fredericklab/capcalc repository overview

⁠Capcalc

capcalc is a suite of python programs used to perform coactivation pattern analysis on time series data. It uses K-Means clustering to find a set of “activation states” that represent the covarying patterns in the data. The container contains the full package with all dependencies.

HTML documentation is here: http://capcalc.readthedocs.io/en/latest/⁠

⁠Ok, I’m sold. What’s in here?

• roidecompose - This program uses an atlas to extract timecourses from a 4D nifti file, producing a text file with the averaged timecourse from each region in the atlas (each integral value in file) in each column. This can be input to capfromtcs. There are various options for normalizing the timecourses.

• capfromtcs - This does the actual CAP calculation, performing a k-means cluster analysis on the set of timecourses to find the best representitive set of “states” in the file. Outputs the states found and the dominant state in each timepoint of the timecourse.

• maptoroi - The inverse of roidecompose. Give it a set of cluster timecourses and a template file, and it maps the values back onto the rois

• statematch - Use this for aligning two state output files. Takes two state timecourse files, and determines which states in the second correspond to which states in the first. Generates a new ‘remapped’ file with the states in the second file expressed as states in the first.