Privacy-Preserving Visualization of Brain Functional Connectivity
Ye Tao1, Anand D Sarwate1, Sandeep Panta2
1Department of Electrical and Computer Engineering at Rutgers, The State University of New Jersey, Piscataway, NJ 08854.
Biorxiv : the Preprint Server for Biology
|October 28, 2024
Summary
Differential privacy protects sensitive neuroimaging data in visualizations. New methods maintain visual quality while ensuring robust privacy for biomedical data analysis.
Area of Science:
- Neuroscience
- Computer Science
- Data Visualization
Background:
- Visualizations of sensitive biomedical data, like neuroimaging, risk exposing personal information.
- Differential privacy offers a robust framework for protecting individual data privacy.
Purpose of the Study:
- To investigate privacy-preserving visualization techniques for neuroimaging data using differential privacy.
- To develop and evaluate methods that balance privacy guarantees with visual utility.
Main Methods:
- Applied differential privacy by perturbing correlation values in neuroimaging data.
- Developed specific workflows for connectogram and seed-based connectivity visualizations.
- Analyzed privacy cost and the effects of pre- and post-processing steps.
Main Results:
- Proposed workflows successfully generated visualizations comparable to non-private versions.
- Qualitative assessments of visualizations were preserved under differential privacy.
- Demonstrated a viable privacy/visual utility tradeoff for neuroimaging data.
Conclusions:
- Differential privacy is a promising approach for securing sensitive information in biomedical data visualizations.
- The developed methods effectively protect privacy without significantly compromising visual interpretability.
