A Two-Stage De-Identification Process for Privacy-Preserving Medical Image Analysis

Area of Science:

• Medical Imaging
• Data Security
• Health Informatics

Background:

• Medical imaging data, particularly DICOM files, are vulnerable to identification and re-identification threats.
• Patient privacy is paramount, necessitating effective de-identification of Personally Identifiable Information (PII).
• Existing de-identification methods for DICOM attributes lack detailed guidance on attribute removal considerations.

Purpose of the Study:

• To address the challenges in medical image de-identification.
• To develop and present a systematic, two-stage de-identification process for DICOM CT scan images.
• To propose future directions for semi-automated or automated DICOM de-identification tools.

Main Methods:

• A two-stage de-identification process was developed for DICOM CT scan images.
• Stage one involves removing PII at the hospital facility via Picture Archiving and Communication System (PACS) export.
• Stage two utilizes a proposed DICOM de-identification tool for attribute-level PII investigation and removal.

Main Results:

• A comprehensive two-stage de-identification methodology for DICOM CT scans was successfully developed.
• The process ensures exhaustive removal of PII through attribute-level analysis.
• A roadmap for future development of automated de-identification tools was outlined.

Conclusions:

• The proposed two-stage process enhances the security and privacy of DICOM medical imaging data.
• Systematic de-identification is crucial for protecting patient PII in medical datasets.
• Further development towards automated tools is recommended for efficient and reliable DICOM de-identification.