Abstract:
HTML
Volumetric measurements may be better than dimensional measurements for monitoring tumor growth or shrinkage in clinical trials of anti-tumor agents. The advantage of volumetric measurements has become more apparent with the increased availability of modalities that provide thin-section images such as helical computed tomography (CT). However, measurement of the volume of tumors remains problematic due to limited image contrast between normal and cancerous tissue and to complexities of tumor shape. Simple methods for tumor segmentation such as region growing or manual contouring are highly subjective. The purpose of this research was to develop improved methodology for measurement of tumor volume with emphasis on the assessment of metastatic lesions in the liver on contrast-enhanced CT. The methodology presented here incorporates the fuzzy c-means (FCM) algorithm that provides an optimal intensity-based segmentation of an image. The FCM segmentation is applied to the image after low-pass filtering to suppress noise and gross manual contouring of the lesion to mask-out obviously extraneous tissue. The gross manual contouring must include both the lesion and about a one-half centimeter-wide swath outside of the lesion and must exclude blood vessels and any extra-hepatic region. FCM segmentation is obtained with 3 clusters (cancerous tissue, normal liver tissue, and masked-out region) and an objective-function exponential parameter of 2.0. All processing was performed with MIPAV software (Center for Information Technology, National Institutes of Health). The methodology was applied to helical CT images obtained with contrast during the portal-venous phase. The study included a total of 11 liver lesions in 7 patients. Each lesion was segmented twice by manual contouring and twice using FCM, all by the same operator. Intra-observer variability using FCM was less than for manual contouring for 9 out of the 11 lesions (p = 0.0137). The mean percent variability was less for FCM (2.9% ± 2.7%) than for manual contouring (7.4% ± 4.4%) for lesions greater larger than 1 cc (p = 0.16, n = 10). FCM volumes were highly correlated with manual contouring volumes (r = 0.99985) In summary, an improvement in the reproducibility of tumor-volume measurement is obtained using the FCM.
Questions about this event email: yimpj@umdnj.edu
Yim PhD, P,
Reduction of Intra-Observer Variability of Tumor-Volume Measurement Using the Fuzzy C-Means Algorithm. Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL.
http://archive.rsna.org/2003/3100485.html
