Comparison of three deconvolution techniques in renography.

The importance of comparing various deconvolution techniques on real patient data lies in the fact that there are two different approaches to that problem. One, which prefers general-purpose techniques, using a matrix algorithm (MA) and fast-Fourier transformation (FFT) with various smoothings, filters, etc., and another which prefers deconvolution techniques which have been modified for the particular application. The modification may be in requiring well defined input or assuming a smooth result of deconvolution. The aim of this paper is to compare the general-purpose deconvolution technique mostly used in analyzing patient data in nuclear medicine using MA and FFT with deconvolution based on Laplace transformation (LT). They have been compared on real patient data renograms in 36 cases. The comparison shows that although the LT technique requires a well defined input and for that reason needs more computing time than the FFT and MA technique, it gives impulse response functions (IRF) without negative values or with negligible ones. Even more, using LT, which has been incorporated in a model of distribution and transport of tracer, without taking blood samples one can assess some other useful clinical parameters such as inverse value of clearance time constant (ICTC) and effective renal plasma flow (ERPF).