373 - Clinical Commissioning and Validation of a Same-Day, MRI-Only Simulation Workflow for Adaptive Linac-Based Non-Coplanar VMAT for SRS/SRT

10:15am - 10:25am ET

Room 107

Presenter(s)

Marvin Kinz, MS - Mass General Brigham, Harvard Medical School, Boston, MA

M. Kinz¹, D. C. Miller¹, T. Ciavattone¹, E. Kaza¹, S. Friesen², V. Nappady Joy³, C. L. Bullens¹, M. Czerminska¹, J. Hesser⁴, A. Sudhyadhom¹, C. V. Guthier², J. S. Bredfeldt¹, and K. Singhrao¹; ¹Department of Radiation Oncology, Brigham and Women’s Hospital, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA, ²Department of Radiation Oncology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA, ³Siemens Healthineers, Forchcheim, Bayern, Germany, ⁴Mannheim Institute for Intelligent Systems in Medicine, Medical Faculty Mannheim; Interdisciplinary Center for Scientific Computing; Central Institute for Computer Engineering; CZS Heidelberg Center for Model Based Ai; Heidelberg University, Heidelberg, Germany

Purpose/Objective(s):

Delays between simulation and stereotactic radiosurgery/-therapy (SRS/SRT) for brain metastases allow for tumor growth and soft tissue changes that require larger target margins and increase risk of adverse events. To mitigate this, we commissioned a same-day workflow combining MRI-only simulation with adaptive non-coplanar volumetric modulated arc therapy (VMAT). We report on workflow design, dosimetric robustness, and end-to-end spatial validation.

Materials/Methods:

The workflow commences with the acquisition of a 3T MRI simulation in the treatment position on delivery day. Target definition is completed on a post-contrast T1-weighted MPRAGE MRI, eliminating multimodality MRI/CT registration uncertainties. For safety during rapid planning, a deep-learning auto-segmentation tool provides a quality assurance (QA) check for target detection. Dose calculation uses an FDA-approved deep-learning synthetic CT (sCT) and a digital model of the MRI-invisible fixation device. A preliminary treatment plan based on a bulk-density overwritten diagnostic MRI is adapted to treatment-day anatomy and re-optimized for VMAT.

Commissioning included retrospective sCT vs. CT dosimetric comparisons and AI sensitivity analyses. Furthermore, end-to-end geometric validation utilized an anthropomorphic phantom and a 6D stereoscopic x-ray image-guidance system to verify targeting accuracy from MRI acquisition through final delivery.

Results:

Dose calculations using sCT and standard CT plans showed agreement with a PTV D95% dose difference of -0.04 ± 0.72% and ?-indices of 99.59 ± 0.36% for 1% local dose deviation within 1 mm distance to agreement. The AI QA tool detected lesions with a sensitivity of 90%, providing a safety net for rapid contouring. End-to-end phantom testing with the 6D image-guidance system confirmed spatial targeting fidelity, demonstrating sub-millimeter and sub-degree agreement between sCT-based and standard CT-based alignments.

Conclusion:

Integrating MRI-only simulation, sCT dose calculation, and AI-driven QA yields a robust same-day intracranial SRS/SRT workflow. The same-day timeline and MRI-only planning eliminate multimodality registration errors and post-simulation tumor dynamics. Additionally, 6D image-guided alignment provides sub-millimeter and sub-degree precision, effectively mitigating intrafraction motion during non-coplanar delivery. Minimizing these combined spatial uncertainties provides the rationale to investigate further margin reduction beyond the standard 1 mm PTV margin supported by our geometric agreement. A prospective clinical trial (NCT07132190) is currently underway to evaluate whether this workflow allows for the safe omission of PTV margins, which would lead to a potential volume reduction of 38 ± 8% compared to 1 mm margins and improve healthy tissue sparing.