The liverplanner system has been developed by Alexander Bornik and his team to address segmentation and visualization problems which occur in the planning stage of liver resections and during the actual liver operations. The project was set up by the Institute for Computer Graphics and Vision at the Graz University of Technology in Austria in collaboration with the Departments of Radiology and Surgery at the University Hospital Graz and with the Department of Electrical and Computer Engineering at the University of Iowa in the United States.
According to the winning EG2003 Medical Prize team, resection is the treatment of choice for patients suffering from liver tumours. The analysis of 2D patient CT images constitutes a tedious and time consuming process for both radiologist and surgeon. This is where the aid of advanced computer graphics technology comes in to support the surgery planning procedure by means of fully automated segmentation of the liver to visualize its vessels and possible tumours and in addition to help assess the surgical approach of liver resection through an interactive interchange of medical sensor data and surgical tools with complex 3D anatomical structures in augmented reality.
In fact, the liver surgery planning system (LSPS) has two main components, as demonstrated by the team: a medical image analysis and an augmented reality (AR) part. The medical image analysis is being used for segmentation of the liver, its vasculature, possible tumours and liver segment approximation. The AR system is utilized for visualization and various kinds of user interaction. The AR hardware set-up as shown by the team consists of a stereoscopic see-through head-mounted display (HMD), an optical tracking system, tracked input devices, and a rendering and tracking workstation.
The see-through HMD allows both surgeon and radiologist to create a realistic interaction between the liver surface, its vessels and tumours generated as virtual objects and the surrounding world. To this purpose, tracked input devices such as a tracked pencil (PEN) and a transparent plexiglass Personal Interaction Panel (PIP) are used to study the virtual objects from all possible angles and distances. The PEN is equipped with buttons to trigger input events like target selection, measurements, and CT snapshot acquisition in 3D space.
For the automated liver and tumour segmentation, the team developed a robust algorithm to address problems related to an accurate separation of the liver from the neighbouring anatomical structures. This is realized through high level shape and appearance knowledge. A real challenge to the team consitutes the segmentation of the smaller vessel branches in the liver. In this regard, a fuzzy connectivity and an adaptive region growing approach have been investigated. The team also announced that experiments with segment partitioning using a fast skeletonization algorithm are now in progress. The results will be published by the end of 2003.
The team admitted that automated segmentation might fail occasionally. To meet this problem, the AR system has two function blocks. The first is implemented for quality assessment and for the editing of the automatically segmented liver structures. The second is for the intrinsic planning of a resection. The system enables the radiologist to display the surface representation of the segmented structures in context with the original patient CT data in cross sections and, if needed, as volume rendering. Once the accuracy of all reconstructed liver structures has been approved by the radiologist, the surgeon can start with the resection planning.
In the resection planning process, the same tools are used as for the preceding phase of visual inspection. The team explained how the LSPS system is able to quantify the total liver volume and the volume of individual liver segments or tumours. There are two different methods for volume calculation: a fast voxelization to assess the liver volume and an algorithm to apply fast geometric operations on the mesh. If a resection is inevitable, the physician draws up a plan based on information gained from the visualization.
Alexander Bornik and his team also noted that an expansion of the core system for intra-operative applications is possible. As such, one platform can be installed for planning as well as support during surgery. The system is also suitable for telemedicine and educational applications. The LSPS system will be validated in co-operation with the medical partners in the project. For more information on the liverplanner, you can consult the Web site of the Institute for Computer Graphics and Vision.