Infantile Epileptic Spasms Syndrome Prognostic Model Data

Caoxue Zuo developed a machine learning model to predict poor seizure outcomes in infants with Infantile Epileptic Spasms Syndrome (IESS). The model, trained on retrospective clinical data from Wuhan Children's Hospital, achieved an AUROC of 0.921 in validation. The dataset was last updated in April 2026.

Use Cases

• Predict poor seizure outcome using features like MRI findings of tuberous sclerosis complex or malformations of cortical development.
• Validate prognostic models using independent risk factors identified via Cox regression analysis.
• Benchmark machine learning algorithms like XGBoost on clinical outcome prediction tasks.
• Interpret model decisions for clinical use via SHAP analysis and decision tree visualization.
• Assess model clinical utility using metrics like decision curve analysis (DCA) on the validation set.

Strengths

• Model achieved a high AUROC of 0.921 in the validation set.
• Study employed a 7:3 train-validation split for robust evaluation.
• Analysis includes multiple performance dimensions: discrimination, calibration, and clinical utility.

Limitations

• Dataset size is small at 31.7 KB, indicating limited scope.
• Data originates from a single hospital (Wuhan Children's Hospital), limiting geographic generalizability.
• Poor seizure outcome was observed in 56% of the cohort, indicating a potential class imbalance.

Provenance

Source

Caoxue Zuo via figshare.

Collection Method

Retrospective study of pediatric patients diagnosed with IESS.

Freshness

Data was last updated in April 2026.

Geography

Wuhan Children's Hospital, China.