Comparison of Spine, Carina, and Tumor as Registration Landmarks for Volumetric Image-Guided Lung Radiotherapy

Purpose

To assess the feasibility, reproducibility, and accuracy of volumetric lung image guidance using different thoracic landmarks for image registration.

Methods and Materials

In 30 lung patients, four independent observers conducted automated and manual image registrations on Day 1 cone-beam computed tomography data sets using the spine, carina, and tumor (720 image registrations). The image registration was timed, and the couch displacements were recorded. The intraclass correlation was used to assess reproducibility, and the Bland-Altman analysis was used to compare the automatic and manual matching methods. Tumor coverage (accuracy) was assessed through grading the tumor position after image matching against the internal target volume and planning target volume.

Results

The image-guided process took an average of 1 min for all techniques, with the exception of manual tumor matching, which took 4 min. Reproducibility was greatest for automatic carina matching (intraclass correlation, 0.90–0.93) and lowest for manual tumor matching (intraclass correlation, 0.07–0.43) in the left-right, superoinferior, and anteroposterior directions, respectively. The Bland-Altman analysis showed no significant difference between the automatic and manual registration methods. The tumor was within the internal target volume 62% and 60% of the time and was outside the internal target volume, but within the planning target volume, 38% and 40% of the time after automatic spine and automatic carina matching, respectively.

Conclusion

For advanced lung cancer, the spine or carina can be used equally for cone-beam computed tomography image registration without compromising target coverage. The carina was more reproducible than the spine, but additional analysis is required to confirm its validation as a tumor surrogate. Soft-tissue registration is unsuitable at present, given the limitations in contrast resolution and the high interobserver variability.

Lung cancer, Lung radiotherapy, Cone-beam computed tomography, CBCT, Treatment verification, Image-guided radiotherapy