Prediction of Long-Term Stroke Recurrence Using Machine Learning Models | JoVE Visualize

Prediction of Long-Term Stroke Recurrence Using Machine Learning Models

Vida Abedi^1,2, Venkatesh Avula¹, Durgesh Chaudhary³

¹Department of Molecular and Functional Genomics, Geisinger Health System, Danville, PA 17822, USA.

Journal of Clinical Medicine|April 3, 2021

View abstract on PubMed

Summary

Machine learning models can predict ischemic stroke recurrence using patient data. Key factors include age, BMI, and lab results, enabling personalized interventions for stroke prevention.

Area of Science:

Neurology
Medical Informatics
Biostatistics

Background:

Assessing long-term ischemic stroke recurrence risk is challenging, with current estimates ranging from 17% to 30%.
Individual-level risk prediction for recurrent ischemic stroke remains unreliable.
This study aimed to develop and optimize machine learning models for predicting stroke recurrence.

Purpose of the Study:

To train machine learning models to predict ischemic stroke recurrence.
To identify key clinical variables associated with stroke recurrence.
To optimize model performance metrics for clinical implementation.

Main Methods:

Utilized patient-level electronic health record data.
Developed 288 models using six interpretable algorithms and four feature selection strategies.
Employed five prediction windows (up to 5 years) and two sampling strategies.

Main Results:

Model performance (AUROC) was stable across prediction windows, with the highest for 1-year (0.79) and lowest for 5-year (0.69).
Age, BMI, and laboratory features (HDL, HbA1c, creatinine) were identified as highly important predictors.
Sampling strategies improved the balance between model specificity and sensitivity.

Conclusions:

Six machine learning algorithms successfully predicted long-term stroke recurrence.
Laboratory-based variables showed strong association with recurrence, suggesting targets for personalized interventions.
Optimized models can serve as decision support tools for targeted interventions within healthcare systems.