Computational Frontiers in Arteriovenous Fistula Maturation: A Review of Fluid Dynamics and Machine Learning Models

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Nephrology

Background:

• Arteriovenous (AV) fistulas are crucial for hemodialysis but have high maturation failure rates (up to 60%).
• Adverse hemodynamic conditions are implicated, but mechanisms are not fully understood.
• Current prediction methods lack comprehensive analysis of fistula hemodynamics.

Purpose of the Study:

• To review the application of computational fluid dynamics (CFD) and machine learning (ML) in understanding AV fistula failure.
• To explore how CFD and ML can predict clinical outcomes like maturation and stenosis.
• To identify limitations and future directions for these computational approaches.

Main Methods:

• Computational fluid dynamics (CFD) modeling to analyze fistula geometry and hemodynamic parameters (e.g., wall shear stress, oscillatory shear index).
• Machine learning (ML) models utilizing clinical, ultrasound, and acoustic data to predict AV fistula outcomes.
• Literature review of existing studies employing CFD and ML in AV fistula research.

Main Results:

• CFD studies link disturbed flow patterns (low WSS, high OSI) to neointimal hyperplasia and stenosis.
• ML models show promise in predicting AV fistula maturation, stenosis, and failure.
• Both CFD and ML show potential for improving understanding and prediction of AV fistula complications.

Conclusions:

• CFD and ML are valuable tools for elucidating AV fistula failure mechanisms and predicting outcomes.
• Clinical utility is currently limited by data size, validation, and model assumptions.
• Future research should focus on integrating CFD and ML with larger, diverse datasets for prospective validation to enhance clinical applicability.