Scatter estimation for a digital radiographic system using convolution filtering.

Abstract

The use of a convolution-filtering method to estimate the scatter distribution in images acquired with a digital subtraction angiography (DSA) imaging system has been studied. Investigation of more than 175 convolution kernels applied to images of anthropomorphic head, chest, and pelvic phantoms using 15-, 25-, and 36-cm fields of view (digitized onto a 512 X 512 pixel image matrix) showed that two-dimensional exponential kernels with a full width at half maximum (FWHM) of 50-150 pixels best reproduced the scatter fields within these images with a root-mean-square percentage error from 4% to 8%. A two-dimensional exponential kernal with a FWHM of 75 pixels in each dimension applied to ten different anatomic presentations and fields of view, resulted in an average root-mean-square percentage error of 6.6% for the ten cases studied. The method should be implementable using an array of small lead beam stops placed in the field of only a single mask image and the above described convolution kernel applied to both mask and postopacification images. The mask beam-stop data are used to scale both mask and postopacification convolution-filtered images. This scaled, convolution-filtered image is then subtracted from the original image to produce a largely scatter-corrected image.