How to automatically anonymise medical images?
Vilnius university hospital Santaros klinikos collectively with Ministry of Health of The Republic of Lithuania are searching for an innovative solution that would automatically remove personal data from medical research images.
- Both in Europe and in Lithuania, there is a growing need for the re-use of health data. In order for health data to be processed, it must be depersonalised, thus ensuring the protection of personal data.
- One type of health data is medical images. To prepare medical images for reuse, it is necessary to mechanically find and hide personal data in each image. Because this process is time-consuming, the practice of anonymising medical images is not widely used.
- Solutions that are already in the market can only anonymise metadata, but many medical devices store personal data at the medical image level (DICOM format) rather than in a separate metadata file. With the development of an automated medical imaging depersonalization tool, a variety of medical images could be prepared for re-use without significant time and human resource costs.
- The solution should automatically anonymise all information related to personal data contained in both the image itself and a separate metadata file, and be adaptable to different medical images generated by different devices. The solution should be easy to use without requiring specific data science competencies.
- Santaros Klinikos can share medical images with false personal data. It can be used to create the image anonymisation tool, so they are not subject to GDPR and can be used without legal restrictions. There are 2,405 such images from 7 different examination methodologies (computed tomography, electrocardiogram, X-ray radiology, etc.).
- More detailed solution requirements are provided in the description of the challenge below.
Solution evaluation criteria
- Innovativeness and relevance of the solution (40% of the total score)
- Commercial potential and competitiveness of the solution (40% of the total score)
- Competences and motivation of the authors of the solution (20% of the total score)
If you have questions about the process, participation, deadlines or you want to get an access to the data mentioned in the challenge description, contact us by e.mail – firstname.lastname@example.org