CT-Based Multi-Scale Fusion Model for Lung Adenocarcinoma Differentiation

Ning Dong's dataset supports a multicenter study on differentiating lung adenocarcinoma subtypes in ground-glass nodules. It includes data from 621 patients across two centers, integrating radiomics, habitat, deep learning, and clinical information. The dataset was last updated on May 12, 2026.

Use Cases

• Developing a multi-scale diagnostic model based on integrated radiomics, habitat, and clinical features.
• Comparing early (feature-level) and late (decision-level) fusion strategies for model performance.
• Applying SHAP interpretability analysis to understand model predictions for lung adenocarcinoma subtypes.
• Training and validating deep learning architectures, such as 2.5D ResNet50, on medical imaging data.

Strengths

• Multicenter design with data from 621 patients, which suggests a substantial sample size.
• Integrates multiple data types: radiomics, habitat, deep learning, and clinical information.
• The described model achieved high performance metrics, including an AUC of 0.988 in the training cohort.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic or institutional bias inherent to the two participating centers.

Provenance

Source

Ning Dong via figshare

Collection Method

Retrospective enrollment of patient data from two medical centers.

Freshness

Last updated 2026-05-12 04:19:28; freshness should be verified.