Simultaneous TMS-EEG in Structure from Motion Dataset Available

We recently collected a large dataset using ant_SPLr_brainconcurrent TMS-EEG in bistable perception. Our preliminary analyses of these data have not revealed many interesting ERPs, but a great deal of analyses can potentially be done on this data, including complexity measures, Granger connectivity analyses on the MRI guided source localised data, looking at pre-stimulus alpha power and phase, ect. Unfortunately, we do not have the time at the moment to look at all these things. However, we would very like to advertise this data to anyone with the necessary expertise, who might want to take a look. You can find a pdf summary of the data here, and there is a description below. If you are interested, please drop me an email. The address can be found under ‘about me’.


Intro:

When visual input has conflicting interpretations, conscious perception can alternate spontaneously between these competing interpretations. This is called bistable perception. Previous neuroimaging studies have indicated the involvement of two right parietal areas in resolving perceptual ambiguity (ant-SPLr and post-SPLr, Euclidian distance about 3 cm between them). Transcranial magnetic stimulation (TMS) studies that interfered with the normal function of these regions suggested that they play opposing roles in this task. Specifically, inhibitory repetitive TMS to ant-SPLr leads to a shortening of percept durations, while inhibitory post-SPLr stimulation leads to a lengthening. In the present study, we investigated this fractionation of parietal function by use of combined TMS with electroencephalography (EEG). Specifically, while participants viewed either a bistable stimulus, a replay stimulus, or during resting-state fixation, we applied single pulse TMS (spTMS) to both locations independently while simultaneously recording EEG.

Participants:

N = 16, age 21-30 yrs, mean 24.94 yrs ± 3.26 S.D.; 10 female, 13 right handed.

Stimuli:

There were three stimuli used in the experiment.

1) The bistable perception stimulus was an ambiguous structure from motion rotating sphere. The sphere moved at a pace where it took 3 seconds for dots to move once to either side and back to it’s starting position. The sphere had a central red fixation dot and was 2.7 angular degrees in diameter.

2) As control replay stimulus, we presented participants with a rotating sphere identical to the bistable stimulus, but disambiguated by a depth percept, which was created by presenting two spheres stereoscopically, where one of the spheres was presented as having moved just ahead of the other, mimicking the slight binocular disparity humans experience when observing 3D objects from two viewpoints. The duration for which the replay sphere would turn either direction were sampled from a gamma distribution obtained from a baseline recording of participants percept durations of the bistable stimulus that was obtained prior to the main experiment.

3) Central fixation consisted just of the red dot without a sphere. For the bistable stimulus and resting state fixation, we also used the stereoscope, but simply presented identical spheres / dots to each eye so that no depth was perceived.

Design:

The main experiment was comprised of six experimental conditions (3 stimuli x 2 TMS sites)

condition 1 = bistable percept – ant-SPLr
condition 2 = replay stimulus – ant-SPLr
condition 3 = resting state fixation – ant-SPLr
condition 4 = bistable percept – post-SPLr
condition 5 = replay stimulus – post-SPLr
condition 6 = resting state fixation – post-SPLr

Each condition appeared three times, hence there were 18 trials in the experiment. Each trial lasted 120 seconds and included 40 spTMS pulses. In trials including stimuli 1 or 2, participants indicated their current percept with continuous button presses. In trials including stimulus 3, participants were asked just to fixate and rest. There were short breaks between all trials, and two larger breaks of 10 minutes after trial 6 and trial 12. The main experiment lasted about 45 minutes and comprised a total of 720 TMS pulses.

TMS:

Single TMS pulses at 0.33 Hz delivered at 90% Resting motor threshold. Pulses were not time-locked to any experimental or behavioural manipulation, but were delivered continuously once every three seconds for the entire duration of the trial. The mean RMT was 76.13% maximum stimulator output (± 5.78 S.D.). Pulses were applied to either the post-SPLr (MNI: x = 38, y = -64, z = 32), or the ant-SPLr (MNI: x = 36, y = -45, z = 51). TMS pulses were delivered using a Magstim Rapid 2 figure of 8 air-cooled coil (70 mm coil width, maximum field strength = 2.2 T, Magstim Ltd, Whitland, United Kingdom). The two parietal locations were localised on the basis of observers’ anatomical MRI scans using the neuronavigation system Visor2 (ANT, Enschede, the Netherlands) utilising an infrared tracking system (Polaris infra-red camera, Northern Digital, Waterloo, Canada). The coil was held manually by the experimenter with its handle pointed directly downwards. The distance between actual coil location and its optimal positioning was kept less than 1.5 mm at all times.

