Predictive Model for Refractory Mycoplasma Pneumonia in Children

419 children from a multicenter retrospective study were used to develop random forest models predicting refractory Mycoplasma pneumoniae pneumonia. The integrated model, combining clinical, imaging, and radiomics features, achieved an AUC of 0.811 in validation. The dataset includes extracted radiomics features from chest CT scans and clinical variables.

Use Cases

• Building predictive models for refractory pneumonia based on clinical variables like D-dimer and fever type.
• Evaluating radiomics feature performance for medical image analysis based on gray-level co-occurrence matrix and wavelet kurtosis.
• Comparing model architectures like random forests for clinical-imaging, radiomics, and integrated approaches.
• Assessing model performance metrics such as AUC, McNemar tests, and net reclassification improvement.

Strengths

• Dataset includes 419 patient cases from a multicenter study.
• Models were validated on separate training, testing, and external validation cohorts (248, 62, and 109 patients).
• Integrated model achieved a specific AUC of 0.811 with a 95% CI of 0.704–0.917.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect geographic/temporal bias inherent to the multicenter retrospective study design.

Provenance

Source

figshare

Collection Method

Multicenter retrospective study; radiomics features extracted from chest CT scans using PyRadiomics.

Freshness

Last updated 2026-03-19 06:25:51; freshness should be verified.