An automated method for comparing motion artifacts in cine four‐dimensional computed tomography images

Guoqiang Cui,Eric W Johnston,Julian C Hong,Billy W Loo,Peter G Maxim,Brian Jew

doi:10.1120/jacmp.v13i6.3838

Abstract

The aim of this study is to develop an automated method to objectively compare motion artifacts in two four‐dimensional computed tomography (4D CT) image sets, and identify the one that would appear to human observers with fewer or smaller artifacts. Our proposed method is based on the difference of the normalized correlation coefficients between edge slices at couch transitions, which we hypothesize may be a suitable metric to identify motion artifacts. We evaluated our method using ten pairs of 4D CT image sets that showed subtle differences in artifacts between images in a pair, which were identifiable by human observers. One set of 4D CT images was sorted using breathing traces in which our clinically implemented 4D CT sorting software miscalculated the respiratory phase, which expectedly led to artifacts in the images. The other set of images consisted of the same images; however, these were sorted using the same breathing traces but with corrected phases. Next we calculated the normalized correlation coefficients between edge slices at all couch transitions for all respiratory phases in both image sets to evaluate for motion artifacts. For nine image set pairs, our method identified the 4D CT sets sorted using the breathing traces with the corrected respiratory phase to result in images with fewer or smaller artifacts, whereas for one image pair, no difference was noted. Two observers independently assessed the accuracy of our method. Both observers identified 9 image sets that were sorted using the breathing traces with corrected respiratory phase as having fewer or smaller artifacts. In summary, using the 4D CT data of ten pairs of 4D CT image sets, we have demonstrated proof of principle that our method is able to replicate the results of two human observers in identifying the image set with fewer or smaller artifacts.PACS number: 87.57.cp; 87.57.N‐

Highlights

Four-dimensional computed tomography (4D CT) imaging is an important tool in radiation oncology
B. 4D CT image acquisition and sorting For this study, we identified ten lung cancer patients that were treated at our institution. 4D CT scans for these patients were acquired on a GE Discovery PET/CT Scanner (General Electric Medical Systems, Waukesha, WI) equipped with the real-time position management (RPM) system (Varian Medical Systems, Palo Alto, CA) for monitoring the patients’ breathing, using previously described acquisition techniques.[19]. Ten reconstructed phase bins were used, yielding 10 full-field volumetric image datasets per breathing cycle for each patient
Our method found the 4D CT image sets sorted using the breathing traces with the recalculated phase values to have fewer or smaller artifacts than the 4D CT image sets sorted using the breathing trace with the originally miscalculated phase values

Summary

Introduction

Four-dimensional computed tomography (4D CT) imaging is an important tool in radiation oncology. The most common method to acquire a 4D CT scan of a patient is to use the CT scanner in cine mode.[1] The time stamps of the reconstructed CT images and the measured respiratory signal of the patient are retrospectively matched. The reconstructed images are sorted either by the phase[2] or by the displacement,(3) which are stacked to create a three-dimensional (3D). The current acquisition and sorting methods led to significant motion artifacts[4,5,6] (artifacts in one study measured > 4 mm in 90% of scans[5]). The occurrence of artifacts is mainly caused by inaccurate determination of the respiratory phase.[4] These artifacts manifest themselves in the CT images as undefined and/or irregular boundaries, degrading image clarity and causing errors in patient contouring and dose calculation.[7,8]

Objectives

Methods

Results

Discussion

Conclusion