Robust Brain MRI Segmentation for 3D Printing Applications

Abstract

Recent advances in computing power and additive manufacturing (3D printing) have now made possible the efficient simulation, optimization, and replication of patient-specific procedures and prosthetics for neurosurgery applications. Two very promising applications are in finite element modeling for brain injury simulation and detection and applying additive manufacturing towards brain analogues or in vivo brain mod-eling. While these applications are very promising, the problem still remains of efficiently segmenting imaging data for use in finite element models or 3D printing. In this project, we put forth a novel algorithm for brain MRI image segmentation that combines statistically-based segmentation techniques with partial differential equation-based methods using neuromechanical models to provide an efficient algorithm for automated brain MRI segmentation. Findings from this project show that combining these segmentation techniques can efficiently segment brain MRI at a level of accuracy required for 3D printing applications. Specifically, we show here that combining nonlinear filtering, k-means clustering, and active contour modeling can produce robust segmentation of brain MRI images. We anticipate that these results will eventually lead to the ability to simulate brain procedures and prosthetics on a patient-specific level by using the segmented images for finite element mesh generation or additive manufacturing processes. When used in conjunction with existing simulation and optimization techniques, image segmentation technology has many far-reaching applications in neurosurgery, and the results from this project have brought some of these applications closer into reach.