Martin Huber PhD, PRESENTER: Nothing to Disclose

Abstract: HTML Purpose: The 3D based evaluation of spiral CT data using present segmentation algorithms can provide more accurate diagnostic information, e.g. volume and density measurements for pulmonary nodules, than common 2D approaches. As CT enhancement techniques are potentially useful in differentiating between benign and malignant lung nodules, a reliable calculation of the mean CT density (in HU) is necessary. We thus investigated the accuracy and reproducibility of quantitative CT density measurements, focusing upon the impact of scanning and reconstruction parameters on the results. Methods and Materials: Scans of a lung phantom, containing spheres with x-ray absorption equivalent to soft tissue (~37 HU), were performed on a MSCT scanner (Sensation 16, Siemens). The spheres ranged in diameter from 3mm to 10mm. To simulate the lung parenchyma, the phantom was filled with shredded cork (~-1000 HU). A low dose protocol was used (120 kV, 16x0.75mm slice collimation, 1mm slice width, 18mm feed/rot, 0.5s rotation time) with 10, 50 and 100 eff. mAs. Reconstruction increments of 0.5mm and 0.8mm and soft and sharp reconstruction kernels were applied. To reduce partial volume effects, an erosion operator was applied to peel off the spheres an outer shell ranging between 0.5mm and 2.0mm; the remaining voxels were used to calculate the mean CT density. Results: The mean CT density was measured with very good reproducibility (fluctuations were <5%). However systematic deviations from the expected mean CT value (37 HU) were observed. The reconstruction kernels had the strongest impact on the results, e.g. the mean density of a 7mm sphere ranged between 6 HU for a soft kernel up to 68 HU for a sharp kernel. The discrepancy between expected mean value and measured ones and the spread within measurements was due to the modular transfer functions used for different kernels. The mean CT density calculation was also affected by the erosion parameter. The influence of the mAs settings on the density measurement was higher (up to 30HU) for smaller spheres (<7mm) than for larger ones, whereas the reconstruction increment had a negligible impact (< 4HU). Conclusion: Reproducible mean CT density measurements for small pulmonary nodules were obtained using a threshold and morphological filtering based segmentation algorithm. The results were influenced by the scanning and reconstruction settings. To compare mean CT density measurements from different CT examinations (e.g. follow-up studies), it is thus necessary to always use the same scan protocol and reconstruction parameters. (M.H., M.M., D.R., G.K., T.S. are employees of Siemens AG.) Questions about this event email: martin.huber@siemens.com

Huber PhD, M, Quantitative Evaluation of Pulmonary Nodules: Impact of Scanning and Reconstruction Parameters on Accuracy and Reproducibility of Mean CT Density Measurements.  Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL. http://archive.rsna.org/2003/3104610.html