Related Experiment Video
Updated: Feb 24, 2026
10:23
Author Spotlight: A Machine-Vision Approach to Transmission Electron Microscopy Workflows, Results Analysis and Data Management
Published on: June 23, 2023
3.6K
Machine Learning-Based Beam Delivery Time Model for Mevion S250i With Hyperscan Technology.
Giorgio Cartechini1, Francesco Giuseppe Cordoni2, Mirko Unipan1
1Maastricht University Medical Centre+, Department of Radiation Oncology (Maastro), GROW School for Oncology and Reproduction, Doctor Tanslaan 12, Maastricht 6229 ET, The Netherlands.
International Journal of Particle Therapy
|February 23, 2026
Summary
We developed the first machine learning model to predict beam delivery time on Mevion systems, crucial for proton therapy efficiency. This model accurately captures system dynamics and aids in advanced treatment planning.
Area of Science:
- Medical Physics
- Machine Learning in Healthcare
- Proton Therapy
Background:
- Accurate beam delivery time (BDT) prediction is vital for proton therapy operational efficiency, 4D dose calculations, and advanced techniques.
- Currently, no machine-specific BDT model exists for Mevion systems, limiting optimization potential.
Purpose of the Study:
- To develop and validate the first machine learning-based BDT model for the Mevion S250i Hyperscan system.
- To analyze feature contributions to BDT prediction using explainable AI.
Main Methods:
- Developed a Random Forest model using institutional machine log files (11 patients, 1120 files).
- Extracted features including spot position, energy layer changes, Adaptive Aperture (AA) movements, and spot charge; inter-pulse time was the target variable.
- Utilized SHAP analysis to quantify feature contributions and validated the model on clinical applications.
Main Results:
- The model achieved mean absolute errors from 0.9 ms (short intervals) to 211 ms (long delays).
- AA movements were the dominant predictor for longer delays (>50 ms), while spot position and pulse charge dominated short intervals.
- The model accurately predicted cumulative delivery times (-1.6% deviation) and maintained dosimetric metrics within delivery variability.
Conclusions:
- Presents the first ML-based BDT model for Mevion S250i, accurately reflecting machine-specific temporal dynamics.
- SHAP analysis provided insights into operational characteristics, highlighting AA, energy layer, and spot position contributions.
- The model shows strong predictive performance for applications like interplay assessment and 4D dose calculation.
