ACGRHA-Net: Accelerated multi-contrast MR imaging with adjacency complementary graph assisted residual hybrid...

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Multi-contrast magnetic resonance (MR) imaging enhances medical diagnosis but suffers from long acquisition times, causing patient discomfort and limiting its use.
• Undersampling k-space data to accelerate MR imaging introduces aliasing artifacts, degrading image quality.

Purpose of the Study:

• To develop an advanced method for reconstructing multi-contrast MR images from undersampled data.
• To mitigate aliasing artifacts and improve the quality of accelerated MR images.

Main Methods:

• Proposes the adjacency complementary graph assisted residual hybrid attention network (ACGRHA-Net).
• Utilizes a fully-sampled auxiliary contrast MR image to learn an adjacency complementary graph, capturing structural similarity between contrasts.
• Integrates a residual hybrid attention module for adaptive feature emphasis and fusion.

Main Results:

• ACGRHA-Net effectively reconstructs multi-contrast MR images from zero-filled data.
• The learned graph captures optimal structural similarity, enhancing detail reconstruction.
• Experimental results demonstrate superior performance over state-of-the-art methods across various sampling patterns and acceleration factors.

Conclusions:

• ACGRHA-Net significantly enhances multi-contrast MR image reconstruction quality.
• The method offers a promising solution for accelerating MR imaging while maintaining high diagnostic accuracy.
• This approach has the potential to improve patient experience and broaden the clinical application of MR imaging.