A workflow to select local tolerance limits by combining statistical process control and error curve model.

Area of Science:

• Medical Physics
• Radiation Oncology
• Quality Assurance

Background:

• Patient-specific quality assurance (QA) is complex and requires local tolerance limits.
• Universal limits may not suit all clinical situations.
• Medical physicists need methods to establish and validate local limits.

Purpose of the Study:

• Develop a comprehensive methodology for determining appropriate local tolerance limits in patient-specific QA.
• Provide a quantitative approach for evaluating error sensitivity of these limits.

Main Methods:

• Simulated multi-leaf collimator (MLC) positional errors in RapidArc plans.
• Extracted six QA metrics (GP10, GP50, µGI50, PTV95, PTV5, PTVmean) from error-free plans.
• Established local tolerance limits using statistical process control.
• Developed error curve models to assess sensitivity of QA metrics to MLC errors.
• Validated tolerance limits using binary classification performance metrics.

Main Results:

• Theoretical detection limits for individual QA metrics ranged from 1.93 mm (PTV95) to 3.52 mm (µGI50).
• Local tolerance limits for PTV95 detected systematic MLC errors >0.6 mm with 76.19% detection rate.
• Combined GP10 and PTV95 tolerance limits achieved an 80.16% detection rate for errors >0.6 mm.

Conclusions:

• The proposed workflow integrates establishment and validation of local tolerance limits for patient-specific QA.
• Offers a practical tool for setting clinically relevant limits.
• Provides a quantitative method for medical physicists to assess error sensitivity.