Program Information
Can We Extract Lung Function Directly From 4D-CT Without Deformable Image Registration?
J Kipritidis1*, M Hofman2 , S Siva2 , J Callahan2 , P Le Roux2 , H Woodruff1 , W Counter1 , N Hardcastle3 , P Keall1 , (1) University of Sydney, Sydney, NSW, Australia (2) Peter MacCallum Cancer Centre, Melbourne, VIC, Australia (3) Royal North Shore Hospital, Sydney, NSW, Australia
Presentations
TU-G-BRA-2 (Tuesday, July 14, 2015) 4:30 PM - 6:00 PM Room: Ballroom A
Purpose:
Dynamic CT ventilation imaging (CT-VI) visualizes air volume changes in the lung by evaluating breathing-induced lung motion using deformable image registration (DIR). Dynamic CT-VI could enable functionally adaptive lung cancer radiation therapy, but its sensitivity to DIR parameters poses challenges for validation. We hypothesize that a direct metric using CT parameters derived from Hounsfield units (HU) alone can provide similar ventilation images without DIR. We compare the accuracy of Direct and Dynamic CT-VIs versus positron emission tomography (PET) images of inhaled ⁶⁸Ga-labelled nanoparticles (‘Galligas’).
Methods:
25 patients with lung cancer underwent Galligas 4D-PET/CT scans prior to radiation therapy. For each patient we produced three CT-VIs. (i) Our novel method, Direct CT-VI, models blood-gas exchange as the product of air and tissue density at each lung voxel based on time-averaged 4D-CT HU values. Dynamic CT-VIs were produced by evaluating: (ii) regional HU changes, and (iii) regional volume changes between the exhale and inhale 4D-CT phase images using a validated B-spline DIR method. We assessed the accuracy of each CT-VI by computing the voxel-wise Spearman correlation with free-breathing Galligas PET, and also performed a visual analysis.
Results:
Surprisingly, Direct CT-VIs exhibited better global correlation with Galligas PET than either of the dynamic CT-VIs. The (mean ± SD) correlations were (0.55 ± 0.16), (0.41 ± 0.22) and (0.29 ± 0.27) for Direct, Dynamic HU-based and Dynamic volume-based CT-VIs respectively. Visual comparison of Direct CT-VI to PET demonstrated similarity for emphysema defects and ventral-to-dorsal gradients, but inability to identify decreased ventilation distal to tumor-obstruction.
Conclusion:
Our data supports the hypothesis that Direct CT-VIs are as accurate as Dynamic CT-VIs in terms of global correlation with Galligas PET. Visual analysis, however, demonstrated that different CT-VI algorithms might have varying accuracy depending on the underlying cause of ventilation abnormality.
Funding Support, Disclosures, and Conflict of Interest: This research was supported by a National Health and Medical Research Council (NHMRC) Australia Fellowship, an Cancer Institute New South Wales Early Career Fellowship 13-ECF-1/15 and NHMRC scholarship APP1038399. No commercial funding was received for this work.
Contact Email: