Room: Exhibit Hall | Forum 9
Purpose: Cardiac motion is not accounted for during thoracic CBCT imaging, thus contributing to motion artifacts that degrade image quality and complicate targeting in adaptive lung radiotherapy. External markers may be used to monitor respiratory and cardiac phase, however extracting this information from CBCT images is a more direct way of relating cardiorespiratory motion to tumour position. This study investigates our hypothesis that variations in CBCT projection pixel intensity arising from cardiorespiratory motion can be used as a measure for the motion of centrally-located tumours.
Methods: Thoracic CBCT scans (5 Hz frequency) of five patients previously treated for locally advanced NSCLC from The Cancer Imaging Archive were analysed in this study. For each scan, implanted fiducial markers were segmented from the projections to quantify their three-dimensional (3D) position as a surrogate for ground-truth tumour motion. An existing algorithm that measures respiratory and cardiac phase intrinsically from CBCT projections was used to quantify the time-dependent mean gray value (MGV) in a region of interest around fixed points in 3D image space subject to cardiorespiratory motion. Frequency spectra of the 3D ground-truth tumour displacement and the MGV signals measured intrinsically from the CBCT images were compared to validate the algorithm's ability to identify low and high frequencies corresponding to respiratory and cardiac motion signals, respectively.
Results: Respiratory and cardiac peaks in the intrinsically-measured frequency spectra were identified for all patient scans in two distinct regions between 0.27â€“0.40 Hz and 1.21â€“1.57 Hz, which correspond with normal values for respiratory and cardiac rates, respectively. Respiratory and cardiac peaks were identified from 5/5 and 4/5 ground-truth spectra, respectively, and agreement between with the intrinsically-measured peaks was within 3%.
Conclusion: Cardiorespiratory signals of centrally-located tumour motion can be identified intrinsically from CBCT projections.
Funding Support, Disclosures, and Conflict of Interest: This research was supported by grant 1123068 which was awarded through the Priority-driven Collaborative Cancer Research Scheme and funded by Cancer Australia. Ricky O'Brien acknowledges the support of a Cancer Institute NSW Career Development fellowship. Paul Keall acknowledges the support of an NHMRC Senior Principal Research Fellowship.
Cone-beam CT, Motion Artifacts, Image Processing