Machine Learning Model for Stroke Identification in Guillain-Barré Syndrome Patients

A retrospective cohort of 260 Guillain-Barré syndrome patients from the Second Affiliated Hospital of Army Medical University between 2015 and 2024. The dataset was used to develop and compare seven machine learning models for identifying concomitant stroke, with an external validation cohort of 60 patients from the First Affiliated Hospital. The neural network model achieved an AUROC of 0.838 in internal validation.

Use Cases

• Training machine learning models for stroke risk stratification based on admission-time clinical predictors.
• Comparing the performance of different ML algorithms (e.g., neural networks) for a binary classification task.
• Analyzing the relationship between model-estimated risk strata and short-term functional outcomes.
• Validating clinical prediction models using an independent external cohort.

Strengths

• Includes 260 patient records from a defined hospital cohort spanning 9 years.
• Model performance is quantified with an AUROC of 0.838 and a 95% confidence interval.
• Validation includes an independent external cohort of 60 patients.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to a single hospital system in China.

Provenance

Source

Second Affiliated Hospital of Army Medical University and First Affiliated Hospital.

Collection Method

Retrospective cohort study; data collected at admission.

Time Range

January 1, 2015 to December 31, 2024.

Freshness

Last updated 2026-05-07 04:44:37; freshness should be verified.

Geography

China (hospital locations unspecified but likely within China).