Scatter Correction in Cone-Beam Breast Computed Tomography: Simulations and Experiments

In Cone-Beam Computed Tomography (CBCT) the X-ray scatter control and reduction is one of the major challenges because CBCT is less immune to scatter than fan-beam CT. In breast volume imaging, studies on Cone-Beam Breast Computed Tomography (CBBCT) have shown the necessity to implement an efficient scatter reduction technique for a successful implementation of a breast CT scan using cone-beam geometry. X-ray scatter reduces image contrast, increases image noise and introduces reconstruction artifacts. A method for scatter evaluation through Monte Carlo simulations is investigated, leading to a scatter correction procedure applied to measured projections via subtraction of the simulated scatter component. Simulations are compared with measurements performed with a CBBCT prototype scanner. This paper presents the evaluation of the method through phantom studies on a cylindrical and on a hemi-ellipsoidal PMMA test object of 120- or 140-mm diameter at its base, simulating the pendant breast. The results indicate that this correction method is effective to reduce and correct X-ray scatter, with no increase of noise in the CT images. The cupping artifact due to scatter was reduced by a factor 3 (from 23% to 7%) for the hemi-ellipsoidal phantom of 140-mm diameter. Correspondingly, the relative noise in the CT slices remains constant to about 3%; the figure of merit for evaluating the correction efficacy was 98% of the ideal case. We discuss the application of this procedure to model breasts characterized in terms of few parameters, as indicated by recent published results of breast anatomy characterization derived from CBBCT patient data, from which a possible parametric breast size model in terms of bra cup size can be set forth. The scatter correction could be applied to projections from patient scans obtained with a breast holder, on the basis of a pre-determined database of simulated scatter distributions corresponding to the breast holder size, shape and estimated v- lume glandular fraction, for given beam quality and scanner geometry.