Author: Kristina Bjegovic, Yong Chen, Gilberto Gonzalez, Yankun Lang, Lei Ren, Leshan Sun, Liangzhong Xiang, Yifei Xu 👨🔬
Affiliation: University of Maryland School of Medicine, University of California, Irvine, University of Oklahoma Health Sciences Center, University of Oklahoma Health Science Center 🌍
Purpose: Proton therapy is becoming an increasingly popular choice for cancer treatment due to its precision, reduced side effects, and effectiveness. While dosimetry is fundamental to the success of proton therapy, no in vivo dosimetry directly on patient in clinical practice. Recently, protoacoustic tomography (PAT) has shown great promise for in vivo dosimetry, however no quantitative dosimetry has been possible. This work aims to develop a complete workflow to use PAT for quantitative dosimetry for proton pencil beam delivery and validate it with experiments using human phantom.
Methods: The proposed workflow includes pressure map reconstruction using a time-reversal (TR) algorithm, limited-view optimization with a physics-informed neural network (PINN), and dose calibration to convert relative pressure maps into absolute dose maps. Validation was performed using both simulation and experimental data. A digital twin model was developed in the k-Wave MATLAB toolbox, incorporating transducer finite-element modeling, proton pulse duration modeling, and ultrasound system impulse response modeling. Simulation datasets generated from the digital twin model with pencil beam treatment plans were used for PINN training. Experimental signals were obtained by directing proton pencil beams into an adult-torso phantom and capturing the data with a 16×16 ultrasonic array integrated with a custom DAQ system.
Results: The proposed method effectively reconstructed single pencil beam 3D volumetric dose maps. Simulation results achieved a 100% pass rate for gamma index evaluation with 3%/3mm criteria, while experimental results showed a 93.4% pass rate under the same conditions. The reconstructions for proton beams with varying energies and pulse numbers aligned well with planning dose maps both qualitatively and quantitatively.
Conclusion: The study highlights the potential of PAT to precisely monitor individual pencil beam doses during proton therapy using radio-frequency signals from a planar ultrasound transducer array, offering significant value for intrafraction treatment adaptation.