Home

PaMiNI

1. se1 OTOC IE 39 Anuepl soJnby JENSIA bioqu121s eul eseq se1 OTOZ IE 39 Anuepl soJnby JENSIA bioqu121s eul eseq se1 OTOZ IE 39 Ajauepl soJnby JENSIA bioqu121s eul eseq se1 OTOZ IE 39 SHOP seine yeni ord uJ91S PPS A sm OTOC E 3 eway ld VLLC ft Excel file valid for the use with PaMiNI Required fields are typeset in bold letters ICroso Exemplary snippet from a M Figure 2 e The 9th column of the first row of an experiment indicates the reference space of the coordinates Use Tal for Talairach and MNT for MNI space If coordinates are in Talairach space PaMiNI automatically converts them to MNI space e From the 10th line of the first row of an experiment the paradigms of the exper iment can be specified Every paradigm can stand in a single column Following columns are read in by PaMiNI as paradigms until an empty column is reached 5 3 mat files of meta analysis data mat files that can be used as input files for PaMiNI have to be in an appropriate format which is commonly used for meta analyses in the Research Center J lich Specifically this format requires The mat file has to contain a 1 x n struct array labeled Experiments where n specifies the number of experiments Every column of the struct array Experiments has to contain another struct array cont
2. 2002 Rama et al 2001 Axmacher et al 2009 Relander et al 2009 Number of activating experiments a ie gt Number of activated clusters Figure 3 The PaMiNI components Pattern Calculation Settings panel In this panel the number of components chosen in the k Opt Selection dialogue is shown and values for the pattern cal culation can be specified This is best done before starting the whole procedure with the Start button but settings can be changed afterwards as well The checkbox Exclude Clusters with STD can be checked if you wish to exclude waste clusters from pattern calculation that have a greater standard deviation than the value specified in the text field This value can be a floating number between 0 and 200 The text field Cluster Membership Probability specifies the probability of an experiment focus belonging to a cluster so the higher you chose this value the closer the foci have to be located to the center of a cluster for being classified to this cluster As it specifies a probability this value has to be a floating number between 0 and 1 The last textfield in this panel is the Minimum Sup port which indicates the minimum number of experiments containing a specific pattern to take this pattern into account The minimum support has to be an integer value between 1 and 1 000 Reset k Opt button After performing the full procedure of Gaussi
3. PaMiNI can basically handle small datasets and even quite big ones For more reliable results the number of experiments should not be far below 40 this also depends on the number of reported foci per experiment For datasets greater than 1 000 the performance of pattern investigation can decrease depending on your system already with lower number of experiments It is recommended to set the minimum support to a higher value when using large datasets or low cluster membership thresholds You can also set the number of repetitions and k Max The number of repetitions indicates how often the mixture modeling is repeated for the same number of compo nents to avoid the modeling stuck in local maxima This value affects the duration of the modeling process If you only want a quick overview of the patterns in your data 12 choose a number of repetitions between 5 and 100 For more reliable results a number of repetitions between 500 and 5 000 should be sufficient depending on the complexity of the underlying dataset A convenient value for k Max strongly depends on the selected dataset This value can dramatically affect the duration of the modeling process since the mixture modeling for a high number of components needs much more time for com putation than the modeling for a low number of clusters 20 can be a good default value for most applications You can see if your choice was sufficient from the graph in the k Opt Select window The
4. 0 5 FWHM each in one file no matter if clusters were excluded by their standard deviation in the Pattern Calculation panel 2 If you want a volume containing all clus ters except for the ones excluded by standard deviation use the Extract Pattern Volume button 14 on the standard pattern in the first row of the Selected Pattern listbox 8 Additionally one single volume file is extracted for every cluster which is not FWHM thresholded These files are labeled with name trunk n where n specifies the number of the cluster Notice that pressing this button can possibly occupy much space on your hard disk Pattern Viewer In the Pattern Viewer a selected pattern is visualized It con tains a cross section viewer with windows for all three dimensions The cross section viewer displays the MNI single subject brain with the colored Gaussian blobs as overlays which indicate the active clusters of the selected pattern To 10 navigate through the brain click the left mouse button in one of the three win dows keep it pressed and drag it to scroll through the two dimensions illustrated in the two other windows The current plane of the two other windows is indicated by the crosshair in the window you clicked on You can also set the crosshair by left clicking on a specific position in the brain The second component of the Pattern Viewer is the Centers of Gravity table which represents the center coordinates or
