Interactive Segmentation Refinement: Towards a Clinically Applicable Universal Segmentation Tool

by

Dr. Alexander Bornik Computer Graphics and Vision Graz University of Technology

Abstract:

Volumetric datasets from CT, MR and other sources facilitate diagnosis as well as treatment planning and therapy monitoring. In order to optimally support physicians it is crucial to extract geometrical model of anatomical structures through segmentation, which can be used for better visualization and calculation of quantitative information. However, the segmentation problem is hard. This is especially true for application in clinical routine, where robustness and time efficiency are needed. Current segmentation techniques fail in terms of robustness, whereas manual or common interactive segmentation techniques don't meet the time constraints. An interactive segmentation refinement system can overcome the problem by reusing erroneous automatic segmentations by providing efficient tools and an intuitive user interface to correct them. An example designed for but not limited to liver segmentation in the context of virtual liver surgery planning will be presented. Furthermore, the segmentation refinement paradigm, potential future application areas as well as limitations will be discussed.

Short CV

• 1996-2001 Graz University of Technology: Telematik
• 2001 MSc: MSc-Thesis: Texture reconstruction from multiple views
• 2001-2007 Graz University of Technology: PhD Student - Virtual Liver Surgery Planning, teaching - computer graphics/VR, responsible for VR hardware
• 2006-Feb 2008 working for the Presenccia EU project - Ubiquitous tracking