AI-generated LAD surrogate strongly correlates with manual contour dose metrics in lung SBRT

An AI-generated left anterior descending (LAD) artery region surrogate showed strong correlation with manual contours for maximum dose (R²=0.93) and V10Gy (R²=0.81) in lung SBRT. Mean heart dose was a poor predictor of LAD dose (R²≈0.4). Retrospective, single-center.

• Retrospective study using deep inspiration breath hold CT scans from 49 patients with early-stage NSCLC, with 96 lung SBRT dose maps registered for analysis.
• Mean heart dose poorly predicted LAD dose (R² ≈ 0.4), highlighting the need for dedicated LAD assessment.
• Limitation: single-center, retrospective design without external validation of the AI segmentation model.

RadPigeon summaries are original and for information only. They are not clinical advice.