Machine Learning Models for Predicting Stroke Outcome After Low-Dose Thrombolysis

202 acute ischemic stroke patient records from a retrospective study conducted between August 1, 2019 and August 31, 2023. The dataset was used to develop and validate machine learning models predicting 90-day functional outcomes for patients receiving low-dose alteplase in an extended time window. The author is Huiru Chen, and the data was last updated on May 8, 2026.

Use Cases

• Training prognostic models for stroke recovery based on clinical features like age and NIHSS score mentioned in the description.
• Comparing the performance of machine learning algorithms like Random Forest for binary medical outcome prediction.
• Interpreting model predictions for clinical risk factors using SHAP analysis on features such as white blood cell count and hemorrhagic transformation.
• Validating predictive models for thrombolysis treatment outcomes in an extended time window.

Strengths

• Dataset contains 202 patient records with a defined 90-day outcome target.
• Model development used a 7:3 train-validation split (142 and 60 patients) and employed five machine learning algorithms.
• The best-performing Random Forest model achieved an AUC of 0.854 in the validation set.

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 temporal and institutional bias inherent to the single retrospective study.

Provenance

Source

figshare, author Huiru Chen.

Collection Method

Retrospective analysis of acute ischemic stroke patients receiving thrombolysis.

Time Range

August 1, 2019 to August 31, 2023.

Freshness

Last updated 2026-05-08 05:55:43; freshness should be verified.

Geography

null