Reduction of respiratory motion artifacts in gadoxetate-enhanced MR with a deep learning-based filter using...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Radiology
• Deep Learning for MRI

Background:

• Gadoxetate disodium-enhanced liver MRI is crucial for diagnosing liver lesions.
• Motion artifacts degrade image quality and can obscure important diagnostic details.
• Existing methods for motion artifact reduction may be insufficient, particularly in challenging cases.

Purpose of the Study:

• To evaluate the effectiveness of a novel deep learning-based filter, motion artifact reduction with convolutional neural network (MARC), for reducing motion artifacts in gadoxetate disodium-enhanced liver MRI.
• To assess the impact of MARC on image quality and lesion conspicuity.
• To determine the clinical utility of MARC, especially in patients with impaired breath-holding capabilities.

Main Methods:

• A retrospective analysis of 192 patients undergoing gadoxetate disodium-enhanced liver MRI was performed.
• A convolutional neural network (MARC) was developed and trained on simulated motion-corrupted MR images.
• Original and MARC-filtered pre-contrast and arterial phase images were evaluated for motion artifacts and lesion conspicuity.

Main Results:

• MARC significantly reduced motion artifact scores in liver MRI (average score 1.97 vs. 2.53, p < 0.001).
• The filter demonstrated substantial improvement in cases with moderate to severe motion artifacts.
• Lesion conspicuity was significantly enhanced by MARC without compromising anatomical detail (p < 0.001).

Conclusions:

• The MARC filter effectively reduces motion artifacts in gadoxetate disodium-enhanced arterial phase liver MRI.
• MARC improves image quality and lesion conspicuity, offering significant clinical value, particularly for patients with breath-holding difficulties.
• This deep learning approach provides a promising tool for enhancing diagnostic accuracy in liver MRI.