Semi-automated pipeline for generating personalised cerebrovascular models

Area of Science:

• Biomedical Engineering
• Computational Fluid Dynamics
• Medical Imaging Analysis

Background:

• Subject-specific cerebrovascular models are crucial for predicting hemodynamics but are often time-consuming to generate.
• Existing methods face challenges in clinical time-sensitivity due to complex model creation processes.

Purpose of the Study:

• To develop a semi-automated pipeline for rapid generation of personalized 0D cerebrovascular models.
• To calibrate these models using a neurofuzzy control scheme for accurate hemodynamic prediction.
• To validate the pipeline using the Circle of Willis (CoW) and demonstrate its clinical utility.

Main Methods:

• A semi-automated pipeline was developed to extract vasculature geometry and blood flow data.
• Bond graph-based models were automatically generated from vessel connectivity and parameters.
• A neurofuzzy control scheme calibrated peripheral resistances to minimize flow distribution discrepancies.
• Subject-specific CoW models were generated and validated against 4D flow MRI data.

Main Results:

• The pipeline achieved a relative error of $$ for flow and $$ for pulsatility in CoW models.
• Higher errors were observed in smaller vessels, indicating areas for future refinement.
• Simulations of vascular occlusion demonstrated the benefit of a fully connected CoW for flow redistribution.

Conclusions:

• The developed pipeline offers rapid, modular generation of personalized cerebrovascular models.
• This approach is a promising tool for both research and clinical applications with variable data and resource constraints.
• The models accurately predict hemodynamics and aid in understanding cerebrovascular responses to conditions like occlusion.