Using synthetic MR images for distortion correction.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Functional MRI (fMRI) data, particularly echo-planar imaging (EPI), suffer significant distortions due to magnetic field inhomogeneities.
• Accurate alignment between EPI and anatomical images (T1w/T2w) is challenging due to contrast differences and EPI distortions.
• Traditional distortion correction relies on field map data, which is often unavailable, unreliable, or difficult to acquire in specific cohorts.

Purpose of the Study:

• To develop and validate a novel, field map-free software package named Synth for distortion correction and cross-modal image registration of fMRI data.
• To provide a robust solution for accurate fMRI data alignment, particularly in datasets with missing or corrupted field map information.

Main Methods:

• Synth utilizes T1w and T2w anatomical images to generate a synthetic, undistorted EPI-like reference image.
• This synthetic image serves as a reference for individual-specific distortion correction of EPI data.
• The method was evaluated on pediatric (ABCD) and adult (MSC, HCP) neuroimaging datasets.

Main Results:

• Synth demonstrates comparable performance to traditional field map-based distortion correction methods.
• In several instances, Synth outperformed existing field map correction approaches.
• The software enables accurate and precise registration of fMRI data, even when field map data is absent or compromised.

Conclusions:

• Synth offers a reliable and effective field map-less approach for correcting EPI distortions and improving cross-modal image registration.
• This method enhances the utility of neuroimaging datasets lacking field map information, particularly in challenging populations like children.
• Synth facilitates more accurate and accessible analysis of fMRI data in diverse research settings.