Adaptive SVD thresholding is shown to be more appropriate for partial brain scans (TR = 1 s) rather than full brain scans (TR = 2 s)
Smith M, Salluzzi M, Frayne R
University of Calgary, University of Calgary
Application of SVD eigen-value thresholds ensure deconvolution stability but remove high frequency signal and noise components; compromising the accuracy of the CBF estimates. Adaptive SVD thresholds have been suggested for minimizing this error. Severe aliasing related distortions should be expected given that the low temporal resolution available with DSC image sequences does not meet the Nyquist sampling criteria for the residue functions. However, the individual aliasing of the sampled values in the arterial and tissue signals showed as more important. Adaptive thresholding proved appropriate for partial (TR = 1 s) but not full (TR = 2 s) brain scans.