Poster — Wed Eve—40: Comparison of CBCT CT Numbers to CT‐Simulator CT Numbers using Mutual Information and Joint Probability Distributions

Abstract

The Varian On‐Board Imaging (OBI) system is an imaging tool used by the London Regional Cancer Program for image‐guided radiation therapy. The system consists of a kilovoltage source (120 kVp), emitting a conical X‐ray beam, and a flat panel amorphous silicon detector. Both source and detector are mounted on the linear accelerator gantry, rotating 360° about the patient creating 3‐dimensional volumetric computed tomography images. Unfortunately, cone‐beam computed tomography (CBCT) images are susceptible to many forms of artifacts resulting primarily from: increased amounts of scatter due to CBCT geometry, spectral X‐ray attenuation, and intra‐scanning organ motion. In order to assess the extent of CBCT image artifacts, we propose a new technique that compares CT Numbers in corresponding tissue regions obtained by the CBCT and Phillips Brilliance CT‐simulator systems. Our approach registers CT images from two imaging modalities using a mutual information (MI) algorithm. Once aligned, quantitative CT Number analysis is performed on the CT images using a 2‐dimensional histogram (or joint probability distribution) of pixel counts for geographically‐matched pairs of CT Numbers. Phantom studies have shown similarity between CBCT and CT‐simulation CT Numbers when the imaged medium is head‐like in diameter (20 cm) and homogeneous (Catphan 504 phantom). Greater discrepancies were observed (400 HU) when CT images of larger phantoms (30 cm diameter, Gammex RMI Tomo phantom) and increased heterogeneity were compared. Presently, quantitative CBCT CT Number inaccuracy precludes the use of CBCT images for whole body treatment planning, and adaptive dose computations.