Optimized Estimation of Scattered Radiation for X-ray Image Improvement: Realistic Simulation

Abstract

Image processing algorithms for compensation of the scattered radiation influence in X-ray imaging are proposed, studied and optimized by numerical simulation. These algorithms include the scattering estimation by convolution (superposition) technique, estimation of kernel functions by Monte Carlo (MC) simulation, the determination of the optimal number and shape of kernel functions and image segmentation. The determination of the number and shape of kernel functions was performed by the MC simulation of the realistic Zubal phantom and the clustering analysis of shape features of kernel functions. Testing simulation study of the algorithms for chest images at 75 keV proves that the optimal number of kernel functions is equal to 8. This number provides the three-fold contrast enhancement without using the anti-scatter grids. The achieved contrast is about 95% of the primary image contrast that exceeds contrast enhancements achieved with anti-scatter grids. An increased number of used kernel functions provides a better image contrast and better resolution of scattered radiation image, but estimation errors also increase due to the segmentation and deconvolution errors.