An artificial intelligence-accelerated 2-minute multi-shot echo planar imaging protocol for comprehensive...

Area of Science:

• Magnetic Resonance Imaging (MRI) Physics
• Artificial Intelligence in Medical Imaging
• Image Reconstruction Techniques

Background:

• Conventional MRI protocols are time-consuming, limiting patient throughput and potentially causing motion artifacts.
• There is a need for rapid MRI techniques that maintain diagnostic image quality across multiple contrasts.

Purpose of the Study:

• To introduce and validate an artificial intelligence (AI)-accelerated multi-shot echo-planar imaging (msEPI) method.
• To achieve high signal-to-noise ratio (SNR), high tissue contrast, low specific absorption rates (SAR), and minimal distortion in 2 minutes.

Main Methods:

• Combined a novel machine learning (ML) scheme to limit g-factor noise amplification and improve SNR.
• Utilized a magnetization transfer preparation module for desirable contrast and high per-shot EPI undersampling factors to reduce distortion.
• Trained ML networks using data from 16 healthy subjects and evaluated performance across various acceleration factors, contrasts, and SNR conditions.

Main Results:

• Optimized msEPI sequences and protocols to balance acquisition efficiency and image quality compared to a 10-minute clinical reference.
• Demonstrated ML reconstruction flexibility across SNR levels and determined optimized regularization through radiological review.
• Validated network generalization to novel pathology in five clinical case studies.

Conclusions:

• The rapid 2-minute msEPI protocol with tunable ML reconstruction offers advantageous trade-offs between speed, SNR, and tissue contrast.
• This AI-accelerated method significantly outperforms the five-fold slower standard clinical reference exam in terms of efficiency and quality.