SDM-PSI tutorial version Jan 2019

SDM-PSI tutorial, version Jan 2019

By Joaquim Radua and Anton Albajes-Eizagirre

The aim of this tutorial is to show how to conduct a meta-analysis using SDM-PSI software. To this end,
you will perform some of the analyses conducted in https://doi.org/10.1192/bjp.bp.108.055046.

Note however that these analyses will be conducted with the updated, SDM-PSI algorithms described in
https://doi.org/10.1016/j.neuroimage.2018.10.077, based on the previous works in http://dx.doi.org/10.1016/j.eurpsy.2011.04.001,
https://doi.org/10.3389/fpsyt.2014.00013 and https://10.42.0.1/10.1016/j.neuroimage.2018.04.065.

We distribute this tutorial in the hope that it will be useful, but without any warranty on the accuracy of
the text or the data.

Before executing the software

We have invested a lot of time and effort to improve the accuracy of the SDM methods and software.
However, it may yield biased estimations if you not consider the following inclusion / exclusion criterion
for peaks when conducting the searches and contacts with the authors:

“While different studies may employ different thresholds, you should ensure that within one
study, the same threshold was used throughout the whole brain”

This is of utmost importance because it is not rare in neuroimaging studies that some regions (e.g., a
priori regions of interest) are more liberally thresholded than the rest of the brain.

Page 1 of 18
SDM-PSI tutorial version Jan 2019

Preparation of the files

SDM allows the combination of statistical maps (in NIfTI format, obtained from e.g., SPM or FSL
software) and peak information (e.g., reported in the papers). For this tutorial, we will only use peak
information, and for your convenience, the text files with this peak information have been already
prepared in the folder containing this PDF.

Look at the names and contents of these text files. Coordinates and t-values of the peaks are written in a
separate text file for each study, and the filename is just a very short identification of the study (e.g., the
name of the first author), plus a dot, plus an identification of the software used in the study and
stereotactic space of the peak coordinates, plus a dot, plus “txt”.

Possible identifications of the software and stereotactic space are:

 *.spm_mni: for SPM studies, reported in MNI space. Very common!

 *.fsl_mni: for FSL studies, reported MNI space. Very common!
 *.other_mni: for studies that used other software, reported MNI space.
 *.spm_brett: for SPM studies, with Brett conversions. Old studies!
 *.fsl_brett: for FSL studies, with Brett conversions.
 *.other_brett: for studies that used other software, with Brett conversions
 *.spm_tal: for SPM studies, reported in Talairach space.
 *.fsl_tal: for FSL studies, reported in Talairach space.
 *.other_tal: for studies that used other software, reported in Talairach space.
 *.no_peaks: whenever there are no peaks, independently of the software and stereotactic space.

These are some of the sample text files:

Carmona.spm_mni.txt
40,39,21,-5.14
53,27,21,-3.77
56,23,20,-3.63
...

Christian.no_peaks.txt

Each line specifies a coordinate and its t statistic. The coordinate is defined by the first three values
(e.g., “40, 39, 21”), and the t statistic by the forth value (e.g., “-5.14”). The extension of the first file is
*.spm_mni.txt, for what these coordinates are understood to be in SPM’s MNI space. The extension of
the second file is *.no_peaks.txt because that study reported no peaks.

Page 2 of 18
SDM-PSI tutorial version Jan 2019

The t statistic should be:

A positive number for: A negative number for:

One-sample fMRI studies: task > baseline task < baseline

(activations) (deactivations)

Two-sample fMRI studies: patients > controls in task > baseline patients < controls in task > baseline
(hyper-activations) (hypo-activations)

patients < controls in task < baseline patients > controls in task < baseline
(failures of deactivation) (hyper-deactivations)

Two-sample VBM / FA studies: patients > controls patients < controls

(increases of volume / FA) (decreases of volume / FA)

In a real meta-analysis, you would have to read carefully the original papers of the studies, and
sometimes you might notice that authors report z scores instead of t statistics. You may straightforwardly
convert z scores into t statistics using the online converter at https://www.sdmproject.com/utilities/?show=Statistics (you
may access this website pressing the [Convert peaks] button within the SDM software).

