Elsevier

Academic Radiology

Volume 13, Issue 9, September 2006, Pages 1082-1092
Academic Radiology

Medical image computing and computer-assisted intervention
Lung Deformation Estimation and Four-dimensional CT Lung Reconstruction

https://doi.org/10.1016/j.acra.2006.05.004Get rights and content

Rationale and Objectives

Four-dimensional (4D) computed tomography (CT) can be used in radiation treatment planning to account for respiratory motion. Current 4D CT techniques have limitations in either spatial or temporal resolution. In addition, most of these techniques rely on auxiliary surrogates to relate the time of the CT scan to the patient’s respiratory phase. We propose a 4D CT method for lung applications to overcome these problems.

Materials and Methods

A set of axial scans are taken at multiple table positions to obtain a series of two-dimensional images while the patient is breathing freely. Each two-dimensional image is registered to a reference CT volume. The deformation of the image with respect to the volume is used to synchronize the image with the respiratory cycle assuming that there is no phase variation along the craniocaudal direction. The reconstructed 4D dataset is a series of deformable transformations of the reference volume.

Results

A synthetic 4D dataset showed that the registration error is less than 5% of the image deformation. A swine study showed that the algorithm can generate better image quality than the image sorting method. A respiratory-gated 4D dataset showed that the algorithm’s result is consistent with the ground truth.

Conclusion

The algorithm can reconstruct good quality 4D images without external surrogates even if the CT scans are acquired under irregular respiratory motion. The algorithm may allow for reduced radiation dose to the patient with a limited loss of image quality. Although the phase variation exists along the craniocaudal direction, the 4D reconstruction is reasonably accurate.

Section snippets

Methods

Our 4D CT lung reconstruction method is illustrated in Fig 1. First, a reference 3D CT volume is obtained under breath hold. Next, a set of continuous CT scans is taken at every table position to obtain a series of two-dimensional (2D) images while the patient is breathing freely. The 2D image series at every table position covers at least one complete respiratory cycle. Using deformable registration, each 2D image is registered to the reference volume to estimate the displacement field of the

Results

Three experiments were conducted to validate the algorithm. These experiments were based on a synthetic 4D CT, a swine study, and a clinical 4D CT, respectively.

Discussion

This article presents a new methodology to reconstruct a 4D lung image from a set of 2D CT scans and a reference CT volume. The temporal resolution of the method is high and the reconstruction provides good-quality images. Based on a synthetic CT data set, the average registration error was less than 5% of the average lung deformation. Results from a swine study also showed that better image quality can be obtained using the algorithm instead of the image sorting method.

The algorithm is

Acknowledgment

The authors gratefully thank Frank Sauer, PhD, Ali Khamene, PhD, and Christophe Chefd’hotel, PhD, at Siemens Corporate Research for providing the lung dataset. The dataset was originally obtained by the EMC in Rotterdam. The authors also thank David Lindisch, RT, for his assistance with the experiments at Georgetown University.

References (19)

There are more references available in the full text version of this article.

Cited by (0)

Supported by U.S. Army grants DAMD17-99-1-9022 and W81XWH-04-1-0078, NSF Engineering Research Center 9731478.

View full text