A collision prediction framework for noncoplanar radiotherapy planning and delivery

Area of Science:

• Medical Physics
• Radiation Oncology
• Computational Geometry

Background:

• Noncoplanar radiotherapy offers dosimetric advantages but faces clinical implementation challenges.
• A critical barrier is the lack of integrated collision prediction tools in clinical workflows.

Purpose of the Study:

• To investigate the feasibility of developing a collision prediction system (CPS) for noncoplanar radiotherapy.
• To create a CPS suitable for seamless integration into clinical practice.

Main Methods:

• Developed a CPS using a geometric model of the linear accelerator (Linac) and patient morphology data from CT scans and 3-D vision cameras.
• Constructed the geometric model using physical dimensions of Linac components (couch, gantry, imaging devices).
• Predicted collisions using vector dot products between Linac components and patient morphology, verified with phantom experiments.

Main Results:

• Achieved high sensitivity (0.95) and specificity (1.00) in 111 collision test cases.
• Predicted collision locations with conservative margins; average differences were 2.3 cm for couch movements and 3.8° for gantry rotation.
• Demonstrated that predicted differences can accommodate interfractional patient positioning variations.

Conclusions:

• Established the feasibility of developing a CPS for noncoplanar radiotherapy using geometric models and vector algebra.
• Outlined a framework for the potential clinical implementation of this collision prediction system.