A Swiss cheese error detection method for real-time EPID-based quality assurance and error prevention

Area of Science:

• Medical Physics
• Radiation Oncology
• Image Analysis

Background:

• Quality assurance (QA) in external beam radiation therapy is crucial for patient safety.
• Volumetric-modulated arc therapy (VMAT) requires precise dose delivery and robust QA methods.
• Electronic portal imaging devices (EPIDs) offer opportunities for during-treatment monitoring.

Purpose of the Study:

• To develop a robust and efficient process for detecting relevant dose errors (≥5%) in external beam radiation therapy.
• To directly indicate the origin of detected dose errors.
• To implement and evaluate this process within an electronic portal imaging device (EPID)-based angle-resolved VMAT QA program for real-time monitoring.

Main Methods:

• A Swiss cheese error detection (SCED) method was developed for cine EPID-based during-treatment QA.
• The SCED method compares reference EPID images with images acquired at 2° gantry angle intervals during VMAT delivery.
• It employs a sequence of checks including aperture, output normalization, image alignment, and pixel intensity (gamma analysis and deviation checks).

Main Results:

• The SCED method detected 95.1% of relevant dose errors within 2° and 100% within 14° of gantry rotation.
• Even with clinically equivalent plan modifications, 89.1% of errors were detected within 2°.
• The SCED method demonstrated a higher detection rate (94.0%-95.8%) compared to standard gamma analysis (82.8%-89.8%) across various noise levels.

Conclusions:

• An EPID-frame-based error detection process (SCED) for VMAT was successfully designed and simulated.
• The SCED method is robust to noise variations and has the potential to detect a majority of relevant dose errors.
• Compared to standard gamma analysis, SCED offers higher detection rates, earlier error identification, and indicates the error source, enhancing VMAT QA.