Presenter(s)
D. Low1, R. Andosca1, C. Miller2, P. Boyle1, M. V. Lauria1, J. P. Neylon3, D. O'Connell3, and D. Moghanaki1; 1Department of Radiation Oncology, University of California, Los Angeles, Los Angeles, CA, 2University of California Los Angeles, Los Angeles, CA, United States, 3UCLA, Department of Radiation Oncology, Los Angeles, CA
Purpose/Objective(s): Cone-beam CT is the main method used to localize and position lung tumors prior to treatment. CBCT image quality is degraded by uncompensated breathing motion, leading to difficulties in position assessment and challenging on-table adaptive strategies. 4DCT can be used to provide motion artifact free images but is limited by breathing irregularity. The 5D motion model has been shown to be more robust to breathing irregularity and has been proposed for motion-compensated CT, but the motion model developed during the CT simulation may not be applicable day to day. We conducted a case study showing whether the 5DCT model was sufficiently stable to provide good image sharpness with 5DCBCT.
Materials/Methods: A patient with lung cancer was prospectively recruited under IRB-approved study. The patient received 30 daily fractions and a standard-of-care CBCT at each fraction (1 minute rotation, no gating). A non-invasive breathing surrogate was recorded during each CBCT and the raw projection data and breathing surrogate data were collected and synchronized. CBCT images were constructed using the mcSART workflow, using the 5DCT motion models acquired at simulation and on fraction days 12 and 22. Image quality was evaluated by measuring the right diaphragm dome blurring for each fraction using each motion model starting on the corresponding fraction day. Tumor size measurements were also conducted to characterize tumor visualization stability. They were analyzed with an estimated error of +/-0.75 mm and an exponential fit was conducted on the measurements and the root-mean squared difference measured to evaluate the day-to-day image fidelity.
Results: The 5DCBCT reconstruction provided sharp diaphragm and tumor surface boundaries for most fractions but one fraction could not be analyzed due to a compromised surrogate file. 22 fractions had a mean 80%/20% diaphragm linear attenuation falloff less than 2 mm, 4 fractions between 2 and 4 mm, and the remaining 4 fractions blurrier than 4 mm. Non-compensated CBCT images were also constructed but were too blurry to quantitatively analyze. The tumor diameter measurement varied from the exponential curve by a root-mean squared of 0.4 mm (R squared 0.981), indicating that the reconstructed images themselves had a level of fidelity at least that size. There were also no significant deviations from the regression trend observed between the three motion models, indicating that the first model adequately represented the tumor size more than 6 weeks later.
Conclusion: 5DCBCT shows promise in providing motion compensated images without extending CBCT time or dose and requiring only a non-invasive breathing surrogate. This could be used to provide breathing amplitude specific images, model on-table tumor motion, and enable on-table adaptive strategies for shrinking tumors, although significant additional validation will be needed.