Real-time AI integration for MR to detect artifacts and guide pulse sequence adaptations.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Artificial Intelligence (AI)
• Medical Spectroscopy

Background:

• Out-of-voxel (OOV) artifacts can compromise the quality of magnetic resonance spectroscopy (MRS) data.
• Real-time artifact detection and correction are crucial for improving diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate a novel AI-integrated MR pulse sequence for real-time detection and correction of OOV artifacts.
• To prospectively update the crusher gradient scheme within the repetition time (TR) to minimize OOV contamination.

Main Methods:

• The PEREGRINE (Per Excitation Real-time Execution & Guided Responses with Integrated Neural-network Evaluation) system utilized convolutional autoencoders for OOV artifact detection.
• Scans were performed on healthy volunteers using edited MRS, comparing AI-off and AI-on conditions.
• The AI system triggered gradient scheme updates based on quantified OOV scores, iterating through 48 permutations.

Main Results:

• PEREGRINE provided real-time OOV scores and updated gradients within each 2-second TR.
• While overall OOV scores showed no difference in the 'Full' condition, the 'Dwell' condition demonstrated significantly lower OOV scores with AI-on compared to AI-off.
• Fit Quality Number (FQN) significantly improved in the AI-on scan, indicating enhanced data quality.

Conclusions:

• PEREGRINE successfully enabled real-time evaluation and reduction of OOV artifacts using an AI-integrated MR sequence.
• The system identified gradient modifications that effectively minimized OOV contamination, leading to improved MRS data quality.