Treatment Plan Complexity Quantification for Predicting Patient-Specific Quality Assurance of Stereotactic Volumetric Modulated Arc Therapy
X Xudong1, Y Ding2, W Wei3, C Ma4, X Wang5*, (1) Hubei Cancer Hospital, Wuhan, ,CN, (2) Hubei Cancer Hospital, Wuhan, ,CN, (3) Hubei Cancer Hospital, Wuhan, CN, (4) Rutgers University, New Brunswick, NJ, (5) Rutgers Cancer Institute of New Jersey, New Brunswick, NJ
Presentations
MO-E115-IePD-F4-6 (Monday, 7/11/2022) 1:15 PM - 1:45 PM [Eastern Time (GMT-4)]
Exhibit Hall | Forum 4
Purpose: To investigate the beam complexity of stereotactic VMAT plans quantitively and predict gamma passing rates (GPRs) using machine learning.
Methods: 301 clinical stereotactic VMAT plans with 594 beams from Varian Edge LINAC were collected. The GPRs were analyzed using Varian’s portal dosimetry with 2%/2 mm criteria. A novel leaf travel modulation complexity score for SBRT (LT_MCSs) complexity metric and other 25 metrics were calculated to investigate the correlation between metrics and GPRs. Random forest and gradient boosting models were developed to train and predict the GPRs based on the extracted complexity features. The threshold values of complexity metric were obtained to predict a given beam to pass or fail from ROC curve analysis.
Results: The three moderately significant values of Spearman’s rank correlation to GPRs were 0.508 (p<0.001), 0.445 (p<0.001), and -0.416 (p<0.001) for proposed metric LT_MCSs, the ratio of the average aperture area over jaw area (AAJA) and index of modulation (M), respectively. Both random forest and gradient boosting achieved 98.4% prediction accuracy with mean absolute error of 1.53% using five-fold cross-validation. LT_MCSs, leaf travelling distance (LT), LT modulation complexity score (LTMCS), M, total MU per beam, mean asymmetry distance (MAD) and edge area metric (EAM) were the top seven most important complexity features. The LT_MCSs metric showed the best performance with AUC value of 0.801, and threshold value of 0.365.
Conclusion: The proposed metric was effective in quantifying the complexity of stereotactic VMAT plans. We have demonstrated that the GPRs could be accurately predicted using machine learning methods based on extracted complexity metrics. The quantification of complexity and machine learning methods have the potential to improve SBRT treatment planning and identify the failure of QA results promptly.
Funding Support, Disclosures, and Conflict of Interest: the National Natural Science Foundation of China (No. 12075095)
Keywords
Quality Assurance, Treatment Planning, Treatment Verification
Taxonomy
Not Applicable / None Entered.
Contact Email