Signal-to-noise ratio effects in quantitative cerebral perfusion using dynamic susceptibility contrast agents

Theoretical and simulation evidence is presented in support of the idea that the optimal manner of determining blood flow from MR perfusion studies is not necessarily obtained by setting experimental conditions to maximize either the arterial input or the measured tissue concentration level for a particular echo time (TE). The noise power in the contrast concentration curve is associated with its peak because of the nonlinear relationship between the contrast concentration and MR signal intensity curves. The optimum signal-to-noise ratio (SNR), SNRC, for a particular contrast concentration curve can be obtained when the experimental concentration level and TE are adjusted to produce an MR intensity curve whose signal loss is 63% of the precontrast MR signal intensity. It is demonstrated that the stability of the singular valued decomposition (SVD) deconvolution approach to determine blood flow parameters is increased when the tissue curve maximum signal loss is in the range of 40-80%. The accuracy and stability of the SVD-determined blood flow parameters are affected by deviations from these optimum conditions in a manner that depends on the mean transit time (MTT) associated with the residue function. It is recommended that the experimental TE value be set so that neither the tissue nor the arterial curves are placed a region of rapidly deteriorating SNRC. (C) 2003 Wiley-Liss, Inc.