help > RE: whole brain voxel2voxel connectivity matrix?
Dec 10, 2016  02:12 AM | Alfonso Nieto-Castanon - Boston University
RE: whole brain voxel2voxel connectivity matrix?
Hi Eran&Pravesh,

This is a good suggestion but keep in mind that when using 2mm^3 voxels/ROIs you are going to end up with something on the order of 160000 ROIs (1.6e5), which also means that ROI-to-ROI correlation matrices are going to have somewhere around 25600000000 (2.56e10) elements, which can take something around 100Gb just to store/represent. A somewhat less demanding approach is to use CONN's implicit representation of voxel-to-voxel correlation matrices in terms of their eigenvalues/eigenvectors, which is what CONN's voxel-to-voxel analyses use (see CONN's 2012 manuscript and http://www.nitrc.org/forum/message.php?m... for a few more details)

Hope this helps
Alfonso
Originally posted by Pravesh Parekh:
Hi Eran,

I believe what you want to do is a kind of ROI-to-ROI analyses where your ROIs are actually individual voxels? What you can do is locate the buildP.m file (conn/utils/otherrois). Edit the line of SideLenghts = [10 12 14 16 18 20] to 2 (assuming that your voxel dimensions are 2x2x2). When you run the script (its going to take a while), you will notice a new ROI file P1844.nii (or some such number).

You can use this file as ROI file for the analysis (it will assume that each "parcel" is an ROI; because the parcel size is equal to voxel size, you are essentially doing a ROI-to-ROI kind of analysis on voxel level).


Hope this helps


Best
Pravesh

Threaded View

TitleAuthorDate
Eran Dayan Dec 9, 2016
Pravesh Parekh Dec 9, 2016
RE: whole brain voxel2voxel connectivity matrix?
Alfonso Nieto-Castanon Dec 10, 2016
Eran Dayan Dec 9, 2016