Purpose: We assessed errors in cerebral blood flow (CBF) obtained from our proposed reference-based method without using arterial input function (AIF) indices in dynamic susceptibility contrast (DSC) magnetic resonance imaging (MRI).
Materials and Methods: We calculated CBF and the referential tissue-related ratio (CBFratio) using numerical simulation and 3 nondeconvolution methods and a deconvolution method of block-circulant singular value decomposition (cSVD). We compared errors with and without simulated noise as parameters of mean transit time (MTT), AIF delay and temporal resolution, and clinical DSC-MRI maps.
Results: The errors in CBF obtained using maximum upslope (US) were smallest among the nondeconvolution methods and almost equivalent to errors in the cSVD method under practical imaging conditions. In addition, errors in the CBFratio obtained using reference-based US (Ref-US) referring to white matter were smallest, even compared to all errors in CBF and CBFratio. The Ref-US method introduced less error than the cSVD method, especially at low flow rates, was further robust against AIF noise and coarse temporal resolution, and was comparably robust against transit delay. In pixel-by-pixel correlations between absolute value maps for US and for cSVD-CBF in clinical DSC-MRI, those correlation coefficients (r) between the 2 maps were stable, r>0.9, despite variation in the slopes of the linear regression line, so the 2 CBFratio maps were visually well correlated in any case.
Conclusion: The Ref-US technique without AIF measurement can become a practical perfusion methodology for DSC-MRI in patients even with acute stroke because it balances robustness with systematic and random errors and it is simply performed.