EEG recording:

The 64 channel Waveguard EEG electrode layout based on the extended international 10-20 system, connected to a TMS compatible Refa-64 amplifier (Refa 8, ANT, Enschede, The Netherlands). Two vertical and two horizontal electrooculogram electrodes are included (channel 65 and 66 in the dataset). Impedances kept below 5 kΩ for all electrodes. Recording was digitised at a sampling rate of 2048 Hz.

Epochs were extracted for time intervals from -500 ms (baseline) to plus 2000 ms around the TMS pulses in each condition, without any baseline correction. The data were visually inspected to reject any single trials contaminated by muscle and ocular artefacts. To keep the data as untouched as possible, no filter, downsampling or re-referencing was used.

TMS artefact rejection:

TMS pulses lead to electrical interference that is several orders of magnitude larger than cognitively evoked EEG signals. This is especially problematic since the recording system requires time to recover from any TMS artefact. In our recordings, the data from TMS onset to about 10 ms after are unusable, and up to about 30 ms after the TMS pulse are highly distorted. To overcome this, we used the automated correction algorithm of the ASA TMS-EEG recording software (ANT, Enschede, The Netherlands), which uses a backward infinite impulse response filter on a fixed interval (-25 ms to 10 ms around TMS pulse) and includes a low-pass filter of 100 Hz on that time interval to capture the signal energy of the TMS artefact, which peaks at 300 Hz. Also within that time interval, a least-squares regression equation is used to reconstruct the signal around the artefact on the assumption that the artefact voltage follows the parameters of exponential functions.

The result of this procedure is an EEG signal that appears clean and artefact clear, except for the time interval -25 ms to 10 ms around TMS pulse, which on some trials includes noise several orders of magnitude greater than the signal. Since this time interval both theoretically and practically uninterpretable, we replaced the signal within that window with a linear equation connecting the voltage at -26 ms with the voltage at +11 ms. Consequently, the signal appears as though around the TMS pulse, the time series data is replaced with a straight line.

The dataset:

Each participant’s data is saved in a separate file. The filename reflects the participant number, e.g. Participant 16 is saved under ’17_for_granger.mat’. The file is a Matlab structure array containing eight fields: six matrices containing neural time series data for each condition, a time stamp vector and some information on the EEG channels.

Data data was extracted from EEGLab’s EEG structure array in the following way:

data.cond1 = EEG_bin{1}.data;
data.cond2 = EEG_bin{2}.data;
data.cond3 = EEG_bin{3}.data;
data.cond4 = EEG_bin{4}.data;
data.cond5 = EEG_bin{5}.data;
data.cond6 = EEG_bin{6}.data;
data.time_stamp = EEG.times;
data.channel_info = EEG.chanlocs;

data.cond
The neural time series data is stored in a separate matrix for each condition. The matrix has the dimensions channel x time x trial, e.g. [66, 5160, 120].
66: There are 66 channels.
5160: Each epoche is just over 2.5 seconds long. At a sampling rate of 2048 Hz, this gives us a total of 5160 time points.
120: We originally recorded 120 trials for each condition for each participant, however, due to artefact rejection, this number is smaller in the dataset, for instance for participant 17, condition 1, there are 110 trials.

data.time_stamp
Each participant’s datafile also includes a vector called time_stamp. It tells you how the sampling moments relate to the timing of the TMS pulse. For instance, time 1 in the data matrix corresponds to -509.77 ms prior to TMS. The TMS pulse occurs at time point 1045, and data is set to the linear function mentioned earlier between time points 993 (-26 ms) and 1066 (+11 ms).

data.channel_info
This contains some info about the individual channels 1 to 66, such as label, location coordinates ect, taken directly from how EEGLab stores these.

Last words:

Electrodes P4 and CP4 lay immediately underneath the TMS coil during the recording and their data are essentially unusable due to noise, especially high frequency noise. For most types of analysis, they probably ought to be removed.

There is also a lot of noise over the mastoid electrodes. Re-referencing to them for ERP type analyses might result in spurious findings.

A serious problem with this dataset is that for unknown reasons and only in about 20% of trials, there is an abrupt change in voltage at about 1500 ms post TMS pulse, as if the EEG system were resetting itself. We speculate that this is due to the EEG recorder reaching the maximum absolute voltage it can record and essentially subtracting a set amount from the raw signal to compensate.

MRI scans are available for all participants on request, e.g. for use in source modelling.

Advertisements

Leave a comment below if you like.

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s