Radiomics and 3D deep learning model on HRCT differentiates minimally invasive from invasive adenocarcinoma in subsolid nodules

Fusion model combining HRCT radiomics and 3D deep learning achieved AUC 0.937 for differentiating minimally invasive from invasive adenocarcinoma in subsolid nodules in internal validation.

• Retrospective single-center study of 374 patients with subsolid nodules (124 minimally invasive adenocarcinoma, 250 invasive adenocarcinoma).
• Radiomics-only model AUC 0.833, deep learning-only AUC 0.887, vs fusion model AUC 0.937.
• Internal validation only; lacks external validation in other cohorts.

Automated summary

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