Modified MRI Anonymization (De-Facing) for Improved MEG Coregistration

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Data Science

Background:

• Accurate source localization of Magnetoencephalography (MEG) and Electroencephalography (EEG) signals necessitates structural Magnetic Resonance Imaging (MRI) for head modeling.
• Sharing sensitive structural MRI data for reproducibility is often hindered by privacy concerns requiring facial anonymization.
• Existing automated de-facing methods can remove crucial facial features, compromising the coregistration accuracy between MRI and MEG/EEG data.

Purpose of the Study:

• To introduce and evaluate a novel automated de-facing technique for structural MRI data that preserves the nose.
• To assess the impact of this 'face-trimming' method on MRI-MEG/EEG coregistration and head model creation for source localization.
• To determine if the proposed method compromises individual identification compared to standard de-facing approaches.

Main Methods:

• Development of an automated de-facing algorithm that specifically retains the nose region of structural MRI scans.
• Comparison of coregistration accuracy between MRI data processed with the new method and standard de-facing techniques.
• Evaluation of the 'face-trimming' method's effect on automated segmentation and surface extraction for head model generation.
• Behavioral assessment to quantify identification risks associated with the 'face-trimming' approach versus standard de-facing.

Main Results:

• The automated de-facing method preserving the nose demonstrated effective MRI-MEG/EEG coregistration, comparable to methods that remove the entire face.
• Behavioral data confirmed that 'face-trimming' does not increase identification risk relative to standard de-facing.
• The proposed method exhibited less impact on automated segmentation and surface extraction processes crucial for head model creation.

Conclusions:

• The novel 'face-trimming' automated de-facing method offers a viable solution for sharing structural MRI data while maintaining high-quality MEG/EEG source localization.
• This approach effectively balances the need for data privacy with the requirements for accurate neuroimaging analysis and reproducible research.
• The 'face-trimming' technique is recommended for structural MRI data intended for forward modeling in MEG/EEG source reconstruction.