Encoding Enhanced Complex CNN for Accurate and Highly Accelerated MRI.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonary Medicine

Background:

• Hyperpolarized noble gas Magnetic Resonance Imaging (MRI) visualizes lung structure and function.
• Long imaging times currently limit clinical applications of lung MRI.
• Deep learning shows promise for accelerating MRI via undersampled data reconstruction.

Purpose of the Study:

• To develop a novel deep learning approach for accelerated pulmonary MRI reconstruction.
• To address limitations of existing Convolutional Neural Networks (CNNs) in k-space data processing.
• To enhance the efficiency and quality of hyperpolarized gas lung MRI.

Main Methods:

• Proposed an encoding-enhanced (EN2) complex CNN for pulmonary MRI.
• Utilized directional convolutions mimicking k-space sampling for efficient feature learning.
• Incorporated complex convolutions and a feature-strengthened modularized unit for improved reconstruction.

Main Results:

• Accurate reconstruction of hyperpolarized 129Xe and 1H lung MRI from 6-fold undersampled data.
• Lung function measurements showed minimal bias compared to fully sampled images.
• Demonstrated the effectiveness of EN2 complex CNN for accelerated pulmonary MRI.

Conclusions:

• The EN2 complex CNN effectively reconstructs undersampled pulmonary MRI.
• The method preserves lung function measurement accuracy.
• This approach holds potential for accelerating lung MRI in research and clinical settings.