A New Scatter Correction Method for KV and MV Cone Beam CT

Abstract

Kilo-voltage(KV) and mega-voltage(MV) cone beam computed tomography (CBCT) play an important role in image-guided adaptive radiotherapy. However, the cupping effect, introduced by Compton scatter contribution due to a large cone angle of CBCT acquisition, significantly degrades the CBCT image quality and hampers its clinical application. The objective of this study is to develop a new CBCT scatter correction method in the image domain. The hypotheses of the proposed method are that the image intensity of different kind of materials in CBCT image should be globally uniform and the cupping effect can be described as a bias field, a low frequency signal across the whole 3-D images. Based on above assumptions, we proposed a maximum a posteriori probability (MAP) framework to estimate the bias field contribution. The well-known fuzzy C-mean model is extended to define the likelihood function of the CBCT image. A markov random field model is used to describe the prior probability of the mixture defined in FCM model. The bias field is characterized by a Gaussian prior probability with a banded covariance matrix representing a low-pass filter. An iterated conditional mode (ICM)-like approach is utilized to minimize the objective function. The corrected CBCT image can be obtained by subtracting the bias field from original CBCT image. The performance of correcting the cupping effect of CBCT image with the proposed MAP framework was tested using both contrast phantoms and clinical CBCT images. In the cylindrical shape-contrast phantom the higher density material is surrounded by the lower density material. In the contrast phantom study, the intensities of higher density material in both KV and MV CBCT images are even lower than that of lower density material due to the cupping effect. The corrected CBCT images showed almost same intensity distribution as the original contrast phantom images. In a real patient study, it is clearly shown that there is a shadow inside the brain soft tissue in the original brain KV and MV CBCT images. The corrected CBCT images significantly improved the uniformity of intensity distribution of brain soft tissue. The proposed image-based scatter correction method showed a promising result to reduce the cupping effect commonly encountered in KV and MV CBCT image.