Efficient Scatter Distribution Estimation and Correction in CBCT Using Concurrent Monte Carlo Fitting
G Bootsma1*, F Verhaegen2, D Jaffray3, (1) University of Toronto, Toronto, ON,(2) Maastro Clinic, Maastricht(3) Princess Margaret Hospital, Toronto, ONSU-F-BRB-4 Sunday 4:00:00 PM - 6:00:00 PM Room: Ballroom B
Purpose: X-ray scatter is a known source of image artifacts, contrast reduction, and CT number inaccuracy in cone-beam CT (CBCT). We present and demonstrate the performance of a novel scatter correction method based on an algorithm that simultaneously combines multiple Monte Carlo (MC) CBCT scatter simulations through the use of a fitting function.
Methods and Materials: The scatter estimation system consists of concurrently run MC CBCT scatter projection simulations that are a sub-sample of the projection angles used in the reconstruction projection set, P, to be corrected. The photons generated by each MC simulation are simultaneously aggregated in an algorithm which computes the scatter detector response, S(i,j,k), for each down-sampled projection location (i,j) and angle (k). S(i,j,k) is fit to a function, FS, and if FS is determined to have a specified goodness of fit value the simulations are terminated. FS is subtracted from P which is subsequently used to create a scatter corrected reconstruction. The scatter correction method was applied to simulated phantoms using a frequency limited sum of sines and cosines as the fitting function. Image quality in the corrected reconstruction was evaluated using metrics looking at contrast, noise, and artifact reduction.
Results: Fitting the scatter distribution to a limited sum of sine and cosine functions, using a low-pass filtered Fast Fourier transform, provides a computationally efficient and accurate fit. Scatter distribution estimates for a 360 image projection set were computed in under one minute. The scatter correction algorithm increased the contrast-to-noise ratio by 46%, reduced the shading artifact by 87%, and decreased the the skin line artifact by 79% in reconstructions of a simulated pelvis phantom.
Conclusion: The algorithm provides an efficient method for estimating and removing the scatter distribution in from CBCT projection images. The results on simulated data show a significant increase in image quality.