Medical visualization and simulation for customizable surgical guides

This thesis revolves around the development of medical visualization tools for the planning of CSG-based surgery. To this end, we performed an extensive computerassisted surgery (CAS) literature study, developed a novel optimization technique for customizable surgical guides (CSG), and introduce three visualization techniques to make the planning more realistic and allow for remote visualization. In Chapter 2 we document the results of an extensive overview study, in which the use of visualization in CAS is analysed. We collected a comprehensive database of visualization relevant CAS publications, and analyse the visualization techniques that are used. We also classify important CAS-related surgical tasks and explain how and why visualization is used. Further, we analysed how surgical plans are transferred to the operating theater. Finally, we discuss how visualization is used in the four most prominent application areas of CAS. Based on this review, we were able to pinpoint interesting new research directions. One of these is the apparent lack of proper tools for CSG-based surgery, a challenge that we addressed in Chapter 3. The optimization of CSG parameters such that the CSG can be docked on bone in an accurate and stable way, is important in the planning of CSG-based surgery. The adjustable nature of the CSG, which allows it to become patient-specific, unfortunately also makes it inherently unstable. Optimizing the configuration by hand leads to poor results as we demonstrated with experiments. In Chapter 3, we therefore solve the problem in sillico. We described a novel planning tool that is able to automatically optimize a CSG for an arbitrary patient. We established this by combining a physical simulator, which models the physical interaction between the CSG and the bone, with a genetic optimization process. With experiments, we were able to prove that our optimization tool produces CSG configurations that lead to accurate and stable docking. In Chapter 4, we address the challenge of enhancing the planning environment with appropriate visualization techniques that help to understand how a CSG is connected to the bone. The state-of-the-art rendering tools in CAS applications are not able to accurately and effectively communicate how the CSG attaches to the bone. However, ambient occlusion (AO) is an illumination technique that is particularly effective at depicting contact between objects, but is generally computationally expensive. Therefore, we developed an efficient version of this algorithm such, that it can be used in the planning pipeline to effectively depict CSG-bone contact. We took the visualization one step further by introducing photo-realistic and physically based volume rendering. Chapter 5 describes Exposure Render, a complete volume rendering framework based on stochastic raytracing, and is able to incorporate a host of otherwise difficult to obtain photorealistic camera, light, and material effects. It is a well known fact that these help to understand shape, depth and size. Therefore, we employed Exposure Render to build a prototype doctor-patient communication system. With this remote visualization system, a doctor can counsel a patient from a distance, or a patient can perform self health management by uploading their tomographic data. In Chapter 6 we optimize the performance of Exposure Render. We introduce visibility sweeps, an efficient method to compute and store visibility information in volume data sets. With this method, it becomes possible to efficiently query approximate global visibility information in a volume data set. We demonstrate that this visibility information can be harnessed to improve the efficiency of the ray sampling processes in Exposure Render, which results in faster convergence. Though we demonstrate the effectiveness of visibility sweeps in the context of stochastic volume rendering, its use stretches beyond this application. Many areas of medical visualization and CAS rely on visibility information, such as automatic view finding in volume data and in various areas of CAS e.g., access, resection and implant planning. In our project it is also relevant because the visibility information can be used to make the physical simulator more realistic, for instance by avoiding docking trajectories that are associated with high risk of tissue damage. The research described in this thesis was part of the project Novel pre-operative planning and intraoperative guidance system for shoulder replacement surgery (10812), funded by the Dutch Technology Foundation.

@PhdThesis { Kro15,
  author       = "Kroes, Thomas",
  title        = "Medical visualization and simulation for customizable surgical guides",
  school       = "Delft University of Technology",
  month        = "September",
  year         = "2015",
  url          = "http://graphics.tudelft.nl/Publications-new/2015/Kro15"
}