A robust morphological algorithm for automatic radiation field extraction and correlation of portal images

Area of Science:

• Medical physics
• Image processing
• Radiation oncology

Background:

• Accurate radiation field boundary extraction is crucial for portal image analysis and treatment verification.
• Traditional edge detection methods are noise-sensitive and computationally intensive.
• Existing techniques struggle with variations in portal image types and quality.

Purpose of the Study:

• To develop a robust and efficient portal-image segmentation algorithm for accurate radiation field boundary extraction.
• To improve the accuracy of portal-simulator image correlation for treatment verification.
• To overcome the limitations of noise sensitivity and computational cost in current methods.

Main Methods:

• A novel algorithm based on morphological edge detection and segmentation techniques was developed.
• The algorithm utilizes two predefined, non-sensitive criteria to automatically search for optimal threshold values.
• A two-stage searching procedure accommodates variations in single and double exposure portal images from different therapy machines.

Main Results:

• The morphological edge detector demonstrated comparable results to optimal edge detectors.
• Portal-simulator image correlation using radiation field mask inertia moments proved more accurate than using contour inertia moments.
• The mask method showed reduced sensitivity to field shape variations and distortions.

Conclusions:

• The presented morphological segmentation algorithm offers an accurate and efficient solution for radiation field boundary extraction.
• The mask-based correlation method enhances the accuracy and simplifies treatment verification processes.
• This approach has the potential to substantially improve the reliability of radiation therapy verifications.