Machine Learning-Based Model for Prediction of Outcomes in Acute Stroke

Area of Science:

• Neurology
• Artificial Intelligence in Medicine
• Biostatistics

Background:

• Predicting long-term outcomes in ischemic stroke is crucial for treatment decisions.
• Machine learning (ML) offers high accuracy and is increasingly used in healthcare.
• Existing prediction models may have limitations in accuracy for ischemic stroke outcomes.

Purpose of the Study:

• To investigate the applicability of ML techniques for predicting long-term outcomes in ischemic stroke patients.
• To compare the predictive performance of different ML models against a standard clinical score.
• To assess the potential of ML to enhance prognostic accuracy in stroke care.

Main Methods:

• Retrospective analysis of a prospective cohort of acute ischemic stroke patients.
• Development and comparison of three ML models: deep neural network (DNN), random forest, and logistic regression.
• Evaluation of model accuracy using the area under the curve (AUC) and comparison with the Acute Stroke Registry and Analysis of Lausanne (ASTRAL) score.

Main Results:

• A total of 2604 patients were analyzed; 78% had favorable outcomes at 3 months.
• The DNN model achieved a significantly higher AUC (0.888) than the ASTRAL score (0.839; P<0.001).
• Random forest (0.857) and logistic regression (0.849) models did not significantly outperform the ASTRAL score.

Conclusions:

• Machine learning algorithms, particularly DNNs, can significantly improve the prediction of long-term outcomes in ischemic stroke.
• ML models hold potential to refine prognostic assessments beyond traditional scoring systems.
• Further research into ML applications can enhance personalized treatment strategies for stroke survivors.