Shape + Texture Matching in Medical Images

Five instances of Liver extracted from MRI data. The black dot indicates a pseudo key point.
Five instances of Liver extracted from MRI data. The black dot indicates a pseudo key point.

In this project, we plan to explore methods to match/align shapes extracted from biomedical images, based on shape and texture-based features. By doing so, we will be able to analyze the variations of these structures over time and compare them with other patients.

We can align two shapes by maximizing the overlap between them under a global transformation. Additionally, we can define alignment based on textural features as well. Initially, we plan to use an optimization-based approach and then use a learning-based approach (i.e. using deep learning models). We will obtain the segmentations of the structures using a pre-trained Voxel2Mesh network[1]. You can find more information about the project in this document.

Contact

Udaranga Wickramasinghe
Sina Honari

Reference

[1] U. Wickramasinghe, E. Remelli, G. Knott, P. Fua. “Voxel2Mesh: 3D Mesh Model Generation from Volumetric Data”, MICCAI 2020

Voxel2Mesh: 3D Mesh Model Generation from Volumetric Data

P. U. Wickramasinghe; E. Remelli; G. Knott; P. Fua 

23rd International Conference On Medical Image Computing & Computer Assisted Intervention

2020

International Conference On Medical Image Computing & Computer Assisted Intervention (MICCAI), Lima, Peru, 4-8 OCTOBER 2020.