Acuros CTS: A fast, linear Boltzmann transport equation solver for computed tomography scatter – Part II: System modeling, scatter correction, and optimization Adam Wang, a) Alexander Maslowski, Philippe Messmer, Mathias Lehmann, Adam Strzelecki, Elaine Yu, Pascal Paysan, Marcus Brehm, Peter Munro, Josh Star-Lack, and Dieter Seghers Varian Medical Systems, Palo Alto, CA 94304, USA (Received 21 September 2017; revised 23 January 2018; accepted for publication 23 February 2018; published 23 March 2018) Purpose: To correct for scatter in kV cone-beam CT (CBCT) projection data on a clinical system using a new tool, Acuros â CTS, that estimates scatter images rapidly and accurately by deterministi- cally solving the linear Boltzmann transport equation. Methods: Phantom and patient CBCT scans were acquired on TrueBeam â radiotherapy machines. A first-pass reconstruction was used to create water and bone density maps of the imaged object, which was updated to include a more accurate representation of the patient couch. The imaging system model accounted for the TrueBeam x-ray source (polychromatic spectrum, beam filtration, bowtie filter, and collimation hardware) and x-ray detection system (antiscatter grid, flat-panel imager). Acuros CTS then used the system and object models to esti- mate the scatter component of each projection image, which was subtracted from the measured projections. The corrected projections were then reconstructed to produce the final result. We examined the tradeoff between run time and accuracy using a Pareto optimization of key param- eters, including the voxel size of the down-sampled object model, the number of pixels in the down-sampled detector, and the number of scatter images (angular down-sampling). All compu- tations and reconstructions were performed on a research workstation containing two graphics processing units (GPUs). In addition, we established a method for selecting a subset of projec- tions for which scatter images were calculated. The projections were selected to minimize inter- polation errors in the remaining projections. Image quality improvement was assessed by measuring the accuracy of the reconstructed phantom and patient images. Results: The Pareto optimization yielded a set of parameters with an average run time of 26 seconds for scatter correction while maintaining high accuracy of scatter estimation. This was achieved in part by means of optimizing the projection angles that were processed, thus favoring the use of more angles in the lateral (i.e., horizontal) direction and fewer angles in the AP direction. In a 40 cm solid water phantom reconstruction, nonuniformities were decreased from 217 HU without scatter correc- tion to 51 HU with conventional (kernel-based) scatter correction to 17 HU with Acuros CTS-based scatter correction. In clinical pelvis scans, nonuniformities in the bladder were reduced from 85 HU with conventional scatter correction to 14 HU with Acuros CTS. Conclusions: Acuros CTS is a promising new tool for fast and accurate scatter correction for CBCT imaging. By carefully modeling the imaging chain and optimizing several parameters, we achieved high correction accuracies with computation times compatible with the clinical workflow. The improvement in image quality enables better soft-tissue visualization and potentially enables applica- tions such as adaptive radiotherapy. © 2018 American Association of Physicists in Medicine [https:// doi.org/10.1002/mp.12849] Key words: cone-beam CT, Monte Carlo, scatter 1. INTRODUCTION In Part I, 1 we pointed to the necessity for a fast and accurate method for scatter estimation in kV cone-beam CT (CBCT) acquisitions. Scatter remains one of the main challenges for CBCT image quality, and we are looking for an improved scatter correction that can be executed in clinically acceptable times without the addition of cumbersome hardware. Part I reported on the development of a deterministic solver of the linear Boltzmann Transport Equation (LBTE) to model pho- ton flow through the imaged object and form a scatter image at the detector. The resulting tool for estimating computed tomography scatter, Acuros â CTS, is also referred to as Acuros for brevity. Acuros utilizes efficient numerical meth- ods that preserve the accuracy of the solution despite high discretization in the spatial, energy, and angular domains. Comparison against Monte Carlo-generated results in digital phantoms (water with lung, air, and bone inserts; pelvis CT 1914 Med. Phys. 45 (5), May 2018 0094-2405/2018/45(5)/1914/12 © 2018 American Association of Physicists in Medicine 1914