Stereotactic Body Radiation Therapy for Low-Intermediate Risk Prostate Cancer—A Single-Institution Initial Analysis of Toxicity

Abstract

therapy. Patients treated by three categories of techniques, including 40 robotic radiation therapy cases, 66 static angle IMRT cases and 94 VMAT cases were included in this study. The dose criteria from the protocol were used to evaluate the original planned rectum and bladder doses. For each patient, a set of modeled normal tissue dose was generated with a machine learning tool, which reads the prostate cancer patients' CT and structure DICOMs, and predicts DVHs for normal structures, i.e. bladder and rectum, by using a 7-beam IMRT plan model from a published institution study. The modeled normal tissue doses were evaluated by the same protocol criteria, and results were compared with the results from the original plans for each technique category. Results: Both the original plan and modeled normal tissue doses met most of the protocol criteria, including D(%)90(%) and D(%)50(%) for bladder and D(%)90(%), D(%)80(%) and D(%)50(%) for rectum. For bladder Dmax, 30% of the original plan failed the criteria, while all the modeled doses met the criteria. The bladder Dmax failure rate is highest for robotic radiosurgery cases (73%) and lowest in IMRT cases (16%). For rectum V(cc)95(%), 8% of the original plan failed the criteria, while 40% of the modeled doses failed the criteria. The rectum V(cc)95(%) failure rate was highest in IMRT cases (12%) and lowest in robotic radiosurgery cases (3%). Conclusion: Trade-off between rectum and bladder doses was observed for different treatment delivery techniques. It is feasible to use machine learning techniques for establishment of radiation therapy dosimetric compliance criteria.