The overall goal of this procedure is to measure connectivity between brain regions based on changes in their distributed activity patterns over time. This is accomplished by first selecting a seed region. Next multi voxel pattern, discriminability is calculated for each time point for the seed region and each searchlight or region of interest.
The final step is to correlate the discriminability time series of the seed region with each searchlight. Ultimately, informational connectivity is used to reveal connectivity between regions that have information contained across patterns of voxels. Well, the main advantage of this technique over others like functional connectivity is that it can detect connectivity using conditions that don't differ in immune response, but do have a unique multivariate signature.
Well, I first had the idea of this method after realizing that many connectivity techniques are ignoring information that's only held in multi voxel activity patterns. Just as we now know that the univariate approaches are completely blind to information held in multi voxel activity patterns To begin this procedure, remove the motion and mean white matter signals from the time series of pre-processed FMRI data by creating a regression model with predictors for motion parameters and mean white matter signal. Then conduct the following analysis on the resulting residuals.
Import the generated residuals into an analysis package such as MATLAB or Python, using, for example, the open source informational connectivity MATLAB toolbox. After that, transform each voxels time series to zco. Separate the datasets time points into independent sets, such as different scanner runs.
Then create a record of the condition labels associated with time points by generating a vector of condition labels. That is end time points long shift the condition labels forward in each run by a number of times to repetition equivalent to five seconds in order to account for the hemodynamic lag between events and recorded FMRI signals. To select a seed region, isolate an anatomical area through selecting voxels in your preferred software or with a labeled atlas functionally localizing a region or selecting a top performing information brain mapping searchlight.
Then compare the MVP of each time. Point to a prototypical MVP for every condition. Using the open source informational connectivity MATLAB toolbox.
Calculate a prototypical MVP for every condition by averaging the time points of each condition in all but one fold. Next, correlate every time points MVP with the mean MVP of each condition from the training data. This will give every time 0.1 correlation value for each condition.
Note that the condition with the highest correlation here would be the prediction of the popular correlation based MVPA classifier.Afterward. Quantify MVP DISCRIMINABILITY for each time point by first identifying the correlation that represents that time point's condition. Then subtract the highest of the remaining correlations.
The result is each time point's MVP discriminability. To conduct a search light analysis, place a three dimensional cluster around each vle. In turn, repeat the previous step for each search light so that every search light has a time series of MVP discriminability values.
Now correlate the seed's MVP Discriminability time series with each searchlight's DISCRIMINABILITY time series. Using spearman's rank correlation, the resulting RS value is the IC between the seed and the searchlight. Then assign each searchlight's IC value to the searchlight's central voxel and write out the resulting individual's brain map.
If the data are not already in standardized space, transform the participant's IC maps into the same space. Optionally smooth the individual's searchlight maps. This figure shows the examples of MVP DISCRIMINABILITY in two regions with synchronous MVP Discriminability without synchronous mean activation.
These data points were collected while the subject viewed visual presentations of manmade objects, which are distinguishable by multi voxel patterns, but not mean responses. And shown here is an example of the IC and FC values between a seed in the left views of form gyrus and searchlights across the brain. The informational and functional connectivity strengths are indicated on the Z axis with respect to each searchlight's mean response on the X axis and MVPA classification accuracy on the Y axis, the top left corner of the IC graph shows informationally connected search lights with high classification performance, but low mean response levels.
The gap in the top left corner of the FC graph shows that these connected search lights are not being detected in a typical FC approach. Here is an example of the connectivity maps. Each row shows the regions that are significantly connected to a seed Significance is determined by a Group T test with minimum cluster size from permutation testing.
The IC results are displayed here using a acne on surface maps produced with free surfer Following this procedure. Other analyses like functional connectivity can be performed in order to ask additional questions such as whether conditions are uniquely connected by multi voxel patterns or if they also are synchronous in their mean responses. So after watching this video, you should have a good understanding of how to measure connectivity between regions based on their multi patterns.
Thanks for watching.
