Spatially interpolated group averaging with 256-channel event-related potential data

• Published in: International Congress series Vol. 1232; pp. 389 - 395
• Main Authors: Suwazono, Shugo, Pflieger, Mark E., Nakada, Tsutomu
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.04.2002
• Subjects: 3D interpolation; Auditory oddball paradigm; High-density EEG; Registration; Topographic mapping; High-density EEG; 3D interpolation; Auditory oddball paradigm; Registration; Topographic mapping
• ISSN: 0531-5131; 1873-6157
• DOI: 10.1016/S0531-5131(01)00847-0

While conventional group averaging assumes that electrode location differences across subjects are small compared with inter-electrode distances, 256-channel event-related potential (ERP) datasets have by necessity much shorter inter-electrode distances (1.7 cm) that can invalidate this assumption even with the meticulous application of the electrodes. To ameliorate this problem, a newly developed spatial interpolation method was applied whereby each subject's potential distribution was interpolated onto a “common space” (3D electrode coordinates averaged across subjects) before group averaging. Using auditory oddball data, a subject's “own space” distribution (mapped using individually digitized 3D electrode coordinates) was compared with the common space distribution, with and without interpolation. Direct mapping of individual ERPs onto the common space (without interpolation) resulted in distorted topographies compared with the “own space” maps. Interpolation minimized this distortion. The net result across subjects was that spatially interpolated group averages distributed more compactly and compared smoothly to the conventional group averages. These findings support the interpretation that interpolation spatially “aligned” each subject's ERP prior to group averaging, whereas the conventional method spatially “blurred” the result by combining individually distorted scalp ERP distributions. Spatial interpolation is therefore recommended for calculating group averages of high density ERP data.