3035 - Quantitative Hounsfield Unit and Dose Calculation Accuracy of Iris CBCT for Head and Neck Radiotherapy
Presenter(s)
A. Keeler1, J. Mason2, T. Joyce2, E. S. Paulson1, M. J. Awan1, M. E. Shukla1, and E. A. Omari3; 1Department of Radiation Oncology, Medical College of Wisconsin, Milwaukee, WI, 2Elekta Limited, Crawley, West Sussex, United Kingdom, 3Medical College of Wisconsin, Milwaukee, WI
Purpose/Objective(s): Conventional c-arm linac CBCT image quality is degraded by scatter, noise, and artifacts, limiting its use in adaptive radiotherapy. In this study, we evaluate the dosimetric and image quality improvements of a novel CBCT reconstruction algorithm for Head and Neck (HN) adaptive radiotherapy.
Materials/Methods: A deep learning (DL) scatter correction combined with polyquant iterative reconstruction algorithm was implemented in a research imaging system. Raw CBCT projections from five HN patients were used to reconstruct monoenergetic CBCT images (Iris CBCT) at 64 keV. Patient selection included those with minimal anatomical differences between the planning CT (pCT) and first treatment fraction clinical CBCT (cCBCT). The clinical pCT contours for six organs-at-risk (OAR); mandible, brachial plexi, brainstem, esophagus, thyroid gland, and parotids, were rigidly copied to the cCBCT and Iris images. The absolute mean differences (MAE) in CT number (Hounsfield Unit, or HU) between the pCT and CBCT OARs were computed, with similarity assessed using Welch’s test. Theoretical calculations give mean energies of 69 keV for the 140 kV pCT spectrum and 63 keV for the 120 kV cCBCT spectrum, suggesting that systematic CT number changes between image types should be within 8% for bone and <1% for soft tissue. Dose accuracy was assessed by recalculating treatment dose on cCBCT and Iris images and comparing to the pCT doses using 3D global gamma analysis with 3%/2mm and 1%/2mm criteria.
Results: Iris OARs showed mean CT number values closer in agreement with the pCT compared to the cCBCT, with MAEs falling from 110-265 HU to 20-38 HU (p<0.05) for soft tissue and from 285 HU to 62 HU for bone (see table). Gamma pass rates showed significant improvements from cCBCT to Iris, improving from 58±17% to 99.4±0.2% (p=0.006) for 3%/2mm criteria and from 40±15% to 96.0±0.9% (p=0.001) for 1%/2mm criteria.
Conclusion: Iris HN CBCT images demonstrate significant improvement in quantitative CT number and dose calculation accuracy compared with cCBCT images. These improvements can facilitate online adaptive replanning using onboard kV CBCT imaging.
| Iris CBCT MAE (HU) | cCBCT MAE (HU) | |
| Mandible | 62±51 | 285±107 |
| Brachial Plexi | 38±23 | 265 ±97 |
| Brainstem | 35±22 | 163±71 |
| Esophagus | 23±15 | 237±91 |
| Thyroid Gland | 33±21 | 168±112 |
| Parotids | 22±26 | 110±70 |