Comparison of blind deconvolution- and Patlak analysis-based methods for determining vascular permeability

Abstract

This study describes a computational algorithm to determine vascular permeability constants from time-lapse imaging data without concurrent knowledge of the arterial input function. The algorithm is based on "blind" deconvolution of imaging data, which were generated with analytical and finite-element models of bidirectional solute transport between a capillary and its surrounding tissue. Compared to the commonly used Patlak analysis, the blind algorithm is substantially more accurate in the presence of solute delay and dispersion. We also compared the performance of the blind algorithm with that of a simpler one that assumed unidirectional transport from capillary to tissue [as described in Truslow et al., Microvasc. Res. 90, 117-120 (2013)]. The algorithm based on bidirectional transport was more accurate than the one based on unidirectional transport for more permeable vessels and smaller extravascular distribution volumes, and less accurate for less permeable vessels and larger extravascular distribution volumes. Our results indicate that blind deconvolution is superior to Patlak analysis for permeability mapping under clinically relevant conditions, and can thus potentially improve the detection of tissue regions with a compromised vascular barrier.

Keywords: Compartmental model; Computed tomography; Magnetic resonance imaging; Microcirculation; Patlak equation.