31 Evaluation of Hounsfield Unit correction method on Cone-Beam CT for dose calculation strategies

Abstract

Introduction Verification of day-to-day dose delivery is possible with 3D-in room imaging modalities as Cone-Beam CT (CBCT). CBCT technique is impacted by artifacts related to the acquisition and reconstruction physical principles, responsible of Hounsfield Units (HU) variability. The purpose of this work is to evaluate CBCT dose calculation performance and precision with HU correction using deformable registration algorithm (Velocity® 3.2, VMS, USA) between CT and CBCT. Methods This study is divided in two axes: (a) an image quality evaluation and HU variability among CTREF (Aquilion LB®, CMS, Holland), CBCTOBI (OBI® v1.6, VMS, USA), CBCTTB (TrueBeam® v2.5, VMS, USA) and corrected CBCT (CBCTDIR,OBI et CBCTDIR,TB) obtained with Velocity® (VMS, USA) using Catphan® 504 (Phantom Laboratory, USA). The analysis is performed with Artiscan® (Aquilab, France). (b) Using CIRS® Pelvis phantom (CIRS, USA), 3DCRT bladder and hip treatment plans and VMAT prostate plan are calculated (AAA v13.5®, Aria®, VMS, USA) on CTREF (120 kV), then calculated with fixed MU’s, with same electronic density-HU curve, on CBCTTB (TrueBeam® v2.5, 125 kV) and CBCTDIR,TB obtained via Velocity®. A comparison of dosimetric parameters (D98%, D50%, D2%) is performed as well as a CT-CBCT dose matrices comparison calculated for each modality using the gamma index obtained with 3DSlicer® (v.4.6). Results Uniformity evaluation on Artiscan® with Catphan® 504 using predefined regions of interest (ROI) is compared. The average deviations of absolute uniformity for ROI between reference (CTREF) and evaluated CBCT’s are in average 24 for CBCTTB, 30 for CBCTOBI, and less than 20 for CBCTDIR TB, OBI. Pixel size is identical among CT and CBCTDIR TB, OBI and distances restitution is infra-millimetric for all modalities evaluated. Attenuation is compared for each insert. HU average differences for all inserts, between CTREF and CBCTOBI,TB/ CBCTDIR OBI,TB modalities, are 1HU for CBCTDIR TB,OBI (min.=0 HU; max.=3HU) and greater than 13HU for CBCTTB,OBI (min.=2 UH; max.=132UH). Largest differences are observed for low and high density regions. Comparison of dosimetric parameters, respectively D98%, D50%, D2%, between CBCT and CT shows deviations of more than 3% in HU regions greater than 200HU and less than 0.5% for water-equivalent or around regions. Among CBCTDIR,TB and CT, average deviations observed are systematically lower than 1% for evaluated organs and percentages of points with a gamma value less than 1 (1%/ 1 mm) are on average 98% for the three locations studied. Conclusions This HU correction method proposed by Velocity 3.2® (VMS, USA) is relevant to account physical uncertainties related to CBCT. This tool is promising and will be evaluated on patient images to confirm these results obtained on anthropomorphic phantom.