Based on ImageX’s building blocks for indexing and federating access to images and reports, deepMed is Medicom’s system for collecting and de-identifying clinical information for research, machine learning, and AI applications.
Medicom complies with and de-identifies all data in accordance with §161.514(a)-(c) of the HIPAA Privacy Rule.
DICOM tags and text containing individually identifiable data are sanitized in deepMed’s automated process - fields are extracted and replaced with randomly generated characters.
Any burned in PHI are removed using a customized implementation where blackout zones are identified on an image defined by a health system user from each modality.
deepMed contains multiple safeguards to ensure that all PHI has been stripped from the DICOM header and removed from images, including:
Images that fail deepMed’s quality control mechanisms are held in Checkpoint for manual review.
Checkpoint groups images based on reason for quality control failure and another identifier (usually modality).
| De-Identification and Anonymization | |
|---|---|
| Conforms with the HIPAA Safe Harbor methodology. | x |
| De-identification is performed locally on customer device. | x |
| DICOM header sanitization. | x |
| Report sanitization. | x |
| Burned-in protected health information (PHI) sanitization. | x |
| Store de-identified data to remote or local destinations. | x |
| Supports longitudinal data collection. | x |
| Supports 3D mammography files. | x |
| Supports re-identification. | x |
| Supports inclusion of other clinical or research documents. | x |
| Integrates with databases (EHR, RIS) and DICOM archives. | x |
| Supports query construction with DICOM and SQL. | x |
| Accessible via a web interface. | x |
| Provide remote de-identification access to partners. | Pending |