Differentiation of Benign and Malignant Neck Pathologies: Preliminary Experience Using Spectral Computed Tomography

作者：Srinivasan Ashok^*; Parker Robert A; Manjunathan Abhishek; Ibrahim Mohannad; Shah Gaurang V; Mukherji Suresh K

来源：Journal of Computer Assisted Tomography, 2013, 37(5): 666-672.

• spectral CT
• neck malignancy
• spectral Hounsfield unit

Purpose: The objective of this study was to evaluate spectral Hounsfield unit (HU) curves and effective Z (atomic number) generated on dual-energy gemstone spectral imaging computed tomography (CT) in the differentiation of benign and malignant neck pathologic findings.
Methods: This was a retrospective review of 38 patients who underwent neck CT on a gemstone spectral imaging dual-energy CT (Light-speed CT750 HD 64-slice CT scanner; GE Medical Systems, Milwaukee, Wis) from November 2009 to June 2012 with identifiable masses. One board-certified radiologist placed regions of interest within the mass (19 benign, 19 malignant) and in paraspinal muscles (PSMs) to create 2 spectral HU curves in each patient. The curve parameters compared between the benign and malignant groups included range (conceptually, the difference between the highest and lowest HU), asymptote, decay, and the differences and ratios (of lesion to PSM) of each of these 3 parameters. A logistic regression model was built with these parameters and effective Z.
Results: The difference in ranges (between lesion and PSM) was the best predictor of malignancy, with a threshold of 75 or greater demonstrating 95% sensitivity, 89% specificity, and 91.8% area under the curve (AUC). Adding other spectral HU parameters and effective Z to the model did not substantially increase the AUC (93.3%, difference between the 2 models not statistically significant, P > 0.25). The effective Z showed a 79.9% AUC with 68% sensitivity and 68% specificity at an 8.80 cutoff.
Conclusions: The spectral HU curve is promising for differentiating benign and malignant neck pathologic findings, with the difference in range between the lesion and PSM showing the best predictive value.

• 出版日期2013-10