Application of Machine Learning to Patient-Specific IMRT Quality Assurance

Abstract

Medical physics is a research field of applying the concepts and methods of physical sciences to medicine, especially radiology including radiation therapy. Intensity modulated radiation therapy (IMRT) is a state of the art technology of radiation therapy which has been developed based on the achievements in medical physics. Managing uncertainty including a detection of unacceptable error is a central task in safe and accurate delivery of IMRT to patients. We developed a machine learning models to automatically detect several errors possibly occur in IMRT dose calculation and IMRT dose delivery system of medical linear accelerator. The models are based on radiomics analysis of X-ray fluence distributions which is a method of extracting a large number of features from medical images. The proposed models showed superior performance to the conventional methods and may expand the possibilities of automatic error detection of IMRT.