A Fully Automated Analysis Pipeline for 4D Flow MRI in the Aorta.

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Image Analysis

Background:

• Four-dimensional (4D) flow MRI is a promising tool for evaluating aortic hemodynamics.
• Traditional manual analysis of 4D flow MRI data is time-consuming, limits reproducibility, and hinders large-cohort studies.

Purpose of the Study:

• To develop and validate a fully automated artificial intelligence (AI) pipeline for 4D flow MRI analysis.
• To assess the pipeline's performance in quantifying hemodynamic parameters and its reproducibility compared to manual analysis.

Main Methods:

• An AI pipeline integrating deep learning networks was developed to automate tasks like background-phase correction, noise masking, velocity anti-aliasing, and aorta segmentation.
• The pipeline quantified hemodynamic parameters including pulse wave velocity (PWV) and flow energetics.
• The pipeline was evaluated on a cohort of 379 subjects, including healthy controls, patients with bicuspid aortic valve (BAV), and pediatric patients with hereditary aortic disease, comparing results to manual analysis.

Main Results:

• The automated pipeline successfully processed 96% of the 4D flow MRI data.
• Pipeline-derived hemodynamic quantifications showed high correlation with manual analysis (peak velocity: r=1.00, PWV: r=0.99, flow energetics: r=0.99).
• The AI pipeline demonstrated superior or comparable agreement to manual analysis, outperforming inter-observer agreement in some cases, and accurately reproduced hemodynamic differences in BAV patients.

Conclusions:

• A fully automated AI pipeline for quantitative 4D flow MRI analysis has been successfully developed and validated.
• This AI framework significantly enhances measurement reliability and efficiency, overcoming limitations of manual processing.
• The pipeline offers a robust solution for large-scale 4D flow MRI studies, enabling more reproducible and accessible hemodynamic assessments.