Quantification of motion artifacts in 4DCT using global Fourier analysis

Abstract

Motion artifacts in 4-Dimensional computed tomography (4DCT) impose a potential problem in tumor and anatomy delineation, causing a systematic error in radiotherapy treatment planning. In patients, frequency and amplitude of motion artifacts appearing in 4DCT are currently manually evaluated by scrolling through all 4DCT images. This is a tedious, time consuming and error-prone process with an outcome that may be difficult for comparison among patients. In this study, we propose an automatic approach to evaluate the motion artifacts in 4DCT using Discrete Fourier transform. Instead of applying Fourier transform in local macro-blocks as done in ordinary signal processing and computer vision applications, we discover that 4DCTs of severe motion artifacts demonstrate special behavioral properties in their spectral domain after a global Fourier transform, which can be exploited to effectively quantify the severity of motion artifacts in 4DCT images via a band-passing analysis. This is the first effort in creating a computerized algorithm to evaluate the motion artifacts with a quantitative and objective evaluation with encouraging performances. Further study will be conducted with more patient data with more rigorous 4-dimensional analysis in the near future.