Project: Working with DICOM data on images: OCR, save, and anonymize #1694
Description
Draft Status
Ready - team will start page creating immediately
Category
DICOM
Key Investigators
- Attila Nagy (University of Szeged, Hungary)
- Gábor Fichtinger (Queen's University, Kingston, ON, Canada)
Project Description
Sometimes CT or MR scans arrive as plain images (JPEG, PNG...), and the data is overlaid on them.
The task is two-fold:
- get the data, because it may store spatial, series and patient information. This is an OCR task.
- get rid of the data, so it can be anonymized
Objective
Our goal was to create an extension prototype that can achieve the above goals, using pure Slicer and python infrastructure.
Approach and Plan
Work on prototype data, test various OCR solutions (Tesseract, EasyOCR), and implement the module with a GUI.
Progress and Next Steps
Module done.
It is able to OCR text in selected ROIs, and then blank out those regions in all of the sliced (technically images).
The modules support reviewing the OCR'ed data, saving it as JSON (and also save it in the scene MRML), and load it back.
The regions can be blanked both with pure black (0,0,0) and pure white (255,255,255) RGB values. The DICOM functionality is not yet extensively tested, and there is also a placeholder where with the use of some regex DICOM data can automatically be imported into the series.
Illustrations
Here is a short video demonstrating the current functionality:
<iframe width="420" height="315" src="https://www.youtube.com/embed/=Ao9UofN0RsY"> </iframe>Background and References
No response