In the context of the colloquium of the Research Training Group Algorithmic Optimization on

Monday, May 15, 2023
16:00 c.t.
lecture hall 10

the following talk will take place:

NLP models for generating synthetic vascular trees.

Marc Steinbach, Leibniz University of Hannover

We present a new framework for generating synthetic vascular trees based on NLP models for optimizing local and global tree geometry. A second novelty is a metaheuristic optimization step for tree topology.

The talk discusses these optimization models and compares the new algorithm with the standard CCO approach for generating vascular trees. It also compares the impact of different model variants on trees with 14 branching levels and about two million nodes.

Validation using a corrosion cast of a human liver shows that the trees generated by our new method match the physiological data better than the trees generated by the standard CCO method.