(Wednesday, 8/1/2018) 4:30 PM - 6:00 PM
Room: Room 207
Purpose: The plan quality of prostate cancer patients treated with pencil-beam scanning (PBS) proton therapy can vary largely due to the extent of disease and the proximity of the organs-at-risk (OARs) to the planning target volume (PTV). We developed a proton knowledge-based planning (pKBP) approach to predict the patient-specific OAR dose-volume histograms (DVHs) in prostate proton PBS cases planned with opposed lateral beams using a novel proton-specific metric.
Methods: Twenty-five patients with prostate cancer treated with PBS proton therapy to 7920 cGy with 180 cGy per fraction were retrospectively identified for this study. Twenty of these patients were randomly selected to build the knowledge-based planning model. A proton-specific metric was developed to obtain the histogram of OAR volume overlapping with the PBS spots binned in the PBS energy layers. The principal components of the overlapping histograms were obtained. The DVHs of the bladder and rectum were also decomposed into principal components. A machine learning approach was used to predict the weights of DVH principal components from the weights of the principal components of the OAR overlapping histograms. The remaining five test patients were used to validate this pKBP DVH prediction model.
Results: The five rectum and bladder DVHs matched well with predicted values. The differences of predicted and planned rectum DVHs are -1.6±3.6% when measured at 1% of dose intervals (792cGy). The differences of predicted and planned bladder DVHs are -1.5±3.7%.
Conclusion: Knowledge based planning that uses a novel proton-specific metric successfully predicted the OAR DVHs in prostate proton PBS treatment plans. This pKBP approach can be used for patient proton plan QA similar to the photon knowledge-based planning tool. We are currently extending the approach to other disease sites such as lung and head and neck.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
