Chemical shift-based prospective k-space anonymization.

Area of Science:

• Medical Imaging
• Data Privacy
• Open Science

Background:

• Publicly available data is crucial for open science, but poses challenges for data privacy and protection regulations.
• Head scans are particularly sensitive due to the potential for facial reconstruction from images.
• Defacing reconstructed images is possible, but not applicable to raw k-space data.

Purpose of the Study:

• To present a novel method for prospective k-space defacing of raw head scan data.
• To address the conflict between open data requirements and data privacy regulations in neuroimaging.
• To enable the publication of anonymized raw k-space data.

Main Methods:

• Developed chemical shift-based prospective k-space anonymization (CHARISMA).
• Utilized an oil-filled mask placed on the subject's face during 7 T MRI acquisition.
• Tested CHARISMA with gradient-echo sequences, varying readout bandwidth to adjust fat shift.
• Employed intensity-based segmentation to assess retrospective unmasking potential.

Main Results:

• A fat shift of 3.3 mm to 4.9 mm demonstrated the most effective masking of subject identification.
• The CHARISMA approach significantly impaired subject identification in gradient-echo k-space data.
• Longer echo times (TEs) inherently impeded retrospective unmasking, independent of CHARISMA.

Conclusions:

• CHARISMA is the first prospective k-space defacing technique.
• This method provides a simple, cost-effective solution for enhancing data privacy in MRI.
• Further validation with additional imaging sequences is warranted.