X-ray scattering point models for breast cone-beam computed tomography

Abstract

The purpose of this work was to determine via simulations the potential use of sim- plified scattering point models in full-field breast Cone-Beam Computed Tomography (CBCT). A many Scattering Point (MSP) per incident beamlet model and a Single Scat- tering Point (SSP) model were tested against Geant4 simulations, as well as against each other. Comparisons were made using both homogeneous as well as heterogeneous phantoms. The homogeneous phantoms were cylinders, 14 cm and 7 cm in diameter and 10.5 cm in length with various fibroglandular(fib):fat compositions. The hetero- geneous phantom was the 14 cm phantom mentioned previously, but modeled as pure fib with a number of smaller embedded cylinders composed of fat. A second configu- ration with the compositions of the main and embedded cylinders swapped was also tested. The simulation used a 60 kV tungsten anode spectrum with a HVL of 3.7 mm Al which irradiated the simulated breast phantom over 300 projections. The detector was modeled as 300 × 300, 1 mm2 energy integrating pixels, with a DQE of unity. Both of the models approximate the cone-beam as a number of individual beamlets (300 × 300 to match with the detector) with scattering points placed along their in- tersections with the phantom. The MSP model incorporated a single scattering point per 1 cm of incident beamlet length within the phantom. The SSP model used an adjustable single scattering point positioned at a fractional depth within the phan- tom. By comparing results from these two scattering point models, values of were determined which would yield SSP model scatter approximations matching those of iii the MSP model. Both models were tested against Geant4 simulations for their ability to adequately estimate the scatter signals upon the detector. The SSP model was also tested for its ability to correct for the cupping artifact in reconstructed images of the heterogeneous phantom, without assuming knowledge of the inner heterogeneous geometry. The Hounsfield Units (HU) obtained with primary photons were 48.5±3.18 and −159 ± 23.0 for fib and fat respectively for one of the heterogeneous phantom con- figurations. Due to the cupping artifact in the reconstructed images which included scatter, these values were −46.6 ± 18.9 and −215 ± 34.0 respectively. Following cor- rections for the single scatter, via the SSP model, the corresponding CT numbers became 52.3 ± 3.67 and −161 ± 23.3. It was encouraging to see that a simple model can minimize the effects of single scattered photons during CBCT of a heterogeneous phantom. The HUs obtained post-scatter correction agreed well with those obtained with primary photons. The preliminary findings encourage further efforts for thoroughly testing these scatter point models’ applicability for obtaining higher quality CBCT images.