Simulation-Free, Single-Visit MR-Guided Adaptive Liver SABR: Workflow Develop and Initial Experience
Abstract
Purpose
To develop and evaluate a simulation-free, single-visit workflow for MRI-guided stereotactic ablative radiotherapy (SABR) for liver tumors using on-table imaging and online adaptive replanning, and to validate an MR-only dose calculation approach based on bulk electron density overrides (ED-BDO).
Methods
A single-institution phase II trial was designed to deliver liver SABR 30-40Gy in one visit without CT simulation. A pre-treatment reference plan was generated from the most recent diagnostic CT. For MR-only dose calculation, a four-class ED-BDO scheme was applied (body, air, lung, and adipose as needed). On the treatment day, Eovist contrast was administered and a breath-hold 3D MRI was acquired for anatomy visualization. Targets and organs at risk (OARs) were contoured, electron density assignments were updated, and an online adaptive plan was generated and delivered with respiratory gating. ED-BDO dosimetric accuracy was retrospectively evaluated in 17 single-fraction liver SABR plans (25–30 Gy) by comparing dose recalculations using CT-derived electron density (ED-CT) versus ED-BDO using DVH endpoints and 3D gamma analysis (1%/1 mm).
Results
ED‑BDO showed strong agreement with ED‑CT for both target and OAR metrics. For PTV dose endpoints, correlations were high (R² = 0.978–0.991), with sub‑Gy maximum absolute differences. PTV V100% differences averaged 0.804% (maximum 1.86%). OAR endpoints (chest wall, bowel, liver minus GTV, spinal cord) also demonstrated high concordance (R² = 0.978–1.000) with sub‑Gy differences. The mean 3D gamma pass rate (1%/1 mm) was 99.1% ± 1.3% (range 95.6%–100.0%). Workflow feasibility was assessed in two volunteer run-throughs, both completed within 45 minutes from vault entry to ready for beam-on.
Conclusion
Simulation-free, single-visit MR-guided adaptive liver SABR is feasible using on-table imaging and online replanning. A streamlined four-class ED-BDO approach (without an explicit bone class) maintains excellent agreement with CT-based dosimetry, supporting efficient MR-only adaptation while preserving clinically relevant target and OAR evaluation.