Scatter point models for breast cone-beam computed tomography: preliminary study

Abstract

Simulations with a single scatter point per incident beamlet (SSP) model applied to homogeneous phantoms could provide scatter signals that can be subtracted from cone beam breast CT projections so as to minimize the effects caused by scattered photons. Consider a heterogeneous 14 cm diameter 10.5 cm long cylindrical fat cylinder with 5 embedded cylinders of fibroglandular (fib) such that the percent volume of fib is 15%. A 60 kV beam delivering a total dose of 7 mGy over 300 projections was used for interrogation. A transmission model and a many scattering point per incident beamlet (MSP) model incorporating a scattering point every 1 cm of depth of beamlet were used to estimate the energy integrated signals (EIS) due to primary (p) and single scattered (s) photons on each pixel. These models were also used to calculate the EIS p and EIS s for a homogeneous phantom of the same size and fib:fat mass fractions. The SSP model was then applied with the scatter point of interaction being at a depth = d dk where d k = length of beamlet k in the phantom and δ was found by matching its peak scatter-to-primary ratio with that obtained with the MSP homogeneous model. The SSP was tested to correct for the effects of the single scatter during cone beam CT of the heterogeneous phantom. The Hounsfield unit deviations from the ideal primary for fib and fat were −95.1 and −56 via no correction for scatter whereas with correction they were 3.8 and −2. It was encouraging to see how a simple model could minimize the effects of single scattered photons during a cone beam CT imaging task. The preliminary findings encourage further efforts for thoroughly testing its applicability for obtaining higher quality images.