Room: Karl Dean Ballroom C
Purpose: The current practical method of plan-quality evaluation by visual or automated examination of dose-volume metrics and visual inspection of 3D dose distribution can be time-consuming for online adaptive replanning. We proposed a method for using 3D dose features to replace the visual 3D dose inspection so that plan-quality evaluation can be fully automated.
Methods: Our method includes calculating three â€œtextureâ€? features from 3D dose distributions of plans to be evaluated and compared: (1) large cold spot (LCS), calculated from dose level size zone matrix (DLSZM), (2) hot contrast (HC), calculate from dose level co-occurrence matrix (DLCM), evaluating hotness of one plan to another plan at the same spatial voxel, and (3) mutual objective function distance (MOFD), evaluating the difference between the test plan and to the goal plan. The superior plan is determined from the differences of these features between the two plans to be compared. The method was used to evaluate 11 pairs of plans created from 5 selected prostate cases. Each pair included a repositioning plan and an adaptive plan generated from a daily CT set. The 11 pairs were categorized into two groups, Group I: adaptive plan superior than repositioning plan (6 pairs), and Group II: repositioning plan superior than adaptive plan (5 pairs), as determined based on the current plan evaluation method.
Results: All metrics differences between superior and inferiors plans were calculated, MOFD were (1.5Â±0.8)Ã—10â?¶ for Group I and (2.81Â±1.1)Ã—10â?µ for Group II, while the LCS differences in PTVs were -0.6Â±0.2 in Group I and -0.18Â±0.07 in Group II. The HC differences in OARs were -13.2Â±3.8 in Group I and -6.9Â±1.9 in Group II.
Conclusion: The proposed dose features extracted from 3D dose distribution along with dose-volume metrics can quickly evaluate plan quality and identify superior plan, thus, can be automated for online replanning.
Not Applicable / None Entered.