Delta-Radiomics Features for Predicting Immunotherapy Pneumonitis in Lung Cancer

A single-center retrospective study of 131 patients with advanced non-small cell lung cancer receiving immune checkpoint inhibitors, among whom 46 developed pneumonitis. Dong Xie published this dataset on figshare in 2026, containing radiomics features extracted from pre- and post-treatment CT scans to build prediction models.

Use Cases

• Training binary classification models for pneumonitis risk based on delta-radiomics features.
• Comparing machine learning algorithm performance (LR, SVM, KNN, XGBoost) for a clinical prediction task.
• Developing integrated clinical-radiomics models using features like interstitial lung disease history and neutrophil/lymphocyte ratio.
• Studying feature selection methods like LASSO regression on high-dimensional radiomics data.

Strengths

• Model performance metrics are provided, including an AUC of 0.92 for the training set and 0.85 for the validation set.
• The study processed 2153 initial radiomics features and selected 22 key delta features.
• The dataset is associated with a detailed methodological description covering patient cohort, imaging, and model validation.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The study notes performance for high-grade pneumonitis could not be evaluated due to limited events (8 patients).

Provenance

Source

figshare

Collection Method

Retrospective extraction of radiomics features from enhanced CT images of primary tumors.

Freshness

Last updated 2026-05-28 06:20:03; freshness should be verified.