Computed Tomography-Based Lung Nodule Volumetry - Do Optimized Reconstructions of Routine Protocols Achieve Similar Accuracy, Reproducibility and Interobserver Variability to that of Special Volumetry Protocols?

Abstract

The aim of this in vitro and ex vivo CT study was to investigate whether the use of a routine thorax protocol (RTP) with optimized reconstruction parameters can provide comparable accuracy, reproducibility and interobserver variability of volumetric analyses to that of a special volumetry protocol (SVP). To assess accuracy, 3 polyurethane (PU) spheres (35 HU; diameters: 4, 6 and 10 mm) were examined with a recommended SVP using a multislice CT (collimation 16 x 0.75 mm, pitch 1.25, 20 mAs, slice thickness 1 mm, increment 0.7 mm, medium kernel) and an optimized RTP (collimation 16 x 1.5 mm, pitch 1.25, 100 mAs, reconstructed slice thickness 2 mm, increment 0.4 mm, sharp kernel). For the assessment of intrascan and interscan reproducibility and interobserver variability, 20 artificial small pulmonary nodules were placed in a dedicated ex vivo chest phantom and examined with identical scan protocols. The artificial lesions consisted of a fat-wax-Lipiodol mixture. Phantoms and ex vivo lesions were examined afterwards using commercial volumetry software. To describe accuracy the relative deviations from the true volumes of the PU phantoms were calculated. For intrascan and interscan reproducibility and interobserver variability, the 95 % normal range (95 % NR) of relative deviations between two measurements was calculated. For the SVP the achieved relative deviations for the 4, 6 and 10 mm PU phantoms were - 14.3 %, - 12.7 % and - 6.8 % and were 4.5 %, - 0.6 % and - 2.6 %, respectively, for the optimized RTP. SVP showed a 95 % NR of 0 - 1.5 % for intrascan and a 95 % NR of - 10.8 - 2.9 % for interscan reproducibility. The 95 % NR for interobserver variability was - 4.3 - 3.3 %. The optimized RTP achieved a 95 % NR of - 3.1 - 4.3 % for intrascan reproducibility and a 95 % NR of - 7.0 - 3.5 % for interscan reproducibility. The 95 % NR for interobserver variability was - 0.4 - 6.8 %. For datasets achieved with an SVP and an optimized RTP, this experimental approach showed comparable accuracy, reproducibility, and interobserver variability to allow for sufficient volumetric analysis of pulmonary lesions.