The Radiation Oncology Feed For radiation oncologists · physicists · RTTs · dosimetrists
Technical relevance brain-metastases SRS

Target-volume model predicted HyTEC-based SRS prescriptions before plan generation

A model using target volume and HyTEC limits estimated automated SRS prescription doses with approximately 1.3% mean absolute error.

RadOncToday · ·1 min read

KEY POINTS

  • Investigators used 3,361 marginless GTV targets from 1,495 clinically approved HyperArc plans in 1,178 patients, generating 30,249 target-specific isodose volumes for model development.
  • The framework links target volume with intermediate-dose fall-off and established HyTEC normal-brain thresholds, allowing a prescription dose to be estimated before generating an SRS plan for 1-, 3-, or 5-fraction treatment.
  • HyTEC-relevant adverse-effect dose levels included 12 and 14 Gy for single-fraction SRS, with biologically corresponding values of 19.6/23.1 Gy for 3 fractions and 24.4/28.8 Gy for 5 fractions.
  • Target volume predicted intermediate-dose volume extremely well across the training data, with power-law R² values of 0.98-0.99. At held-out relative isodose levels, median prediction residual was only 0.2%, although the 95% interval extended from −12.8% to +27.0%.
  • The resulting prescription-dose equations closely matched numerical solutions. Root-mean-square error was 0.28 Gy for single-fraction, 0.46 Gy for 3-fraction and 0.57 Gy for 5-fraction SRS, with a mean absolute percentage error of 1.3%.
  • The proposed use is essentially inverse planning before planning: rather than generating several candidate prescriptions and checking whether V12/V14-type limits are met afterward, clinicians could estimate a starting dose from target volume and a chosen adverse-effect level.
  • The model was derived from a single institution and one automated HyperArc planning strategy. It did not account for aggregate normal-brain dose from multiple closely spaced lesions, prior cranial RT or highly irregular target geometry, and held-out testing was internal rather than independent external validation.

CLINICAL TAKEAWAY

Automated SRS may be predictable enough that normal-brain risk constraints can inform prescription selection before optimization rather than only being checked afterward. The framework is technically elegant, but centres should not transplant the equations directly to other planning systems or manual workflows without local validation.

SOURCE

Physics and Imaging in Radiation Oncology

Browse more research Suggest a correction