A new protocol for multiple muscle mapping using nTMS

Background: Single-pulse transcranial magnetic stimulation is a safe and non-invasive tool for in14 vestigating cortical representation of muscles in the primary motor cortex. While non15 navigated TMS has been successfully applied to simultaneously induce motor-evoked 16 potentials (MEPs) in multiple muscles, a more rigorous assessment of the correspond17 ing cortical representation can greatly benefit from navigated transcranial magnetic 18 stimulation (nTMS). 19


Introduction
In traditional TMS motor mapping procedures, a grid is printed on a cap worn by the subject. This grid 48 serves to manoeuvre the coil and stimulate at the adequate position [10,11]. Navigated TMS (nTMS) 49 has been developed to substitute the cap, yet the grid-based positions remain. The grid can be dis-50 played on an anatomical scan (typically an MRI) allowing to position the coil according to neuroana-51 tomy. This reduces the error from relative movements of cap and head [12]. That advance has been 52 improved by also accounting for the orientation of the coil, next to the mere positioning [13]. The 53 most recent addition in nTMS systems has been positioning with instantly assessed electromyography 54 (EMG), to display MEP parameters (almost) online on the stimulated sites of the cortex. By this, one 55 can, e.g., a create map of MEP amplitudes for a certain muscle during an experiment (rather than 56 during post-processing) [14]. 57 These advances in technology have led to improvements in experimental protocols. As said, for TMS 58 without neural navigation, a grid on a cap serves to guide coil positioning [10]. In nTMS, the grid can 59 be readily repositioned around the so-called hotspot of a target muscle, i.e. the centre of the area of 60 interest for subsequent assessments, e.g., with different stimulation intensities [15,16]  Not only did this alternative approach reduce the time needed to map a single muscle, it turned out 67 to be as reliable as the conventional procedures; see also Cavaleri,Schabrun,Chipchase [20].
Identifying the cortical map of multiple muscles in one experimental run will be difficult -if at all 69 possible -without neural navigation [10]. Yet, studies exploiting nTMS for that sake are few and far 70 between -see [21] for an exception where up to four muscles were targeted. The focus on single 71 muscle mapping arguably stems from the fact that -in most nTMS protocols -stimulation intensities 72 at a specific percentage of a single muscle's RMT are deemed important. Detecting RMTs for several 73 muscles is laborious because it often involves offline EMG analyses. Quite recently, however, it has 74 been shown that the RMT of a small hand muscle (there, the first dorsal interosseous, FDI) may be 75 similar to the RMT of all upper extremity muscles [16]. 76 All these developments led us to design a new nTMS-based protocol for multiple muscles on the pre-77 central gyrus (primary motor cortex). Here, we illustrate its feasibility, validity, and reliability. As will 78 be shown, our mapping protocol significantly reduces operation times and drastically simplifies pro-79 cedures. By the same token, it comes with proper validity and reliability. The protocol can provide 80 information about the cortical representations of multiple muscles and the degree to which they over-81 lap. For this, we also submit a new way to define and measure the so-called active area. Through our 82 research, we anticipate changing the nTMS paradigm in research and clinic: from grid-to gyrus-based 83 and from single to multiple muscle mapping. 84

