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

Mitigating effects of respiratory motion during image guided radiotherapy (IGRT) is important especially during thoracic and abdomen scanning protocols such as cone-beam CT (CBCT) imaging, However, the lack of `ground-truth' in validating new algorithms has always been a challenge. The objective of this study is to outline the development of a novel 4D digital phantom for simulation of respiratory motion effects during CBCT image reconstruction based on Varian On-Board Imager (OBI): Half-Fan (HF) operating mode geometry. A set of actual 4D Magnetic Resonance (MR) data was used to develop the digital phantom. Firstly, the MR data sequence was extended to mimic a standard CBCT imaging acquisition protocol. Then, the images were segmented into several organs of interest and assigned with respective CT attenuation values. Subsequently, 2D project ions of the developed digital phantom were simulated using the Varian OBI geometry, A Poisson noise model was also incorporated to the projection data to realistically simulate quantum noise that is present in an actual clinical environment. Three types of projections were then reconstructed using the standard 3D Feldkamp-Davis-Kress (FDK) algorithm, projections: without noise, with noise, and with noise and reconstructed with an additional Hann filter, As validation, the reconstructed images were compared against a single-frame of the developed phantom; quantitatively, using normalized root mean squared error (NRMSE) and qualitatively, using difference images. The results indicated that the phantom managed to display a consistent trend in modeling the effects of respiratory motion on the reconstructed images. On average, the NRMSE values for all three reconstructed images within the entire field-of-view (FOV) were evaluated to be approximately 29.07 +/- 0.22%. Nonetheless, the difference images indicated a large en-or in areas largely affected by respiratory motion, The NRMSE of a region-of-interest (ROI) near the affected area was evaluated as 51.26% that constitute to a significant +22.19% difference.