In case of studies not reporting any measure related to effect size (t statistic, z score, p value, etcetera),
you should write a “p” for positive peaks and an “n” for negative peaks. The SDM software conducts a
pre-analysis to provide an effect size for these peaks.

IMPORTANT: Statistical maps in a standard stereotactic are preferred to any peak information file. If
you are able to obtain these images, use the [Convert images] button within the SDM software to
prepare them for the analysis.

Page 3 of 18
SDM-PSI tutorial version Jan 2019

Preparation of the SDM software

In this step, you should first specify a working folder for the meta-analysis, and afterwards create an
SDM table specifying at least the names of the studies and their sample sizes. We prepared the latter for
you in this tutorial.

IMPORTANT: To prevent errors, run SDM software from a local disk (rather than from a network drive).

 Start the graphical user interface (GUI) SDM software:

 Linux users: to start the software click a file named “SdmPsiGui” / “SdmPsiGui.desktop” in the
SDM software folder. If the program does not execute follow, the instructions to change file
permissions at https://www.sdmproject.com/software/?show=Linux

 Mac OSX users: to start the software click a file named "SdmPsiGui" / “SdmPsiGui.app”. If the
program does not execute follow, the instructions to change file permissions at
https://www.sdmproject.com/software/?show=Mac

 Windows users: to start the software click a file named “SdmPsiGui” / “SdmPsiGui.exe” in the
SDM software folder. If the program does not execute follow, the instructions to change file
permissions at https://www.sdmproject.com/software/?show=Windows

 To specify the working folder for the meta-analysis, click the [Change meta-analysis] button, look for
the “home/tutorial” folder containing this PDF (within the SDM software folder), and click [Choose].

A dialog similar to the following one should appear:

Page 4 of 18
SDM-PSI tutorial version Jan 2019

 If SDM software did not found MRIcron, you can manually specify its location with the following
steps: open the [Tools] menu, select [Preferences], select the [Brain viewer] tab, and change the
[Brain viewer executable] (look for the MRIcron program, typically something like “C:\Program
Files\MRIcron\mricron.exe” in Windows). Then click [Open] and [OK].

A dialog similar to the following one should appear:

 To create or edit the SDM table, click the button [SDM table editor] button.

A window similar to the following one should appear:

Page 5 of 18
SDM-PSI tutorial version Jan 2019

Each row in the SDM table specifies one study. In this example, the first column (“study”) sets the
identification of the study (the same than in the text files). The second and third columns specify the size
of the patients’ sample (“n1”) and of the controls’ sample (“n2”). The fourth column (“t_thr”) specifies the
t-value threshold of statistical significance used in each study, which you might sometimes find in the
manuscript or its figures (e.g. something such as "t > 4.3"). If you are unsure, a conservative option
might be typing the t-value corresponding to p=0.001 uncorrected (about 3.1, larger in smaller studies).
However, if authors applied cluster-based statistics, we suggest using the t-value threshold used to
create clusters, which may be <3.1.

The 5th-8th columns specify optional global gray matter values, and the 9th-10th columns optional
variables. You could also add a special optional column, called “threshold”, to specify the threshold type
(e.g., “uncorrected” vs “corrected”) used in each study. In case of meta-analyses that only involve
healthy controls, the second column (“n1”) should be the size of the samples, and you should not call
any column “n2”.

How to do it using the terminal

To change the directory, use the command “cd”. E.g., to follow this tutorial and assuming that your
SDM software folder is within your Documents folder, you should type something similar to:

cd Documents/sdm/home/tutorial (Linux or Mac OSX)

cd Documents\sdm\home\tutorial (Windows)

Afterwards you should create or modify a text file named “sdm_table.txt” with the information of the
SDM table using any text or spreadsheet editor (the file has been already created for this tutorial, but
please open and inspect it). To open a text editor from the terminal you may type something similar to:

gedit sdm_table.txt (Linux)

Note: you may use any other text editor included in your
distribution, e.g., “kwrite”, “mousepad” and etcetera.

open –a TextEdit sdm_table.txt (Mac OsX)

Note: if the file “sdm_table.txt” doesn’t exist, you can create
it typing:
echo "" > sdm_table.txt

notepad.exe sdm_table.txt (Windows)

To minimize the risk of errors when SDM reads the file “sdm_table.txt”, do not type any character other
than numbers and simple letters (apart from the tabs to separate the columns).

