Bilal Ahmed MD, Presenter: Nothing to Disclose

Rene Korn, Abstract Co-Author: Employee, Definiens AG, Munich, Germany

Markus Kietzmann, Abstract Co-Author: Employee, Definiens AG, Munich, Germany

Johann Kim, Abstract Co-Author: Employee, Definiens AG, Munich, Germany

Guenter T. Schmidt, Abstract Co-Author: Employee, Definiens AG, Munich, Germany

Masoom A. Haider MD, Abstract Co-Author: Nothing to Disclose

Gerd K. Binnig, Abstract Co-Author: Employee, Definiens AG, Munich, Germany

00030490-DMT et al, Abstract Co-Author: Nothing to Disclose

Liver volume is routinely used in liver surgery and volumetric measurement of metastatic disease is useful in assessing tumor response to therapy. Manual volumetry of liver is a tedious, time-consuming and subjective process. This has led to a growing interest in the development of fast and accurate liver segmentation methods. The purpose of this study is to retrospectively evaluate the accuracy of liver and liver metastases volume measured by a fully-automated software platform tailored to segmentation of medical images developed by the Definiens Cognition Network Technology®. The segmentation algorithm quantifies the liver and its lesions in a fully automatic manner by identifying context information from the lung, spine, ribs and gall bladder.  

Institutional Review Board approval was obtained. A total of 30 consecutive portal phase CTs from distinct patients who had liver metastases were obtained. Scans were performed at 1mm collimations on a 320 slice CT scanner. All CTs were manually segmented to define reference standard boundaries for liver metastases and liver by author BA. Five representative cases were selected as a training set for algorithm refinement. The remaining 25 cases were used to test the resultant segmentation algorithm. Automated segmentations generated were compared to the expert generated references and deviations were measured.  

The correlation coefficient between the measured and estimated volumes obtained from Definiens segmentation algorithm was 0.864 ± 0.04 (range, 0.777 – 0.935). The mean volumetric overlap was 77% ± 6 (range, 64 – 88%). Mean liver volume measured by the automatic segmentation algorithm and manual segmentation was 1820 cm³ ± 788 (range, 710 – 4565 cm³) and 1806 cm³ ± 685 (range, 934 – 3460 cm³), respectively. The mean total time for the automatic segmentation was 4.8 minutes ± 1.2 on E6750 @2.66GHz Intel Core™ 2 Duo CPU with 3.25 GB RAM.  

An evaluation of the automated segmentation algorithm developed by Definiens shows promising results for segmentation of liver and liver metastases.  

Automatic, accurate and fast liver and liver tumor volumetry has important applications in clinical medicine, including liver surgery and assessment of tumour response to therapy.  

Ahmed, B, Korn, R, Kietzmann, M, Kim, J, Schmidt, G, Haider, M, Binnig, G, et al, 0, Automated CT-based Liver and Metastases Volume Assessment.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8016546.html