Accelerated cardiac diffusion tensor imaging using deep neural network.

Area of Science:

• Medical Imaging
• Biophysics
• Artificial Intelligence

Background:

• Cardiac diffusion tensor imaging (DTI) is crucial for assessing myocardial microstructure noninvasively.
• Current DTI methods suffer from long scan times, limiting clinical applicability.

Purpose of the Study:

• To develop a novel deep learning framework (FG-Net) for accelerated cardiac DTI.
• To generate high-quality DTI parameter maps using a reduced number of diffusion-weighted images (DWIs).

Main Methods:

• Developed FG-Net, a deep learning framework combining image generation and tensor fitting.
• FG-Net generates inter-directional DWIs from six input DWIs to enhance information.
• Evaluated FG-Net on ex vivo human heart datasets.

Main Results:

• FG-Net accurately generated fractional anisotropy, mean diffusivity, and helix angle maps from only six DWIs.
• Quantification error for DTI parameters was less than 5%.
• FG-Net demonstrated superior performance over conventional tensor fitting and black-box network fitting.

Conclusions:

• FG-Net enables rapid, high-quality cardiac DTI parameter mapping.
• The framework shows potential for improving the clinical translation of fast cardiac DTI.
• FG-Net is robust across different b-values for accurate DTI parameter estimation.