Synthetic CT artifacts in MR‐only brain radiotherapy: Clinical implementation experience
Hyeri Lee, Ahmet S. Ayan, Sasha Beyer, Dukagjin Blakaj, Dominic DiCostanzo, John C. Grecula, Joshua Palmer, Raju Raval, Raj Singh, Evan Thomas, Michael Weldon, Xiangyu Yang, Wesley Zoller, Nilendu GuptaAbstract
Background
Magnetic resonance imaging (MRI) provides superior soft tissue contrast compared with CT and is essential for delineation of targets and organs at risk in brain radiotherapy. Because CT is traditionally required for dose calculation in radiation therapy, synthetic CT (sCT) generation from MRI has enabled the development of MR‐only workflows. Prior studies have primarily focused on dosimetric equivalence between CT‐ and sCT‐based treatment plans, with limited emphasis on practical implementation challenges encountered during clinical use.
Purpose
This study aimed to share clinical experience implementing an MR‐only brain radiotherapy workflow, with a focus on practical challenges encountered.
Methods
Twenty‐eight patients underwent a dedicated MR simulation protocol for sCT generation in addition to standard‐of‐care planning CT simulation. MRI acquisition was performed using an immobilization mask with a dedicated head coil positioned over the mask. The sCT protocol consisted of four sequences: T1 VIBE Dixon, T2 SPACE, PETRA, and TOF‐MRA. Clinical treatment plans generated using the planning CT (pCT) for photon therapy were recalculated on the corresponding sCT datasets.
Results
Twenty‐eight patients were scanned, and 25 synthetic CT datasets were successfully generated. Nine sCT datasets required rigid re‐registration because of inter‐sequence patient motion. Findings were categorized into workflow‐related sCT challenges, including inter‐sequence motion, suboptimal coil placement, and external localization marker interference, and algorithm‐related imaging artifacts, including HU misclassification at tissue/bone/air interfaces and sCT degradation in post‐operative anatomy with surgical meshes.
Conclusion
Clinical implementation of a brain MR‐only workflow identified two major categories of challenge: workflow‐related sCT perturbations and algorithm‐related imaging artifacts. This distinction has important implications for mitigation strategies. Case‐specific sCT review, structured eligibility screening, and prospective site commissioning are essential for safe clinical deployment. Practical commissioning and per‐patient QA checklists are provided as supplementary materials.