Knowledge-Based VMAT Planning of Total Marrow Irradiation as a Part of Conditioning Regimen Before Allo-SCT in a Prospective Phase I/II Study

Abstract

Total marrow irradiation (TMI) involves optimization of extremely large target volumes and requires extended time for treatment planning and delivery. Although volumetric modulated arc therapy (VMAT) achieves substantial reduction in the delivery time, the planning process still presents a challenge with multiple isocenters and multiple overlapping arcs to adequately cover sub-target volumes of head and neck, chest, and pelvis. We developed and applied a knowledge-based planning (KBP) model to set optimization objectives, and evaluated its performance for VMAT-TMI. We created 3 model libraries for head and neck, chest, and pelvis plans from 20 patients treated on a prospective study examining the utility of TMI. Plans utilized a 3-isocenter, 13-arc VMAT technique. Head and neck and pelvis plans were optimized utilizing the chest plans as the base dose to prevent hot spots around the field junctions. For the KBP model, they were re-optimized without the base-dose plan and were added to the corresponding library to minimize scatters in dose-volume histogram (DVH) plots of target volumes. The KBP models were validated on a cohort of 6 recent patients who were not used for the training. Optimization objectives for organs at risk (OARs) away from the junction regions were placed along the predicted DVHs. For OARs overlapping two sub plans, sums of predicted mean dose values from each sub plan were used as objectives. The KBP database generated achievable mean dose values along with the prediction ranges of DVHs for each OAR. Performance of KBP generated within 2∼3 iterative optimizations was comparable to clinical plans that used to take a few days of effort with indefinite number of iterations. KBP provided target coverages (95% volume covered by the prescribed dose of 9 Gy) and the homogeneity index (<1.4) in accordance with our institutional practice. Both clinical plans and KBP had mean dose values in the range of 104%-111% of the prescribed dose for planning target volumes, 23%-29% for lenses and oral cavity, and 42%-63% for brain, heart, lungs, bowels, liver, kidneys, eyes. Our study demonstrates that KBP can be implemented for Linac-based multiple-isocenter TMI plans by creating a model for each sub plan. VMAT-TMI powered by KBP can potentially support uniform dosimetric quality among different users and clinics, and can reduce treatment planning time and effort by providing a goal attainable within a few iterative optimizations.Tabled 1Abstract 2665; Table; Comparison of mean dose values to OAR (% of prescribed dose)bowelsbraineyesheartkidneysliverlenseslungsoral cavityclinical plans48±660±444±351±647±651±529±263±425±6KBP42±657±545±349±348±651±229±159±323±2 Open table in a new tab