Estimates of the image quality in accordance with radiation dose for pediatric imaging using deep learning CT: A phantom study

Abstract

Pediatric patients are at higher risk of radiation-induced damage than adults because cells of children tend to be sensitive to radiation. Therefore, the purpose of this study is to evaluated the dose reduction and image quality in pediatric computed tomography (CT) using deep learning reconstruction (DLR). Pediatric and adult phantom images were acquired with various tube voltages (kV), radiation doses, and image reconstruction methods. Tube voltages of 80, 100, and 120 kV were used to evaluate the image quality such as contrast-to-noise ratio (CNR). Radiation dose was set based on volume CT dose index (CTDIvol) and did not exceed the diagnostic reference level for abdominal CT scans in children and adults. To confirm the noise reduction effect of DLR, images acquired with DLR were compared with filtered back projection (FBP) and iterative reconstruction (IR) images. In the results on the effect of tube voltage, a 100 kV improved the CNR in both pediatric and adult phantom images of DLR. In the results according to the effect of the image reconstruction method, DLR showed almost superior results in terms of image quality than FBP and IR at the CTDIvol. Under the same imaging conditions, the CNR of pediatric phantom images was higher than that of adult phantom images. According to the results, DLR can improve image quality by reducing noise during abdominal CT scans in children. In addition, since the same image quality can be maintained even at a lower dose than adult CT scans, DLR will be useful for CT scans in children.