Machine-Specific Statistical Process Control (SPC) Thresholds for MR-Linac Daily QA Trending
Abstract
Purpose
Daily QA tolerances are often implemented as fixed pass/fail thresholds, yet MR-Linac performance can be facility-specific and may evolve over time. This study used statistical process control (SPC) to establish global, machine-specific 2σ warning and 3σ action levels for Daily QA3 metrics to support longitudinal trending and early identification of baseline drift.
Methods
Daily QA3 records from a clinical MR-Linac were retrospectively reviewed (n=567, 2023-03-08 to 2026-01-26). Baseline-referenced (“relative”) metrics included output (%), beam-shape constancy (%), and axial/transverse symmetry (%). For each metric, Individuals–Moving Range (I–MR) SPC was applied across the full dataset. The center line (CL) was defined as the overall mean. Following standard SPC convention, 2σ warning and 3σ action limits were calculated using I–MR constants as CL ± 1.77·MR̄ and CL ± 2.66·MR̄, respectively, where MR̄ is the mean moving range.
Results
Global SPC produced a concise set of machine-specific thresholds for daily trending. For output, the CL was 0.543%, with 2σ limits of 0.421–0.666% and 3σ limits of 0.359–0.728%. For beam-shape constancy, the CL was 0.335%, with 2σ limits of 0.239–0.431% and 3σ limits of 0.191–0.479%. For axial symmetry, the CL was 0.098%, with 2σ limits of −0.131–0.328% and 3σ limits of −0.247–0.444%. For transverse symmetry, the CL was −0.018%, with 2σ limits of −0.367–0.331% and 3σ limits of −0.543–0.507%. Together, these 2σ/3σ bands provide a standardized escalation framework (warning vs action) for identifying deviations from typical daily variability and for recognizing non-random behavior consistent with baseline drift during multi-year monitoring.
Conclusion
Global I–MR SPC provides practical, machine-specific 2σ warning and 3σ action levels for MR-Linac Daily QA, complementing tolerance-based QA with standardized trending and earlier identification of performance change. The approach is easy to implement and can guide institutional thresholds and periodic baseline review.