MC² -Net: motion correction network for multi-contrast brain MRI.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Motion artifacts significantly degrade the quality of multi-contrast brain MRI scans.
• Accurate motion correction is crucial for reliable diagnosis and analysis of brain imaging data.

Purpose of the Study:

• To develop and validate a deep learning-based motion correction network for multi-contrast brain MRI.
• To address in-plane rigid motion artifacts in brain MR images.

Main Methods:

• A two-part approach involving unsupervised image alignment using a CNN and supervised motion correction.
• Image alignment minimizes normalized cross-correlation loss and maximizes normalized mutual information.
• Motion correction network trained using structural similarity and VGG loss, with simulated motion-corrupted datasets.

Main Results:

• The network successfully corrected simulated motion artifacts, showing significant increases in structural similarity and normalized mutual information for various MRI contrasts.
• Performance was enhanced with image alignment and artifact-free input images for other contrasts.
• The method quantitatively outperformed existing deep learning techniques and demonstrated potential in real-world clinical settings with healthy subjects.

Conclusions:

• A novel deep learning-based motion correction method for multi-contrast MRI has been successfully developed.
• Experimental results validate the effectiveness and clinical potential of the proposed technique.