AI-Based STroke Risk fActor Classification and Treatment (ABSTRACT) study

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Prediction Models

Background:

• Stroke is a major global cause of death and disability, with significant economic impact.
• A substantial portion of stroke patients lack identifiable risk factors, necessitating improved risk prediction.
• The AI-Based STroke Risk fActor Classification and Treatment (ABSTRACT) study aims to enhance stroke risk assessment.

Purpose of the Study:

• To develop three distinct machine learning (ML) models for stroke risk prediction using different data types: brain imaging (CT/MRI), cardiovascular data (ECG/echocardiography), and clinical/historical data.
• To conduct explainability analyses to uncover novel stroke risk factors.
• To calibrate predictive models with real-world probabilities and create a unified ensemble model.

Main Methods:

• A retrospective observational cohort study involving 9,155 stroke patients and 109,581 controls from southwest England.
• Data extraction from hospital and general practice records, including CT/MRI, ECG, echocardiography, laboratory tests, ultrasound, and medical history.
• Application of machine learning techniques for stroke risk prediction and novel risk factor identification.

Main Results:

• Phase I focuses on the protocol development for creating a multimodal stroke prediction model.
• The study outlines data handling procedures aligned with UK ethical governance.
• Strategies for data pre-processing and model training are detailed.

Conclusions:

• ABSTRACT Phase I establishes a framework for developing an AI-driven multimodal stroke prediction model.
• The protocol details ethical data handling and pre-processing for machine learning model training.
• This work lays the foundation for improved stroke risk stratification and personalized treatment strategies.