Optimizing contrast-enhanced abdominal MRI: A comparative study of deep learning and standard VIBE techniques

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Conventional coronal Volume Interpolated Breath-hold Examination (VIBE) sequences are crucial for upper abdominal MRI.
• Faster acquisition times are desirable to improve patient comfort and reduce motion artifacts.

Purpose of the Study:

• To validate a deep learning (DL) reconstruction technique for faster post-contrast enhanced coronal VIBE sequences.
• To compare the image quality of DL-reconstructed VIBE (VIBE_DL) with standard VIBE (VIBE_SD).

Main Methods:

• A prospective study of 151 patients undergoing 3T upper abdominal MRI.
• Acquisition of both VIBE_DL and VIBE_SD coronal T1 fat-suppressed sequences.
• Independent radiologist evaluation of image quality parameters and lesion detection using a 4-point Likert scale.

Main Results:

• VIBE_DL acquisition time was significantly reduced (10.2s vs 22.3s).
• VIBE_DL demonstrated superior overall image quality, reduced noise, enhanced liver edge and vessel sharpness, and better lesion conspicuity.
• VIBE_DL detected 99.3% of lesions compared to 94.9% for VIBE_SD, with increased peripheral artifacts.

Conclusions:

• Deep learning reconstruction significantly outperforms standard VIBE in image quality and lesion detection.
• The DL technique reduces acquisition time, showing promise for enhancing contrast-enhanced abdominal MRI.
• This validates DL as a valuable tool for improving MRI efficiency and diagnostic accuracy.