The Application of Embedded and Tubular Structure to Tissue Identification for the Computation of Patient-Specific Neurosurgical Simulation Models

Abstract

This paper presents a method for identifying tissue classes of the head from co-registered MR and CT, given a small set of training points of each class, in a manner that exploits the embedded or tubular structure, as well as the contiguity, of each tissue. Tissue maps produced by this method are then applied to a suite of patient-specific anatomical models for the simulation of trans-nasal pituitary surgery, through subsequent tissue-guided surface and volume meshing. The method presented here overcomes shortcomings of a previous method of ours that exploited the embedded structure and contiguity of tissues to produce a tissue map that accounted for intra-cranial, intra-orbital and extra-cranial regions. Specifically, the assumption of embedded structure breaks down for critical tissues such as vasculature and cranial nerves, which often straddle two regions. This paper presents a method that first identifies critical tissues, beginning with a Minimal Path (MP) through user-specified points from each critical structure, generally coinciding with loci of strongly negative outward Gradient Flux, known as sink points. The rest of the critical tissue voxels can then be identified by proceeding radially from the GF-weighted minimal path through sink points. The final step is the identification of the remaining tissues, making use of embedded structure and contiguity, based on our previous method.