Reducing Contrast Agent Dose in Cardiovascular MR Angiography with Deep Learning.

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Contrast-enhanced magnetic resonance angiography (MRA) is crucial for diagnosing cardiovascular conditions.
• Gadolinium-based contrast agents (GBCAs) used in MRA pose risks of dose-related adverse effects.
• Reducing GBCA dosage is essential for patient safety in MRA procedures.

Purpose of the Study:

• To develop and evaluate a deep learning (DL) method for significantly reducing GBCA dose in MRA by up to 80%.
• To assess the image quality and diagnostic performance of DL-enhanced low-dose MRA (ELD-MRA) compared to standard low-dose (LD-MRA) and high-dose MRA (HD-MRA).

Main Methods:

• A neural network was trained using retrospective data to enhance LD 3D MRA, then tested on prospective LD MRA data from 40 congenital heart disease patients.
• Image quality metrics including signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and edge sharpness were quantitatively assessed.
• Perceptual image quality, diagnostic confidence, and diagnostic accuracy (sensitivity/specificity) were evaluated and compared across LD-MRA, ELD-MRA, and HD-MRA.

Main Results:

• ELD-MRA demonstrated significantly improved SNR, CNR, and edge sharpness compared to LD-MRA, with comparable results to HD-MRA.
• Perceptual sharpness was slightly lower in ELD-MRA than HD-MRA, but overall perceptual contrast was comparable.
• Sensitivity and specificity for ELD-MRA were 0.882/0.960, approaching HD-MRA, and diagnostic confidence was significantly higher than LD-MRA.
• Vessel diameter measurements showed comparable agreement with HD-MRA for both LD-MRA and ELD-MRA.

Conclusions:

• Deep learning effectively enhances image quality in low-dose cardiovascular MRA, improving contrast and sharpness.
• The developed DL method allows for an 80% reduction in GBCA dose while preserving diagnostic image quality and confidence.
• This approach holds promise for safer and more effective cardiovascular MRA by minimizing contrast agent risks.