2166 - Development of an Online Adaptive Decision Making Tool for Helical Delivery Platforms
Presenter(s)
O. D. H. Luu, A. Amjad, J. Garcia Alvarez, M. Huang, E. S. Paulson, and G. P. Chen; Department of Radiation Oncology, Medical College of Wisconsin, Milwaukee, WI
Purpose/Objective(s): With kV CT images available, helical delivery platforms facilitate online adaptive workflows. This study aimed to provide an independent decision-making framework for adaptive planning on such systems.
Materials/Methods: Three prostate patients treated on treatment delivery system, each with a total planned dose of 70 Gy delivered in 28 fractions, were examined in this work. Deep learing auto-segmentation (Manteia) was performed on each daily kV CT. Segmentation accuracy was verified by comparing to manually constructed ground truth bladder and rectum contours using Dice similarity coefficient (DSC) and mean distance to agreement (MDA). Daily dose was calculated using SureCalc (MIM software), with reference DICOM RT plan modified to have daily isocenter shift, and compared against that from PreciseART, using 3D gamma index analysis. Level 1 gamma criteria of TG218 were used. Dose-Volume Histogram (DVH) parameters, V(4000 cGy), V(6500 cGy) and V(7000 cGy) for both bladder and rectum were evaluated for each fraction, and their deviation from the planning goal, with 35, 10 and 3%, respectively, was used as an indication of the necessity of adaptive planning.
Results: Overall, bladder and rectum DSC exceeded 0.7 and MDA fell under 5mm for all patients and fractions studied. The dose calculated with SureCalc agrees with that of PreciseART very well, the gamma passing rate was 99.95 ± 0.08%, with a minimum of 99.54%. DVH parameters for the three cases are summarized in the Table. For most fractions, the bladder and rectum volumes receiving 40 Gy and 65 Gy met dose-volume constraints. In certain fractions, the bladder exceeded the 70 Gy constraints. Online adaptive planning was recommended for 35 out of 59 fractions for the three patients studied.
Conclusion: We developed an independent framework that can trigger online adaptive planning based on daily anatomy change. This can help establish guidelines for the potential online adaptive workflow on helical delivery platforms.
Abstract 2166 - Table 1| Case | 1 | 2 | 3 | ||||
| OARs | Bladder | Rectum | Bladder | Rectum | Bladder | Rectum | |
| Volume (cc) | Original Plan | 390.1 | 59.6 | 99.2 | 183.0 | 279.8 | 39.9 |
| Range | 193.9 - 428.5 | 28.1 – 73.6 | 54.3 – 368.3 | 46.2 – 176.7 | 120.7 – 228.9 | 32.8 – 68.5 | |
| Average±Std | 295.4 ± 63.6 | 47.7 ± 12.2 | 158.9 ± 79.5 | 91.8 ± 50.5 | 169.1 ± 26.8 | 46.7 ± 10.5 | |
| V(4000 cGy) | Original Plan | 92.7 | 14.9 | 32.5 | 58.6 | 65.5 | 13.3 |
| Range | 28.4 – 61.89 | 8.0 – 26.6 | 23.3 – 59.67 | 11.2 – 53.8 | 55.6 – 83.1 | 4.7 – 21.8 | |
| Average±Std | 51.4 ± 8.6 | 13.8 ± 4.3 | 40.5 ± 10.5 | 27.8 ± 14.8 | 66.2 ± 7.6 | 9.1 ± 5.0 | |
| V(6500 cGy) | Original Plan | 35.8 | 5.4 | 7.8 | 12.8 | 16.1 | 3.2 |
| Range | 5.45 – 23.5 | 1.2 – 10.8 | 1.5 – 17.3 | 0.6 – 11.8 | 18.5 – 42.7 | 0.1 – 6.5 | |
| Average±Std | 15.0 ± 4.2 | 4.0 ± 2.6 | 9.7 ± 5.1 | 4.2 ± 3.5 | 28.7 ± 6.2 | 1.8 ± 1.8 | |
| V(7000 cGy) | Original Plan | 22.1 | 3.0 | 2.0 | 2.9 | 4.7 | 0.9 |
| Range | 2.2 – 12.4 | 0.2 – 7.7 | 0.0 – 3.6 | 0.0 – 1.1 | 8.3 – 30.6 | 0.0 -3.1 | |
| Average±Std | 7.5 ± 2.5 | 2.0 ± 1.9 | 1.6 ± 1.2 | 0.4 ± 0.3 | 17.6 ± 5.6 | 1.0 ± 0.6 | |