Characterization of different reconstruction techniques on computer-aided system for detection of pulmonary nodules in lung from low-dose CT protocol

Abstract

In this study, we aim to investigate the effect of different reconstruction techniques on computer-aided system (CAD) pulmonary nodules in low-dose CT. A prospective collection of 380 patients with low-dose chest CT pulmonary nodules was performed. Filtered back projection, quantum noise reduction, 3D adaptive iteration, and cloud iterative techniques were used to reconstruct the original data. The average signal-to-noise ratio (SNR) of each technique was measured. Compared with the contrast-to-noise ratio (CNR), we analyzed the sensitivity and false-positive number of the lung nodules by CAD in each group of techniques. The SNR and CNR are sequentially improved according to the filtered back projection, quantum noise reduction, three-dimensional adaptive iteration, and cloud iterative technique (P < 0.05). The sensitivity increased with the improvement of the objective image quality, while the false positive rate decreases. There are differences in the sensitivity of various nature nodules under different reconstruction techniques (P < 0.05). Although the CAD has good performance in detecting various lung nodules under different reconstruction techniques, there are some differences in energy efficiency under different reconstruction techniques. Therefore, the robustness of CAD still needs to be further solved. Multi-center large sample validation is still required to evaluate the energy efficiency of pulmonary nodule CAD.