Post-Stroke Epilepsy Prediction Data from a 21,459-Patient Cohort

A retrospective cohort study of 21,459 stroke patients, including 936 who developed post-stroke epilepsy, used to develop an interpretable machine-learning framework. The dataset was created by Lijun Wu and last updated on 2026-05-21. It was designed to address severe class imbalance for clinical screening.

Use Cases

• Developing imbalance-aware prediction models based on neurological severity and lab markers mentioned in the description
• Training high-sensitivity screening tools based on the described dual-tier framework
• Interpreting feature importance for clinical outcomes using SHAP values as described
• Benchmarking model performance on imbalanced medical data based on reported metrics like macro-AUC and F1-score

Strengths

• Dataset is derived from a cohort of 21,459 patients, providing a substantial sample size.
• Includes 936 confirmed cases of post-stroke epilepsy, enabling analysis of a specific complication.
• Model performance metrics are reported, including a macro-AUC of 0.996 and sensitivity up to 0.971.
• Key clinical predictors like neurological severity, hypertension, lactate, D-dimer, and AST are identified.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the underlying data tables is unknown, which may limit suitability assessment.
• The dataset is very small (21.5 KB), suggesting it may contain summary statistics or a model description rather than the full patient-level data.

Provenance

Source

figshare, author Lijun Wu

Collection Method

Retrospective cohort study

Freshness

Last updated 2026-05-21 04:19:23; freshness should be verified.