NA-MIC Project Weeks
Back to Projects List
Quality Control Model for Brain Surfaces
Key Investigators
- Florian Davaux (University of North Carolina, USA)
- Juan Carlos Prieto (University of North Carolina, USA)
- Lucie Dole (University of North Carolina, USA)
- Martin Styner (University of North Carolina, USA)
Project Description
ShapeAXI is a shape analysis package that regroups many AI networks which use analysis via transformer networks or 2D convolutional neural networks. This package is available on Pypi and has been developed using Python and MONAI framework. The objective of ShapeAXI is to provide different architectures that can be used by anyone using his own data.
One of this network, called SaxiRing, has been used on the Adolescent Brain Cognitive Development (ABCD) data as a quality control (QC) model. One of the outputs of this architecture is a visual explanation from the regions of an input image that are most influential for the model’s decision.
The project would be to create the extension of this QC model and the visualization on 3D Slicer.
Objective
- Build and deploy the extension on 3D-slicer for the QC model and the visualization (GRAD-CAM)
- The end result would be to have a new 3D Slicer extension ready to be used for anyone who wants to use the QC model on his own data
Approach and Plan
- Create the extension into 3D-Slicer
- Implement the Extension Logic (organise the code, develop the Logic Module, develop the User Interface (UI))
- Integrate the QC model
- Integrate the GRAD-CAM
- Distribute the extension
Progress and Next Steps
- The model is ready (already trained)
- GRAD-CAM with MONAI is ready too
Next steps :
- Make sure that all preliminary steps have no issue
- Start creating the extension
- Thinking about the best UI to improve the accessibility
Illustrations
QC Model Results
Example of GRAD-CAM in 3D-Slicer