Meeting Banner
Abstract #3070

CBF Quantification in the Face of Dispersion and Separation of Its Effects from Normal Cerebral Haemodynamics: A Comparison of Deconvolution Methods in DSC-MRI

Amit Mehndiratta1, Fernando Calamante2, Bradley J. MacIntosh3, David E. Crane3, Stephen J. Payne1, Michael A. Chappell1

1Institute of Biomedical Engineering, University of Oxford, Oxford, Oxfordshire, United Kingdom; 2Melbourne Brain Centre - Austin campus, Brain Research Institute, Heidelberg, Victoria, Australia; 3Medical Physics, University of Toronto, Toronto, ON, Canada

Bolus dispersion is a significant problem in perfusion imaging because of remote AIF measurement. It is challenging to accurately estimate CBF in presence of dispersion; in this study our motivation was to investigate the accuracy in CBF estimate with a number of deconvolution methods, and also investigate the possibility to separate the effects of dispersion from cerebral haemodynamics using these techniques. The ambiguity between true residue function and dispersion could not be resolved with any of the deconvolution methods, but accurate estimates of cerebral perfusion can be achieved with an effectively non-parametric CPI approach even in presence of dispersion.

Abstract PDF Presentation Video

Keywords

accuracy accurate achieve achieved actual added address allowed ambiguity amplified amplitude applied approaches arrival arterial assessed assuming assumption audience better biomedical blood bolus brain calculating campus cerebral characteristics characterize clearly coefficient coefficients complete conditions considered constant constrained control converted corrected course crane cubic decomposition deconvolution degree delay determination dispersed dispersion distant distinguish effectively errors estimation evaluated even existence exponential extended face findings fitting fixed flow form function functions gamma generated ignored improved in vivo initial institute insufficient interpolation interpretable introduces kernel kernels known local longer made medical might model models moderate noise offer operation original oxford parametric particularly past pathology peak performance perfusion physics practical practice precise priori problem proposed quantification reasonably recently regression relation required residue resolved response restriction separate separation shape simulated simulations singular site smooth solutions specifically spline still subject sufficient superior table target terms though tissue transport true typically underestimation unity unknown validated variable variants varied vary vascular whilst