table of contents
LIPSIA     Correlation and PPI Analysis
vlcorr: Correlation analysis

A correlation analysis of fMRI time series can be performed using 'vlcorr'. The anatomical and functional data set has to be specified using the command line options '-in' and '-raw', respectively. Anatomical and raw data must match. 'vlcorr' gives correlation values for a single subject only. For statistical significance and group studies, the correlation analysis has to be performed using 'vcolorglm' (see 'vgetcovariates' section below).

vlcorr -in ana.v -raw tbsmMK3T.v

In the following example, correlation analyses was performed with reference voxel 23 59 29 (raw data voxel coordinates).


vgetcovariates: Correlation analysis

A correlation analysis can also be performed using 'vcolorglm' or 'vwhiteglm'. The appropriate design file can be created with 'vgetcovariates' using preprocessed fMRI raw data. When a special region is chosen (by specifying the coordinates), the program 'vgetcovariates' produces a simple design file with a regressor. The regressor contains the time course of the selected voxel (or the average time course of the selected voxels and neighbouring voxels, if specified).

To select the reference voxel, the program 'vlview' can be used. Make sure that the same type of coordinates are selected in 'vlview' and 'vgetcovariates' (e.g. voxel coordinates). The procedure is as follows:

vlview -in mr413_t1_pl.v -zmap zmap.v -raw tbsmMK3T.v

vgetcovariates -in tbsmMK3T.v -out design.v -addr 20 35 30

vcolorglm -in tbsmMK3T.v -out gtbsmMK3T.v -design design.v

vgetcontrast -in gtbsmMK3T.v -out cgtbsmMK3T.v -con 1 0

In our example, the result is the following:

Instead of 'vcolorglm', 'vwhiteglm' can be also used to perform a correlation analysis in the following way:

vgetcovariates -in tbsmMK3T.v -out design.v -addr 20 35 30

vwhiteglm -in tbsmMK3T.v -out ctbsmMK3T.v -design design.v -con 1 0

vgetcovariates: PPI analysis

In Lipsia, the PPI is realised as it was developed by Friston et al. (1997). The event-related PPI-approach (by Gittelman et al.) is not implemented in Lipsia.

A PPI (psychophysiological) analysis can be performed by specification of an additional psychological variable (e.g. attention). This psycological variable has to be specified in a one-column text file (e.g. with 1 for attention and -1 for no attention). The number of rows (items) must coincide with the number of timesteps of the experimental session. If no value is availabe for a timestep, the line can be filled with a capital X in the text file.

In the columns of the text file (which specifies the psychological variable), there should be always at lest 3 consecutive entries of '1' or '-1', respectively, because (in Lipsia) the PPI is working only between 'blocks' of conditions and not for short 'events'.

vgetcovariates -in tbsmMK3T.v -out design.v -addr 20 35 30 -regr psycho.txt -ppi true

Now either

vcolorglm -in tbsmMK3T.v -out gtbsmMK3T.v -design design.v
vgetcontrast -in gtbsmMK3T.v -out cgtbsmMK3T.v -con 0 0 1 0

or

vwhiteglm -in tbsmMK3T.v -out ctbsmMK3T.v -design design.v -con 0 0 1 0

'vgetcovariates' is able to generate an output as text file if '-outtype txt' is specified. This feature can be used if a design was already generated with 'vgendesign'. Then the PPI covariates must be appended to the existing design file. In this procedure, conditions and the PPI covariates can be specified in a single design matrix:

vgendesign -in design1.txt -out design1.v -tr 2 -ntimesteps 648

vgetcovariates -in tbsmMK3T.v -out design2.txt -outtype txt -addr 20 35 30 -regr psycho.txt -ppi true

vaddcovariates -in design1.v -out design_add.v -file design2.txt

Now either

vcolorglm -in tbsmMK3T.v -out gtbsmMK3T.v -design design_add.v
vgetcontrast -in gtbsmMK3T.v -out cgtbsmMK3T.v -con 0 ... 0 1 0

or

vwhiteglm -in tbsmMK3T.v -out ctbsmMK3T.v -design design_add.v -con 0 ... 0 1 0

Reference for PPI analyes: Friston KJ, Buechel C, Funk GR, Morris J, Rolls E, and Dolan RJ (1997). Psychophysiological and modulatory interactions in neuroimaging. Neuroimage 6, 218-229.

Group statistics using contrast images:
A group statistic can be performed on the basis of individual contrast images. In this case, the input images must be contrast images of individual subjects that are registered and normalized. Contrast images are obtained by applying 'vgetcontrast' or 'vwhiteglm' with the option '-type conimg'.
Problems

Correlation and PPI analysis might give useless results when the raw data of the selected voxel contain (A) drifts, (B) steps, or (C) spikes. Most of drifts can be removed using a temporal high pass filter in 'vpreprocess'. However, steps and spikes can not be removed.

Parameters of 'vgetcovariates':
-help
Prints usage information.
-in
... Input file. Default: (none)
-out
... Output file. Default: (none)
-outtype [ vista | txt ]
Type of output. Default: vista
-addr
Address of point to be processed. Required.
-type [ single | 8adj | 26adj ]
Type of region for regressor. Default: single
-system [ voxel | mm | talairach ]
Type of coordinate system. Default: voxel
-norm [ true | false ]
Normalize all regressors. Default: true
-report
Report file with voxel timecourse.
-minval
Signal threshold. Required.
-regr
Additional (psychological) regressor. Default:
-conv [ true | false ]
Convolve additional (psychological) regressor with HRF. Default: false Additional (psychological) regressor.
-ppi [ true | false ]
PPI analysis. Default: false
-orth [ true | false ]
Orthogonalize PPI interaction term. Default: true
Parameters of 'vlcorr':
-help
Prints usage information.
-in
Input file (anatomical data set). Default: (none)
-raw
Raw data file. Required.
-des
Design file. Default: (none)
-verbose
Verbose messages. Default: 0


Max Planck Institute for Human Cognitive and Brain Sciences. Further Information: lipsia@cbs.mpg.de
Copyright © 2007 Max Planck Institute for Human Cognitive and Brain Sciences. All rights reserved.