2133 - Evaluating an Algorithm for CBCT Correction for Adaptive Planning
Presenter(s)
C. Sawyer1, J. J. Caudell1, A. O. Naghavi1, M. Qayyum2, V. Feygelman1, and K. Latifi1; 1H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL, 2RaySearch Laboratories AB, Stockholm, Sweden
Purpose/Objective(s): To increase the efficiency of adaptive radiotherapy we tested a cone beam computed tomography (CBCT) correction algorithm to evaluate the use of daily CBCTs for planning and eliminate the need for a computed tomography (CT) re-simulation.
Materials/Methods: A lung phantom was scanned with a CT SIM and CBCT on two different clinical treatment linacs. The CBCTs were processed through a correction algorithm included in the treatment planning system (TPS). The algorithm simultaneously tries to identify and remove artifacts while also scaling and smoothing the image intensity closer to that of an initial CT to generate corrected CBCTs. Voxels outside of the CBCT’s field of view are replaced with voxels deformed from an initial planning CT. A plan was generated for the CT image and then the plan was recalculated on the corrected CBCTs. The same workflow was followed for 5 previously adapted head and neck and 5 sarcoma patients. Their adaptive plans were then recalculated on the corrected CBCTs to evaluate the differences between the two dose volumes. Dose differences were then analyzed for all of these plans using a 3%/2mm gamma analysis.
Results:
Both plans on the CBCTs of the phantom had high matching dose per voxel according to the gamma analysis. Six of the recalculated plans also had over 95% (3%/2mm) gamma passing rate. Seven out of ten were over 90% passing. The remaining three plans all had less than 85% agreement and would not pass clinical quality assessment in our clinic. In the cases with low gamma passing rates small field of view of the CBCT clips part of the target leading to missregistration between the planning CT and CBCT that introduces errors in the corrected CBCT.Conclusion: The CBCT correction algorithm demonstrates a promising new opportunity to eliminate the need for re-simulation and enable quicker turnaround for the offline adaptive plans. While Gamma analysis was excellent on phantoms, it failed in 3/10 patient cases. It is important that the field of view of CBCT at the very least includes the target otherwise the missing information can introduce errors in the CBCT correction process.
Abstract 2133 - Table 1: Gamma analysis of corrected CBCT vs. delivered adaptive plan (3%/2mm)| CBCT Plan vs. CT Plan | Matching(voxels) | High(voxels) | Low(voxels) | 3%/2mm Gamma |
| Phantom CBCT | 160095 | 102 | 0 | 99.9 |
| Phantom CBCT 2 | 159681 | 0 | 0 | 100 |
| H&N_1 | 278141 | 18846 | 49988 | 80.2 |
| H&N_2 | 305888 | 9948 | 14746 | 92.5 |
| H&N_3 | 122908 | 599 | 476 | 99.1 |
| H&N_4 | 537420 | 17576 | 4567 | 96 |
| H&N_5 | 184415 | 48 | 2669 | 98.5 |
| Sarcoma_1 | 170354 | 11962 | 62077 | 69.7 |
| Sarcoma_2 | 2160962 | 11759 | 31051 | 98.1 |
| Sarcoma_3 | 86932 | 10721 | 5656 | 84.1 |
| Sarcoma_4 | 1462236 | 19384 | 32639 | 96.6 |
| Sarcoma_5 | 643791 | 7960 | 4334 | 98.1 |