Estimating Cumulative Radiation Dose to Organs at Risk for Glioblastoma Multiforme Cases via a Radiotherapy Dose Accumulation Routine

Abstract

Purpose/Objective(s)To evaluate the cumulative dose to organs at risk (OAR) for Glioblastoma multiforme (GBM) patients with multiple treatment courses via a purpose-developed RAdiotherapy Dose Accumulation Routine (RADAR), and to compare the results with manual estimations from the special medical physics consult (SPC).Materials/MethodsRADAR as a treatment planning system plugin was created to automatically transfer dose from previous radiation treatment plans to the current planning CT based on approved rigid (RR) or deformable image registration (DIR), convert the physical dose in each voxel to equivalent dose in 2Gy fractions (EQD2), and subsequently accumulate EQD2 doses from all relevant treatments. RADAR provides a novel option to take the registration and contour uncertainty into considerations by conservatively replacing dose to a given OAR voxel with a local maximum dose searched within an ellipsoid of the voxel. Ten GBM cases treated with 2 or more radiation courses between 2017 to 2020 were selected to analyze the registration/contour uncertainties and compare OAR dose from SPC and RADAR retrospectively. Brainstem, optic chiasm, and both optic nerves on previous treatment scans were propagated to the current CT scan based on RR and DIR. For each structure, mean distance to agreement (MDA) was measured between propagated previous and current OAR contours, and used for assigning search radius of ellipsoids accounting for uncertainties. An α/β of 2 for OARs and 10 for gross tumor volume were used for EQD2 conversion. The maximum dose (Dmax) and dose to 5% volumes (D05) from the accumulated EQD2 dose was evaluated based on the structures in the most recent plan and compared to SPC estimated dose. The SPC dose was documented by adding dose reported to the OAR based on the structure set in each of the previous plans without considering dose location.ResultsThe mean, standard deviation of the MDA of brainstem, optic chiasm, left and right optic nerve were used to determine a search radius of 2.0 mm, 2.5 mm, 2.5 mm, and 2.5 mm for these OAR, respectively. Comparing the manually estimated EQD2 dose from SPC and RADAR, the dose differences of -23Gy up to 40 Gy were found. RADAR used current structure set for OAR evaluation whereas SPC used the structure set from each course. Inconsistencies in contouring can be attributed to the large dose difference for some cases. For 5 of 10 test cases, the RADAR-derived OARs dose was higher than departmental guidelines and would potentially trigger a re-plan.ConclusionRADAR is capable to automatically convert the physical dose to EQD2 dose with various treatment courses by rigid or deformable image registration, take into account the registration and contour uncertainties and evaluate accumulated dose more accurately and efficiently than manual estimation. RADAR improves the workflow of decision making when optimizing plans with multiple previous treatments. To evaluate the cumulative dose to organs at risk (OAR) for Glioblastoma multiforme (GBM) patients with multiple treatment courses via a purpose-developed RAdiotherapy Dose Accumulation Routine (RADAR), and to compare the results with manual estimations from the special medical physics consult (SPC). RADAR as a treatment planning system plugin was created to automatically transfer dose from previous radiation treatment plans to the current planning CT based on approved rigid (RR) or deformable image registration (DIR), convert the physical dose in each voxel to equivalent dose in 2Gy fractions (EQD2), and subsequently accumulate EQD2 doses from all relevant treatments. RADAR provides a novel option to take the registration and contour uncertainty into considerations by conservatively replacing dose to a given OAR voxel with a local maximum dose searched within an ellipsoid of the voxel. Ten GBM cases treated with 2 or more radiation courses between 2017 to 2020 were selected to analyze the registration/contour uncertainties and compare OAR dose from SPC and RADAR retrospectively. Brainstem, optic chiasm, and both optic nerves on previous treatment scans were propagated to the current CT scan based on RR and DIR. For each structure, mean distance to agreement (MDA) was measured between propagated previous and current OAR contours, and used for assigning search radius of ellipsoids accounting for uncertainties. An α/β of 2 for OARs and 10 for gross tumor volume were used for EQD2 conversion. The maximum dose (Dmax) and dose to 5% volumes (D05) from the accumulated EQD2 dose was evaluated based on the structures in the most recent plan and compared to SPC estimated dose. The SPC dose was documented by adding dose reported to the OAR based on the structure set in each of the previous plans without considering dose location. The mean, standard deviation of the MDA of brainstem, optic chiasm, left and right optic nerve were used to determine a search radius of 2.0 mm, 2.5 mm, 2.5 mm, and 2.5 mm for these OAR, respectively. Comparing the manually estimated EQD2 dose from SPC and RADAR, the dose differences of -23Gy up to 40 Gy were found. RADAR used current structure set for OAR evaluation whereas SPC used the structure set from each course. Inconsistencies in contouring can be attributed to the large dose difference for some cases. For 5 of 10 test cases, the RADAR-derived OARs dose was higher than departmental guidelines and would potentially trigger a re-plan. RADAR is capable to automatically convert the physical dose to EQD2 dose with various treatment courses by rigid or deformable image registration, take into account the registration and contour uncertainties and evaluate accumulated dose more accurately and efficiently than manual estimation. RADAR improves the workflow of decision making when optimizing plans with multiple previous treatments.