ESTRO 2021 Abstract Book

S136

ESTRO 2021

Conclusion This study demonstrates that CNNs can reliably detect relevant changes in realistically simulated PGI data and classify most of the underlying sources of treatment deviations. While validation on measured patient data is needed, our study highlights the potential of a reliable, automatic interpretation of PGI data, which is highly desired for broad clinical application and a prerequisite for the inclusion of PGI in an automated feedback loop for online adaptation. OC-0205 Thermoacoustic Range Verification During Delivery of a Clinical Plan to a Abdominal Imaging Phantom S. Patch 1,2 , C. Nguyen 1 , R. Labarbe 3 , G. Janssens 3 , J. Lambert 4 , M. Cohilis 5 , K. Souris 5 , S. Ono 6 , T. Lynch 7 1 UW-Milwaukee, Physics, Milwaukee, USA; 2 Acoustic Range Estimates, All, Milwaukee, USA; 3 Ion Beam Applications, Research Science, Louvain-La-Neuve, Belgium; 4 The Rutherford, Physics, Newcastle, United Kingdom; 5 UCLouvain, Medical Physics, Louvain-la-Neuve, Belgium; 6 Computerized Imaging Reference Systems, Radiation Therapy, Norfolk, USA; 7 Computerized Imaging Reference Systems, Ultrasound, Norfolk, USA Purpose or Objective The purpose of this phantom study is to demonstrate that thermoacoustic range verification could be performed clinically. Thermoacoustic emissions generated in an anatomical multimodality imaging phantom during delivery of a clinical plan are compared to simulated emissions to estimate range shifts compared to the treatment plan. Materials and Methods A single-field 12-layer proton pencil beam scanning (PBS) treatment plan prescribing 6 Gy/fraction was delivered to a triple modality (CT, MRI, and US) abdominal imaging phantom made of hydrogels (CIRS 057a). A superconducting synchrocyclotron delivered ~2 cGy/pulse at an average dose rate of approximately 20 Gy/sec, with 0.5% duty cycle . High-dose spots received approximately 40 cGy. Data was acquired by four acoustic receivers rigidly affixed to a linear ultrasound array. Acoustic receivers (transducer + amplifier) tuned to this application provided 15-25 dB amplification relative to 1 mV/Pa over 10- 100 kHz. Receivers 1-2 were located distal to the treatment volume, whereas 3-4 were lateral. Receivers’ room coordinates were computed relative to the ultrasound image plane after co-registration to the planning CT volume. For each prescribed beamlet, a MCsquare Monte Carlo simulation of energy density provided initial pressure from which thermoacoustic emissions were computed using k-Wave. Emissions from beams that stopped in soft tissue were bandlimited below 100 kHz~15 mm. To overcome the diffraction limit, range shifts were computed from time shifts between simulated and measured emissions. Results Shifts were small for high-dose beamlets that stopped in soft tissue. Signals acquired by channels 1-2 yielded shifts of -0.2±0.7 mm relative to Monte Carlo simulations for high dose spots (~40 cGy) in the second layer (see Fig. & Table). Additionally, for beam energy ≥125 MeV, thermoacoustic emissions qualitatively tracked lateral motion of pristine beams in a layered gelatin phantom, and time shifts induced by changing phantom layers were self-consistent within nanoseconds. Table 1. Time and range shifts between measured and simulated data compared to initial pressure per proton pulse and cumulative dose within the planned treatment spot.