Monte Carlo evaluation of scatter mitigation strategies in cone‐beam CT

Abstract

Purpose: To investigate the efficacy of two widely used scatter mitigation methods: antiscatter grids ͑ASGs͒ and beam modulating with bowtie filters ͑BTFs͒, in combination with subtractive scatter correction or zeroth order normalization phantom calibration, for improving image noise, contrast, contrast-to-noise ratio ͑CNR͒, and image uniformity for on-board cone-beam CT ͑CBCT͒ imaging systems used for image-guided radiation therapy. Methods: PTRAN Monte Carlo CBCT x-ray projections of head and pelvic phantoms were calculated for combinations of beam-modulation and scatter rejection methods and images were reconstructed by in-house developed software. In addition, a simple one-dimensional analytic model was developed to predict scatter-to-primary ratio ͑SPR͒ and CNR as a function of cylindrical phantom thickness, ASG transmission, and beam modulation with bow-tie filters. Results: ASGs were found to have slightly negative or no effect on head phantom image CNR and to modestly improve CNR ͑10%-20%͒ in pelvic phantom images. However, scatter subtraction and norm-phantom calibration perform better when applied on data acquired with ASGs. Scatter subtraction improves CT number accuracy, but increases noise, and in high SPR/low primary-photon transmission scenarios can dramatically reduce CNR and introduce streaking artifacts. The BTF is found to reduce SPR and image noise, resulting in a better trade-off between CNR and imaging dose, but introduces a circular band artifact. Conclusions: Our study shows that ASGs have a modest positive impact in pelvic scans and negative in head scans, scatter subtraction improves the HU accuracy but reduces CNR, while BTF has a clearly positive effect.

Monte Carlo study of the effects of system geometry and antiscatter grids on cone‐beam CT scatter distributions

“…Other authors have shown that for some configurations, bowties are beneficial in terms of SPR reduction and image uniformity. 43,47,54,60 Bowties were omitted from the current study because they are not commonly used in all configurations studied here, to avoid further increasing the already large parameter space, and to maintain the focus of the current work on effects of system geometry in and of itself. Similarly, the simulations did not account for the effects of finite focal spot size.…”

“…There is therefore no simple, nominal choice of filter for each configuration that would permit a fair comparison across this range of geometry, object size, and kVp. Rather, the design of bowtie filters optimal to a particular configuration is a continuing area of work, with results indicating potential advantages in x-ray scatter for some configurations (e.g., breast CT 43 and Linac CBCT 47,60 ). For simplicity in an already broad space of parameters considered below, the simulations considered scenarios without a bowtie.…”

“…The grid was approximated as locally parallel, oblique lamellae oriented at an angle corresponding to the angle of the focused grid at the location of photon incidence. Each photon was assigned a weight given by the analytical expression of Day and Dance 47,56 for grid transmission as a function of photon incidence angle, photon energy, grid cell materials and dimensions, and local grid angle. The photon was subsequently scored in the detector with a weight representing the product of energydependent probability of absorption in the scintillator (quantum detection efficiency, g 1 ), and energy-dependent optical gain in the scintillator (g 2 ).…”

“…In kilovoltage (kV) CBCT for image-guided radiotherapy (IGRT), Siewerdsen et al 46 showed that although grids significantly reduced cupping and streak artifacts, image uniformity, and CT number accuracy, they did not offer an improvement in CNR per unit square root dose and offered little or no qualitative improvement in soft-tissue visibility (prostate). Similarly, Lazos and Williamson 47 used MC simulations to show that antiscatter grids in CBCT for IGRT gave little or no improvement in CNR for head and neck imaging and a 10%-20% improvement in CNR for pelvic imaging at a fixed dose.…”

A digitally reconstructed radiograph algorithm calculated from first principles

“…To do this we used an in-house Monte Carlo (MC) algorithm capable of simulating scatter-only projections of an object. 26,33 A Monte Carlo scatter projection can be represented as…”