Scanned particle-beam tracking with beam correction based on predictive volumetric imaging: A simulation study

Tracking irradiation to moving targets in spot-scanning particle therapy, which corrects the spot position and energy in real-time, may decrease treatment time and increase accuracy. However, because of the temporal performance of the system, clinical translation remains challenging. Processing time, including image acquisition, volumetric image synthesis, correction assessment, and system response, is required to control the actual treatment system. These processing delays cause millimeter-order discrepancies due to tumor motion. Predicting future states may compensate for this latency. However, research on predicting volumetric images required for energy correction assessment has not been reported.

This study aimed to investigate the dosimetric effectiveness of particle-beam tracking irradiation according to predictive volumetric imaging under various latency conditions.

Surrogate-driven volumetric image synthesis is combined with surrogate position prediction in the predictive volumetric imaging technique. A linear regression model in volumetric imaging that can derive internal deformation from surrogate displacement is established for each voxel from a four-dimensional computed tomography (4DCT) dataset in the modeling process. A volumetric image is predictively synthesized during the imaging process using the surrogate position predicted by a pretrained long short-term memory network. This predictively synthesized image enables the prospective assessment of beam parameter correction, including spot position and energy. In this study, 4DCT datasets and time-series trajectory data of the internal marker from three patients each with lung, liver, and pancreatic cancers were utilized for the dosimetric simulation. An intensity-modulated proton therapy plan was generated for each patient. Dosimetric simulations were conducted assuming the latencies of 133.3, 266.6, and 400.0 ms. Assessments included (1) tracking irradiation without latency as a benchmark, (2) tracking irradiation with latency but without prediction, and (3) tracking irradiation with latency and prediction. Further, dose-volume histograms and dose metrics of the clinical target volume (CTV) were compared.

Doses in tracking with prediction were comparable to those in the benchmark. Differences in D99%, D95%, and D5% of the CTV in the lungs between the treatment plan and tracking irradiation without prediction exceeded 5% at all latencies. Differences in D95% and D5% in tracking irradiation with prediction were less than 5% in most cases. Differences in D99%, D95%, and D5% in the liver and pancreas exceeded 5% at a latency of 400.0 ms without prediction but remained below 3% with prediction. Doses to organs at risk showed only minor deviations from the treatment plan in tracking irradiation.

The proposed tracking irradiation technique based on predictive volumetric imaging in spot-scanning particle therapy demonstrated tracking doses comparable to doses in the treatment plan across all latency conditions in the lung, liver, and pancreas. Further research and development of treatment devices and treatment planning protocols are warranted for the proposed tracking irradiation technique to become an effective motion management technique in terms of both dosimetric accuracy and treatment efficiency.