Automated High-Order Shimming for Neuroimaging Studies

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Magnetic Resonance Spectroscopy (MRS)
• Medical Imaging and Spectroscopy

Background:

• B0 inhomogeneity is a major challenge in high-field MRI and MRS, causing image distortions and signal loss.
• Current high-order shimming methods often require manual and time-consuming region of interest (ROI) selection.

Purpose of the Study:

• To develop and evaluate an automated high-order shimming (autoHOS) method for efficient B0 inhomogeneity correction.
• To eliminate the need for manual ROI selection in high-order shimming procedures.

Main Methods:

• Proposed an automated high-order shimming (autoHOS) method integrating deep learning for brain extraction and image-based shimming.
• Automated real-time brain extraction defined the field map ROI for the shimming algorithm.
• Evaluated autoHOS using in vivo echo-planar imaging (EPI) and spectroscopic studies at 3T and 7T.

Main Results:

• AutoHOS significantly reduced EPI image distortion and narrowed MRS spectral lineshapes compared to linear and manual high-order shimming.
• Demonstrated improved image and spectral quality at both 3T and 7T field strengths.
• Successfully corrected B0 inhomogeneity without requiring additional user interaction.

Conclusions:

• The automated high-order shimming (autoHOS) method effectively corrects B0 inhomogeneity in MRI and MRS.
• AutoHOS offers a significant advancement by automating ROI selection, saving time and improving consistency.
• This deep learning-based approach enhances overall image and spectral quality in high-field applications.