NA-MIC Project Weeks

Back to Projects List

GPU Nonlinear Registration

Key Investigators

• Houssem Gueziri (Montreal Neurological Institute and Hopital, Canada)
• Mohammadreza Eskandari (Montreal Neurological Institute and Hopital, Canada)

Presenter location: In-person

Project Description

This project aims to add a multimodal nonlinear registration plugin to Slicer. This work extends the functionality of GPU Rigid Registration project (which was initially part of Ibis Neuronav). This project is based on Multi-Modal Image Registration Based on Gradient Orientations of Minimal Uncertainty.

Objective

1. Be able to nonlinearly register a source image to a target image from different modalities

Approach and Plan

1. Figure out what needs to be changed compared to rigid registration
2. Add required components
3. Test

Progress

1. Theoretical foundation of the method has been investigated
2. Previous implementation has been reviewed
3. Successfully integrated CPU implementation of non-rigid registration in Ibis

Next setps

1. Replace metric value computation with OpenCL code

Background and References

[1] De Nigris, D., et al., 2014. SymBA: Diffeomorphic Registration Based on Gradient Orientation Alignment and Boundary Proximity of Sparsely Selected Voxels. In Biomedical Image Registration: 6th International Workshop, WBIR 2014, London, UK, July 7-8, 2014. Proceedings 6 (pp. 21-30). link

• http://ibisneuronav.org
• https://github.com/IbisNeuronav/Ibis

Illustrations

Before registration:

After registration: