Room: Track 1
Purpose:
This longitudinal study aims to establish the minimum error bars from positioning, scanner variationa and measurement software for quantitative imaging and volumetry in serial lung nodule computed tomography(CT) scans.
Methods:
Six synthetic lung nodules of different sizes, shapes and attenuation characteristics were placed in an anthropomorphic chest phantom. The nodules were affixed to the lung vasculature. Two pairs of matching elliptical nodules were included with one proximal to the midline and one near the chest wall. Two spherical nodules were also included. The phantom was scanned on the same multi-detector row CT once a week for 8 months with standard chest and screening protocols and reconstructed with a variety of kernels and algorithms. Automatic exposure control was used for the standard chest CT and a fixed technique was used for the screening CT. The phantom was positioned at isocenter freehand for each scan to simulate patient position variability.
A lung nodule volumetry application was used to measure the nodule on each reconstruction (N=310). Automated detection and measurement settings were configured based on the lesion location and attenuation. The volumes were recorded and analyzed for percent bias(PB) and standard deviation from mean.
Results:
The measurements of the 2 nodules with 488mm3 volumes and -630HU had average s from all reconstructions of 15.1(central) and 33.4mm3(peripheral) and average absolute PB of 9.7 and 3.3%. The software program used was not able to reliably segment the 5000mm3 -800HU nodules. The 2 spherical nodules of 904mm3 had average sigmas of 22mm3 (-630HU) and 6.6mm3(100HU) with 5.6 and 2.4 average absolute PB over all reconstructions.
Conclusion:
Errors increased for lower attenuation nodules and increased for the matched nodules placed at the periphery compared to the center of the scan region. The error in volumetry due to scan to scan variation and measurement error are small.
Not Applicable / None Entered.
Not Applicable / None Entered.