Prediction of the Characteristics of Moving Lung Tumor in CBCT Imaging Using Virtual CBCT Image Simulated From 4-D CT Dataset
X Li*, T Li, Y Yang, Y Zhang, D Heron, M Huq, University of Pittsburgh Medical Center, Pittsburgh, PASU-E-J-124 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: Due to the slow scanning speed of on-board CBCT imaging, the characteristics of moving lung tumor acquired by such system could be complicated. Recent phantom studies have shown that the maximum intensity projection (MIP) image, which was used for target delineation in treatment planning, might overestimate the target volume determined by CBCT imaging. The purpose of this study is to propose a new strategy to predict the characteristics of moving lung tumor in CBCT imaging.
Methods: Six non-small-cell-lung cancer (NSCLC) patients were retrospectively selected in this study. All patients underwent 4D CT scan using GE LightSpeed scanner under audio coaching. The respiratory motion of all patients ranged from 7.5mm to 12.5mm. For each patient, ten sets of phase-sorted CT images and a MIP image including all respiratory phases were acquired. To simulate the respiratory motion effect in CBCT imaging, 4-D projection images were generated using ray-tracing algorithm based on the 3-D CT images of the corresponding phase at each gantry angle. Filtered-back-projection algorithm was used to reconstruct this simulated virtual CBCT (SVCBCT) image. The targets were then contoured at the MIP image, SVCBCT image and the clinic CBCT image acquired at first treatment, respectively.
Results:For the three cases volume differences between SVCBCT and MIP image were less than 5%. For the other three cases the target volume determined by SVCBCT image were 30%~62% less than that of MIP image, and the distance of target center between those two images could be up to 6mm. In comparison, for all patients, the volume differences between SVCBCT and CBCT were all less than 4%.
Conclusion:The characteristics of a moving lung tumor determined by CBCT imaging are highly dependent on the specific respiratory motion pattern of each patient, and could be well predicted by SVCBCT image.