3D Modeling–Based Collision Analysis for Verification of Non-Coplanar Stereotactic Radiosurgery Treatments
Abstract
Purpose
Automation and geometric verification tools are increasingly important in radiotherapy for non-coplanar stereotactic treatments. This study implements a patient-specific 3D modeling–based method to generate gantry–couch collision maps from clinical DICOM data for stereotactic radiosurgery (SRS) treatments.
Methods
A patient-specific collision analysis workflow was implemented using a Python-based (v3.13) graphical user interface that guides the user through DICOM loading, patient model generation, three-dimensional scene construction, and collision and clearance analysis. The patient external contour and clinical isocenter were automatically extracted using 3D Slicer (v5.8.1). The treatment delivery geometry corresponding to a Varian TrueBeam STx linear accelerator was imported into a three-dimensional environment using Blender (v4.2.3), based on a previously reported 3D collision framework extended to incorporate the Brainlab Cranial 4Pi immobilization system, cone holder, and conical collimator. Collision analysis was performed by systematically sweeping the gantry and couch angles using bounding volume hierarchy methods, generating two-dimensional gantry–couch collision maps with overlaid planned non-coplanar arc trajectories. A confusion matrix was developed for seven couch angles to verify collision zones.
Results
Patient-specific gantry–couch collision maps were successfully generated from clinical plans. For two trigeminal neuralgia cases with right- and left-sided isocenter offsets evaluated at seven couch angles, similar global collision patterns were observed with localized differences, and the incorporation of a conservative safety margin at the couch and body eliminated false-negative collision predictions in the confusion matrix analysis. The superposition of planned non-coplanar arc trajectories and minimum distance evaluation enabled identification of angular regions suitable for safe treatment delivery and configurations with reduced clearance.
Conclusion
A patient-specific 3D modeling framework was developed to generate collision maps and minimum clearance information from clinical DICOM data. The system may assist in identifying collision-free angular regions to support collision verification in non-coplanar SRS treatments