Core-based portal image registration for automatic radiotherapy treatment verification

Area of Science:

• Medical Physics
• Radiation Oncology
• Computer Vision

Background:

• Portal imaging is crucial for radiation therapy quality assurance, monitoring setup reproducibility.
• Traditional portal imaging methods have limitations in detecting patient setup errors, potentially compromising treatment accuracy.
• Advancements in 3D treatment planning necessitate improved error detection methods.

Purpose of the Study:

• To evaluate automatic image registration using core-based image analysis for on-line screening of electronic portal imaging device (EPID) images.
• To improve the accuracy and efficiency of detecting patient setup errors in radiation therapy.
• To assess the suitability of core-based image analysis as a tool for real-time quality assurance.

Main Methods:

• Developed a robust method for automatic portal image registration based on core analysis.
• Utilized digitally reconstructed portal radiographs (DRPRs) with known positioning errors for method validation.
• Assessed the accuracy and precision of the registration method using simulated patient data.

Main Results:

• Automatic core-based analysis achieved high accuracy in detecting positioning errors on DRPRs.
• Translation errors were reported within 1 mm of actual values (mean absolute error 0.3 mm).
• Rotation errors were reported within 0.6 degrees of actual values (mean absolute error 0.18 degrees).

Conclusions:

• Automatic core-based registration shows promise as an on-line screening tool for radiation therapy.
• The method is suitable for detecting and quantifying patient setup errors with high accuracy.
• This technique can enhance the quality assurance of radiation therapy delivery.