Optimal image reconstruction for detection and characterization of small pulmonary nodules during low-dose CT

Abstract

To optimize the slice thickness/overlap parameters for image reconstruction and to study the effect of iterative reconstruction (IR) on detectability and characterization of small non-calcified pulmonary nodules during low-dose thoracic CT. Data was obtained from computer simulations, phantom, and patient CTs. Simulations and phantom CTs were performed with 9 nodules (5, 8, and 10mm with 100, -630, and -800 HU). Patient data were based on 11 ground glass opacities (GGO) and 9 solid nodules. For each analysis the nodules were reconstructed with filtered back projection and IR algorithms using 10 different combinations of slice thickness/overlap (0.5-5mm). The attenuation (CT#) and the contrast to noise ratio (CNR) were measured. Spearman's coefficient was used to correlate the error in CT# measurements and slice thickness. Paired Student's t test was used to measure the significance of the errors. CNR measurements: CNR increases with increasing slice thickness/overlap for large nodules and peaks at 4.0/2.0mm for smaller ones. Use of IR increases the CNR of GGOs by 60%. CT# measurements: Increasing slice thickness/overlap above 3.0/1.5mm results in decreased CT# measurement accuracy. Optimal detection of small pulmonary nodules requires slice thickness/overlap of 4.0/2.0mm. Slice thickness/overlap of 2.0/2.0mm is required for optimal nodule characterization. IR improves conspicuity of small ground glass nodules through a significant increase in nodule CNR. • Slice thickness/overlap affects the accuracy of pulmonary nodule detection and characterization. • Slice thickness ≥3mm increases the risk of misclassifying small nodules. • Optimal nodule detection during low-dose CT requires 4.0/2.0-mm reconstructions. • Optimal nodule characterization during low-dose CT requires 2.0/2.0-mm reconstructions. • Iterative reconstruction improves the CNR of ground glass nodules by 60%.