Baiyu Chen, Presenter: Nothing to Disclose

Samuel Richard PhD, BSC, Abstract Co-Author: Nothing to Disclose

Huiman Barnhart PhD, Abstract Co-Author: Nothing to Disclose

James G. Colsher PhD, Abstract Co-Author: Employee, General Electric Company

Maxwell Amurao PhD, Abstract Co-Author: Nothing to Disclose

Ehsan Samei PhD, Abstract Co-Author: Advisory Board, Ion Beam Applications, SA Consultant, Siemens AG

Our purpose is to determine the effects of reconstruction algorithm, dose, and slice thickness on the accuracy (bias) and precision (variance) of 3D assessment.

An anthropomorphous thoracic phantom consisting of realistic pulmonary vessels and synthetic nodules of two sizes (3/16’’, 3/8’’) was scanned using a Discovery CT750 HD scanner (GE). The scan parameters were chosen based on a routine chest protocol in Duke Hospital: 40mm collimation, 120 kVp, 1.375:1 pitch, 0.5s rotation time, and 28.0 noise index. Five more dose levels, 75%, 50%, 25%, 10%, and 3% were investigated by changing the mA setting. Each protocol was repeated five times and reconstructed into two slice thicknesses of 0.625 mm and 2.5 mm using Filtered Back-projection (FBP), a Adaptive Statistical Iterative Reconstruction (ASiR), and a Model-based Iterative Reconstruction (MBiR). The volume of synthetic nodules was quantified from the reconstructed CT images using a clinical segmentation software. The measured nodule volumes were then employed for accuracy and precision calculation in terms of percentage bias (PB) and repeatability coefficient (RC).

Assessment demonstrated that between FBP and ASiR, where noise reduction is the primary difference, there was little impact on accuracy and precision. Very low dose showed better accuracy and worse precision, but quickly plateaued at higher dose(PB3%=6.9%, PB>25%=7.9+/-0.5%, RC 3%=9.3, RC>25%=5.9+/-0.5%). Slice thickness showed no significant effect on accuracy, and contrasting effects on precision for small and big nodules: for small nodule, 0.625 mm always had better precision; for large nodule, 2.5 mm was advantageous (RC=5+/-2%, RCsmall-2.5=13+/-5%, RCbig-0.625=5+/-1%, RCbig-2.5=2+/-1%). Analysis is ongoing for MBiR where the primary difference extends to additional facets of image quality.

This study evaluated the effects of various imaging parameters on lung nodule volume quantification, showing the dependence on dose level and slice thickness. It provides possible guidelines for protocol optimization and for tasks comparisons between images acquired with different protocols.

A detailed knowledge of the accuracy and precision dependence on CT imaging parameters help determining whether a measured volume change of lesion is due to errors or an actual shrinkage/growth.

Chen, B, Richard, S, Barnhart, H, Colsher, J, Amurao, M, Samei, E, Volumetric Quantification of Lung Nodule in CT: Effects of Reconstruction Algorithm (FBP, ASiR, and MBiR), Dose and Slice Thickness.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9012854.html