Purpose: Conventional pre-treatment thoracic 4DCBCT scans take 1320 projections over 240 seconds. The ADAPT clinical trial (NCT04070586) uses first-in-world implementations of motion compensated acquisition and reconstruction to reduce scans to just 200 projections over 20 breaths (40 to 160 seconds). We present results of the first 20 of 30 trial patients and investigate our motion compensated acquisition and reconstruction relative to conventional 4DCBCT.
Methods: The ADAPT trial enrolled 30 lung cancer radiation therapy patients for additional imaging in their first 2 treatment fractions. For each fraction, they receive a conventional 4DCBCT, followed by a 200 projection motion compensated 4DCBCT. In motion compensated acquisition, the imaging hardware is adapted to patient respiration in real time to ensure even data spread across the respiratory cycle. In motion compensated reconstruction, the Motion Compensated McKinnon Bates (MCMKB) algorithm estimates patient anatomy and motion so that data acquired at every phase contributes to reconstruction of each phase. Acquisition performance was quantified by scan dose and time, as well as Root-Mean-Square-Deviation (RMSD) between our real time adaptive gantry control against perfectly even breathing and data spread. Reconstruction performance was quantified with Contrast-to-Noise-Ratio (CNR), Structural SIMilarity Index (SSIM) to conventional and Tissue Interface Sharpness (TIS) at the diaphragm and tumour boundaries.
Results: Relative to conventional 4DCBCT, motion compensated 4DCBCT delivered 85% less dose, was on average 64% faster, had RMSD of 1.8⁰ from ideal, images had 35% better CNR, 99% structural similarity, 47% better diaphragm TIS and 56% better tumour TIS.
Conclusion: The ADAPT clinical trial provides real world evidence that motion compensated acquisition and reconstruction can be implemented on standard linacs to perform thoracic 4DCBCT with 85% less dose, 64% faster and with improved image quality.
Funding Support, Disclosures, and Conflict of Interest: This research was supported by NHMRC project grant #1138899 and grant #1123068 which was awarded through the Priority-driven Collaborative Cancer Research Scheme and funded by Cancer Australia. Ricky O'Brien acknowledges support of a Cancer Institute NSW Career Development fellowship. Paul Keall acknowledges support of an NHMRC Senior Principal Research Fellowship.