Obscuring surface anatomy in volumetric imaging data.

Area of Science:

• Medical Imaging
• Radiology
• Neuroimaging

Background:

• Sharing medical images like MRI and CT scans raises patient privacy concerns due to identifying anatomical features.
• Existing de-identification methods often remove significant data, potentially impacting research utility.

Purpose of the Study:

• To develop and evaluate a novel anatomical surface modification technique for de-identifying medical images.
• To assess the impact of this surface modification on image statistics and neuroimaging processing tools.

Main Methods:

• Developed 'normalized anterior filtering,' a method that extracts, modifies (stretches, flattens, smooths), and reapplies a boundary layer of anatomical surfaces.
• Applied the method to high-resolution MR and CT scans.
• Compared the performance of common skull stripping and MR gain field correction tools on both original and de-identified data.

Main Results:

• The normalized anterior filtering method successfully obscured sensitive anatomical details.
• Initial comparisons indicate minimal impact on the output of standard neuroimaging processing tools.
• The method avoids voxel removal, preserving more original data compared to common approaches.

Conclusions:

• Normalized anterior filtering offers a promising approach for de-identifying MR and CT images while preserving data quality for research.
• This technique can enhance the secure sharing of medical imaging data.
• Further investigation into the effects on various processing tools is warranted.