Advances in machine learning applications for cardiovascular 4D flow MRI.

Area of Science:

• Cardiovascular Imaging
• Medical Physics
• Biomedical Engineering

Background:

• Four-dimensional flow MRI (4D flow MRI) is a non-invasive technique for visualizing and quantifying blood flow.
• Derived hemodynamic parameters are crucial for diagnosing cardiovascular diseases.
• Current limitations include long acquisition times, limited resolution, and extensive post-processing.

Purpose of the Study:

• To review recent advancements in machine learning (ML) applied to 4D flow MRI.
• To highlight ML's role in overcoming technical limitations of 4D flow MRI.
• To focus on ML applications in data generation, post-processing, and blood flow evaluation.

Main Methods:

• Review of ML techniques for 4D flow MRI image reconstruction and acceleration.
• Application of ML for super-resolution to integrate CFD simulations with 4D flow MRI data.
• Utilization of ML for automating post-processing tasks like phase correction, anti-aliasing, and segmentation (e.g., U-Net).

Main Results:

• ML significantly improves the speed and accuracy of 4D flow MRI data acquisition.
• ML-based super-resolution enhances the realism of velocity measurements.
• ML automates laborious post-processing steps, including vessel delineation and phase correction.

Conclusions:

• Machine learning offers powerful solutions to enhance 4D flow MRI capabilities.
• ML advancements are critical for improving diagnostic accuracy and efficiency in cardiovascular imaging.
• Future research should focus on further integrating ML for comprehensive blood flow analysis.