Remove burned-in PHI from DICOM images.
Ultrasound, screen captures and older scanners often draw the patient name, ID or date into the pixels. dcmage runs OCR on the image in your browser or the Mac app, proposes redaction boxes, and writes a new file with those regions set to zero.
Find text burned into the pixels and black it out.
Metadata rules cannot reach a name or record number that the scanner drew into the image itself. OCR proposes a box around each line of text, you review and add boxes, and the export writes zeros into every frame of the file.
- On-device OCR Tesseract runs locally, with its model served by dcmage rather than a third-party CDN.
- Manual regions draw rectangles with the Redact tool for text OCR misses.
- Every frame regions are applied to all frames of a multi-frame file, after you confirm.
- Derived copy the export is a new file; the source file is not modified.
How it works
- 01
Open the file in Anon mode
Load the DICOM file and switch to Anon. Choose the confidentiality profile for the header at the same time.
- 02
Detect text
Click Detect text. OCR reads the current frame and adds a proposed region for each line of text it finds.
- 03
Review and add regions
Delete regions you do not want and draw extra boxes with the Redact tool for anything OCR missed.
- 04
Apply and export
Click Apply & Export. The regions are set to zero in every frame of a new file, and the original stays unchanged.
Checks that still need judgment
- OCR is a proposal, not a guarantee. It reads English text on the current frame only, so check every frame and draw boxes for text it missed, including small, faint or rotated text.
- Compressed pixel data cannot be redacted. Decompress the series first from the QA tab.
- The Burned In Annotation attribute (0028,0301) is not updated. Set it in the tag editor if the receiving system relies on it.