3D Delivered Dose Assessment for SBRT Treatment Based On Estimated Volumetric Time-Varying Images From a Single Projection Image
P Mishra1*, J Seco2, R Li3, S St. James1, R Berbeco1, C Williams1, J Lewis1, (1) Brigham and Women's Hospital, Dana-Farber Cancer Center, Harvard Medical School, Boston, MA, (2) Mass General Hospital; Harvard Medical, Boston , MA, (3) Stanford University, Stanford, CATU-E-141-10 Tuesday 2:00PM - 3:50PM Room: 141
Purpose: To develop a method of verifying delivered dose based on time-varying 3D treatment images estimated from single kV projection images and a 4DCT-based patient-specific motion model.
Methods: Delivered dose calculation is performed in two steps. In the first step 3D fluoroscopic (time-varying) images are generated from single 2D kV projection images. In the second step a set of 3D fluoroscopic images are used to calculate the delivered dose. 3D fluoroscopic images are estimate by creating a patient-specific lung motion model through deformable registration of 4DCT, and then iteratively optimization the parameters of the motion model. Using a modified digital XCAT phantom, the accuracy of the delivered dose calculated with our method was compared to the dose that would be calculated during routine 4DCT-based treatment planning and to the actual delivered dose. The planning dose was calculated on an average intensity projection (AIP) image. The actual delivered dose is calculated based on the defined motion of the digital XCAT phantom. The same treatment plan was delivered to all three phantoms. The treatment plan was developed from the AIP and designed as a 3x18 Gy SBRT plan.
Results: The dose calculated on 3D fluoroscopic images closely matches to the actual delivered dose. Both the calculated and delivered doses vary from AIP-based planned dose. D95 tumor dose values of planned, delivered and calculated dose are 54.0, 59.0 and 58.4 Gy. Corresponding V20 lung values are 4.5%, 4.8% and 4.8% respectively.
Conclusions: We introduced a novel method for delivered dose calculation. Our algorithm successfully generates a 3D fluoroscopic image from a single kV images and then uses these time-varying 3D images to accurately calculate delivered dose.
Funding Support, Disclosures, and Conflict of Interest: The project described was supported, in part, by an RSNA Research Scholar Grant and Award Number R21CA156068 from the National Cancer Institute. The content is solely the responsibility of the authors and does not necessarily represent the official views of the RSNA, National Cancer Institute or the National Institutes of Health.