A Novel Classification Method of Hybrid Proton PBS Plans using DVH based Metrics
Advisor(s)
Draik Hecksel, Steven Laub, Mark Pankuch, Aditya Panchal; Northwestern Medicine
Location
Room IN2
Start Date
26-4-2019 1:40 PM
End Date
26-4-2019 2:05 PM
Abstract
Proton Pencil-Beam Scanning treatment plans are optimized using Single-Field Uniform Dose (SFUD), Multi-Field Optimization (MFO), or a combination of the two techniques into a Hybrid plan. In this study, we develop a method to evaluate plans using metrics applied to field-specific differential dose volume histograms (DVHs) from various treatment areas.
An application was developed to create normalized differential DVHs of the primary target volume for each field in a proton PBS treatment plan, and used five metrics to create a final ranking system for 235 patients plans. The results were then compared to their initially selected optimization technique, compared across treatment locations, and ran through statistical and machine-learning algorithms to test the validity of the ranking criteria.
Out of the 235 patient plans, our system reclassified 33 plans as MFO, 57 Hybrid, and 145 SFUD. Statistical analyses using ANOVA and T-test assuming unequal variances showed all values to be significantly different, and various clustering and re-classification methods proved our ranking system to be a more accurate representation of the treatment plans than the initial automatic optimization.
Related Image
Click on the image above to open a full resolution pdf
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
A Novel Classification Method of Hybrid Proton PBS Plans using DVH based Metrics
Room IN2
Proton Pencil-Beam Scanning treatment plans are optimized using Single-Field Uniform Dose (SFUD), Multi-Field Optimization (MFO), or a combination of the two techniques into a Hybrid plan. In this study, we develop a method to evaluate plans using metrics applied to field-specific differential dose volume histograms (DVHs) from various treatment areas.
An application was developed to create normalized differential DVHs of the primary target volume for each field in a proton PBS treatment plan, and used five metrics to create a final ranking system for 235 patients plans. The results were then compared to their initially selected optimization technique, compared across treatment locations, and ran through statistical and machine-learning algorithms to test the validity of the ranking criteria.
Out of the 235 patient plans, our system reclassified 33 plans as MFO, 57 Hybrid, and 145 SFUD. Statistical analyses using ANOVA and T-test assuming unequal variances showed all values to be significantly different, and various clustering and re-classification methods proved our ranking system to be a more accurate representation of the treatment plans than the initial automatic optimization.