253 - Report of 5DCT for 108 Lung Tumor CT Simulations
Presenter(s)
D. Low, C. Miller, D. O'Connell, M. Lauria, J. P. Neylon, L. Naumann, R. Andosca, P. Boyle, R. R. Savjani, J. Deng, and D. Moghanaki; Department of Radiation Oncology, University of California, Los Angeles, Los Angeles, CA
Purpose/Objective(s): The limitations of 4DCT have been known for >20 years in that it provides unreliable, qualitative and error-ridden CT simulation images of free-breathing patients, hampering progress in the automation and standardization of free-breathing CT simulation workflows. To overcome this challenge, an alternative simulation technique termed 5DCT (labelled for describing breathing motion using breathing amplitude and rate in addition to x, y, and z coordinates) was developed and successfully implemented at our institution in 2019. Herein we report the performance of 5DCT CT simulation on 108 patients with lung tumors.
Materials/Methods: Internal gross tumor volumes (IGTV) were generated from 5DCT simulation images and used as ROIs for tumor-based statistics. The breathing surrogate was correlated to diaphragm dome position, yielding the amplitude and rate to be the diaphragm position and speed. Breathing irregularity, in terms of amplitude and period, and breathing speed were evaluated. The IGTV motion magnitudes, defined by the 5th to 95th percentile motion, were calculated as well as the root-mean-squared IGTV breathing motion model residuals. The free-breathing scans (n=25 per patient) used in the 5DCT process provided the ground truths and were reconstructed to review the 5DCT workflow accuracy. These were divided into 3 categories, 0-2 mm, 2-4 mm and >4 mm errors to compare the ground truth to the model residuals.
Results: Individual patient breathing patterns varied widely. The average mean breathing amplitude was 16.0 ± 6.2 mm with a range of 5.0 to 40.2 mm (SD: 3.9 mm ± 2.8 mm with a range of 0.8 mm to 13.4 mm). The average relative breathing irregularity (std/mean) was 23.8% ± 13.9% with a range of 6.5% to 83.7%. The IGTV motion magnitudes ranged from 0.9 mm to 46.0 mm with a mean of 6.6 mm ± 6.4 mm. The IGTV RMS model residual errors ranged from 0.3 mm to 3.4 mm with a mean of 1.3 mm ± 0.6 mm. The minimum motion model error relative to the mean amplitude was 5.7% and relative to the 10th-90th percentile amplitude was only 1.6%, and the median relative error was only 7.7%. The IGTV motion model errors correlated poorly with the breathing speed (r2 = 0.31, p=0.001) and relative breathing amplitude variation. There was a statistically significant correlation between the IGTV motion model accuracy (comparing against ground truth) and precision (model residuals. Pearson Correlation Coefficient = 0.53, p<0.001).
Conclusion: This is the first report of lung tumor 5DCT motion model errors in a large cohort. The 5DCT workflow allowed for the generation of sorting-artifact free images and generated CT scans at a range of breathing amplitudes to allow the clinic to define accurate IGTVs. In addition, the clinic was notified of the motion model precision and accuracy to allow them to increase ITVs as necessary. The 5DCT workflow was found to be robust and failed in only the most irregular breathing patients. This work is leading to the development of downstream automation and workflows such as 5D Cone-Beam CT.