Simulation of cerebrovascular perfusion using realistic synthetic tree generation

Abstract

Ischemic stroke and vascular disease are among the leading causes of death worldwide. The development of accurate models of the cerebrovascular system capable of simulating hemodynamics and perfusion promises accelerated development of treatment techniques to counter these diseases. Further, the generation of patient-specific models strives towards individual assessment of diseased states and personalised planning and prediction of treatments and their outcomes. In this work, a processing pipeline for the patient-specific quantification of cerebrovascular perfusion is developed. The major arteries are segmented from magnet resonance images (MRI). Further, the tissue-growth based tree generation is used to extend vessel networks beyond the resolution of the segmentation by the artificial generation of smaller vessels. To investigate the anatomical accuracy of the generated synthetic tree structures, morphological parameters are compared against experimental data obtained from 7T MRI, 9.4T MRI, and dissection data. Segment-specific parameters were found to depend highly on the tissue-mesh resolution. Bifurcation related data is resembled well, however the inability of MRI based segmentation to resolve small branches hinders the comparability. Using the generated vasculature, cerebral blood flow, pressure, and perfusion are estimated. A 1D approximation of the Navier Stokes equations is employed in the vascular model and coupled to a Darcy flow model at the outlets. Different boundary conditions as well as coupling techniques are employed and compared. The simulated tissue perfusion is compared against measured perfusion data. A qualitative similarity is found, however the simulation fails to resolve the steep gradients in perfusion values observed in the measured perfusion.