Magnetic resonance perfusion quantification: the advantages of frequency-domain modeling and the impact of partial volume effects
In dynamic-susceptibility contrast (DSC) magnetic resonance (MR) perfusion imaging, the cerebral blood flow (CBF) is estimated from the tissue residue function obtained through deconvolution of the arterial input function (AIF) and the tissue contrast concentration function. The reliability of CBF estimates obtained by deconvolution is sensitive to various distortions including high-frequency noise amplification and signal under-sampling. In the presence of noise or contrast recirculation, the most commonly used time-domain singular-value decomposition (SVD) -based algorithms introduce biases into the CBF estimates, the degree of which varying with tissue mean transit time. In addition, the arterial concentration signal measurements are distorted due to partial-volume effects (PVE). By modeling the residue function in the frequency domain, it is shown that the desired signal can be recovered from amidst the above distortions to provide improved CBF estimation accuracy. The advantages and applicability of this novel approach are explored by characterizing the residue function through a simple frequency-domain Lorentzian model (FDLM). As the performance of the FDLM method is model-dependent, this approach does not represent an ideal frequency-domain modeling technique; however, FDLM is shown to be applicable to several clinically relevant tissue models, and resulted in decreased CBF estimation error compared to SVD. The results also suggest that the calibration factor needed to obtain absolute CBF estimates varies with the deconvolution algorithm, and that the FDLM calibration factor is less sensitive to tissue mean transit time than that for SVD-based approaches. In addition, in the presence of PVE, the accuracy of CBF estimates has been shown to deteriorate at a much faster rate than increases in the degree of PVE in AIF measurement. To correct for PVE, the nonlinear relationship between changes in MR signal intensity and arterial concentration must be considered. A new PVE-estimation and correction method was proposed, in which the MR signal baseline intensities were used to estimate the partial-volume fraction, from which it was possible to correct for AIF distortions caused by PVE. Practical application of this method requires more reliable source data, and hence improvement in image quality.
Bibliography: p. 122-130
Chen, J. J. (2004). Magnetic resonance perfusion quantification: the advantages of frequency-domain modeling and the impact of partial volume effects (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/19847