(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: Critical issues in CBCT-based adaptive radiotherapy and dose-of-the-day calculations are inaccuracy of HU values and omission of relevant anatomy in CBCT images. This study presents a procedure using Deformable Image Registration (DIR) to correct both problems.
Methods: 21 patients with 5 different treatment sites/techniques were assessed, including Head&Neck (HN), liver, lung, breast, and spotlight breast. CBCTs were resampled and padded with air to match the planning CT (pCT) voxel spacing and image array size. An open-source DIR algorithm was used to rigidly align and then deform the pCT to the CBCT, with a cylindrical mask indicating the registration volume. For spotlight cases, rigid alignment was performed in a TPS and transferred to the DIR algorithm. Registration cost function used mutual information and a bending energy penalty; the transformation employed a cubic B-spline basis. Deformed images were visually inspected for similarity to CBCT and pCT, and the spatial Jacobian was assessed to determine if physically-possible deformation had occurred.
Results: This method takes less than 30 minutes to complete per patient. For HN, it was able to correct for dental artifacts and expand the FOV to include dosimetrically-impactful anatomy such as shoulders. For liver, it corrected artifacts from gas buildup while returning the entire pCT FOV. For spotlight breast, it expanded the FOV to include the full pCT anatomy, while reflecting anatomical differences captured by CBCT. However, this deformation displayed some artifacts along parts of the CBCT FOV edge.
Conclusion: The proposed procedure provides a practical and flexible methodology for dose reconstruction at each fraction, requiring no additional software licensing or re-simulation scan for different treatment sites/techniques.
Keywords
Not Applicable / None Entered.
Taxonomy
IM/TH- Image Registration: Multi-modality registration
Contact Email
