor mapping with lower SI but without clinically relevant constraints for motor map extent or location [174,175]. Particularly in patients with brain tumors, the rMT can be frequently high in the tumor-affected hemisphere, and accurate mapping with a lower SI related to a lower determined rMT could permit successful use of nTMS even in the most demanding cases. This may also be highly relevant for other applications of nTMS in patients with brain tumors, such as language mapping [197–201]. In such applications, a higher SI is often used and stimulation is more widespread and, thus, can entail discomfort that negatively interferes with the mapping outcome. However, studies using pp-nTMS for other purposes than motor mapping are currently lacking.

From a methodological perspective, the presently used motor mapping approach is focused on determining the volume, area, coil locations, or corresponding cortical EF maximum locations associated with motor responses. However, as the EF induced by nTMS is not ideally focused, the spread of the EF also stimulates adjacent areas and, thus, when stimulating an area in the cortex adjacent to the target muscle representation area, the target muscle may activate even though the representation area was not targeted. Hence, quantitative mapping at present cannot precisely capture the true representation size of the motor representation area, and may over- or underestimate it. Previous studies have shown that inclusion of the EF information in generating the motor maps may aid in localizing the motor representation [202–204]. Specifically, minimum norm estimation (MNE) has recently been employed for estimating the true representation area of muscles by accounting for the EF spread and the input–output characteristics of the MEP values [204]. Figure 14 demonstrates the application of MNE in a single subject in comparison with the outlined cortical maximum EF locations with associated positive responses. A quite similar method to MNE was presented by Weise et al., also utilizing the input–output characteristics of the MEPs [203]. The practical difference between these methods arises from the number of coil locations applied and how the input–output characteristics are estimated spatially.

**Figure 14.** The MNE method (**left**) utilizes the spread of the EF at each stimulation point in addition to the MEP amplitude to estimate the source of motor activation, as opposed to the more conventional motor mapping approach not considering the spread of the EF associated with each stimulus (**right**) [204]. The active stimulation sites in the conventional motor map were outlined using the spline interpolation method [27].

In essence, one challenge for more accurate clinical motor mapping with nTMS in the future may be the accurate inclusion of the EF information to estimate the source of the induced MEPs. Currently, there are no clinically validated tools classified as medical devices available for source estimation utilizing EF spread, and the tools that are currently available require special skills and are not feasible for clinical routines. However, the development seems to be heading in the right direction. With tools such as the SimNIBS, EF simulations have been made quite straightforward, albeit the pipelines feasible for clinical mapping applications are still lacking [205,206]. Furthermore, the shape of the motor map may be evaluated by the aspect ratio (i.e., the ratio between map extension along the EF direction and perpendicular to it) [195]. This means that if the aspect ratio is 1, the shape of the motor map is approximated as circular, if the ratio is >1, the map is elongated along the EF direction, and with an aspect ratio <1, the motor map is elongated to the direction perpendicular to the EF direction. With single-pulse nTMS, the aspect ratio tends to be >1 [174,175,195].

Despite increasing acceptance of nTMS motor mapping in clinical routine, it cannot be emphasized enough that the value of the method stands and falls with its accuracy. In this regard, many parameters such as the location of the motor hotspot and CoG, and area and volume of the motor map are associated with the applied SI, which needs to be determined with highest diligence. Therefore, with the increase in applications of nTMS, methods for ensuring feasible accuracy become more and more important. In the context of quantitative mapping, awareness of the relevant parameters and control over them is warranted to assure best practices and reliable mapping outcomes. Specifically, quantitative mapping has the potential to derive parameters related to the motor maps of the patients that could impact diagnostics, prognostics, and follow-up examinations by enabling spatial and spatio-temporal metrics related to cortical motor function. Major future development efforts should be put to understanding the correlations between motor

mapping, interpretation of stimulation targets, and resulting responses in relation to their origin in the cortex by consideration of the physical effects of induced EFs.

#### **5. Conclusions**

The technique of nTMS is increasingly used particularly for preoperative motor mapping in patients harboring brain tumors, which is due to its sufficient accuracy and reliability in a clinical setup. The combination of nTMS motor mapping with tractography as well as the option of serial mapping over time profoundly expands its role beyond a mere surgical planning tool. Development of quantitative motor mapping can include further applications while the accuracy of current mapping modalities can be improved by standardized protocols and increased consideration of EF information.

**Author Contributions:** All authors declare that they have made substantial contributions to the conception and design of the work and have drafted the work or substantively revised it. All authors approved the submitted version. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Academy of Finland (grant no: 322423). L.S. was funded by UEF in a research project funded by Business Finland (grant no: 2956/31/2018) and State Research Funding (grant no: 1689/2020).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** The authors would like to thank Axel Schroeder for his support during the generation of figures to illustrate the motor mapping procedure by nTMS and fiber tracking.

**Conflicts of Interest:** P.J. has a shared patent with, and has received consulting fees from Nexstim Plc (Helsinki, Finland), a manufacturer of navigated TMS systems. N.S. received honoraria from Nexstim Plc (Helsinki, Finland). S.M.K. is a consultant for Nexstim Plc (Helsinki, Finland) and Spineart Deutschland GmbH (Frankfurt, Germany) and received honoraria from Medtronic (Meerbusch, Germany) and Carl Zeiss Meditec (Oberkochen, Germany). S.M.K. received research grants and is a consultant for Brainlab AG (Munich, Germany). The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Abbreviations**



#### **References**


*Article*
