Building Treatment Planning Knowledge Base In RT Jackie Wu, PhD FAAPM Professor Disclosure Research Grant: NIH/NCI Master Research Grant: Varian Medical Systems License Agreement: Varian Medical Systems Speaker Agreement: Varian Medical Systems Duke University Medical Center Department of Radiation Oncology DUKE University Radiation Oncology DUKE University Radiation Oncology Building Planning Knowledge Base for RT Goal: Model Best Treatment Option(s) for A Patient Based on His/Her Unique Anatomy, Medical Condition and Consensus Data Target and OAR delineation Dose prescription/fractionation Treatment plan/parameter design Treatment plan evaluation/selection Outcome and plan dosimetry DUKE University Radiation Oncology 1 Building Planning Knowledge Base for RT Building Planning Knowledge Base: All Component of Tx. Plan Design Planning Knowledge Sources Past data Guidelines, clinical trials, peer reviewed papers, normal tissue complication models Experience Planning Knowledge Sources Plans, quality assessment, patient anatomy, machine characteristics, calibration, treatment outcomes Formalized knowledge Quantitative vs. qualitative Volume, overlaps, etc Guidelines, outcome data, etc Continuous vs. discrete Beam angles, Tradeoffs, etc Numerical vs. Descriptive Cord Distance to OAR (DOH): PCS OAR PTV PTV volume Fraction of OAR volume overlapping PTV (overlap volume) Case with 2 Spinal volumes Case 2 Case 1 Cord Fraction of OAR volume outside the treatment fields (out-of-field volume) PTV Tightness of the geometric enclosure of PTV surrounding OAR Curvature of specific OAR Percent Dose (%) In the head of physicians, physicists, dosimetrists, therapists DUKE University Radiation Oncology Distance to target histogram (DTH): Case 1PCS 1 Spinal Cord and PTV DVHs Case Spinal Percent Volume (%) Building Planning Knowledge Base for RT DUKE University Radiation Oncology DUKE University Radiation Oncology 2 Building Planning Knowledge Base: All Component of Tx. Plan Design Building Planning Knowledge Base: All Component of Tx. Plan Design Rectum _ _ HN Patient and dose prescription Model Predict parotid D50 by “standard model” Plan N Y Meet Tradeoff Criteria? Predict parotid DVHs by “standard model” DUKE University Radiation Oncology Building Planning Knowledge Base: All Component of Tx. Plan Design R Which side has lower D50? Predict R parotid DVH model” DUKE University Radiation Oncology by “combined L Predict L parotid DVH by “combined model” DUKE University Radiation Oncology 3 Building Planning Knowledge Base: All Component of Tx. Plan Design DUKE University Radiation Oncology Building Planning Knowledge Base: All Component of Tx. Plan Design DUKE University Radiation Oncology Building Planning Knowledge Base: All Component of Tx. Plan Design DUKE University Radiation Oncology Courtesy Of Y. Ge @UNCC 4 Building Planning Knowledge Base: Integrated Decision Support Tool Prediction Building Planning Knowledge Base: Multi-institution, Multi-modality Model Building Planning Knowledge Base: Multi-institution, Multi-modality Model Guidelines Parotid DVH Institution 2 Not contributing to model Gui Contralateral Lung DVH _ _ Plan vs. Tomotherapy Model vs. IMRT Model Model Plan Institution 3 Contributing to model DUKE University Radiation Oncology Courtesy Of Y. Ge @UNCC DUKE University Radiation Oncology DUKE University Radiation Oncology 5 Building Planning Knowledge Base: Platform Design Knowledge Base Training Data base of High Data base of DataDatabase base ofHigh High of quality treatment quality treatment quality treatment Expert plan plan plan Knowledge & Treatment Cases Patient & Plan Design Features Building Planning Knowledge Base: Platform Design Dose Parameter Features Output Features Yi Input Features Xi Machine Learning Knowledge Base Y1= F1(X) Y2=F2(X) Knowledge Base Application Feature New Characterization DUKE University Radiation Oncology Patient Xnew Knowledge Base Application Y=f(X) Dose Parameter Features Ynew Feature New Characterization DUKE University Radiation Oncology Patient Xnew Y=f(X) Dose Parameter Features Ynew Acknowledgement Yaorong Ge, PhD FangFang Yin, PhD Lulin Yuan, PhD Jianfei Li, PhD Ying Xiao, PhD Jun Lian, PhD Taoran Li, PhD Simin Lu, MS Yang Sheng, MS DUKE University Radiation Oncology 6 References AAPM2014: SU-F-BRD-9: Bouquet Templates AAPM2014: TU-C-17A-11: Lu et al, Progressive Knowledge Modeling for Pelvic IMRT/VMAT Treatment Planning Yuan et al, Lung IMRT Planning Using Standardized Beam AAPM2014: TH-A-9A-1: Liu et al, Active Optical Flow Model: Predicting Voxel-Level Dose Prediction in Spine SBRT AAPM 2014: SU-E-T-527: Treatment Planning Lian et al, Prior Knowledge Guided TomoTherapy Lian et al, Modeling the dosimetry of organ-at-risk in head and neck IMRT planning: An intertechnique and interinstitutional study, Medical Physics 2013, 40(12) Yuan et al, Incorporating single-side sparing in models for predicting parotid dose sparing in head and neck IMRT, Medical Physics 2014, 41(2) Liu et al, An Active Optical Flow Model for Dose Prediction in Spinal SBRT Plans, MICCAI Workshop on Computational Methods and Clinical Applications for Spine Imaging 2014 Zhu et al, A planning quality evaluation tool for prostate adaptive IMRT based on THANK YOU! machine learning, 2011, 38(2) Yuan et al, Quantitative analysis of the factors which affect the inter-patient DUKE University Radiation Oncology organ-at-risk dose sparing variation in IMRT plans, Medical Physics 2012, 39(11) DUKE University Radiation Oncology 7