A Physiologic Adaptive Radiation Therapy Pipeline for Glioblastoma by Daily Multiparametric MRI and Machine Learning

点击翻译按钮获取中文摘要

Project Summary/Abstract Glioblastoma is the most common cancer originating in the brain with ~12,000 new diagnoses per year in the U.S.A. and median survival about 18 months. A common dilemma for the patient and treatment team is that the clinical MRI one month after radiation therapy (RT) shows growth of unclear significance in up to 50% of patients. Patients with true progression (TP) of non-responding tumor continue to progress on serial MRIs and usually die within 9 months. TP is usually determined 6 or more months after completion of RT, often when it is too late to intervene. The mission of the National Cancer Institute is to improve survival of patients with cancer. Our goal, and an unmet need, is to identify glioblastoma patients with TP early during treatment and implement aggressive second-line therapy to improve survival. This proposal describes an innovative approach to identify early glioblastoma TP by improving neuroimaging and image processing on MRIdian, a new combination MRI and RT (MRI-RT) device from ViewRay, Inc where patients undergo MRI daily as part of their RT. Our preliminary data with MRIdian is the first to demonstrate daily glioblastoma growth on MRI in patients during RT. By developing physiologic MRI techniques on MRIdian (Aim 1), we seek to identify TP when there is growth during RT (Aim 2), and intensify RT to that TP (Aim 3/future). The most sensitive and specific commonly applied clinical MRI techniques for identifying TP after RT correlate with tumor physiology: diffusion (cellularity), perfusion (hypoxia), and spectroscopy (metabolism). These are collectively termed mpMRI. The hypothesis of aim 1 is that the academic-industrial partnership between Miami and ViewRay can develop mpMRI for daily measurements during RT on MRIdian. The hypothesis of aim 2 is that the images from daily mpMRI during MRI-RT in glioblastoma patients can be processed by machine learning and radiomics techniques to automatically detect glioblastoma growth and predict long-term outcome. Aim 3 then combines aims 1-2 to test a prospective workflow to intensify RT to TP when TP is first identified during RT based on mpMRI elucidated trends in tumor physiology, so called “physiologic adaptive RT” (PART). The PART workflow will be developed by Miami and Viewray and integrated into MRIdian. The advantage of using single platform MRIdian is that the PART workflow will be distributed by Viewray to the over 60 MRIdian centers. This easy clinical translation will permit us to proceed with multi-institutional trials of early RT dose escalation to improve survival of poorly responding glioblastomas.