This guide provides the basic information for people who want to use the PLS Analysis for fMRI package. Basic Concepts -------------- Session Profile: contains the essential information about the fMRI experiment. The profile will be used for converting scanned volumes into an appropriate matrix format (see below Datamat section), performing PLS analysis, and displaying results. Each subject should have its own session profile in the same experiment. Information contains in the profile: - Session Description: experiment description - PLS Data Directory: the directory in which the PLS generated data files are stored. - Datamat Prefix: the prefix string to be added to the data file generated by the PLS process. It is used to distinguish the data files between different subjects and experiments. - Conditions: > the number of different conditions in the experiment. > the names of conditions. - Run Informations: > the total number of runs in the experiment. > number of scans for each runs > data directory for the run > data files for the run > event onsets of each conditions for the run ** Onsets are specified in terms of scan number, which starting from 0. If a onset is located between two scans, decimal number can be used for the onset. However, it will be converted to the closest smaller integer number. Datamat: data matrix that stores the scanned volumes. Data are rearranged in a 2D matrix for PLS analysis. Datamat is created by choosing "Create ST Datamat" option from the "Datamat" menu item in the Session Profile window. There are two types of data matrix generated by the process: 1) volume datamat - (a.k.a. datamat) The scanned volume are stacked together to form a 2D matrix. Each row of the matrix contains 1 scanned volume, and the columns are the voxels of the volumes. To save space, only the brain voxels will be stored in the matrix. The non-brain voxels are defined as the voxels that are below certain thresholds, which is specified during datamat creation. The datamat are stored in the file with name pattern of '???_datamat.mat', where '???' is the datamat prefix stored in the session profile. Each run has one datamat associated with it. 2) spatial/temporal datamat - (a.k.a. st_datmat) Created from the volume datamats. For each event onset in a run, several consecutive rows of the datamat, that corresponds to the scanned volumes for the event, are stringed together to form a vector. The average of the vectors of the same condition is used to create a row in the st_datamat. Depending on the choice of the user, the average vector may be computed from the vectors across all runs or within a single run. If single subject analysis is performed, "average within a single run" should be used. The number of rows to be used to generate a row of st_datamat is same as the WINDOW SIZE specified by the user during datamat creation. The st_datamat are stored in the file with name pattern of '???_st_datamat.mat', where ??? is the datamat prefix stored in the session profile. Volume datamats will be removed after the st_datamat is created. Each session profile has one st_datamat associated with it. ------------------------------------------------------------------------ PLS Analysis: specify options to run the analysis. Session Profiles - specify the data to be analyzed. Several st_datamats can be combined together by choosing the corresponding profiles in the same group. The rows of the st_datamats in the same group are averaged to form a single row of the group st_datamat to represent the same condition. Thus, there will be total of CxG rows of in the for the group st_datamat, where C is the number of conditions, and G is the number of groups. For the single subject data analysis, which only allow one gorup, no averaging is performed, all st_datamats are stacked together to form a group st_datamat. Check the "Group Analysis" box to perform group analysis, i.e. averaging all data within groups. Contrasts - specify the contrast to be used. None (Grand Mean Deviation) - st_datamat for each group are subtracted from its group mean for each condition. The PLS results is very similar to that of using Helmert matrix. Helmert Matrix - use Helmert Matrix as the contrasts. Using contrast file - use user specify contrast, which can created by using "Contrast" menu option of the PLS Analysis window. Number of Permutation - specify the number of permutation to be performed. Number of Bootstrap - specify the number of bootstrap to be performed. For group analysis, there must be at least 3 subjects in each group. For nongroup analysis, there must be at least 3 subjects (or repetitions) for each conditions. ------------------------------------------------------------------------ PLS Result: display the result of PLS analysis. If bootstrap has been performed, the bootstrap ratios are displayed on the window. Otherwise, the brain LV values are shown. To switch between BrainLV and Bootstrap Ratio display, use the "View" menu option. Adjust LV Index: LV index - specify which latent variable to be shown Adjust dispaly slices: First Slice - the first slice to be shown Step - the step between two display slices Last Slice - the maximum last slice to be shown Adjust dispaly color: Threshold Value/Ratio - the cut off threshold, any absolute values/ratios below the threshold are shown in grey color. Min Value/Ratio - the mininum value/rato to be displayed, any value/ratio below the min. value will be shown in the color of the mininum value, i.e. dark blue. Max Value/Ratio - the maximum value/rato to be displayed, any value/ratio above the max. value will be shown in the color of the maximum value, i.e. saturated red. ** After adjust any of the above value, press "PLOT" button to update the display. Voxel Location: XYZ - the absolute voxel coordinate of the picked point. XYZmm - the voxel coordinate (in mm) of the picked point, the coordinates are specified relative to the origin of the first volume data. ------------------------------------------------------------------------ Note - The program assumes the volume data that are processed together have been transformed to the same volume space. Therefore, any coregistration and normalization process should be done before using the PLS analysis. (Third party software, such as SPM and AIR can be used) The reporting voxel locations are using the convention same as the SPM, which is left is Left, right is Right, top is Anterior, bottom is Posterior, slice number are running from Inferior to Superior. If the data is normalized against the SPM template, the reporting location is identical to that of the SPM. ------------------------------------------------------------------------ TIPS - Input Onset: Onsets can be inputed by calling a MATLAB statement. If the first character of the onset edit line is '@', the rest of the line will be interpreted as a MATLAB command which return a vector for the onsets. Before the line being executed as MATLAB statement, any words of '{run#}' and '{cond#}' in the statement will be replaced by the current run number and condition number, respectively. To repeat the last MATLAB statement, put '@@' in the onset edit line. Filename convention: There are no restriction on the file name for the session profile, contrast, and result. However, since the program assumes the following file name pattern, PLSsession_*.mat - for session profile PLScontrast_*.mat - for design contrast PLSresult_*.mat - for pls result PLScluster_*.mat - for active cluster report It will make your life easier if you follow the above convention. ------------------------------------------------------------------------