Evaluation of target localization accuracy for image-guided radiation therapy by 3D and 4D cone-beam CT in the presence of respiratory motion: a phantom study

Abstract

We evaluate the target definition accuracy of four-dimensional CT (4D-CT) simulation and target localization accuracy of 3D or 4D cone-beam CT (CBCT) in the presence of respiration. The target motion is modelled by a sine curve or a cos6 curve. The target volumes, shapes, and positions obtained from the 4D-CT simulation are compared with the static CT image and theoretical values of the phantom. Reference average intensity projection (AIP) and maximum intensity projection (MIP) images for target localization are generated from the 4D-CT simulation. Localization involves aligning the AIP/MIP to 3D cone-beam CT (3D-CBCT) or 4D cone-beam CT (4D-CBCT), and localization accuracy is evaluated from the difference in target position between the reference AIP/MIP image and 3D-CBCT/4D-CBCT measurements. 4D-CBCT also allows measurement of the target motion standard deviation (SD) and excursion (EX). The SD and EX errors are calculated with respect to the theoretical value of the phantom. The target volume and position accuracies obtained via 4D-CT at each phase are within 3.0% and 2.5 mm, respectively, of the static and theoretical values for the sine and cos6 curves. The target localization errors for 3D-CBCT are within 1.0 mm regardless of the EX variation and reference image for the sine curve, whereas the errors for the cos6 curve increase from 0.1 to 5.1 mm with increasing EX variation. In contrast, the 4D-CBCT localization errors are within 1.0 mm regardless of EX variation, reference image, and motion pattern. In addition, SD and EX errors, respectively, range from −1.3 to 0.1 mm and −2.2 to 0.1 mm (AIP) and from −4.4 to −2.7 mm and −13.5 to −4.2 mm (MIP). 4D-CBCT for AIP is more accurate than that for MIP. Target localization is simple and accurate for 3D-CBCT and 4D-CBCT with the AIP. However, 3D-CBCT is more inaccurate than 4D-CBCT when considering EX variations with the cos6 curve.