Predictive quality assurance of a linear accelerator based on the machine performance check application using...

Area of Science:

• Medical Physics
• Radiation Oncology
• Quality Assurance

Background:

• Linear accelerators (linacs) require rigorous quality assurance (QA) to ensure optimal performance and patient safety.
• Current QA methods may not always predict potential issues proactively, leading to unexpected downtime.

Purpose of the Study:

• To develop and demonstrate the feasibility of a predictive QA system for linacs.
• To enable proactive preventative maintenance by forecasting linac performance based on Machine Performance Check (MPC) data.

Main Methods:

• Utilized statistical process control and autoregressive integrated moving average (ARIMA) modeling on daily MPC data.
• Trained the model on 85% of data and tested predictions (one-step and six-step ahead) on the remaining 15%.
• Evaluated model accuracy using root-mean-square error, absolute error, and average accuracy rate.

Main Results:

• The predictive model demonstrated high accuracy, with average errors less than 0.05 for all parameters.
• The system achieved an average accuracy rate above 85.00% in identifying normal and warning stages of linac performance.

Conclusions:

• A predictive QA system based on MPC data is feasible and effective.
• Implementing this system can facilitate preventative maintenance, enhance linac performance, and minimize unscheduled downtime.