Calibration-Based Machine Log Analysis for Automated Quality Assurance In Proton Beam Therapy
Abstract
Purpose
Traditional patient-specific QA in proton therapy requires detector-based measurements that consume beam time and delay workflow. This study presents a calibration-based machine log analysis system enabling automated, measurement-free verification of delivered dose distributions through coordinate transformation and treatment planning system integration.
Methods
A coordinate calibration framework was established by analyzing >21,000 spot positions across four treatment rooms. Linear regression derived transformation matrices (R² > 0.99) converting machine log coordinates to treatment planning coordinates. An automated Python pipeline extracts spot positions and monitor units from delivery logs, applies room-specific calibrations, and generates "measured plans" via RayStation API for dose recalculation. Dosimetric accuracy was evaluated by comparing planned versus log-reconstructed dose-volume histograms across multiple anatomical sites.
Results
Excellent coordinate transformation accuracy was achieved (X-direction R² = 0.998, Y-direction R² = 0.999). Automated processing successfully reconstructed delivered doses for lung, prostate, and breast cases without manual intervention. Target dose metrics demonstrated sub-percent agreement: D99 < 1%, D95 < 0.5%, mean dose < 0.1%. Critical organ doses agreed within 2% for heart, spinal cord, and rectum. Processing time averaged 5 minutes per patient without requiring additional beam time.
Conclusion
Calibrated machine log analysis provides accurate, efficient quality assurance without detector setup or beam time requirements. Multi-site validation demonstrates clinical feasibility for same-day verification of all treated fractions. This approach represents a paradigm shift toward fully automated, log-based PSQA complementing or replacing conventional measurement-based methods.