Related Experiment Video
Updated: Mar 31, 2026
08:34
Proton Therapy Delivery and Its Clinical Application in Select Solid Tumor Malignancies
Published on: February 6, 2019
21.3K
Enabling time-aware treatment plan evaluation for clinical proton pencil beam scanning systems
Arturs Meijers1, Marvin Nick Reimold1, Pietro Pisciotta2
1Center for Proton Therapy, Paul Scherrer Institute, Forschungsstrasse 111, Villigen, Switzerland.
Summary
This study introduces a machine-learning framework to predict proton pencil beam scanning (PBS) delivery timing, enabling advanced cancer therapies like FLASH. The model accurately predicts delivery times, paving the way for optimized treatment efficiency.
Area of Science:
- Medical Physics
- Computational Biology
- Radiotherapy Technology
Background:
- Clinical Treatment Planning Systems (TPS) for proton pencil beam scanning (PBS) lack temporal modeling capabilities, hindering advanced applications such as FLASH therapy and 4D dose calculations.
- Accurate temporal modeling is crucial for emerging radiotherapy techniques that rely on precise timing of radiation delivery.
Purpose of the Study:
- To develop and validate a machine-learning framework capable of predicting machine-specific proton PBS delivery timing using standard DICOM-RT plan data.
- To enable time-aware treatment plan evaluation for optimizing radiotherapy efficiency and facilitating next-generation treatment modalities.
Main Methods:
- A component-based model utilizing Random Forest regressors was developed to predict individual delivery time components: spot delivery, spot transition, and energy layer switching.
- The framework was trained on machine log files and validated on two distinct proton therapy systems (IBA ProteusPlus and Varian ProBeam), incorporating system-specific pre-processing.
- Machine-specific pre-processing was implemented to handle proprietary logic, such as spot reordering, ensuring adaptability across different systems.
Main Results:
- The machine-learning models demonstrated high accuracy in predicting spot delivery (R² > 0.98) and spot transition (R² > 0.95) times on both validated systems.
- Energy layer switching time prediction introduced the primary source of error, resulting in a slight underestimation of total treatment field time (approximately 3-5%).
- Gamma analysis comparing predicted dose rate maps with log-file-based maps showed excellent agreement, with pass rates exceeding 97% (0.5%/2mm criteria).
Conclusions:
- The developed framework is robust and adaptable for predicting proton pencil beam scanning delivery timing.
- Enabling time-aware plan evaluation is foundational for optimizing treatment efficiency in proton therapy.
- This predictive capability supports the development and implementation of advanced, dose-rate-dependent radiotherapy modalities.
