Experimental validation of a Fisher information–based predictive framework for dose and time optimization in PET-guided online adaptive proton therapy

<i>Objective.</i>The range determination uncertainty (σest) based on positron emission tomography (PET) imaging, which stems from the Poisson statistics of the detected signal, can be theoretically predicted using Fisher information. This study aims to experimentally validate a Fisher information-based predictive framework that optimizes the irradiation dose and measurement time required for reliable range verification in PET-guided online adaptive proton therapy.<i>Approach.</i>First, we defined a precision criterion of1.5σest<2mmfor reliable range verification. Then, using polyethylene, water, and a head and neck phantom, we determined the minimum measurement time-calculated in 2 s increments-required to satisfy this criterion at given irradiation doses (0.5 Gy and 0.1 Gy) based on Fisher information. For each condition, 5000 PET images were generated from the measurement datasets, and the maximum likelihood estimation method was independently applied to each to determine the standard deviation of the measured range (σmeas). Finally, the values ofσmeaswere compared with those ofσestto validate the predictive framework.<i>Main results.</i>The values ofσmeasandσestshowed consistent agreement (within approximately 0.5 mm), regardless of target properties, dose levels, and measurement times. Furthermore, the measured range uncertainty satisfied the pre-defined precision criterion of1.5σmeas<2mmunder almost all of the tested conditions.<i>Significance.</i>This study provides the first experimental validation of the Fisher information-based predictive framework for PET-based range verification. The findings offer a rationale for integrating this framework into PET-guided online adaptive proton therapy, which will potentially enable reliable range verification with the minimum pre-irradiation dose and measurement time.