Experimental procedures 120
Participants were comfortably seated in an armchair, relaxing muscles of hands and arms. The exper-121 iment consisted of two identical sessions, which were separated by one hour to test for test-retest 122 reliability of our outcomes -electrodes were kept fixed to minimise placement errors. The interval of 123 one hour only was set to prevent drying of the conductive electrolyte gel. Each session contained three 124 parts: (1) Testing the RMTs for FDI, EDC and FCR; there, we stimulated thirty points near the omega-125 shaped area of the precentral gyrus to identify the respective hotspots defined as the coordinates for 126 which the largest peak-to-peak amplitudes in the corresponding EMG signals could be observed. This 127 served to determine the RMTs for the three target muscles following [24] (i.e. the stimulator output 128 at which peak-to-peak amplitudes were higher than 50μV in five out of ten stimuli). (2) We conducted 129 two sets of stimulations to map the representations of all eight muscles using three intensities each, 130 namely 105% RMT of FDI, EDC and FCR, respectively. For the first set, we adopted the conventional 131 grid-based method [25] using a square grid (5 cm × 5 cm) with either the FDI-, EDC-or FCR-hotspot as 132 the centre and applied 80 stimuli. This was immediately followed by a pseudorandom positioning over 133 the whole gyrus (40 stimuli), yielding a total of 80+40=120 stimuli -in the following we consider the 134 total of 120 stimuli for our gyrus-based approach while the grid mapping contained only the first 80 135 stimuli.
(3) We analysed the last 40 stimulations and estimated the hotspots of the other five muscles 136 (ADM, APB, FPB, FDS and ECR) and determined their RMTs. 137 Offline data processing 138 During the measurement, peak-to-peak amplitudes and latencies of the MEPs were estimated. The 139 MEP was defined as the range between the minimum and maximum peak of the EMG signal, and the 140 latency as the onset of the MEP signal (https://github.com/marlow17/surfaceanalysis). The onset was 141 defined as the point in time at which the signal exceeded mean ± 1.96×sd of the signal baseline 142 (100 ms prior to up to the moment of stimulus). 143 For every muscle and stimulation, we determined whether a MEP was elicited; see Supplementary 144 Material for details. Whenever MEPs were present, the corresponding parameters were included 145 when computing the mean MEP amplitude and latency, the centre-of-gravity of the stimulation sur-146 face ( $,&,' ) and the size of the active area, i.e., the size of that surface. We defined the centre-of-147 gravity as follows [26]: 148 Here, -represents the peak-to-peak amplitude for stimulation at position ( -, -, -). The defi-150 nition of the active area is more involved, but we provide more details in the Supplementary Material. 151 All the analyses were performed separately for the 3 × 2 × 2 cases: three different muscle-specific in-152 tensities, for grid-and gyrus-based mapping, and for both sessions. The different steps are illustrated 153 in Figure 2.  Between session reliability 199 Table 1 Table S2 and Table S3   In Figure 4, we illustrate the cortical representations of the eight muscles in Figure 4

Reliability of gyrus-based mapping at different intensities 208
The reliability of gyrus-based mapping at different stimulation intensities is summarised in Table 2. 209 The MEP-responses (across parameters) showed an excellent agreement in intensity 105% RMT of 210 EDC, and the ICCs of latency performed almost excellent consistent for a stimulation intensity of 105% 211 RMT of EDC and FCR. The ICCs for in $,&,' at all the intensities were in the range of good to excel-212 lent. 213 Table 2.
The number of muscles in the four levels of ICCs cross the three intensities. ICC was interpreted as follows: <0.5 = poor, 0.5-0.64 = moderate, 0.65-0.79= good and >=0.8 was excellent. We did not find any significant effects of Intensity or Session (nor of their interaction) on $,&,' (see 214 also Table S4 in the Supplementary Material). However, the size of the active area for each muscle 215 differed significantly across intensities, without significant effects of Session or Intensity × Session in-216 teraction. The post-hoc pairwise comparisons revealed that the active area tested in the intensity of 217 105% RMT of FCR was markedly larger than in the other intensities; cf. Table 3. For several muscles, 218 the amplitude and latency varied significantly across the intensities (see Table S4 in the Supplementary  219 Material for details). 220

