ROC-based evaluation of tight gamma criteria for detecting MLC and jaw positioning errors in brain SRS using SRS MapCHECK

Precise patient-specific quality assurance (QA) is critical for brain stereotactic radiosurgery (SRS). This study evaluated tight gamma criteria using the SRS MapCHECK high-resolution dual-layer detector.

Seventeen volumetric modulated arc therapy (VMAT) plans delivered on an Elekta Versa HD were measured under error-free conditions and after intentional 0.5-mm multi-leaf collimator (MLC) or 1.0-mm Y-jaw shifts. Gamma passing rates (GPRs) were calculated using nine criteria at low-dose thresholds of 10%, 20%, and 30%. Pooled and error-type-stratified receiver operating characteristic (ROC) analyses were performed, and confidence intervals for the area under the curve (AUC) were estimated using patient-level cluster bootstrap resampling.

The conventional 3%/2 mm criterion showed limited pooled discrimination (AUC < 0.65). The 1%/0.5 mm criterion yielded the numerically highest pooled AUC of 0.830, whereas the 3%/0.5 mm criterion yielded an AUC of 0.796, with sensitivity and specificity of 0.765 at a cutoff of 88.8%. With the 3%/0.5 mm criterion, the AUCs were 0.872 for MLC aperture-changing errors, 0.581 for MLC translations, 0.871 for jaw aperture-changing errors, and 0.861 for jaw translations. MLC translations produced only minor dose-volumetric changes, whereas aperture-changing errors had larger effects on target coverage and normal-tissue dose.

For Elekta Versa HD brain SRS QA using the SRS MapCHECK system, the 3%/0.5 mm criterion may be considered for institution-specific use, particularly for aperture-changing errors, after local validation.