A CT-based multi-scale fusion model with SHAP interpretation for preoperative differentiation between lung adenocarcinoma in situ/minimally invasive adenocarcinoma and invasive adenocarcinoma: a multicenter study - Summary - MDSpire
Advertisement
A CT-based multi-scale fusion model with SHAP interpretation for preoperative differentiation between lung adenocarcinoma in situ/minimally invasive adenocarcinoma and invasive adenocarcinoma: a multicenter study
To develop a multi-scale model integrating radiomics, habitat, deep learning, and clinical information for preoperative precise differentiation between lung adenocarcinoma in situ/minimally invasive adenocarcinoma and invasive adenocarcinoma in ground-glass nodules (GGNs), ultimately improving patient outcomes.
Key Findings:
The early fusion model achieved AUCs of 0.988, 0.903, and 0.872 in training, internal validation, and external testing cohorts, respectively, indicating high diagnostic accuracy.
The 2.5D ResNet50 architecture provided the best performance among deep learning models.
Interpretation:
The multi-scale early fusion model significantly enhances preoperative diagnosis accuracy compared to existing methods, aiding in individualized treatment decisions for GGNs.
Limitations:
The study is retrospective and may have selection bias, potentially affecting the generalizability of the findings.
The model's performance may vary in different clinical settings.
Conclusion:
The developed model offers a robust framework for the preoperative differentiation of GGNs, potentially improving patient outcomes.
