To obtain accurate and precise EEG electrode locations with minimal scan time and no additional experimental setup through a UTE sequence optimised specifically to highlight materials from which the electrodes are composed (1).

Sequence parameters:

All experiments were performed on a clinical 3T system (Ingenia, Philips Healthcare, Best, the Netherlands), using a 32-channel head coil for signal reception. Ultra-short TE (UTE) imaging was performed using an axial 3D radial stack-of-stars sequence with parameters: repetition time (TR) = 8ms, echo time (TE) = 0.14ms, flip angle = 10 degrees, field-of-view (FoV) = 240 mm x 240 mm x 170 mm, spatial resolution = 2 mm x 2 mm x 2 mm, acquisition time = 3min29sec. To evaluate the impact of spatial resolution on the electrode position identification process, the same sequence was also repeated on one subject with isotropic spatial resolutions of 3 mm (acquisition time = 1min33sec), 2.5 mm (acquisition time = 2min14sec) and 1.5 mm (acquisition time = 6min11sec).Being a low-flip angle gradient-echo sequence, this UTE sequence leads to a specific absorption rate (SAR) and a RF duty cycle that are very comparable to ones obtained for a typical 3D T1-weighted gradient-echo acquisition, thus making the UTE sequence safe for use with an EEG cap present.

An axial anatomical 3D T1-weighted gradient-echo sequence was also performed with parameters: TR = 7.9ms, TE = 3.5ms, inversion time (TI) = 950ms, flip angle = 8 degrees, FoV = 240 mm x 240 mm x 150 mm, spatial resolution = 1 mm x 1 mm x 1 mm, SENSE factor =2.4, acquisition time = 6min01sec.

Image Processing:

UTE: Raw UTE images were upsampled to 1mm isotropic. A head mask was defined using in-house software, and scalp layers were created at distances 0-6mm from the surface of the head mask by iteratively dilating the mask. Scalp layers were averaged in increments of 2mm, to create three separate layers at 0-2mm, 2-4mm, and 4-6mm from the scalp. The three layers were projected to a pancake view (1), and an RGB image was created by combining the three layers (red=0-2mm, green=2-4mm, blue=4-6mm) (figure 1).

T1: a similar procedure to the UTE was performed, except the layers were defined from 0-3mm inwards from the surface of the scalp (red=2-3mm, green=1-2mm, blue=0-1mm), in order to isolate the gel artifacts (2).

Labeling: Hand labeling was done on the pancake projection with a single mouse click by centering a crosshair over the electrode location. After all electrodes have been labeled, the coordinates are transformed back to volumetric space.

Reproducibility: UTE resolves electrode locations in a subject independent manner, while gel artifacts vary from subject to subject (figure 2). The inter subject variability in gel artifacts on the T1 was not due to image processing deficiencies, as visual inspection of the raw T1 showed that in one subject the gel artifacts were almost entirely absent (figure 3).

Voxel sizes: all electrodes are visible from 1.5mm to 2.5mm, but become more difficult to resolve at 3mm (figure 4).

Validation: visual inspection of back-projected locations show that hand labeled positions closely coincide with electrode locations in MRI space (figure 5).

While we are not the first to use MRI to resolve EEG electrode locations, the variability of the T1 gel artifacts in (2) may be due to a dephasing effect stemming from magnetic susceptibility differences at the gel-skin interface. If this is the case, the T1 gel artifacts would be expected to vary in a subject and scan specific manner, based on the amount of applied gel, subject hair thickness, subject skin composition, and possibly other parameters. Our method, on the other hand, directly images the EEG electrodes and hence provides electrode locations in a subject independent manner.

We tested four separate resolutions (3mm, 2.5mm, 2mm and 1.5mm isotropic), and while it is possible to see most electrodes even at 3mm isotropic, we suggest keeping resolution below 2.5mm in order to obtain the most precise coordinates possible. The 2.5mm resolution scan is 2 minutes and 14 seconds, which is a negligible time increase in the context of most EEG-FMRI studies.

In summary, a fast and reliable method to resolve EEG electrode locations in a subject independent manner with no additional setup time and minimal scan time is presented. The entire pipeline, from acquisition to hand labeling takes less than 10 minutes. We have created an image processing script to enable hand labeling, and encourage the addition of UTE to all simultaneous EEG-FMRI protocols.