ATTENTION/WARNING - NE PAS DÉPOSER ICI/DO NOT SUBMIT HERE

Ceci est la version de TEST de DIAL.mem. Veuillez ne pas soumettre votre mémoire sur ce site mais bien à l'URL suivante: 'https://thesis.dial.uclouvain.be'.
This is the TEST version of DIAL.mem. Please use the following URL to submit your master thesis: 'https://thesis.dial.uclouvain.be'.
 

Detection and removal of texts burned in medical images using deep learning

Szelagowski, Nicolas
(2023)

Files

Szelagowski_21661800_2023.pdf
  • Open access
  • Adobe PDF
  • 25.31 MB
Download

Details

Supervisors
Jodogne, Sébastien
Degree label
Master [120] : ingénieur civil en science des données, à finalité spécialisée
Abstract
This master thesis addresses a critical yet under-explored aspect of medical imaging: the removal of burned-in text that can bias AI algorithms and threaten patient confidentiality. While anonymization of DICOM tags is a well-established practice, the removal of patient identifiers burned into the pixels of medical images often poses a challenge due to the lack of accessible, open-source tools. To address this, this study presents a new approach involving the application of a scene text detection algorithm called TextBoxes. Using a synthetic dataset derived from The Cancer Imaging Archive, we trained different models and obtained impressive results. This achievement led to the creation of MedTextCleaner (MTC), an open-source and user-friendly plugin for Orthanc. MTC, incorporating our deep learning model, automates the de-identification process by suggesting potential text regions within an image for removal and allowing for manual user validation and adjustment. The integration of MedTextCleaner into Orthanc represents a valuable development in DICOM image anonymization, which can be especially useful for teaching and research applications.

Citations

  • APA

Szelagowski, N. (2023). Detection and removal of texts burned in medical images using deep learning. https://dial-mem.test.bib.ucl.ac.be/handle/2078.2/32872