Page 6 of 18
SDM-PSI tutorial version Jan 2019

“Globals” analysis

Prior to the voxel-based meta-analysis, you may want to conduct an analysis of the global gray matter
volumes. To this end, the following variables must be defined in the SDM table: “mean1” and “mean2”
(global gray matter means), and “sd1” and “sd2” (global gray matter standard deviations).

 To conduct the “globals” analysis, click the button [Globals], and click [OK].

A dialog similar to the following one should appear:

This will create and automatically open a web-like file named “globals_MyGlobals.htm” with standard
meta-analytic measures for global gray matter. The most important measures are the mean (Hedge’s g
along with its corresponding Z and P values and the confidence interval) and the analysis of
heterogeneity (τ and its corresponding Q and P values).

Page 7 of 18
SDM-PSI tutorial version Jan 2019

Pre-processing

In this step, SDM will use the peaks’ text files to voxelwise recreate the lower and upper bounds of the
possible effect-size values of the studies.

 To pre-process the peaks, click the button [Preprocessing], select the [VBM - gray matter] modality,
and click [OK].

A dialog similar to the following one should appear:

This will create a folder named “pp” that will contain:

 A web-like file named “pp.htm”, which the software will open automatically. You should check that
the absolute maximum and minimum peaks reported in the summary of “pp.htm” roughly
correspond to those reported in the original manuscripts. Pay special attention to check the side
(left vs. right).

 Four “*.nii.gz” brain image files for each study: two for the lower and upper bounds of the possible
effect size and two for the lower and upper bound of the possible t-values.

It will also create the following system files:

 An image with the meta-analytic mask (“sdm_mask.nii.gz”).

 A file for study settings named “sdmpsi_params.xml”, which may be useful for modifying some
parameters when working from the console. Please be aware that the integrity of this xml file is

Page 8 of 18
SDM-PSI tutorial version Jan 2019

required to SDM-PSI to be able to execute normally. Therefore, we strongly suggest the user to
manipulate this file with the utmost care.

 A file stamping the preprocessing step as done ("sdm_maps.txt"). In absence of this file, the GUI
will require the user to perform the preprocessing step before running Mean or Linear Model
steps.

How to do it using the terminal

To pre-process using the gray matter correlation template, with full anisotropy (1.0) and 20mm FWHM,
within the gray matter mask and using a voxel size of 2mm, type the following:

sdm pp gray_matter,1.0,20,gray_matter,2

Important: here and in any subsequent call to SDM software you have to replace sdm by the path of
the SDM software, e.g.

~/Documents/sdm/sdm pp gray_matter,1.0,20,gray_matter,2 (Linux)

~/Documents/sdm/sdm_mac pp gray_matter,1.0,20,gray_matter,2 (OSX)

C:\Users\john\Documents\sdm\sdm.bat pp gray_matter,1.0,20,gray_matter,2 (Windows)

Page 9 of 18
SDM-PSI tutorial version Jan 2019

Mean analysis

Now it is time to conduct the mean analysis, which is usually the main (but not the only!) outcome of a
meta-analysis. In this tutorial, the mean analysis represents the weighted mean difference in regional
grey matter between patients with OCD and healthy controls.

 To conduct the mean analysis, click the button [Mean], specify a name for this analysis (we will call it
“MyMean”) and click [OK].

A dialog similar to the following one should appear:

This will create a folder named “analysis_MyMean” with a folder for the multiple imputations (“mi”), a
folder for the beta coefficients ("betaMaps"), the mean map (“_g”) with its variance (“_var”) and z-value
(“_z”), and the between-study heterogeneity maps (τ2, H2, I2 and Q test). It will also create a folder
named "log" with internal files describing which kernel has been applied to each of the voxels in the
imputations process.

