Development of a 4D phantom for respiratory motion modeling during Cone-Beam CT (CBCT) imaging on the Varian On-Board Imager (OBI)

Abstract

Cone-beam computed tomography (CBCT) imaging is widely used in image guided radiotherapy. Current protocols assume the patient to be static during imaging but, respiratory motion can affect the reconstructed CBCT images. The objective of this study is to simulate respiratory motion effects on a CBCT image using a digital phantom constructed from actual 4D Magnetic Resonance Imaging (MRI) data. Firstly, the length of the 4D MR data was extended to 1-minute, mimicking a typical CBCT imaging acquisition period over a 360-degree scan. The MR images are then segmented and converted to CT attenuation values. The Varian On-Board Imager geometry was used in the simulation where Poisson noise was added to realistically model imaging noise. Simulated projections were reconstructed using the standard Feldkamp-Davis-Kress algorithm. The images were then compared against the developed phantom at two respiratory positions: end-inspiration, and end-expiration, using normalized root mean squared error (NRMSE) and difference images. The results were encouraging with a NRMSE increment of 0.59% when noise was added into the phantom, whilst a −0.51% decrease when an additional Hann filter was used during reconstruction. The respiratory motion effects were successfully modeled when the results showed a 0.55\% difference between the different positions. This is also observed from the difference images by selecting a region-of-interest with NRMSE values evaluated as 33.84% and 51.26% respectively, constituting a significant 17.42% difference. Therefore, it is evident that respiratory motion affects image reconstruction quality. This indicates the currently practiced protocol and reconstruction algorithm must consider respiratory motion to ensure accuracy.