Digital twin integrating clinical, morphological and hemodynamic data to identify stroke risk factors

Area of Science:

• Cardiology
• Medical Imaging
• Computational Biology

Background:

• Ischemic stroke, a major cause of mortality, is often linked to atrial fibrillation (AF).
• Current stroke risk assessments (e.g., CHA₂DS₂-VASc) primarily use clinical data, neglecting crucial anatomical and flow dynamics within the left atrium (LA) and left atrial appendage (LAA).
• Limited studies have integrated detailed LA morphology and hemodynamics with clinical data for comprehensive risk evaluation.

Purpose of the Study:

• To develop and validate a Digital Twin framework for enhanced thrombus risk stratification in atrial fibrillation patients.
• To integrate LA morphology, hemodynamics, and clinical data for improved patient phenotyping.
• To identify patient subgroups with varying stroke risk based on integrated data.

Main Methods:

• A Digital Twin framework was developed, integrating statistical and mechanistic models.
• Unsupervised Multiple Kernel Learning was applied to data from 130 AF patients.
• The model combined left atrial (LA) morphology, hemodynamics, and clinical data, including B-type natriuretic peptide levels.

Main Results:

• The integrated approach successfully stratified patients into three distinct phenogroups.
• The highest-risk group was characterized by larger atrial dimensions, complex LAA anatomy, and elevated B-type natriuretic peptide.
• This phenogroup demonstrated a higher propensity for thrombus formation.

Conclusions:

• Digital Twin models offer a powerful tool for assessing thrombus formation risk in AF patients.
• Integrating morphological and hemodynamic data significantly improves patient stratification compared to clinical scores alone.
• Further research is warranted to refine these models for precise stroke prediction.