Discussion 221
We designed a protocol to investigate the cortical representation of multiple muscles using navigated 222 single-pulse TMS. Conventional grid-based mapping and our gyrus-based mapping were performed 223 using three stimulation intensities in two consecutive sessions. We determined RMTs of eight muscles 224 as well as MEP amplitudes and latencies, the centre-of-gravity and the size of the corresponding active 225 areas. We found that our new protocol is as valid and as reliable as the commonly applied grid-based 226 approach but appears much more feasible. Our protocol reduces assessment times and simplifies ex-perimental procedures. In our experiment, it took only about 10 minutes to map eight muscles simul-228 taneously (5 second intervals × 120 points / 60), while for the conventional grid-based approach with 229 serial muscle assessment, at least 80 minutes would be spent. Experimenting over such a long period 230 is prone to navigational errors and aggravates participants' fatigue. Moreover, by construction sepa-231 rate measurements will not provide reliable insight about overlapping cortical muscle representations. 232 One of the problems in designing multiple muscles mapping is that the RMT of a single muscle is con-233 sidered a reference when setting the stimulation intensity. We found that the RMTs of the hand and 234 forearm muscles considered are indeed marginally different (EDC-FCR and FDI-FCR). The average 235 RMTs indicated that the difference between RMTs is small (not more than 3.1% of stimulator output). 236 Intensities of 105%, 110% to 120% [27] RMT have been widely used in motor mapping [8,27,28], 237 suggesting that the here-observed difference is acceptable if not negligible. Hence, forearm and hand 238 muscles might be pooled in a group of muscles with "similar RMTs" and may indeed be evaluated at 239 the same intensity. 240 The validity testing clearly revealed that our gyrus-based mapping agrees with the grid-based standard, 241 consistent with a previous study on pseudorandom stimulus positioning [20,29]. The variation of 242 $,&,' values between the two protocols was restricted to a range of about -4mm to 4mm. The 243 differences in size of the active area between protocols are more considerable, and the between-244 subjects variability is clearly noteworthy. One explanation for this is that any outlier stimulation site 245 may strongly affect the estimate of the active area. Moreover, the individual MRIs profoundly differed 246 in size. While the first calls for more statistical evaluation and (spatial) outlier detection, the latter will 247 soon be addressed by projecting the individual MRI to, e.g., the MNI template [13]. 248 The reliability of amplitude and latency were excellent or good for all the muscles and all the param-249 eters analysed in APB and ECR achieved excellent to good ICCs, indicating the required reliability of 250 our approach. The ICC values of the size of the active area in FDI, ADM, EDC and FCR were moderate 251 to poor. This may be related to the factors. First, in some participants, the RMTs of these muscles were 252 higher than the intensities we set. That is, the intensity of stimulation was too low to activate the 253 corresponding cortical areas. Second, one must realise that even a figure-of-eight coil comes with a 254 widespread focus. Hence, TMS activates not isolated but also nearby areas. Yet, the estimated centres-255 of-gravity appeared very consistent and should be considered reliable parameters in motor mapping, 256 in line with previous studies [20,21]. 257 A gyrus-based protocol with nTMS for multiple muscles allows for not just assessing multiple muscles 258 in parallel, especially when stepping away from mere "active point" drawing [30] to estimating muscle-specific active areas in detail. Isolated muscles are not mapped to isolated cortical areas, but one de-260 termines the cortical representation of a set of muscles. Results show that these cortical representa-261 tion manifest overlaps that diverges across the muscles [31]; cf. Figure 4. This comes particularly to 262 the fore when analysing cortical surfaces in 3D. There are several methods to determine mapping 263 areas using nTMS [19,32] that stand out for their computational ease. However, by ignoring the 3D 264 representation of surface folds, one runs the risk of missing important information about active areas 265 in the gyri vis-à-vis the sulci. We used a 3D-alternative (as explained in the Supplementary Material) 266 that overcomes these potential shortcomings. This also provides a welcome visualisation of the active 267 area on the cortex (for single or multiples muscles), which might serve as a convenient tool. In future 268 research, we will exploit its capacity to unravel the nature of the overlap region of active areas of 269 multiple muscles. 270 We selected three intensities when eliciting MEPs, namely 105% RMT of FDI as a representative of the 271 hand muscle, EDC as extensor and FCR for flexor muscles. Despite the small variation in RMTs of all 272 muscles, one may still consider this sub-optimal. Image-based RMT prediction may be an excellent 273 way to address this issue. This should involve modelling the electric field distributions of TMS, which 274 may also serve to quantify overlapping cortical representation with even higher precision than cur-275 rently possible. 276

Conclusion 277
We designed a time-efficient protocol for identifying the cortical representation of multiple muscles 278 in parallel. Our procedure takes only about ten minutes per subject when including eight muscles. The 279 results of the active area's centre-of-gravity at three distinct stimulation intensities can be considered 280 very reliable. All the here-considered outcome parameters confirm the validity of the procedure when 281 compared to a conventional, grid-based approach with muscle-specific resting motor threshold-scaled 282 stimulation intensities. 283