TMS SMART: Transcranial Magnetic Stimulation Scalp Mapping of Annoyance Ratings and Twitches

TMS-SMART (Meteyard & Holmes, 2018) is a method to select stimulation sites for human TMS studies which allows you to control for the subjective annoyance, pain, and muscle twitches, as well as for visible twitches and reaction time changes associated with single-pulse TMS at 43 different sites on the scalp. This website allows you to browse the data visually, and to retrieve suggested control site locations.Holmes & Meteyard (2018) also made a similar comparison between our annoyance ratings and a small meta-analysis of other studies. And we are also compiling a database of adverse events in TMS studies.

(this website works best in a desktop browser; we will improve it for mobile devices on the next major upgrade...)

Maps - Interactive maps - Background - Methods - Glossary - Data - References

Methods

Top - Participants - TMS - Tasks - Design and procedure - Statistical analysis

Participants

Twenty healthy participants were recruited from the local population of staff and students at the University of Reading. There were two sub-groups of ten participants. One sub-group (mean±SD age = 25.2±3.5 years; 6 female; 8 right-handed; mean±SD resting motor threshold = 48.5±6.4% of maximum stimulator output) performed the Eriksen 'flanker' task. The second subgroup (mean±SD age = 31.6±5.5 years; 6 female; 10 right-handed; mean±SD resting motor threshold = 51.9±7.7%) performed a choice reaction time task. Participants were paid £20 for their time.

TMS

Apparatus

A PowerMag Research100 stimulator and a 10cm diameter figure-8 (or 'butterfly') coil was used. Magnetic pulse were stimuli were biphasic. TMS intensity was set at 50% of the maximum stimulator output, which, was the mean threshold for evoking muscle respones (MEPs) in intrinsic hand muscles using this equipment in previous studies.

Scalp locations

We aimed for maximum coverage of the entire scalp, so that most potential TMS locations would be covered. We started by transferring locations from a 32 channel electroencephalography (EEG) cap onto a tight Lycra swimming cap. These 32 locations were registered onto a T1 structural scan (of author NPH) using BrainSight Frameless Stereotaxy (Rogue Resolutions, Cardiff, Wales). This registration made clear that the occipital, temporal, and cerebellar areas were not sufficiently covered by the standard EEG montage, so we added 5 sites per hemishpere and one midline site (italicised in the table below). This gave a total of 24 locations in each hemisphere (5 along the midline, and 19 lateral). Odd numbered locations are always in the left hemisphere, even numbers in the right. These numbers were used during the experiment to support single-blinding. The additional locations were added by extrapolating lines from the EEG cap more posteriorly and laterally, and placing locations approximately 6cm apart on these lines.

Table 1: Scalp locations, approximate MNI coordinates on the brain surface, and approximate anatomical locations

Coil orientations

Since coil orientation can influence the scalp sensations caused by TMS, four cardinal orientations were tested, with North taken as the direction facing the participant (i.e., posterior-anterior). Each location and orientation was stimulated biphasically using four coil orientations 45° apart. The orientations are South-North (180° or 0° clockwise, parallel to the midline, parasagittal), South West-North East (225° or 45° clockwise), East-West (90° or 270°, parallel to the ear-to-ear axis, i.e., coronal), and South East-North West (135° or 315° clockwise). The coil handle was usually aligned pointing towards the back of the head and away from the face. It was reversed along the above axes if this position was not possible (i.e., when the shoulder was in the way for inferior and lateral sites).

Tasks & stimuli

All stimuli were presented using Matlab and Psychtoolbox 3 (Brainard, 1997) in black text on a white background, font size 20, screen resolution 1920 x 1080 pixels. Visual stimuli were presented using a CRT monitor with 75 Hz refresh rate.

Choice reaction time task

This task was a basic spatial-compatibility, choice RT task designed to be simpler than the flanker task (below). Stimuli were the left and right chevrons (< and >). A fixation cross was presented centrally. Three chevrons appeared on the left or right of the fixation cross, displaced 100 pixels horizontally from the central fixation cross. These could be left or right facing, as above. Therefore, there were two congruent conditions (<<< +, + >>>) and two incongruent conditions (>>> +, + <<<). Participants were instructed to press C with one finger if the chevron appeared on the left, and N with another finger if it appeared on the right - spatially-compatible responses. Speed and accuracy were both encouraged. When the data were inspected, three participants had responded to arrow direction rather than the side of the screen on which the arrow appeared. As this still constituted a choice RT task with spatial compatibility (i.e., the congruence conditions were still valid), the behavioural task was not the primary purpose of the present report, and their data were comparable to other participants' (mean RT, proportion correct, and size of congruency effect, see Appendix 1), these three participants' data were included in the final analysis.

Flanker task (Eriksen & Eriksen, 1974)

Stimuli were the left and right chevrons (< and >). Three chevrons were presented centrally, and faced left (<<<) or right (>>>). Flanker chevrons facing left or right were presented either side, displaced 100 pixels horizontally from the centre of the screen. Therefore, there were two congruent conditions (<<< <<< <<<, >>> >>> >>>) and two incongruent conditions (<<< >>> <<<, >>> <<< >>>). Participants were instructed to press the C key with one finger if the central chevrons pointed left, and N with another finger if they pointed right. Speed and accuracy were both encouraged.

