NA-MIC Project Weeks
Back to Projects List
Defacing head and neck CT scans while preserving lymph nodes
Key Investigators
- Steve Pieper (Isomics, Inc., USA)
- Ron Kikinis (BWH, USA)
Project Description
For the Lymph Node Quantification project we want to analyze and share head and neck scans but don’t want to share PHI.
Objective
- Use generic segmentation tools (CADS, SuperSynth, etc.)
- Identify facial feature landmarks. Find existing models or examples that identify features like corners of eyes and mouth
- Figure out which features can be reliably detected that define a reasonable identity removal but leaving the nodes in the cheeks and jaw
- Decide what bluring or removal to do in the identified face area
Approach and Plan
- Figure out if existing MR defacing tools can be used for this, perhaps by mapping the CT to MR
- Explore the state of the art in Facial animation systems
- A more recent paper referencing the Milchenko paper Deidentification of CT head images
Progress and Next Steps
Mockups
Area to be defaced
- Closed curve snapped to model surface
- Dynamic modeling applied to suface wrap solidfy model
- Facial blur
- Extra blur
Open Questions
- What are the state-of-the-art definitions for facial deidentitication and how do they intersect with our requirements?
- What primitives exist in the Slicer ecosystem that can be used as building blocks, such as landmark detection, image manipulation, and segmentation to make a robust pipeline.
- Can we use exising infrastructure or is it better to build something custom.
Illustrations
We will use the CTHead sample data for experiments.