Preserving privacy in medical images while still enabling AI-driven research: A comprehensive review

Abstract

The application of AI algorithms in medical image processing requires access to a large amount of data containing protected health information (PHI). Preserving individuals’ privacy is not only an ethical issue, but it is also dictated by personal privacy laws such as HIPAA or GDPR.Health authorities and hospitals should be aware that is not possible to fully anonymize medical images without losing their research utility, meaning that some level of risk for potentially reidentifying patient information will be present in any case. On the other hand, researchers and software developers should be informed about de-identification or anonymization approaches and should consider them as part of any solution.Review papers published in this area are mainly focused on preserving privacy in structured medical data or exploring defense mechanisms and approaches against adversarial attacks. This paper takes into consideration unstructured medical data and provides an overview of the: techniques and tools adopted for medical image de-identification or anonymization; faced limitations in ensuring patient privacy; and researchers' future directions.

Ähnliche Arbeiten