Ratings Task

Participants were asked to rate the annoyance, pain, and peripheral muscle twitches in the head, face and neck caused by TMS. Each was rated on a scale of 0 to 10, with 0 being the minimum and 10 the maximum. For annoyance ("How annoying were the TMS pulses?"), anchor points (printed on a sheet of paper for reference) were "not at all annoying" (0) to "highly distracted and unable to complete the task" (10). For pain ("If painful, how intense was the pain from the TMS pulses?"), anchors were "no pain at all" (0) to "the worst pain they could tolerate during the experiment" (10). We specified pain intensity, so that participants would focus on the sensory aspects of the pain (intensity, quality) rather than their emotional response (i.e., pain unpleasantness) that would already be partly captured by the rating of annoyance (Melzack & Casey, 1968). For twitches ("If there were any twitches, how strong were the muscle twitches from the TMS pulse?"), anchors were "no twitches" (0) to "a very strong cramp" (10). In addition to the subjective ratings provided by the participants, a second experimenter recorded whether any visible twitches were observed in the face, head or neck that were coincident with the TMS pulse (0 to 5 twitches observed in each block). To facilitate this, a mirror was angled in front of the participant so the researcher could sit or stand behind them and observe.

Design and procedure

The Lycra cap with scalp locations was fitted to the participant with Cz aligned to their vertex. This alignment was periodically checked during the experiment. Resting motor threshold was first established for each participant (Rossini et al., 1994).

For the behavioural experiment, stimulation strength was set at 50% of the maximum stimulator output (biphasic). This was chosen so that stimulation intensity was constant and the same across all participants. As confirmed by our RMT data, this level was selected to be approximately at threshold (based on unpublished observations using the same equipment), and therefore unlikely to cause significant cortical excitation, but still at the approximate intensity used in TMS studies. One hemisphere per participant was stimulated at all scalp locations during the experiment. Scalp location was fully randomised, and coil orientation was nested within location and then randomised. Half of the participants had their left, and half their right hemisphere stimulated. Participants provided button press responses using the hand ipsilateral to the TMS, to minimise any chance of cortical stimulation from the TMS influencing hand motor control and response times. Single pulse TMS was delivered on each trial using a figure-of-eight coil.

Each task had identical timings and trial and block structure. Participants completed 60 practice trials with no TMS (coil disabled, away from head; 3 blocks of 20 trials, with each block split into 4 mini-blocks of 5 trials to reflect the actual experiment). They were then introduced to the rating task and response scales. TMS pulses delivered during the RMT procedure were used as reference pulses for participants' ratings. During the experiment, participants completed 5 experimental trials for each combination of coil orientation and scalp location. The experiment was structured so that a location was randomly selected, and then each of 4 randomly-ordered coil orientations was completed for that location (4 blocks of 5 trials for each location). In each block of 5 trials, a fixation cross appeared for a randomly selected duration between 500 and 1000ms, followed by the imperative stimulus. A single TMS pulse was delivered at the same time as the imperative stimulus appeared on screen. Stimuli appeared on screen until the participant responded or until 2 seconds, whichever was sooner. There was an inter-trial-interval of 2 seconds. In each block of 5 trials, participants saw one of each trial type (congruent left, incongruent left, congruent right, incongruent right) and a randomly selected 5th trial. After completing the 5 trials of the task, they were asked to give ratings for the annoyance, pain and twitches caused by the preceding 5 TMS pulses. The total number of trials per participant was 480 (5 trials x 4 orientations x 24 locations). In between each block of 5 trials, participants were able to take a break for as long as they wished.

Statistical analysis

Ratings data were taken from all 20 participants. RTs were taken from correct trials only. We used correct trials from the practice to derive a measure of baseline RT for the task (the first 20 trials were discarded as true practice trials - we assumed that RT would be stable after these initial trials). The mean RT was taken for each trial type (congruent left, incongruent left, congruent right, incongruent right) and this was subtracted from the RT for that trial type during the TMS experiment, on a trial-by-trial basis. This gave us the difference in RT for a specific trial type, when the participant was undergoing TMS. Therefore, all analyses of RTs look at the how predictor variables change the impact of TMS on RTs. Inspection of initial linear mixed-effects models showed that residuals were not normally distributed, caused by a rightward skew typical in reaction time data (Baayen & Milin, 2010; Ratcliff, 1993). Following Baayen & Milin (2010) data trimming and transformations, followed by model inspection, were explored to best normalize residuals for the models. The best results were achieved with a single cut-off value, excluding all reaction time differences longer than 400ms (the minimum reaction time difference was -389ms, so an upper cut-off was selected to mirror the minimum value). This removed 2.9% of the data.

We coded homologous scalp locations across left and right hemispheres together to look at the effect of scalp location on RT. For example, FP1 (left) and FP2 (right) were coded together, the two ATL sites were coded together, and so on. This is based on the assumption that the scalp is broadly symmetrical and the peripheral effects of TMS at a site on the left hemisphere are similar to those effects at the same site on the right hemisphere.

Analysis focused on three aspects of the data. The first was a descriptive analysis of ratings across scalp locations and orientations, giving 'heat maps' of where TMS is the most annoying, painful, and causes the most twitches. The second was to predict differences in RTs by scalp location and orientation, in order to explore whether scalp location and coil orientation have a significant effect on RT during simple behavioural tasks. The third was to predict differences in RT from the ratings provided by participants, in order to explore whether the subjective and peripheral effects of TMS influence behavioural performance. The second and third analyses were completed using linear mixed-effect models implemented in R (R Core Team, 2014); using the packages lme4 (Bates, Maechler, Bolker, & Walker, 2014), multcomp (Hothorn, Bretz, & Westfall, 2008), lmerTest (Kuznetsova, Brockhoff, & Bojesen, 2014), ggplot2 (Wickham, 2009) and corrplot (Wei, 2013). For these models, the reference level for scalp location was set to Cz/vertex. Multiple comparisons were not completed to evaluate each scalp location against each other location, as the 24 locations made this approach unfeasible. Using Cz/vertex as the reference level illustrates variations in the effect of TMS across scalp locations and variation relative to a commonly used control.

Top - Participants - TMS - Tasks - Design and procedure - Statistical analysis