5. BIC curve should show a pit in the middle and an ascent for high numbers of clusters If the BIC does not ascend for high numbers of clusters you should restart the mixture calculation using a higher k Max 7 2 Setting up for Pattern calculation Since the Gaussian mixture modeling tries to accomodate all foci reported in the exper iments there will usually always be waste clusters which have a big expansion but less neurobiological relevance Therefore it is normally useful to check the Exclude Clus ters with STD check box to disregard these clusters for the calculation of frequent patterns Neurobiologically meaningful values for the respective standard deviation are difficult to determine but values between 15 and 30 mm should be quite reasonable From experience the results with values between 20 and 25 were quite good depending on the used data set The Cluster Membership Probability indicates the minimum level for the probability of finding a Gaussian mixture component given a specific focus This is a measure for how rigidly the classification of foci to the clusters is done Empirically it is best kept at 0 95 which also conforms with the usual values for significance testing If you want to investigate the patterns with a more liberal classification of the foci broader cluster boundaries decrease the value if you want to have a more rigorous classification narrow cluster boundaries increase the value closer to
6. choice of interesting patterns and collected them in the Selected Pattens listbox use the Pattern Information panel 17 to get further in formation on the underlying experiments of the patterns You can also get even more information on a specific experiment by clicking on it in order to open the Experiment Information window see figure 5 You should also use the Print Selected Patterns button 13 to receive txt file of the patterns you are interested in and their corre sponding information If you want to receive volume data of a pattern or of all clusters found in the analysis in the NIfTI format for further external usage use the buttons Extract Pattern Volume 14 and Extract All Cluster Volumes 15 respectively 7 6 Words of warning It is important to correctly interpret the results produced by PaMiNI That is it is necessary to take the neurobiological context into account when investigating the results PaMiNI is a Data Mining tool or more specifically a Pattern Mining tool which means that it should extract possibly meaningful information out of large datasets to aid an observer by getting insights The insight has to be derived by the user and it cannot come from the program since it operates only on the data without having information of the neurobiological background So do not expect a result given by PaMiNI to be a fundamental statement in brain research but let PaMiNI support you
7. initialized by pressing the Start button for information on the k Opt Selection window see below 2 18 1 54 f 9 10 8 PaMiNI by Julian Caspers CT File bout m Total Experiments Input File C Program Files MATLABIR2008a work PaMiNl m Total Activations 2662 Repetitions 1000 K Max 20 Start ctedk Opt 16 V Exclude Clusters with STD 20 Minimum Support Caic Patterns Patterns gt 5 LES Pattdrns of Interest ATOT TT m 0 Ennty Listas Listbox 0101101110001111__ 0 0101101111000111__ 0100 0101101111001011__ E 010p rint Selected Patterns 9 gt E 10100 0100100011000001 ___ 0001101111001110 3 3 0100100111001111 3 0101100111001101 C Batract Pattern Volume w N 1 4 Cluster Membership Probability 0 95 01011011010000100 7 0001100011001111 0100101111001100 RD 0000100011000000 0100100011000101 6 Pattern Information Pattern 0100100011000001 Number of Clusters Number of Experiments Activating Experiments n Veltman et al 2003 Stern et al 2000 Veltman et al 2005 Mainero et al 2004 Kumari et al 2006 Meisenzahl et al 2006 IRicciardi et al 2006 Sanchez carrion et al 2008 McNab et al 2008 Elzinga et al 2007 Cerasa et al 2008 Forn et al 2007 Garavan et al 2000 Cross et al 2007 Ragland et al 2002 Linden et al 2003 Mayer et al 2007 Pessoa et al
