3147 - From Three-Dimensional QA Measurement to Clinically Relevant Margin Reduction: Projection-Based Flexmap Correction Using 3D-Starshot

Purpose/Objective(s): Submillimeter geometric accuracy is essential for intracranial stereotactic radiosurgery (SRS), where small geometric uncertainties translate into clinically meaningful dose variations. Conventional Winston–Lutz testing provides limited three-dimensional (3D) geometric characterization of linear accelerator and does not directly inform margin selection. We previously implemented a 3D starshot (3DSS) technique enabling volumetric evaluation of radiation isocenter behavior and developed a margin formulation from the 3DSS. However, translation into geometric correction and its implication for clinical margin determination remain unclear. This study aimed to develop a method that directly incorporates 3DSS-derived irradiation position errors into Flexmap correction and evaluate its impact on clinically relevant geometric uncertainty and margin requirements.

Materials/Methods: A 3D rigid-body transformation algorithm was developed to map irradiation position errors measured in treatment coordinates onto the projection geometry of the megavoltage portal imager, enabling projection-consistent correction of the Flexmap sinogram. Validation measurements were performed using a CT gel dosimeter (dGEL, Triangle Products) on a VersaHD linear accelerator (Elekta) with XVI imaging. The 3D offset between radiation and CBCT imaging isocenters was quantified using 3DSS measurements before and after Flexmap correction. PTV margins were estimated using measured 3D irradiation position offsets combined with representative residual setup errors after image guidance for cranial SRS treatments. A representative intracranial SRS planning scenario (10 brain metastases, total GTV volume 0.418 cm³) was replanned using calculated margins before and after correction, and normal brain V12Gy and corresponding model-based normal tissue complication probability (NTCP) was compared.

Results: Before correction, the 3D radiation–imaging isocenter offset was 0.75 mm. After projection-based Flexmap correction, this was reduced to 0.27 mm (64% reduction). Maximum beam deviation relative to the imaging isocenter decreased from 1.22 mm to 0.68 mm. Estimated PTV margins were reduced, particularly in the superior–inferior direction (2.7 mm to 2.1 mm). In the representative multi-target SRS plan, normal brain V12Gy decreased from 23.4 cm³ to 21.2 cm³, corresponding to a reduction in model-estimated NTCP from 10.1% to 9.0%.

Conclusion: Direct incorporation of 3D irradiation position errors derived from 3DSS measurements into Flexmap correction enabled measurable reduction of radiation–imaging isocenter discrepancies. This approach provides a practical pathway to translate advanced geometric QA measurements into clinically meaningful uncertainty reduction and margin optimization in high-precision SRS. The proposed framework may support data-driven margin strategies in multi-target SRS workflows requiring submillimeter geometric accuracy.