Collision prediction for intracranial stereotactic radiosurgery planning: An easy-to-implement analytical solution.

Area of Science:

• Medical Physics
• Radiation Oncology
• Machine Learning in Healthcare

Background:

• Gantry collisions pose a significant risk in linac-based stereotactic radiosurgery (SRS).
• Lack of collision screening can lead to compromised treatment planning, delays, and incomplete therapies.
• Accurate prediction of collisions is crucial for safe and efficient SRS delivery.

Purpose of the Study:

• To develop and validate a software tool for predicting gantry collisions in linac-based SRS.
• To provide a solution that enhances safety and efficiency in SRS planning.
• To address the limitations of current collision screening methods.

Main Methods:

• Identified three collision types: gantry-couch mount, gantry-couch, and gantry-patient.
• Developed trigonometric formulas to calculate collision distances relative to the gantry rotation axis.
• Created a software code using machine measurements, patient dimensions, and collimator information to predict collision-free angles and identify collision sectors.

Main Results:

• The software accurately predicted collision absence in 19 configurations.
• For the remaining 61 configurations, the mean absolute error in predicted gantry collision angle was 0.86 degrees.
• The tool demonstrated high accuracy in predicting collisions on a TrueBEAM machine with BrainLab accessories.

Conclusions:

• The developed software accurately predicts collisions in linac-based intracranial SRS.
• The tool is easily implementable in any radiotherapy facility.
• This software enhances the safety and efficiency of SRS planning and delivery.