8. least one pattern but not all patterns of a multiple pattern scatter for example by removing one pattern from the listbox with button 10 or by adding a single pattern from the Interesting Patterns listbox with button 9 a left click on the specific pattern will fill up the Selected Patterns listbox with the remaining patterns If all patterns of a scatter are already in the listbox a left click on this scatter will remove all patterns from the listbox and the scatter color will be switched back to black When the Show Related Patterns button 11 is pressed the related patterns of the pattern selected in the listbox are indicated in the scatter plot by red scatters If scatter is selected and related its color will be magenta Interesting Patterns listbox After pattern calculation this listbox provides a list of potentially interesting patterns sorted by their interestingness Each row first represents the binary vector of the pattern then its number of activated clusters and its number of supporting experiments at the end The measure of interestingness depends on the number of activated clusters of the pattern the number of experiments that contain the pattern and its closedness This is the difference between the number of supporting experiments of the specific pattern and the support of its best supported super pattern that entirely contains the spe cific pattern A high degree of closedness indicates that a patt
9. more specifically the means of the Gaussian distributions for each cluster with their z y and z component The fourth column indicates the blob color which is used for the overlay in the cross section viewer for the specific cluster Rows of clusters that are not active in the shown pattern remain empty If you click on a row of the table the planes and crosshairs of the cross section viewer will change to the selected center coordinate To visualize a pattern you can either make a selection in the Selected Pattern listbox 8 or perform a right click on a scatter in the Pattern Distribution scatter plot 6 17 Pattern Information panel In this panel you can get information on the pat tern selected in the Selected Patterns listbox 8 This includes the pattern vector itself the number of clusters activated by the pattern the number of ex periments that contain the pattern and a list of these experiments that support the pattern You can click on an experiment in this list and a window appears that gives you more information on the selected experiment see Figure 5 This Experiment Information window contains the author of the experiment the year of publication the paradigms used for the experiment the number of subjects and a list of all reported peak foci The fourth column of this list indicates to which cluster of the analysis the peak coordinate is assigned 18 File menu The File
10. scatter in the Pattern Distribution scatter plot 6 in blue If the pattern is already in the Selected Patterns listbox a message dialogue will appear indicating that the pattern has already been added and the specific pattern will be selected and visualized in the Selected Patterns listbox Remove Selected Pattern button This button removes the pattern from the Selected Patterns listbox 8 that is selected After removing the selection of the listbox will be set to the first pattern Since the standard pattern is always kept in the listbox an error message dialogue will appear if you try to remove the first pattern If the removed pattern is the only pattern represented by one scatter in the Pattern Distribution scatter plot 6 or the last one in the listbox of multiple patterns represented by one scatter the specific scatter color will be turned back to black If there are still patterns of a multiple pattern scatter in the listbox the scatter color remains blue Show Related Patterns button When pressing this button all related pat terns of the pattern selected in the Selected Patterns listbox 8 are shown in the Pattern distribution scatter plot 6 by turning the color of the respective scatters to red Related means that a pattern is either a sub pattern that is a pattern that s entirely included in the selected pattern or a super pattern that is a pattern
11. stored e The first row of an experiment has to contain the authors name and the year of publication in the first column Best use the format Author yyyy e The second column of the first row of an experiment has to contain the number of subjects Eq J9A Ajlquap SPIOM Asoypne c Tuosueduio2 Eq J9A Aylquap SPpJOM Asoypne c Tuosueduuioo2 JEquaA Ajauepl SPIOM Aoyipne c Tuosueduuio2 Eq J9A Aauepl SPIOM Aoyipne c Tuosueduio2 JEq4aA Ajauepl SPIOM Aoyipne c Tuosueduio2 JEq4aA Ajauepl SPIOM Aojipne c Tuosueduuio2 Eq J9A Ajauepl SPIOM Aojipne c Tuosueduio2 0 3002 lt Se 013U09 lt Se 013U09 lt Se 043U09 lt Se 013U09 lt JS 013U09 lt Se 013U09 lt JS 6002 e 32 19puejoy Z6ZZ 6002 IE 32 Jepue ed T6ZZ 6002 Ie 32 Jepue es 0672 600Z e 32 Jepue ed 68ZZ 6002 IE 32 Jepue ed 88Z2 600Z Ie 32 Japuejay Z8Z2 600Z Ie 32 Jepue eu 98Z2 1eq1 A A11u9p SpJOM Asoypne ZTuosiedwos 013U09 lt ySe 600 19 Jopuelsy 584Z Anuepl soJnby JENSIA bioqu121s eul eseq se1 OTOZ IE 39 Anuepl soJnby JENSIA bioqu121s eul eseq se1 OTOZ IE 39 Aiquept soJnby JENSIA bioqu121s eul eseq se1 OTOC IE 39 Anuepl soJnby JENSIA bioqu121s eul eseq se1 OTOZ IE 39 Aiquept soJnby JENSIA bioqu121s eul eseq se1 OTOZ IE 39 Anuepl soJnby JENSIA D4equae3S eul eseq
