1Samsung R & D Institute, Bangalore, India, 2Medical System Lab, Samsung Electronics, Suwon, Korea, Republic of
The standard approaches for performing the deconvolution in
post-processing of DSC-MRI data is singular value decomposition (SVD) and
Frequency-Domain Deconvolution (FDD), which are known to be relatively less
accurate and requires a careful choice of threshold to obtain physiologically
meaningful results. In this work, a method that utilizes QR decomposition to
perform the deconvolution is proposed. The QR method is a well-known
dimensionality reduction technique for solving ill-conditioned linear system of
equations and known to have better numerical properties in obtaining a
regularized solution. It is shown that the proposed method provides superior
results compared to the standard deconvolution methods.
