First experience of autonomous, un-supervised treatment planning integrated in adaptive MR-guided radiotherapy and delivered to a patient with prostate cancer

Abstract

Background and purpose: Currently clinical radiotherapy (RT) planning consists of a multi-step routine procedure requiring human interaction which often results in a time-consuming and fragmented process with limited robustness. Here we present an autonomous un-supervised treatment planning approach, integrated as basis for online adaptive magnetic resonance guided RT (MRgRT), which was delivered to a prostate cancer patient as a first-in-human experience. Materials and methods: For an intermediate risk prostate cancer patient OARs and targets were automatically segmented using a deep learning-based software and logical volume operators. A baseline plan for the 1.5 T MR-Linac (20x3 Gy) was automatically generated using particle swarm optimization (PSO) without any human interaction. Plan quality was evaluated by predefined dosimetric criteria including appropriate tolerances. Online plan adaptation during clinical MRgRT was defined as first checkpoint for human interaction. Results: OARs and targets were successfully segmented (3 min) and used for automatic plan optimization (300 min). The autonomous generated plan satisfied 12/16 dosimetric criteria, however all remained within tolerance. Without prior human validation, this baseline plan was successfully used during online MRgRT plan adaptation, where 14/16 criteria were fulfilled. As postulated, human interaction was necessary only during plan adaptation. Conclusion: Autonomous, un-supervised data preparation and treatment planning was first-in-human shown to be feasible for adaptive MRgRT and successfully applied. The checkpoint for first human intervention was at the time of online MRgRT plan adaptation. Autonomous planning reduced the time delay between simulation and start of RT and may thus allow for real-time MRgRT applications in the future.

Automated Planning for Prostate Stereotactic Body Radiation Therapy on the 1.5 T MR-Linac

“…The main difference between that study and the present study, other than Bijman and colleagues’ use of Erasmus-iCycle instead of mCycle, was the increased dose complexity that is required for pelvic treatments when passing from standard fractionation, where OAR sparing typically translates into mean dose reduction, to severe hypofractionation, where plan approval results from an optimal compromise between avoidance of the hotspot and dose coverage to the several overlaps of the target with the critical OARs (ie, rectum, bladder, urethra, and penile bulb). In another study on 1 patient with PC treated by 60 Gy in 20 fractions, 24 an offline autoplanning solution for the 1.5 T MR-linac was tested, where an in-house made optimizer generated the fluence map, which was then passed as input to the standard segmenter of Monaco. The offline plan was then used as a reference plan in the daily ATS workflow; however, the usual optimization tools of online Monaco were adopted to generate any adaptive plan.…”

Clinical evaluation of autonomous, unsupervised planning integrated in MR-guided radiotherapy for prostate cancer

“…The autonomous workflow consisted of a deep learning-based annotation software (ARTplan 1.7.1, TheraPanacea, Paris, France) [2] and our in-house developed automatic PSO planning tool [21,22]. The planning CT was automatically sent to the annotation software and transferred afterwards to the treatment planning system (TPS) (MonacoÓ 5.40, Elekta AB, Stockholm, Sweden).…”

“…Xia et al recently proposed full automation of the whole treatment preparation chain for rectal cancer [20]. We demonstrated clinical and technical feasibility of autonomous, un-supervised treatment plan-ning for MR-guided RT (MRgRT) of prostate cancer in a recent firstin-human application [21]. Auto-contouring was performed with a deep learning model, whereas for auto-planning a particle swarm optimization (PSO) was utilized [22].…”

“…Auto-contouring was performed with a deep learning model, whereas for auto-planning a particle swarm optimization (PSO) was utilized [22]. Such autonomous workflow can create baseline plans for subsequent online adaptive MRgRT where daily plan adaptation serves as check-point for human interaction to ensure quality and approval by a radiation oncologist and medical physicist [21]. Even though proof of conceptual and clinical feasibility is a milestone in the implementation of fully autonomous workflows, a broader evaluation of the clinical usability, integrity and quality of the individual workflow steps is a major prerequisite for scalability of autonomous RT plan generation.…”

Stellenwert des 1,5-T-MR-Linearbeschleunigers für die primäre Therapie des Prostatakarzinoms

“…Diesbezüglich sind zeitnah Verkürzungen der täglichen Therapiedauer pro Patient zu erwarten [ 36 ]. In einer „Proof-of-principle“-Studie konnte bereits eine vollautomatisierte Konturierung und Bestrahlungsplanung durchgeführt und diesen Plan online-adaptiv genutzt werden [ 37 ].…”