yuzhounh / slic_2 Goto Github PK

SLIC: a whole brain parcellation toolbox
Copyright (C) 2016 Jing Wang

The SLIC toolbox contains five whole brain parcellation approaches that 
operates on resting-state fMRI data. Three of them are reproduced from the 
Ncut-based approaches in (Craddock et al., 2012, HBM) and (Shen et al., 
2013, Neuroimage). The remaining two are the mean SLIC and two-level SLIC 
approaches, which combine Ncut and SLIC to perform whole brain 
parcellation. By running this demo, you could reproduce the major results 
in our paper. See the Readme_plus file for further information. This 
project is also shared on NITRC, https://www.nitrc.org/projects/slic/. 

Usage:
1. Download the preprocessed fMRI data from NITRC, and then uncompress
the data. 
   https://www.nitrc.org/frs/?group_id=1030
2. Run main.m to play this demo. It takes about 10 hours on a server with
   40 CPUs and 256 GB memory. 

Changes:
1. Don't discard the eigenvectors corresponding to the trivial eigenvalues 
   (<10^-4) anymore because this step is not necessary. 
2. Set the error tolerance to 1e-3 and set the maximum iteration number to
   100 for iterations.
3. Store usefull information in sInfo.mat.

Related Codes:
SLIC, https://github.com/yuzhounh/SLIC
SLIC_2, https://github.com/yuzhounh/SLIC_2

Reference:
Jing Wang, Haixian Wang. A supervoxel-based method for groupwise whole 
brain parcellation with resting-state fMRI data. Frontiers in Human 
Neuroscience. DOI: 10.3389/fnhum.2016.00659

Contact information:
Jing Wang
[email protected]
[email protected]

2018-6-20 15:11:46