How to Read a Functional Connectivity Chart
Overview¶
Before we brainstorm using the CONN toolbox, a short demonstration of functional connectivity using a simpler, easy-to-understand method may help prepare you for using a more sophisticated bundle. I prefer to apply the fMRI analysis package FSL for this kind of demonstration, although both AFNI and SPM are able to do functional connectivity analyses as well. Before going on, I recommend that you piece of work through the FSL tutorials to familiarize yourself with how FSL works, what a time-series is, and how to extract information from regions of interest (ROIs). If you already know how to do a functional connectivity assay in FSL and simply want to get started with the CONN toolbox, click the Next button down below.
Note
This tutorial will utilize a sample dataset that you can download hither. The file size is well-nigh 20MB.
Functional Connectivity Analysis with FSL¶
As you saw in the FSL tutorial, each voxel of an fMRI dataset contains a time-series, or sequence of Assuming activation measurements that are concatenated together. The sampling rate of the fourth dimension-series is the Fourth dimension to Repetition (TR), which for most fMRI studies is effectually 2-3 seconds. For each voxel, therefore, the BOLD activation is measured at each TR, and this continues for a catamenia of time specified by the user. This catamenia of time when the scanner is measuring the BOLD indicate is called a run (or session, in SPM), and a typical experiment is composed of several runs.
To examine the fourth dimension-series of the sample dataset, navigate to where you downloaded the TimeSeries_ExampleData.nii.gz file, and type:
fsleyes TimeSeries_ExampleData . nii . gz This will open the file in the fsleyes viewer. From the card at the top of your window, click on View -> Time series . This will open a fourth dimension-series console at the bottom of the fsleyes viewer. To make the numbers easier to understand, and to enable the states to compare time-series from distant voxels, click on the "Plotting mode" dropdown menu and select "Normalised". This will scale the time-series relative to a mean of 0.
A sample time-serial from a voxel in the dorsal inductive cingulate cortex.
At present, click on a random voxel, and so from the card at the top of the screen select Tools -> Seed correlation (Pearson) . The voxel y'all take currently selected volition be a seed voxel; the time-serial of this seed voxel will be used as a reference. This will generate a connectivity map between your seed voxel and every other voxel in the brain, with hotter colors (i.e., more intense shades of cerise and yellowish) representing greater correlation. Libation colors (i.east., shades of bluish), past contrast, correspond where there is negative correlation.
You will also notice that this has created a new file in the Overlay list, TimeSeries_ExampleData/correlation . If you highlight this image by clicking on information technology, you lot tin can so set the "Min." and "Max." thresholds to only show those voxels that correlate with the seed region at or higher than a certain level. Fix the "Min" field to 0.4 , and take a await at which voxels remain after this threshold.
The functional connectivity map after thresholding at r > 0.four. Notation that the voxel nosotros used as our seed region will accept a perfect r=1.0 correlation with itself, and that neighboring voxels are besides highly correlated with it. There are a couple of smaller clusters farther away from the dACC that correlate with the seed region, simply they are relatively weak.
Notation
If you've read the Cluster Correction appendix, y'all volition call back that one of the justifications for cluster correction is that a given voxel is not completely independent from its neighbors. How practise y'all see that principle being reflected in this correlation map?
At present let'south encounter whether we can replicate one of the most well-known functional connectivity networks, the so-chosen Default Manner Network. This network is a pattern of correlated regions, primarily the ventromedial prefrontal cortex and the posterior cingulate cortex. Utilise the "Voxel location" fields to enter values of X=32, Y=53, and Z=sixteen, and use this voxel as a seed for another functional connectivity assay. What practise you notice about the resulting correlation map?
The Default Mode Network, showing high correlation between the ventromedial prefrontal cortex and posterior cingulate cortex nodes. This map can be saved as its own prototype past clicking on the deejay icon next to the correlation map that has been generated. Yous can then convert this to a z-score using Fisher's r-to-z transform, and utilise those transformed images equally input into a group-level assay. The reasons for this transform, which is done automatically in the CONN toolbox, will exist discussed in a later affiliate on 1st-level analysis.
Functional Connectivity Assay with Neurosynth¶
Another useful tool for understanding functional connectivity is Neurosynth. Neurosynth is primarily a meta-assay tool, only it can besides exist used to quickly generate functional connectivity maps for seed regions in a template brain. If you click on the "Locations" tab, a new window opens that says "Functional connectivity and coactivation maps". Click on any coordinate in the brain, so click on "What'southward here?" Afterward a few moments, a correlation map will be generated.
For example, let's say that we wanted to generate a functional connectivity map for the default way network (described to a higher place). If you click on the coordinates [0, 44, -14] and click the "What'southward here?" push button, yous should see pregnant clusters in both the region that y'all clicked, and also in the posterior cingulate cortex. These maps are generated based on a sample of a m subjects nerveless past Yeo et al. (2011), and contain the typical networks yous would see in any sample.
As an practise, effort setting the coordinates to i of the nodes of the salience network, for example - [0 26 32], which is roughly within the anterior cingulate cortex - and see whether it correlates with the bilateral insula. Try setting the seed in other regions, such as the motor cortex or auditory cortex. What other regions would you await would correlate with those seed regions? Why?
Demonstration of how to generate functional connectivity maps with Neurosynth. Note that yous can threshold the image to just show correlations higher up a certain value, and you lot tin can also download the correlation map to use as a mask.
Video¶
Click here for a brief overview of the history of functional connectivity, and how to do a unproblematic functional connectivity analysis in FSL.
Next Steps¶
Now that yous have an idea of what functional connectivity is and how nosotros generate correlation maps, y'all are prepare to begin learning virtually the CONN toolbox. Click Next to learn about how to starting time download the data nosotros will be analyzing with CONN.
How to Read a Functional Connectivity Chart
Source: https://andysbrainbook.readthedocs.io/en/latest/FunctionalConnectivity/CONN_ShortCourse/CONN_01_FSL_Demo.html