An Eigendecomposition Approach to Weighted Graph Matching Problems
IEEE Transactions on Pattern Analysis and Machine Intelligence
Feature-based correspondence: an eigenvector approach
Image and Vision Computing - Special issue: BMVC 1991
Robust Point Matching for Nonrigid Shapes by Preserving Local Neighborhood Structures
IEEE Transactions on Pattern Analysis and Machine Intelligence
Robust 3D Shape Correspondence in the Spectral Domain
SMI '06 Proceedings of the IEEE International Conference on Shape Modeling and Applications 2006
Global medical shape analysis using the Laplace-Beltrami spectrum
MICCAI'07 Proceedings of the 10th international conference on Medical image computing and computer-assisted intervention - Volume Part I
International Journal of Computer Vision
Discrete Calculus: Applied Analysis on Graphs for Computational Science
Discrete Calculus: Applied Analysis on Graphs for Computational Science
Point Set Registration: Coherent Point Drift
IEEE Transactions on Pattern Analysis and Machine Intelligence
Spectral demons --- image registration via global spectral correspondence
ECCV'12 Proceedings of the 12th European conference on Computer Vision - Volume Part II
Diffeomorphic spectral matching of cortical surfaces
IPMI'13 Proceedings of the 23rd international conference on Information Processing in Medical Imaging
Spectral Log-Demons: Diffeomorphic Image Registration with Very Large Deformations
International Journal of Computer Vision
| Hi-index | 0.00 |
Brain matching is an important problem in neuroimaging studies. Current surface-based methods for cortex matching and atlasing, although quite accurate, can require long computation times. Here we propose an approach based on spectral correspondence, where spectra of graphs derived from the surface model meshes are matched. Cerebral cortex matching problems can thus benefit from the tremendous speed advantage of spectral methods, which are able to calculate a cortex matching in seconds rather than hours. Moreover, spectral methods are extended in order to use additional information that can improve matching. Additional information, such as sulcal depth, surface curvature, and cortical thickness can be represented in a flexible way into graph node weights (rather than only into graph edge weights) and as extra embedded coordinates. In control experiments, cortex matching becomes almost perfect when using additional information. With real data from 12 subjects, the results of 288 correspondence maps are 88% equivalent to (and strongly correlated with) the correspondences computed with FreeSurfer, a leading computational tool used for cerebral cortex matching. Our fast and flexible spectral correspondence method could open new possibilities for brain studies that involve different types of information and that were previously limited by the computational burden.