Generation of Organ-scale Synthetic Vasculature via Mathematical Optimization
Please login to view abstract download link
The functional assessment of organs is essential for increasing the success of treatment strategies. Detailed information on the organ’s vasculature could critically improve this assessment and our understanding of vascular development. In this talk, we present a new framework for generating the complete vasculature of an organ synthetically. Our main contribution is the formulation of a nonlinear optimization problem (NLP) with super-linear time complexity. In this context, we discuss the addition of an extended problem, capturing the non-Newtonian behavior of blood and constraining pressure and shear stress of vessels to valid physical values. In addition, we show how the problem of multiple, non-intersecting trees and non-convex perfusion domains can be naturally included into the framework. We compare our results against benchmarks of multiple anatomic regions of brain tissue and show that our framework outperforms current state-of-the-art algorithms by an order of magnitude. Furthermore, we generate the complete liver vasculature with over 5,000,000 vessels and compare them against measurements in real liver organs.