 To correct for multiple comparisons, click the button [FWE correction], specify the number of CPU
threads to use, and click [OK].

We strongly suggest you to increase the number of CPU threads to use to a value close to the maximum
available on your machine (you can see this maximum value at the "preferences" dialog of SdmPsiGui).
However, one way or another, it will take a lot of time!

Do anything else while the software works…

Take also a coffee…

Page 10 of 18
SDM-PSI tutorial version Jan 2019

After a long, long time, it will finished. It will have created a folder for the distribution of the maximum
statistics (“fwe”) and the maps of corrected p-values (“corrp_*”).

 To threshold and see the results, click the button [Threshold], select the map of TFCE-corrected
values and click [OK].

A dialog similar to the following one should appear:

This will create and automatically open a web-like file named something like
“MyMean_z_p0.05000_10.htm” with several statistics, coordinates and brain regional breakdowns, and it
will start the MRIcron program to inspect visually them. In addition, it will create the following images:

 MyMean_z_p0.05000_10.nii.gz (positive statistically significant differences)

 MyMean_z_p0.05000_10_p.nii.gz (p-values of the positive differences)

 MyMean_z_p0.05000_10_neg.nii.gz (negative statistically significant differences).

 MyMean_z_p0.05000_10_neg_p.nii.gz (p-values of the negative differences).

 A folder with masks for each blob and peak.

How to do it using the terminal

To calculate the mean and threshold it using a p-value = 0.05, and 10 voxels extent, type the following
(at this point you may want to very carefully modify the file sdmpsi_params.xml and raise the number of
CPU threads to use, variable):

sdm MyMean=mi 50
sdm perm 1000,MyMean
sdm threshold analysis_MyMean/corrp_tfce,analysis_MyMean/MyMean_z,0.05,10

Page 11 of 18
SDM-PSI tutorial version Jan 2019

Assessment of heterogeneity and potential publication bias

We strongly recommend extracting values from relevant peaks, inspecting the corresponding I2 statistics
(or other heterogeneity estimates) and check their funnel plots. You may also use extracted values to
create meta-regression plots with Microsoft Excel, R or similar software.

You should first create a mask that includes the voxel or region from where you want to extract the
values, and then extract these values using the mask. Fortunately, the “Thresholding” automatically
creates the masks for the peaks.

 To create the mask, click the [Create a mask] button, select [MNI coordinate], click [OK], type the
coordinate (X = -24, Y = 10, Z = -2), and click [OK].

A dialog similar to the following one should appear:

This will create a file named “analysis_MyMean/masks/mask_-24_10_-2.nii.gz” which contains the mask.
Note that you can copy this file to the folder of another meta-analysis in order to avoid creating it again.

Page 12 of 18
SDM-PSI tutorial version Jan 2019

 To extract the values using this mask, click the [Extract] button, select model "MyMean", select “-
24_10_-2”, and click [OK].

A dialog similar to the following one should appear:

This will create and automatically open a web-like file named “analysis_MyMean/extracts/extract_-
24_10_-2.htm” with the gray matter values of each map in this voxel. In addition, it will create a simple
text file for programming purposes.

How to do it using the terminal

To create the mask and extract the values, type the following:

sdm -24_10_-2 = mask MyMean, coordinate, -24, 10, -2

sdm extract MyMean,-24_10_-2

To create a mask of an atlas’ structure, please use the codes at the end of this tutorial (which could
change in future versions of SDM!). E.g., to create a mask of the left anterior commissure and name it
“ac”, type the following:

sdm ac = mask MyMean, label, 117

Page 13 of 18
SDM-PSI tutorial version Jan 2019

Subgroup analysis of adult samples

This is similar to the mean analysis, with the exception that you will specify the “adults” filter in order that
only studies with adult samples are included in the analysis.

 To conduct the subgroup analysis, click the button [Mean], specify a name for this analysis (we will
call it “myAdults”), select the “adults” filter, and click [OK].

