Imaging a moving lung tumor with megavoltage cone beam computed tomography

Authors

  • Gayou Olivier,

    1. Department of Radiation Oncology, Allegheny General Hospital, Pittsburgh, Pennsylvania 15212 and Allegheny Campus, Temple University School of Medicine, Pittsburgh, Pennsylvania 15212
    Search for more papers by this author
    • a)

      Author to whom correspondence should be addressed. Electronic mail: ogayou@wpahs.org; Telephone: (412) 359-4058; Fax: (412) 359-3171.

  • Colonias Athanasios

    1. Department of Radiation Oncology, Allegheny General Hospital, Pittsburgh, Pennsylvania 15212 and Allegheny Campus, Temple University School of Medicine, Pittsburgh, Pennsylvania 15212
    Search for more papers by this author

Abstract

Purpose:

Respiratory motion may affect the accuracy of image guidance of radiation treatment of lung cancer. A cone beam computed tomography (CBCT) image spans several breathing cycles, resulting in a blurred object with a theoretical size equal to the sum of tumor size and breathing motion. However, several factors may affect this theoretical relationship. The objective of this study was to analyze the effect of tumor motion on megavoltage (MV)-CBCT images, by comparing target sizes on simulation and pretreatment images of a large cohort of lung cancer patients.

Methods:

Ninety-three MV-CBCT images from 17 patients were analyzed. Internal target volumes were contoured on each MV-CBCT dataset [internal target volume (ITVCB)]. Their extent in each dimension was compared to that of two volumes contoured on simulation 4-dimensional computed tomography (4D-CT) images: the combination of the tumor contours of each phase of the 4D-CT (ITV4D) and the volume contoured on the average CT calculated from the 4D-CT phases (ITVave). Tumor size and breathing amplitude were assessed by contouring the tumor on each CBCT raw projection where it could be unambiguously identified. The effect of breathing amplitude on the quality of the MV-CBCT image reconstruction was analyzed.

Results:

The mean differences between the sizes of ITVCB and ITV4D were −1.6 ± 3.3 mm (p < 0.001), −2.4 ± 3.1 mm (p < 0.001), and −7.2 ± 5.3 mm (p < 0.001) in the anterior/posterior (AP), left/right (LR), and superior/inferior (SI) directions, respectively, showing that MV-CBCT underestimates the full target size. The corresponding mean differences between ITVCB and ITVave were 0.3 ± 2.6 mm (p = 0.307), 0.0 ± 2.4 mm (p = 0.86), and −4.0 ± 4.3 mm (p < 0.001), indicating that the average CT image is more representative of what is visible on MV-CBCT in the AP and LR directions. In the SI directions, differences between ITVCB and ITVave could be separated into two groups based on tumor motion: −3.2 ± 3.2 mm for tumor motion less than 15 mm and −10.9 ± 6.3 mm for tumor motion greater than 15 mm. Deviations of measured target extents from their theoretical values derived from tumor size and motion were correlated with motion amplitude similarly for both MV-CBCT and average CT images, suggesting that the two images were subject to similar motion artifacts for motion less than 15 mm.

Conclusions:

MV-CBCT images are affected by tumor motion and tend to under-represent the full target volume. For tumor motion up to 15 mm, the volume contoured on average CT is comparable to that contoured on the MV-CBCT. Therefore, the average CT should be used in image registration for localization purposes, and the standard 5 mm PTV margin seems adequate. For tumor motion greater than 15 mm, an additional setup margin may need to be used to account for the increased uncertainty in tumor localization.

Ancillary