| |
DOG Congress Home
Invitation
Organization, Deadlines
Overview of the Congress
Scientific Programme
Opening Ceremony
Ceremony 150 Years of DOG
Thursday, 20.September
Friday, 21.September
Saturday, 22.September
Sunday, 23.September
Poster Sessions
Symposia
Courses
Satellite Programme
Information
Social Programme
Sponsors, Exhibitors
DOG Homepage
|
|
AbstractFR.18.07 Virtual electrode rotation with the multifocal VEP Bach M. Sektion Funktionelle Sehforschung, Universitäts-Augenklinik Freiburg Objective: The mfVEP is known to vary markedly between visual field locations in amplitude and polarity. This is due to the convoluted folding of the cortex and the ensuing variation of dipole projection on the recording electrodes. The problem has been addressed by increasing the number of channels, and selecting the largest trace across channels for each location (WTA = winner takes all-algorithm). Two problems remain: the polarity of the traces is somewhat arbitrary (not all traces obey the inversion rule at the horizontal meridian), and the statistical properties of the WTA results are unclear, since its derivation involves a non-linear step. Methods: A novel way to combine traces across n (two or more) channels, alike to a virtual rotation, aligns the electrodes to the local dipole by computing a weighted sum of all channels. The weights [-1, +1] are subject to boundary conditions derived from n-dimensional spherical coordinates. The specific weights are selected to maximise a response measure, namely the cross-correlation with a template. The largest trace across all channels is chosen as template, this trace is flipped to a standard polarity. The cross-correlation extends through a time window of ±10 ms, which allows latency mapping. Results: In 12 subjects with 3 channels each, the mean SNR of the rotated channel mix across all traces was 32% better compared to the SNR of a WTA channel combination. In all traces with a significant SNR the trace polarity was P100 up after the rotation. Conclusions: The virtual rotation method to combine mfVEP channels produces about 30% of response enhancement. Perhaps more important is the regularisation of the trace polarity. This allows to apply all standard peak measures to the mfVEP and to average across trace pools without the noise-prone quadratic RMS values.
|
|