Optimisation and Validation of Dynamic Susceptibility Contrast MRI Perfusion Measurements

Abstract: The studies presented in this thesis concern the optimisation and evaluation of the dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) technique for the assessment of perfusion-related parameters of the brain, such as cerebral blood flow (CBF), cerebral blood volume (CBV) and mean transit time (MTT). Several methodological factors influence these measurements, for example, contrast-agent administration, arterial input function (AIF) registration, choice of deconvolution algorithm and the choice of pulse-sequence parameters. In the first study, a comparison of two different deconvolution techniques was made, i.e., one based on the fast Fourier transform (FT) and the other on singular value decomposition (SVD). The primary result of this study was that CBF estimates obtained by FT-based deconvolution were lower than the CBF values resulting from SVD-based deconvolution. This is in agreement with the results presented in previous publications, demonstrating that the use of FT-based deconvolution underestimates high blood-flow rates (at short MTT). In the second study, perfusion parameters were calculated from simulated data corresponding to different experimental conditions. For example, variations in signal-to-noise ratio (SNR), temporal resolution, AIF shape, signal drop and cut-off level in the truncated SVD deconvolution were investigated. The main conclusions were that the echo time requires optimisation to ensure sufficient signal drop in combination with reasonable baseline SNR, and that a broad input function can lead to underestimation of the CBF. Regional AIFs (rAIFs) were the subject of the third investigation. By using factor analysis of dynamic studies in combination with principal component analysis, rAIFs were obtained and the CBF was calculated by using the rAIF located closest to each tissue voxel. The conclusions drawn from the study were that the use of rAIFs reduced dispersion effects which can lead to CBF underestimation. In the fourth study, CBF was measured in absolute terms in 20 volunteers using Xe-133 SPECT and DSC-MRI. An AIF time-integral correction was introduced in order to improve the absolute CBF quantification in DSC-MRI. Average whole-brain estimates as well as regional CBF values in grey matter (GM) and white matter (WM) were obtained, and the results from the two modalities were compared. For the whole-brain average, the linear relationship was found to be CBF(MRI)=2.4?CBF(Xe)-7.9 [CBF given in units of ml/(min 100 g)], with a correlation coefficient of r=0.76.