12. 1 The Minimum Support value which specifies the minimum number of experiments that have to contain a pattern to further take this pattern into account can be set to 1 or 2 for small datasets up to sets with approximately 500 experiments depending on the performance of your system since a small minimum support does not affect or distort the investigation of the patterns Patterns that are only represented in one experiment are normally not of great interest and the minimum support can be set to 2 without loss of information For larger datasets the minimum support should be adapted since higher values accelerate the calculation of frequent patterns and improve the performence when investigating the patterns because of a decreased number of found frequent patterns 0 5 to 2 of the total number of experiments included in the dataset are a good estimate for the minimum support to ensure the least information is lost Note that the Cluster Membership Probability can also affect the performance of your system since low threshold values lead to a higher number of assigned coordinates and hence a higher number of possibly frequent patterns Therefore you should also adapt the Minimum Support for lower membership thresholds 13 1 3 The PaMiNI calculation procedure After clicking the Start button the mixture modeling procedure will begin As already mentioned this can take some time especially for high numbers of
13. PaMiNI User manual Julian Caspers December 20 2011 Contents 1 Introduction 1 2 System requirements 1 3 Installation 2 4 Getting started 2 5 Input file requirements 3 5 1 txt files from Sleuth 3 5 2 xls files of meta analysis data 0 000 eee ee 3 5 3 mat files of meta analysis data a eee eee 5 6 Components of PaMiNI 5 7 How to use PaMiNI 11 7 1 Setting up for Gaussian mixture modeling 12 7 2 Setting up for Pattern calculation 13 7 3 The PaMiNI calculation procedure cnn 14 T Change settings s ob dox oe de dei a de Momo muB OW Soe GHD 14 7 0 Investigate the patterns 14 7 6 Words of warning 15 1 Introduction PaMiNI which stands for Pattern Mining in NeuroImaging is MATLAB applica tion for finding frequent neural patterns in sets of neuroimaging experiments It uses a Gaussian mixture modeling approach to classify the experiments coordinates into clus ters and detects frequent combinations of these clusters in the given experiments The frequent patterns can be observed in PaMiNI with a built in interactive cross section viewer on the MNI single subject brain Information to the underlying experiments for each pattern are provided in the application including name and year of the study the used paradigms participating subjects and
14. aining the following variables e A string variable labeled Author containing the author and the year Best use the format Author yyyy e A 1x p cell array labeled Cond with the paradigms as string variables in each column p indicates the number of specified conditions e A numeric variable 1 x 1 Subjects which indicates the number of subjects e A 3 x c numeric matrix labeled XYZmm which contains the reported foci as MNI coordinates The first second and third row respectively contain the z y and z coordinates of the c specified foci in each column Notice that all coordinates have to be specified in MNI space A conversion from Talairach to MNI space will not be performed by PaMiNI 6 Components of PaMiNI The components of PaMiNI are shwon in figure 3 These are 1 General Settings panel These are the first settings that have to be made when the program is started The input file can be specified using the button next to the text field labeled which opens a file selection dialogue Furthermore the number of repetitions for the Gaussian mixture modeling as well as the number of clusters upto which the modeling should be performed k Max can be specified Repititions has to be an integer value between 1 and 1 000 000 k Max has to be an integer between 1 and 100 When everything is set right the mixture modeling procedure and the following steps of pattern calculation can be
