Room: AAPM ePoster Library
Purpose: To evaluate two atlas-based methods for PTV contouring for total marrow irradiation (TMI).
Methods: Forty patient contours (20 male, 20 female) from cases previously treated with TMI were loaded into the Eclipse Smart Segmentation (version 13.6) and Velocity (version 3.2.1) contouring atlases. Contours extended from the top of the skull to approximately mid-thigh, which represents the region treated with VMAT fields at our center. Five male test patients were then used to create automated PTV contours, PTV(Eclipse) and PTV(Velocity), based on deformation of contours from matches found within the atlases. These automated PTVs were then compared to the physician-drawn PTV contour, PTV(RO), and evaluated for accuracy and efficiency.
Results: The overlap between the atlas-based contours and PTV(RO) was calculated for each patient using Boolean operators. In all cases, PTV(Eclipse) covered more than 80% of PTV(RO) (average: 83.8%, range: 81.5% - 88.8%) while the overlap between PTV(Velocity) and PTV(RO) ranged from 76.8% to 92.3% (average: 82.6%). The software that provided a better match was not consistent over the five patients, as Eclipse was better in 3/5 cases, while Velocity was better in the other 2. In 4 out of 5 cases, PTV(Velocity) had a larger volume of over-contouring than PTV(Eclipse), signifying more necessary edits, and in 3/5 cases it was by more than double.
Conclusion: Atlas-based contours from both software resulted in similar overlap with approved physician contours, however PTV(Eclipse) was more efficient as it had significantly less over-contouring, indicating decreased time required for editing.
Contour Extraction, Deformation, Treatment Planning
TH- External Beam- Photons: Treatment planning using machine learning/Knowledge Based Planning/automation