Discussion of Discriminating Multiple Sources of SEF Overlapping in Time using SVD Method
Bongsoo Kim, Y. Uchikawa · IEEJ Transactions on Fundamentals and Materials · 2003
This paper describes a new method with singular value decomposition (SVD) for discriminating multiple sources in the brain overlapping in time. We carried out a 3-D vector measurement of somatosensory evoked field (SEF) with the electric stimulus to the right thumb, using a 3-D second-order gradiometer connected to 39-channel SQUIDs, which can detect the magnetic components perpendicular to the scalp (Br) and tangential to the scalp (Bθ, Bφ) simultaneously. The SVD method was applied to each spatio-temporal data, having two frequency ranges (1-40Hz, 15-40Hz), with the time window of 25ms and it was shifted every 2.5ms in analysis. Two dominant singular values were obtained for each of three components, respectively. It was found that magnetic field distribution with 15-40Hz was different to that of 1-40Hz at around 92ms latency corresponding to the SII activity overlapped in the SI activity. It was also confirmed a good agreement between isofield contour maps of measured magnetic field and theoretical maps calculated by signal subspaces with SVD analysis. We conclude this proposed method using dynamic SVD method with different frequency ranges is useful for discriminating multiple sources overlapping in time.