Related Experiment Video
Updated: May 12, 2026
07:31
Characterization of Recombination Effects in a Liquid Ionization Chamber Used for the Dosimetry of a Radiosurgical Accelerator
Published on: May 9, 2014
11.9K
Radiotherapy class-solution to correct an energy-dependent optically stimulated luminescence film dosimeter
Marco Caprioli1, Arnaud Colijn2, Laurence Delombaerde3
1Department of Oncology, KU Leuven, Leuven, Belgium.
Medical Physics
|June 4, 2025
Summary
This study developed an automated method using clustering to correct optically stimulated luminescence (OSL) film dosimetry for radiotherapy, improving dose accuracy in patient-specific quality assurance (PSQA). The new approach outperformed manual classifications.
Area of Science:
- Radiotherapy dosimetry
- Medical physics
- Radiation oncology
Background:
- Patient-Specific Quality Assurance (PSQA) in radiotherapy requires high-resolution dosimetry for intensity-modulated radiotherapy (IMRT) and volumetric-modulated arc therapy (VMAT).
- Optically stimulated luminescence (OSL) film dosimeters offer submillimeter resolution but have energy-dependent responses requiring correction.
- Previous corrections were class-specific, assuming uniform OSL energy response within treatment categories.
Purpose of the Study:
- To explore class-specific corrections for OSL film dosimetry using a comprehensive radiotherapy treatment dataset.
- To develop new treatment classes based on quantitative parameter similarity, eliminating the need for subjective user-based classifications.
- To objectively assign treatments to classes using K-means clustering and compare with manual and random assignments.
Main Methods:
- A dataset of 101 IMRT/VMAT plans from Varian linacs was analyzed.
- K-means clustering utilized twelve quantitative parameters, reduced to principal components, to form objective treatment classes.
- OSL film (BaFBr:Eu2+) was calibrated and used for dose measurements in a phantom; class-specific corrections were derived and dosimetric performance evaluated.
Main Results:
- K-means clustering identified eight distinct treatment classes, explaining 75% of data variance.
- Automated clustering showed no significant similarity to manual classifications (ARI < 0.01), highlighting interoperator variability in manual methods.
- Class-specific corrections using automated clustering significantly improved dosimetric accuracy, reducing mean dose differences to -0.2% ± 2.0% within D20% and D50%, with 88% and 74% of treatments below 3% difference, respectively.
Conclusions:
- Eight objective treatment classes derived from clustering effectively corrected the energy-dependent response of OSL films for PSQA.
- Automated, quantitative classification methods for OSL dosimetry yielded superior dosimetric results compared to subjective, qualitative manual classifications.
