Comparative evaluation of noise texture and images of a synthetic lung nodule using energy-integrating and photon-counting CT

Noise texture and magnitude are well known characteristics that impact the detectability of lesions in CT images, especially when the main task is to identify small and low contrast lesions. The purpose of the present work was to assess how NPS properties impact visual perception of low contrast lung nodules across various CT image acquisition protocols. This was achieved by evaluating the quantitative characteristics of the NPS using a previously-proposed NPS parameterization (peak frequency of the NPS, standard deviation (or “sigma”) of a fitted half-Gaussian through the downslope of the NPS, and noise magnitude) and comparing them to the appearance of a synthetic ground-glass nodule inserted into a lung phantom. This phantom was imaged using an energy-integrating (EICT) and a photon-counting (PCCT) CT systems. Different dose levels, reconstruction algorithms, and kernels were used. For the EICT images, at the two lower dose levels studied (0.4 mGy and 0.2 mGy), the nodule appears speckled with its edge visualization severely reduced. Using similar dose levels in both systems, the use of deep-learning-based reconstruction (DLR) resulted in improved noise properties and better visualization in comparison to that with a hybrid iterative reconstruction. The improvement of the nodule edge definition using DLR/lung combination is evident in PCCT images. For the same dose levels and reconstruction algorithm/kernel combination, in general, PCCT resulted in lower noise magnitude and higher peak frequency and sigma than EICT images. Therefore, PCCT/DLR/lung combination demonstrated improved capability to characterize the edges of the low contrast nodule.