Evaluation of model-independent deconvolution techniques to estimate lung parenchymal perfusion

Dynamic contrast enhanced (DCE) MRI has become a useful technique for measuring perfusion in many applications. !e theoretical basis for perfusion quanti"cation is the central volume principle, which requires deconvolution of the measured arterial input and tissue concentration curves to derive a residue function (R) and mean blood perfusion. Deconvolution methods generally di#er in assumptions of the global shape of R, computational stability and oscillations in estimated R. !erefore, several deconvolution methods to estimate R and blood perfusion are evaluated in this study. Among these are model-dependent and model-independent techniques. All methods were applied to series of Monte Carlo simulations to evaluate the accuracy to reproduce di#erent underlying shapes of R and blood perfusion. Of the model-independent approaches the use of B-splines with Tikhonov regularization had a reasonable accuracy in perfusion estimations and was less biased than all model-dependent approaches. !is technique is most promising for application to experimental data.