Home of the FAME RSA Analysis for the PSU CAN Lab
rootdir = /gpfs/group/nad12/default/nad12/FAME8/RSA
Subdirectory holding preprocessing scripts. In order to run our multivariate analyses, we need to grab the unsmoothed, non-normalized functional images.
wildcard_preprocess.m = batch preprocessing script from KyleSPMToolbox. The scripts reads in the raw data that @kkurkela transfered over to his scratch directory and runs a full preprocessing pipeline on it.
wildcard_parameters12.m = the parameters used for preprocessing. Realignment (SPM12 defaults) --> Slicetiming --> Coregistration (SPM12 defaults) --> Segmentation (SPM12 defaults) --> Normalization (SPM12 defaults) --> Smoothing (6mm FWHM).
Subdirectory holding general linear modeling (glm) scripts. In order to run our multivariate analyses, we need to run Least Square- All (LSA) single trial models in order to get an estimate of the BOLD activation pattern elicited for each trial at encoding and retreival.
SpecifyRetrievalModel.m = a heavily modified version of the SpecifyModel.m script from KyleSPMToolbox, this scripts directly creates a LSA single trial model for Retrieval. Each trial is input as a seperate trial type, given a descriptive name detailing all of the information for that trial. For example:
imagename-backpack(1)_visualcategory-backpack_response-remember_trialtype-target_enctype-scrambled
Given these trial type names, we should be able to determine which trial type this trial would fall into (A RecHit backpack from the scrambled condition).
SpecifyEncodingModel.m = a heavily modified version of the SpecifyModel.m script from KyleSPMToolbox, this scripts directly creates a LSA single trial model for Encoding. Each trial is input as a seperate trial type, given a descriptive name detailing all of the information for that trial. For example:
imagename-backpack(1)_visualcategory-backpack_response-pleasent_gistPosition-1_enctype-scrambled
Given these trial type names, we should be able to determine which trial type this trial would fall into (e.g., a backpack from the scrambled condition that was presented first and was labeled as pleasent).
SpecifyGistEncModel2.m = a heavily modified version of the SpecifyModel.m script from KyleSPMToolbox, this scripts created a special "gist" model at encoding. The gist model consists of a trial type for each visualcategory, in order to estimate the "gist" neural pattern from encoding for, for example, backpacks. We are left with 93 regressors; once for each visual category in the experiment.
EstimateModel.m = a slighly modified version of the EstimateModel.m script from KyleSPMToolbox, this script takes the multiple condition files from the previous three scripts and estimates them.
Subdirectory holding the Encoding-Retrieval Similarity (ERS) scripts. These scripts run the actual multivariate analyses we are interested in.
run_ers_searchlight.m = CoSMoMVPA based script for running an "Item-Level" ERS searchlight analysis on the FAME data. The script reads in the spm_T*.nii images from each subject's Encoding and Retrieval LSA glms and calculates the mean of the correlation values on the diagnol of the corrlation matrix created between each trial's encoding and retrieval pattern for a specified trial type. What results is a whole brain *.nii file for each of (RecHits, FamHits, Miss) for each subject. These *.nii files are then submitted to a second level, random effects analysis.
run_ers_gist_searchlight.m = CoSMoMVPA based script for running a "Gist-Level" ERS searchlight analysis on the FAME data. The script reads in the spm_T*.nii images from each subject's Retrieval LSA glm and Encoding "gist" glm and calculates the mean correlation between Retrieval trials and their corresponding encoding "gist" category. What results is a whole brain *.nii file for each of (RecHits, FamHits, Miss, RecRelFA, FamRelFA, RelCR) for each subject. These *.nii files are then submitted to a second level, random effects analysis.
run_ers_global_searchlight.m = CoSMoMVPA based script for running an "Global-Level" ERS searchlight analysis on the FAME data. The script reads in the spm_T*.nii images from each subject's Encoding and Retrieval LSA glms and calculates the mean of the correlation values of each Retrieval trial and all Encoding trials. What results is a whole brain *.nii file for every possible trial type for each subject. These *.nii files are then submitted to a second level, random effects analysis.
run_ers_roi.m = an older script, designed to run the "Item-level" ERS searchlight on a specified ROI (instead of a full-brain searchlight).
kyles_cosmo_ers_measure.m and correlation_summary_measure.m = "measure" functions @kkurkela wrote, based heavily on the default measure functions in the CoSMoMVPA software package. These functions calculate a specified mean. See the "measure" function description on the CoSMoMVPA website.