Fully automated, multi-criterial planning for Volumetric Modulated Arc Therapy – An international multi-center validation for prostate cancer

Publication date: Available online 30 June 2018
Source:Radiotherapy and Oncology
Author(s): Ben Heijmen, Peter Voet, Dennie Fransen, Joan Penninkhof, Maaike Milder, Hafid Akhiat, Pierluigi Bonomo, Marta Casati, Dietmar Georg, Gregor Goldner, Ann Henry, John Lilley, Frank Lohr, Livia Marrazzo, Stefania Pallotta, Roberto Pellegrini, Yvette Seppenwoolde, Gabriele Simontacchi, Volker Steil, Florian Stieler, Stuart Wilson, Sebastiaan Breedveld
Background and purposeReported plan quality improvements with autoplanning of radiotherapy of the prostate and seminal vesicles are poor. A system for automated multi-criterial planning has been validated for this treatment in a large international multi-center study. The system is configured with training plans using a mechanism that strives for quality improvements relative to those plans.Material and methodsEach of the four participating centers included thirty manually generated clinical Volumetric Modulated Arc Therapy prostate plans (manVMAT). Ten plans were used for autoplanning training. The other twenty were compared with an automatically generated plan (autoVMAT). Plan evaluations considered dosimetric plan parameters and blinded side-by-side plan comparisons by clinicians.ResultsWith equivalent Planning Target Volume (PTV) V95%, D2%, D98%, and dose homogeneity autoVMAT was overall superior for rectum with median differences of 3.4 Gy (p < 0.001) in Dmean, 4.0% (p < 0.001) in V60Gy, and 1.5% (p = 0.001) in V75Gy, and for bladder Dmean (0.9 Gy, p < 0.001). Also the clinicians' plan comparisons pointed at an overall preference for autoVMAT. Advantages of autoVMAT were highly treatment center- and patient-specific with overall ranges for differences in rectum Dmean and V60Gy of [−4,12] Gy and [−2,15]%, respectively.ConclusionObserved advantages of autoplanning were clinically relevant and larger than reported in the literature. The latter is likely related to the multi-criterial nature of the applied autoplanning algorithm, with for each center a dedicated configuration that aims at plan improvements relative to its (clinical) training plans. Large variations among patients in differences between manVMAT and autoVMAT point at inconsistencies in manual planning.



https://ift.tt/2NdlN5l