Combining EEG and MRI Techniques - Quantitative EEG Analysis Methods and Clinical Applications - page 319

12.1.1 Coregistration
Coregistration brings EEG electrodes and MRI into the same three-dimensional
spatial reference system. Electrode locations can be inferred from their labels or
measured using a three-dimensional pointing/tracking device (digitizer), together
with anatomical landmarks (fiducials) or the subject's headshape.
12.1.1.1 Label-Based
Because of the advent of electrode caps and the increased number of EEG channels,
hardly any lab implements the International 10-20 System [2] or one of the pro-
posed modifications thereof (10% system [3], 5% system [4]) by actually measuring
and subdividing distances on the skull, based on the locations of the nasion, the
inion, and the preauricular points. The electrode labeling scheme brought forth by
these systems, however, is in wide use. As a consequence, it can make sense to infer
three-dimensional electrode locations based on their labels, especially if a
digitization is not available or if data is to be pooled or averaged across subjects on a
channel-by-channel basis.
For identifying labeled electrodes in the MRI, one could identify the electrodes
actually used in a list of stock locations, in conjunction with matching fiducials and
one of the methods described in the next section. Alternatively, one of the measur-
ing schemes defined in the literature could be employed, but performed by a com-
puter algorithm on the segmented skin of the subject's MRI and based on fiducials
identified in the MRI beforehand [5]. In either case, only approximate electrode
locations can be determined.
12.1.1.2 Landmark-Based
Three or more landmarks are digitized together with the electrodes. Landmarks
should be chosen according to two criteria. (1) They should be identifiable on sub-
ject's head as well as in MRI. Anatomical landmarks or MRI-visible markers may
be used. (2) Robust coregistration is required. The smallest coregistration error
occurs for locations close to the landmark's center of gravity. Most EEG labs use
three fiducials: the nasion and two points near the ears (ear points). Because the
preauricular points used for the 10-20 system are hard to identify in MRI, the
tragus, the lower end of the intertragic notch, or the incisura anterior auris are often
a better choice. After landmarks have been located in the MRI, digitizer and image
coordinates can be matched. For this purpose, least-squares fitting can be used.
Because of the one-to-one correspondence between landmarks, it is sufficient to
solve the related orthogonal Procrustes problem [6].
However, if nasion and ear points are used, a landmark-based coordinate sys-
tem is preferable. Such a coordinate system can be defined by placing the origin on a
line connecting left and right ear points, using an x axis through the right ear point
and a y axis through the nasion, with the z axis pointing upward, orthogonal to
both x and y . The advantage of this approach is threefold. (1) The coordinate sys-
tem definition is robust against landmark mismatch along the x and y axes, which
can occur due to pressure applied during digitization and uncertainties regarding
the skin-air boundary in the MRI. (2) Regardless of which information (digitized