Functional Hypoxia Imaging–Based Joint Prediction of Efficacy and Toxicity In Proton Flash Radiotherapy
Abstract
Purpose
To establish an FMISO-PET–guided framework to estimate tumor control probability (TCP) and normal tissue complication probability (NTCP) in proton FLASH radiotherapy for efficacy prediction and clinical decision support.
Methods
An oxygen-depletion–based model was developed to calculate voxel-level FLASH sparing effectiveness (FSE) under varying oxygen partial pressures (pO2) and dose rates. FMISO-PET uptake was converted to voxel-wise pO2 using an SUV–pO2 mapping model. The FLASH physical dose (DFLASH) was converted to the equivalent dose at conventional dose rate (Deq) via FSE, followed by EQD2 conversion. Generalized equivalent uniform dose (gEUD) was used to integrate biological dose distributions for targets and organs at risk (OARs) for TCP and NTCP assessment. The framework was applied to NSCLC patients to predict TCP and symptomatic radiation pneumonitis risk (CTCAE grade≥2).
Results
(1) FSE showed strong oxygen dependence: >1.7 in well-oxygenated normal tissues (pO2≈40 mmHg) and ≈1.0 in hypoxic tumor tissues.(2) Under FLASH conditions (>40 Gy/s), modeling predicted an ~90% relative reduction in NTCP compared with conventional dose rate, with an ~3% relative decrease in TCP.(3) In two NSCLC cases, TCP decreased by 1–21% under FLASH. Ipsilateral lung NTCP decreased by ~83% in the peripheral case and ~88% in the central case; heart NTCP decreased by ~87% in the central case. Greater sparing was observed for central lesions due to larger normal tissue exposure.
Conclusion
This method quantifies the FLASH effect and integrates it with clinical prediction models, providing a feasible framework for biological effect assessment and personalized optimization of proton FLASH radiotherapy.