How to do it using the terminal

To calculate the subgroup mean and threshold it, type the following:

sdm myAdults = mi 50,,adults

sdm perm 1000,myAdults
sdm threshold analysis_myAdults/corrp_tfce,analysis_myAdults/myAdults_z,0.05, 10

Page 14 of 18
SDM-PSI tutorial version Jan 2019

Meta-regression by YBOCS

The last analysis of this tutorial will be a meta-regression of voxel values across the studies by the
YBOCS of the corresponding patients’ samples.

 To conduct the regression analysis, click the button [Linear model], specify a name for this analysis
(e.g., “ybocs”), select “YBOCS” as the first variable, set its contrast value to “1” and click [OK].

A dialog similar to the following one should appear:

This will create a new folder named "analysis_ybocs/" for the meta-regression by the YBOCS
variable. Run then the FWE correction and threshold the resulting maps.

How to do it using the terminal

To calculate the regression and threshold its output maps, type the following:

sdm ybocs = lm_mi YBOCS,0+1,50

sdm perm 1000,ybocs
sdm threshold analysis_ybocs/corrp_tfce,analysis_ybocs/ybocs_z,0.05,10

Page 15 of 18
SDM-PSI tutorial version Jan 2019

Atlas codes (subject to change)

Anterior commissure 117 Left fronto-marginal tract 156

Cerebellum, vermic lobule I / II 109 Left fusiform gyrus 55
Cerebellum, vermic lobule III 110 Left gyrus rectus 27
Cerebellum, vermic lobule IV / V 111 Left hand inferior U tract 158
Cerebellum, vermic lobule VI 112 Left hand middle U tract 160
Cerebellum, vermic lobule VII 113 Left hand superior U tract 162
Cerebellum, vermic lobule VIII 114 Left heschl gyrus 79
Cerebellum, vermic lobule IX 115 Left hippocampus 37
Cerebellum, vermic lobule X 116 Left inferior frontal gyrus, opercular part 11
Corpus callosum 132 Left inferior frontal gyrus, orbital part 15
Left amygdala 41 Left inferior frontal gyrus, triangular part 13
Left angular gyrus 65 Left inferior network, inferior fronto-occipital fasciculus 164
Left anterior cingulate / paracingulate gyri 31 Left inferior network, inferior longitudinal fasciculus 166
Left anterior thalamic projections 118 Left inferior network, uncinate fasciculus 129
Left arcuate network, anterior segment 120 Left inferior occipital gyrus 53
Left arcuate network, long segment 122 Left inferior parietal (excluding supramarginal and angular)
61
gyri
Left arcuate network, posterior segment 124 Left inferior temporal gyrus 89
Left calcarine fissure / surrounding cortex 43 Left insula 29
Left caudate nucleus 71 Left lenticular nucleus, pallidum 75
Left cerebellum, crus I 91 Left lenticular nucleus, putamen 73
Left cerebellum, crus II 93 Left lingual gyrus 47
Left cerebellum, hemispheric lobule III 95 Left median cingulate / paracingulate gyri 33
Left cerebellum, hemispheric lobule IV / V 97 Left median network, cingulum 127
Left cerebellum, hemispheric lobule VI 99 Left middle frontal gyrus 7
Left cerebellum, hemispheric lobule VIIB 101 Left middle frontal gyrus, orbital part 9
Left cerebellum, hemispheric lobule VIII 103 Left middle occipital gyrus 51
Left cerebellum, hemispheric lobule IX 105 Left middle temporal gyrus 85
Left cerebellum, hemispheric lobule X 107 Left olfactory cortex 21
Left cortico-spinal projections 133 Left optic radiations 168
Left cuneus cortex 45 Left paracentral lobule 69
Left face U tract 135 Left paracentral U tract 170
Left frontal aslant tract 137 Left parahippocampal gyrus 39
Left frontal inferior longitudinal fasciculus 140 Left pons 172
Left frontal orbito-polar tract 142 Left postcentral gyrus 57
Left frontal superior longitudinal 144 Left posterior cingulate gyrus 35
Left fronto-insular tract 1 146 Left precentral gyrus 1
Left fronto-insular tract 2 148 Left precuneus 67
Left fronto-insular tract 3 150 Left rolandic operculum 17
Left fronto-insular tract 4 152 Left striatum 126
Left fronto-insular tract 5 154 Left superior frontal gyrus, dorsolateral 3

