ENCOG™
Data Anonymization Tool To De-Identify Your Data With AI
- It is data anonymization tool powered by AI that keeps all clinically relevant information while removing and protecting Protected Health Information (PHI), auditing activities, and maintaining a chain of custody.
- PHI is removed from all attributes, including where it is burned in and private tags.
- ENCOG uses computer vision and Natural Language Processing (NLP) to differentiate PHI from relevant clinical information.
- Supported modalities include MR, CT, XR, and ultrasound.
- ENCOG maintains the historical associations of a patient study by applying a consistent shift to dates, preserves the identifiers that are needed for associating patient data, and defines redaction rules by DICOM attribute or value representation.
- Data can be re-identified using a decrypt key – this can ONLY be done where the deidentification occurred. The decrypt key is owned by the organization, ensuring security is maintained.
Anonymize MR, CT, XR & Ultrasound Images With The Power Of AI
- Names, dates, locations, & other types of Personal Health Information (PHI) can be removed from DICOM attributes (metadata) and pixel data (burn-ins)
- Where PHI elements are used for associating data to each other, identifiers can be hashed to unique IDs to preserve relationships.
- Dates or time stamps can be shifted to preserve longitudinal patient data.
- Deployed within the Enlitic framework, ENCOG will be able to query the PACS/VNA for studies to be Deidentified
- ENCOG supports the sending of deidentified studies to another DICOM source or to a file on a disk (local or network)
Removing Burned-in PHI from Medical Images
- Designed to recognize PHI overlays present in the image (Pixel) data, commonly called “burn-ins’
- Identifies overlaid text in an image (pixel data) by exploiting the consistent width of high-contrast strokes within text.
- “Content-aware” Computer Vision algorithms preserve non-PHI text that have clinical significance, such as:
- Laterality marker (‘R’ / ‘L’)
- Patient Position (‘Supine’/ ‘Erect’)
- Projection (‘AP’ / ‘PA’)
With the Combination of ENDEX and ENCOG Data Becomes Extremely Valuable
The basic components of a real-world database require data to be deidentified to protect patient data. Add standardization of different datasets from multiple sources to better create clinically relevant cohorts and the real-world database becomes highly valuable. The opportunity to monetize this deidentified, standardized data with clinical research organizations, clinical trials or research organization increases the value of your data significantly. When combining ENDEX for data standardization and ENCOG for your anonymization, you achieve clinical context not achievable with today’s naming conventions from all modalities. In addition, your get a high level of security through AI driven anonymization and consistent naming for your data – making data curation easy and fast.
