Binsheng Zhao DSc, Presenter: Nothing to Disclose

Lawrence Howard Schwartz MD, Abstract Co-Author: Nothing to Disclose

Li Jiang MS, Abstract Co-Author: Nothing to Disclose

Jane Anna Clare Colville MBBCH, Abstract Co-Author: Nothing to Disclose

Liang Wang MD, Abstract Co-Author: Nothing to Disclose

Robert Andrew Lefkowitz MD, Abstract Co-Author: Nothing to Disclose

John P. Kalaigian BS, Abstract Co-Author: Nothing to Disclose

et al, Abstract Co-Author: Nothing to Disclose

To develop and validate a computerized method for automatic measurements of hepatic metastasis tumors on contrast-enhanced CT images.

A shape-constrained region-growing technique has been developed to automatically delineate liver tumors on the contrast enhanced CT images. The algorithm first divided the range of pixel attenuations into several sub-ranges and then applied different inclusion/exclusion strategies to the to-be-processed pixels whose attenuations falling into different sub-ranges. In addition, a local shape, global shape and gravity shift index were introduced into the iteration of the algorithm to prevent the lesion growing from leaking into its surrounding tissues of similar densities or textures. 54 representative lesions in 15 patients with the liver metastases from colorectal primary tumors were automatically delineated using this algorithm. Subsequently, uni-dimension, bi-dimension, and area were automatically calculated for each of the segmented lesions. 3 independent Radiologists manually measured the same lesions once and one of them repeated the measurements in another 2 separate sessions. Statistical T-test was used to compare measurement difference and coefficient of variation was used to assess variability.

The mean uni-, bi-dimension, and area of lesion size were 43.1 mm (+/- 3.7 std), 1926.3 mm2 (+/- 325 std), and 1282.3 mm2 (+/- 211 std) measured using the automated technique, and 39.9 (+/- 3.4 std), 1637.3 (+/- 274 std), and 1146.1 (+/- 188 std) by 3 Radiologists’ average measurements, respectively. There was no statistically significant difference between the measurements obtained by each of the Radiologists and by the computer. Inter-observer variability for the 3 Radiologists and intra-observer variability for the single radiologist was statistically greater than the variability between each radiologist and computer for both bidimensional and area measurements.

Measurements performed on hepatic metastases are equivalent for lesions that are extracted automatically and lesions that are outlined manually. Computer segmentation may lead to less variability in measurement as compared to manual measurements performed by Radiologists.

Zhao, B, Schwartz, L, Jiang, L, Colville, J, Wang, L, Lefkowitz, R, Kalaigian, J, et al, , Automatic Measurements of Hepatic Metastases on Contrast-enhanced CT Scan: Comparison with Radiologists Manual Results.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4411590.html