AI cuts cancer radiation planning time by automating tumor targeting
Researchers automated the complex process of planning personalized radiation therapy for head and neck cancer, using AI to match doses to tumor biology. While the initial automation took longer than manual planning, the workflow eliminates human variability and could standardize care across institutions—a significant advantage for cancer centers managing high patient volumes.
Originaltitel: Automated class-solution planning for biologically guided radiotherapy: a comparison with manual planning in head and neck cancer
PURPOSE: To develop an automated class-solution treatment-planning workflow for biologically guided dose-painting based on combined FDG- and FMISO-PET in head and neck cancer (HNC), and to compare its performance with manual planning. MATERIAL AND METHODS: The workflow incorporating image-processing and treatment planning via a class-solution template was implemented in RayStation-10B-R and applied to patients imaged with FDG- and FMISO-PET/CT. The workflow converted FMISO- and FDG-PET uptake into oxygen partial pressure and clonogenic cell-density distributions, respectively. Accordingly, simultaneous integrated boost plans aiming at 95% tumour control probability (TCP) and using a dose-painting-by-contours approach for TV1, TV2, the GTV, and the hypoxic target volume (HTV), were created. For nine patients, automated and manual plans were compared using equivalent dose in 2-Gy fractions (EQD2)-based target metrics, organ-at-risk (OAR) doses, plan-complexity parameters, planning time, TCP and normal tissue complication probability (NTCP). RESULTS: in the TV1-TV2 and HTV. Manual planning required ∼ 1 h, whereas automated planning required ∼ 5 h with no user interaction. CONCLUSIONS: A scripting-based, biologically guided class-solution for dose-painting in HNC is feasible and achieves plan quality and radiobiological outcomes comparable to manual planning, providing a platform for standardised and adaptive radiotherapy workflows.