Machine learning-based patient classification system for adults with stroke: A systematic review.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Decision Support Systems

Background:

• Stroke classification is crucial for timely and effective treatment.
• Machine learning (ML) offers potential for improving stroke patient stratification.
• Existing evidence on ML-based stroke classification systems requires systematic evaluation.

Approach:

• A systematic review was conducted following PRISMA guidelines.
• Searched PubMed, MEDLINE, Web of Science, and CINAHL Plus Full Text (January 2015 - February 2021).
• Included twelve studies utilizing fifteen distinct ML algorithms.

Key Points:

• Support Vector Machines (SVM), Random Forests (RF), and Decision Trees (DT) were the most frequently used ML algorithms.
• Age and gender were the most common input features for ML models.
• Forty-four different input features were identified across the included studies.

Conclusions:

• No single ML algorithm demonstrated universal superiority for stroke patient classification.
• Optimal algorithm selection is dependent on the specific characteristics of the input data.
• Further research is needed to identify best practices for ML application in stroke stratification.