Predicting 90-Day Prognosis in Ischemic Stroke Patients Post Thrombolysis Using Machine Learning

Area of Science:

• Neurology
• Medical Informatics
• Data Science

Background:

• Ischemic stroke is a leading cause of disability.
• Thrombolysis improves outcomes but requires careful patient selection.
• Accurate prognosis prediction is crucial for optimizing post-thrombolysis care.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting ischemic stroke prognosis 90 days post-thrombolysis.
• To identify key predictors of patient outcomes using SHAP analysis.

Main Methods:

• Utilized data from Qatar's stroke registry (2014-2022) including 723 ischemic stroke patients treated with thrombolysis.
• Assessed clinical variables: demographics, stroke severity, comorbidities, lab results, vital signs, and complications.
• Compared five machine learning models, with Support Vector Machine (SVM) performance evaluated using AUC and SHAP analysis.

Main Results:

• The SVM model achieved an Area Under the Curve (AUC) of 0.72 for prognosis prediction.
• Significant predictors included stroke severity, admission blood pressure, hypertension, coronary artery disease, stroke subtype (undetermined origin), and hospital-acquired urinary tract infections.

Conclusions:

• Machine learning, specifically the SVM model, shows promise for improving early prognosis prediction in ischemic stroke patients undergoing thrombolysis.
• This approach can support clinicians in developing personalized treatment strategies.
• Further research incorporating comprehensive data is encouraged to enhance personalized stroke care and survivor quality of life.