The Virtual Surgical Pelvis: Anatomy Visualization for Education and Surgical Planning

This thesis deals with visualizing anatomical data for medical education and surgical planning purposes. To this end, we have developed a detailed virtual atlas, the Virtual Surgical Pelvis (VSP),which unifies surgically relevant knowledge on pelvic anatomy. We provide methods to share the knowledge contained in the VSP for educational purposes, and to visualize the VSP in the context of individual patients for pre-operative planning purposes.

Chapter 2 deals with a representation developed to unify spatial and non-spatial
anatomical knowledge. Via this representation, it is possible to store, access and visualize these heterogeneous datasets through a shared coordinate system. This allows us to construct the VSP atlas, a process which we describe in detail in Chapter 3, where we also detail the application potential of the VSP. We present several examples of the VSP mapped to clinical pre-operative MRI scans, as examples of how the VSP can be used to enrich clinical data with surgically relevant information that is not available from the scans themselves.

To share the VSP for educational purposes, we present an online tool, the Online Anatomical Human (OAH) in Chapter 4. OAH runs directly in the browser and can be used to explore the complex relation between 2D and 3D anatomy. Furthermore, annotations can be added directly on the 3D structures for quizzing purposes, or to enrich the VSP further with annotations performed by experts. The OAH was successfully deployed in aMassiveOpenOnline Course (MOOC), where thousands of studentsworldwide used the application to study pelvic anatomy.
While the VSP is based on multiple datasets, it does not include all potential topological anatomical variations in branching structures such as vessels and nerves. Illustrations and text are traditionally used by medical specialists to study these variations, but it is difficult to compare complex variations in such illustrations. Therefore, in Chapter 5 we present an interactive visualization application for anatomical variations, which allows the user to compare and explore variations of branching structures interactively for educational purposes. With methods inspired by graph theory, users can intuitively select groups of variations, based on a similarity measure, and compare local differences.

In Chapter 6, we present a state-of-the-art report on multimodal medical visualization. We describe the basics of medical image acquisition, and the clinical workflow for dealing with such data. We discuss suitable rendering and visualization techniques appropriate for rendering multiple modalities. The core contribution of this work is a taxonomy based on the multimodal medical visualization applications so far, the visualization techniques they employ, and the medical domain context. Additionally, we provide an outlook on open problems and potential future research directions.
To make the VSP patient-specific and to enrich the VSP with more datasets, registration is needed. Unfortunately, current registration software is often difficult to use for non-medical-imaging-experts. In Chapter 7 we present a new registration application, RegistrationShop, that allows user to register 3D medical image datasets based on 3D visualizations and simple interactive transformation tools. Based on real-time visual feedback via comparative visualization techniques, users can inspect the current registration result and iteratively improve the alignment. Besides basic interactive transformation tools, we propose a novel way of placing corresponding landmark-pairs in 3D volumes.

After combining the VSP atlas with patient-specific pre-operative MRI scans, we visualize the results in an interactive application for surgical planning aimed at pelvic oncological procedures, entitled PelVis, which is described in Chapter 8. We present visualization methods to represent context, target, and risk structures for surgical planning of the Total Mesorectal Excision (TME) procedure. We employ distance-based and occlusion management techniques to represent the patient-specific pathology and anatomy. Furthermore, we visualize the confidence in the registration outcome in relation to the distance of the target structure to the risk zones.
The research described in this thesis was supported by the Dutch Technology Foundation STW via project 10903: “High-definition Atlas-based surgical planning for Pelvic Surgery”.