Room: AAPM ePoster Library
Purpose: To develop a knowledge-based planning (KBP) model to automate treatment planning for craniospinal irradiation and evaluate clinical-suitability by comparison with a traditionally developed clinical plan (CP).
Methods: Volumetric Modulated Arc Therapy (VMAT) CSI previously delivered plans were used to build a RapidPlan model in Aria v15.6 (Varian Medical Systems). In each, segmented organs at risk (OAR) included esophagus, lungs, kidneys, heart, thyroid and bowel. Once the model was created, a KBP was created for an additional CSI patient case outside the training set (Rx=3600cGy). Optimization of the KBP was limited to the six OARs listed above with RapidPlan generated constraints and priorities. The KBP was renormalized to 99% to ensure equivalent PTV coverage. KBP and CP dosimetry was then compared.
Results: PTV D98%, D95%, and heterogeneity index (e.g. (D2%-D98%)/Rx) were 97.4%, 99.6%, 9.7%, and 97.8%, 99.7%, 11.3% for the KBP and CP, respectively. The mean dose to esophagus, lungs, kidneys, heart, thyroid and bowel was (1079 cGy, 761 cGy, 535 cGy, 708 cGy, 768 cGy, 735 cGy) for the KBP and (1331 cGy, 728 cGy, 513 cGy, 721 cGy, 845 cGy and 639 cGy) for CP. The stomach, bronchial tree and liver were not included in the model and resulted in mean doses of (682 cGy, 1502 cGy and 593 cGy) for KBP and (738 cGy, 1261 cGy and 562 cGy) for CP, respectively. Total MU was 345 (KBP) and 471 (CP).
Conclusion: similar dosimetry of the KBP and CP has demonstrated the utility of KBP to generate clinically deliverable plans for large PTVs including CSI with a 25% reduction in MUs. Considering OARs outside the model (e.g. stomach, bronchial and liver), KBP was able to achieve comparable or slightly better sparing.
TH- External Beam- Photons: Treatment planning using machine learning/Knowledge Based Planning/automation