Deep Learning Assisted Outer Volume Removal for Highly-Accelerated Real-Time Dynamic MRI.

Area of Science:

• Medical Imaging
• Cardiovascular MRI
• Image Reconstruction

Background:

• Real-time (RT) dynamic MRI is crucial for assessing cardiac function and motion.
• Conventional cardiac MRI requires breath-holding and ECG-gating, which is challenging for some patients.
• High acceleration rates in RT cine MRI are limited by aliasing artifacts from surrounding tissues.

Purpose of the Study:

• To develop a novel outer volume removal (OVR) method for artifact reduction in RT cine MRI.
• To enable higher acceleration rates in cardiac MRI while maintaining diagnostic image quality.
• To provide a post-processing solution that does not require modifications to the MRI acquisition.

Main Methods:

• A post-processing outer volume removal (OVR) technique was developed.
• A deep learning (DL) model was trained to estimate and remove outer volume signals from k-space data.
• A physics-driven DL (PD-DL) reconstruction with an OVR-specific loss function was used.

Main Results:

• The proposed OVR method effectively reduced aliasing artifacts from non-cardiac regions.
• High-acceleration RT cine MRI achieved image quality comparable to clinical baselines.
• The method outperformed conventional reconstruction techniques both qualitatively and quantitatively.

Conclusions:

• The novel OVR method is a practical and effective solution for artifact reduction in RT cine MRI.
• This approach facilitates higher acceleration rates, improving efficiency and patient comfort.
• The technique preserves diagnostic image quality, enhancing the utility of free-breathing cardiac MRI.