Cognition Network Technology for a Fully Automated 3D Segmentation of Liver
Günter Schmidt, Maria Athelogou, René Korn, Gerd K. Binnig · 2007
Abstract. Accurate and fully automated segmentation and quantification of organs are necessary prerequisites towards operational computer aided diagnosis and therapy control systems. The Definiens Cognition Network Technology presented in this paper provides a framework for rapid development and scalable execution of image analysis solutions. The Cognition Network Technology combines pixel processing techniques with iterative, context-based object or segment generation and classification processes using a semantic knowledge base. We applied this technology to detect automatically all major human organs in 3D CT data. On basis of 10 test data sets we show first quantitative results for liver detection compared to manual segmentations provided by medical experts. Initial analysis achieved a mean overlap error of 16 % which provides a good starting point to improve the implemented knowledge network by liver disease specific rules. 1