Yulia Dodonova, Mikhail Belyaev, Anna Tkachev, Dmitry Petrov, Leonid Zhukov. Kernel Classification Of Connectomes Based On Earth Mover’s Distance Between Graph Spectra, in BACON: Workshop on Brain Analysis using Connectivity Networks / MICCAI 2016

Ch. 5. P. 1–10.

Dodonova Y., Belyaev M., Tkachev A., Petrov D., Zhukov L. E.

In this paper, we tackle a problem of predicting phenotypes from structural connectomes. We propose that normalized Laplacian spectra can capture structural properties of brain networks, and hence graph spectral distributions are useful for a task of connectome-based classication. We introduce a kernel that is based on earth mover's distance (EMD) between spectral distributions of brain networks. We access performance of an SVM classier with the proposed kernel for a task of classication of autism spectrum disorder versus typical development based on a publicly available dataset. Classication quality (area
under the ROC-curve) obtained with the EMD-based kernel on spectral distributions is 0.71, which is higher than that based on simpler graph embedding methods.
 

Keywords: машинное обучение, Machine learning for medicine, коннектомы, connectomes

In book

Proceedings of the 19th International Conference on Medical Image Computing and Computer Assisted Intervention, October 17-21, 2016, Athens, Greece, Springer

Athens: Springer, 2016.