Page 16 of 18
SDM-PSI tutorial version Jan 2019

Left superior frontal gyrus, medial 23 Right fronto-insular tract 3 151

Left superior frontal gyrus, medial orbital 25 Right fronto-insular tract 4 153
Left superior frontal gyrus, orbital part 5 Right fronto-insular tract 5 155
Left superior longitudinal fasciculus I 174 Right fronto-marginal tract 157
Left superior longitudinal fasciculus II 176 Right fusiform gyrus 56
Left superior longitudinal fasciculus III 178 Right gyrus rectus 28
Left superior occipital gyrus 49 Right hand inferior U tract 159
Left superior parietal gyrus 59 Right hand middle U tract 161
Left superior temporal gyrus 81 Right hand superior U tract 163
Left supplementary motor area 19 Right heschl gyrus 80
Left supramarginal gyrus 63 Right hippocampus 38
Left temporal pole, middle temporal gyrus 87 Right inferior frontal gyrus, opercular part 12
Left temporal pole, superior temporal gyrus 83 Right inferior frontal gyrus, orbital part 16
Left thalamus 77 Right inferior frontal gyrus, triangular part 14
Middle cerebellar peduncles 139 Right inferior network, inferior fronto-occipital fasciculus 165
Right amygdala 42 Right inferior network, inferior longitudinal fasciculus 167
Right angular gyrus 66 Right inferior network, uncinate fasciculus 131
Right anterior cingulate / paracingulate gyri 32 Right inferior occipital gyrus 54
Right anterior thalamic projections 119 Right inferior parietal (excluding supramarginal and
62
angular) gyri
Right arcuate network, anterior segment 121 Right inferior temporal gyrus 90
Right arcuate network, long segment 123 Right insula 30
Right arcuate network, posterior segment 125 Right lenticular nucleus, pallidum 76
Right calcarine fissure / surrounding cortex 44 Right lenticular nucleus, putamen 74
Right caudate nucleus 72 Right lingual gyrus 48
Right cerebellum, crus I 92 Right median cingulate / paracingulate gyri 34
Right cerebellum, crus II 94 Right median network, cingulum 130
Right cerebellum, hemispheric lobule III 96 Right middle frontal gyrus 8
Right cerebellum, hemispheric lobule IV / V 98 Right middle frontal gyrus, orbital part 10
Right cerebellum, hemispheric lobule VI 100 Right middle occipital gyrus 52
Right cerebellum, hemispheric lobule VIIB 102 Right middle temporal gyrus 86
Right cerebellum, hemispheric lobule VIII 104 Right olfactory cortex 22
Right cerebellum, hemispheric lobule IX 106 Right optic radiations 169
Right cerebellum, hemispheric lobule X 108 Right paracentral lobule 70
Right cortico-spinal projections 134 Right paracentral U tract 171
Right cuneus cortex 46 Right parahippocampal gyrus 40
Right face U tract 136 Right pons 173
Right frontal aslant tract 138 Right postcentral gyrus 58
Right frontal inferior longitudinal fasciculus 141 Right posterior cingulate gyrus 36
Right frontal orbito-polar tract 143 Right precentral gyrus 2
Right frontal superior longitudinal 145 Right precuneus 68
Right fronto-insular tract 1 147 Right rolandic operculum 18
Right fronto-insular tract 2 149 Right striatum 128

Page 17 of 18
SDM-PSI tutorial version Jan 2019

Right superior frontal gyrus, dorsolateral 4 Right superior parietal gyrus 60

Right superior frontal gyrus, medial 24 Right superior temporal gyrus 82
Right superior frontal gyrus, medial orbital 26 Right supplementary motor area 20
Right superior frontal gyrus, orbital part 6 Right supramarginal gyrus 64
Right superior longitudinal fasciculus I 175 Right temporal pole, middle temporal gyrus 88
Right superior longitudinal fasciculus II 177 Right temporal pole, superior temporal gyrus 84
Right superior longitudinal fasciculus III 179 Right thalamus 78
Right superior occipital gyrus 50

Page 18 of 18