3D Intratumoral Heterogeneity Scores for Lung Adenocarcinoma from Multicenter CT Study

A multicenter retrospective study from three medical centers includes data from 1,301 patients with lung adenocarcinoma (LUAD). The dataset was used to develop a stacking ensemble model integrating three-dimensional intratumoral heterogeneity scores with clinicoradiologic features for preoperative prediction of visceral pleural invasion. The model achieved an area under the curve of 0.878, with the 3D ITH score identified as the most influential predictor.

Use Cases

• Developing machine learning models for preoperative prediction of visceral pleural invasion based on 3D intratumoral heterogeneity scores.
• Comparing the performance of ensemble models against base classifiers like XGBoost for clinical prediction tasks.
• Interpreting feature importance in medical prediction models using SHAP analysis based on radiomic and clinicoradiologic features.

Strengths

• Data originates from a multicenter study involving 1,301 patients, providing a substantial sample size.
• Model performance is validated with a fixed external validation cohort from a separate medical center.
• The final stacking ensemble model achieved a reported AUC of 0.878.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small (34.0 KB), suggesting it contains summary or model results rather than raw patient data.

Provenance

Source

figshare

Collection Method

Multicenter retrospective study.

Freshness

Last updated 2026-05-12 04:19:51; freshness should be verified.