Hoen-Oh Shin MD, Presenter: Nothing to Disclose

Lars Bornemann MS, Abstract Co-Author: Nothing to Disclose

Volker Dicken PhD, Abstract Co-Author: Nothing to Disclose

Jan-Martin Kuhnigk, Abstract Co-Author: Nothing to Disclose

Dag Wormanns MD, Abstract Co-Author: Nothing to Disclose

For assessment of tumor size, calculation of volume is more precise than uni- and bilateral measurements. We present a new segmentation algorithm for volumetric assessment of hypodense hepatic tumors.

A multi-center study was performed on contrast-enhanced computed tomography studies of 21 patients with 129 hypodense lesions of the liver (127 metastasis, 2 cysts). The volume of each lesion was determined twice by a first reader and by two other readers using the software. Tumor size ranged from 6 to 49 mm (median 14.5mm) with a median density of 65 HU (range 20-98 HU). All CT scans were performed using a slice thickness of 1 – 3 mm. A new hybrid segmentation algorithm based on local histograms, thresholding and morphologic filtering was applied for determination of tumor size. Observer variability of segmented total volume as well as the number of user interactions and the time required for segmentation were assessed.

Median time for automatic segmentation was 1.5 sec. Calculation time increased with lesion size. However, it was less than 3 sec. per lesion on current personal computers for all evaluated lesions. 88 % of all measurements were successful, only 31% of those required minor user interaction. Only for 6 lesions no acceptable segmentation was obtained. For 75th / 90th quantile of all lesions, the standard deviation of measured volume was 12% / 27% corresponding to 4 % / 9 % deviation in diameter.

Volumetric measurement of tumor size is feasible in clinical routine work using the proposed algorithm.

Shin, H, Bornemann, L, Dicken, V, Kuhnigk, J, Wormanns, D, Volumetric Analysis of Hypodense Liver Tumors in Multi-Detector Row CT: Evaluation of a New Segmentation Algorithm.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4420199.html