Deep learning on whole-lung CT predicts air space spread in lung adenocarcinoma

A deep learning model using whole-lung CT achieved an AUROC of 0.90 (95% CI 0.86-0.94) for preoperative prediction of spread through air spaces (STAS) in lung adenocarcinoma, outperforming clinical (0.72) and radiomics (0.65) models.

• DAFNet provided fully automated end-to-end analysis without manual tumor delineation, using adaptive gate fusion and DINOv3 pre-training.
• The multi-center retrospective study included 1164 patients (511 STAS-positive, 653 negative) split 70:30 into training and test sets.

Automated summary

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