3D Segmentation and Visualization of Human Brain CT Images for Surgical Training - A VTK Approach

Abstract

Simulated surgical planning and training has been proved to be effective in enhancing the performance of surgical operations. 3D medical image modeling and visualization is therefore gaining increasing attention in the research community as the navigation using 3D DICOM data provides a more realistic planning and training environment. However, medical applications often have specific targets, e.g. TB, cancer and tumor. Algorithms developed for the respective applications are designed based on the characteristics, like shape and intensity, of the target. In order to analyze the target for symptom diagnostic or monitoring purposes, the target region needs to be segmented out from its background so that the algorithm output will not be adversely affected by irrelevant signals close to the target. In this paper, we focus on segmentation and visualization, with a use case of developing a 3D environment for Ventricular puncture operation planning and training. The difference between our work and other segmentation techniques is that we need to segment not only one target, but also the path along the surgical tool inserted into the brain. This creates challenges to the algorithm design because a set of segmentation parameters may be effective for one region, but may not be effective for another due to the different data region contrasts, densities, shapes and so on. Segmentation is only an initial step but is necessary in order to conduct the actual surgical training. While many researchers or clinicians waste effort in generating segmentation results, our contribution lies in our VTK approach, which is fast to implement so that the users can focus on the core process. Our experimental results demonstrate the feasibility of providing a realistic 3D visualization and interactive environment for surgical planning and training.