15. an mixture modeling and pattern calculation where you have to chose an optimal number of clusters you can reset the number of clusters k Opt for which you wish to investigate the pattern distribution After pressing the button a window will ap pear where you can select k Opt see figure 4 This window will also appear after pressing the Start button when the Gaussian mixture modeling is finished since the Start button induces the complete process of mixture modeling and pattern calculation The window shows a graph that plots the Bayesian Information Cri terion BIC against the number of clusters of the different solutions The BIC gives you a hint what could be the optimal number of clusters The lower the BIC the better the Gaussian distributions were fitted to the data The lowest BIC value will be marked in the graph by a red circle and the corresponding number of clusters will initially be selected in the drop down box Even if the BIC might give a good suggestion for k Opt you can select any number of clusters in the drop down box from 1 upto the k Max you specified in the General Settings panel 1 If you made your choice just press the OK button or close the window and PaMiNI will proceed with the pattern calculation The currently selected k Opt can be seen in the pattern calculation panel 2 just above the Reset k Opt button E Select number of clusters x 10 79 Figure 4 The k Opt Sele
16. clusters A waitbar shows the calculation progress and indicates for which number of clusters the Gaussian distributions are currently modeled If you try to cancel the calculation you have to wait until the calculation of the current number of clusters is finished then the procedure will terminate After the Gaussian mixture modeling is finished the k Opt Selection Window ap pears see figure 4 Choose your desired number of clusters and click OK so that this the application calculates the overlays for the cross section viewer This can take some time as well The following calculation of the cluster vectors should proceed much faster The progression of both of these calculations is also visualized in a waitbar and can be termintated in the same way as the calculation of the Gaussian mixtures The last step of the calculation process is the computation of frequent patterns This can take much time for large datasets and the progress is not displayed you only see the sandglass mouse icon The duration of pattern calculation can be decreased by incrementing the Minimum Support value in the Pattern Calculation panel 2 So if you are using large datasets remember to set the minimum support to a proper value to avoid long waiting time 7 4 Change settings You can change the settings of the Gaussian mixture modeling and the pattern calcula tion at any time The calculation procedure will begin at different points depen
17. ction window 4 Calc Patterns button After performing the full procedure of Gaussian mixture modeling and pattern calculation you can recalculate the pattern distributions with new values specified in the pattern calculation panel 2 by pressing this button Dataset Information These two labels give you information about the dataset which was read in for the Gaussian mixture calculation Le it specifies the number of experiments in the dataset and the total number of activation foci Pattern Distribution scatter plot This diagram indicates the distribution of the patterns It illustrates the number of supporting experiments that contain a pattern on the y axis against the number of activated clusters of this pattern on the x axis Every scatter in the plot indicates at least one pattern If two pat terns have the same number of active clusters and the same number of supporting experiments they are represented by the same scatter If you only want to investigate a pattern in the Pattern Viewer 16 right click on the specific scatter If the scatter represents multiple patterns you can switch between the visualization of these patterns by multiple right clicks You can add the pattern s represented by a scatter to the Selected Patterns listbox 8 by simply left clicking on the specific scatter If at least one pattern of a scatter is in this listbox the scatter color will be blue If the listbox contains at
18. ding on what you want to change e If you want to use a new input file or want to set new values for the number of repetitions or k Max you have to restart the full calculation procedure by clicking the Start button in panel 1 e If you want to investigate the patterns for a different number of clusters but on the same Gaussian mixtures just click the Reset k Opt button 3 PaMiNI will recalculate the overlays for the cross section viewer the cluster vectors and the patterns The patterns will be calculated using the values currently specified in the Pattern Calculation panel 2 e If you want to recalculate the patterns with different values in the Pattern Cal culation panel 2 but with the same number of clusters on the same Gaussian mixtures simply click the Calc Patterns button 4 This will only re initiate the frequent pattern calculation 1 5 Investigate the patterns After the calculation procedure is completed you can investigate the patterns The standard vector will already be added to the Selected Patterns listbox 8 and visu 14 alized in the Pattern Viewer 16 Use the Pattern Viewer to get an overview of all clusters found in the data for the specified k Opt Then check out the Interesting Patterns listbox 7 and add patterns you are interested in to the Selected Patterns listbox by using the Add Patterns button 9 with the label Investigate t
