Room: Exhibit Hall | Forum 2
Purpose: We evaluated the effects of three factors that may effect on 4D Cone Beam CT (4D-CBCT)-ventilation images (VICBCT): (1) Linac image processing, (2) 4D-CBCT scan speed, and (3) deformable image registration (DIR) strategy.
Methods: First, 4D-CT and 4D-CBCT scans were performed on an in-house deformable lung phantom using planning CT scanner, XVI [Versa HD (VHD); Elekta], and OBI [TrueBeam (TB); Varian]. 4D-CBCT scans were acquired using different gantry rotation speeds (0.6, 1.0, 2.0, and 3.0 deg/s for VHD 2.0 and 3.0 deg/s for TB). Next, DIR strategies were performed between peak-exhale and peak-inhale images: one was intensity-based DIR (DIRIN) and the other was structure-based DIR (DIRST). VICBCT and 4D-CT ventilation images (VICT) were derived by DIR using the Jacobian determinant-base (Jac) and changes in HU value-base (HU) metrics. Finally, we evaluated correlations between VICBCT and each factor, and VICT (as ground truth) values were calculated using voxel-wise Spearmanâ€™s rank-correlation (rspear).
Results: The maximum differences of rspear for both Linacs were 0.58 (VHDJac â€“ TBJac), and 0.15 (VHDHU â€“ TBHU). For each scan speed, rspear values were 0.68 for 0.6 deg/s, 0.28 for 1.0 deg/s, 0.21 for 2.0 deg/s, and 0.23 for 3.0 deg/s (VHDJac). Further, rspear were 0.10 for 2.0 deg/s and 0.10 for 3.0 deg/s (TBJac). For each Linac, average rspear (range) were 0.51 for DIRST,ã€€0.13 for DIRIN (VHD), and 0.25 for DIRST, 0.10 for DIRIN (TB).ã€€The strongest rspear: 0.85 was observed for VHDJac (using DIRST).
Conclusion: We investigated the factors that affected 4D-CBCT ventilation images using an in-house deformable lung phantom. Our results showed that the effect caused by three factors on VICBCT could be different for two Linacs respectively. Further studies are needed to evaluate VICBCT and VICT using patient-generated images.