19. ern is stable and that the number of supporting experiments radically decreases when any component is added to the pattern If you want to change the specification of interestingness you can do this by editing the m file evalInterestingPatterns m where the function is implemented and which you find in the PaMiNI directory 10 11 12 Selected Patterns listbox This listbox represents the patterns that were se lected in the Pattern Distribution scatter plot 6 or by adding patterns from the Interesting Patterns listbox by pressing button 9 The patterns are repre sented by their binary vector representation where every digit represents a cluster A 1 indicates that this cluster is active in the pattern while a 0 indicates that it is not active The first row always represents the pattern where all clusters are active except for the ones that where excluded by their standard deviation in the Pattern Calculation panel 2 The excluded clusters are indicated by an x in the vector If you click on a pattern in the listbox it will be visualized in the Pat tern Viewer 16 and information to the pattern and its underlying experiments are given in the Pattern Information panel 17 Add Interesting Pattern to Listbox button This button will add the pattern that is selected in the Interesting Patterns listbox 7 to the Selected Patterns listbox 8 and further mark the respective
20. hese patterns in the Pattern Viewer too You can also add patterns to the listbox from the Pattern Distribution scatter plot 6 by left clicking on a scatter that might be interesting for you Notice that this can add multiple patterns to the Selected Patterns listbox You can remove single patterns of the listbox by selecting them and pressing the Remove Pattern button 10 If the list in the listbox gets too confusing use the Empty Listbox button 12 to remove all patterns from the listbox except for the standard pattern A really usefull feature to find the most relevant patterns and to evaluate the relevance of a pattern is the Show Related button 11 Select a pattern in the Selected Patterns listbox and press this button to indicate all sub patterns and super patterns of the selected pattern in the Pattern distribution scatter plot The related patterns are indicated in red If a related pattern is already in the Selected Patterns listbox its scatter color will be magenta To investigate which pattern hides behind a scatter use a right click on this scatter to visualize the underlying pattern in the Pattern Viewer If the scatter represents multiple patterns you can switch between the visualization of these patterns by right clicking on the scatter multiple times If you found an interesting pattern in the related patterns just add it to the listbox by left clicking the scatter If you have found the right
21. menu contains three items Save Load and Exit After a complete processing of Gaussian mixture modeling and pattern calculation you can save your workspace in a mat file using the Save menu item If you click on it a file choose dialogue opens to specify the mat file where your workspace should be stored You can use this feature to reinvestigate your results in a later moment in time or to exchange these files with other users to communicate your results Notice that the saving of a workspace might occupy much space on your hard disk To load a saved workspace simply click the Load menu item at any time and select a former saved mat file in the opening file choose dialogue You can only load mat files which contain a saved workspace otherwise an error message dialogue will appear The last menu item in the File menu is the Exit item which closes PaMiNI How to use PaMiNI This section will give an overview on how you can use PaMiNI 11 E Experiment Info Author Z emus et al Year 2007 Paradigm task gt control n back visual letter identity verbal access verification Subjects Activations Fa Cluster 50 16 Figure 5 The Experiment Information window 7 1 Setting up for Gaussian mixture modeling As already described in the Getting started section 4 simply add a valid input file see section 5 containing a dataset of experiments by clicking the button
22. rent Directory to the PaMiNI directory Then type PaMiNI into the Command Window of MATLAB Now the graphical user interface of PaMiNI should open which should look roughly like Figure 1 B Paint by Julian Caspers File About General Settings Input File Repetitions 100 K Max m 25 Sat Pattern Settings Selected k Opt Total Experiments Total Activations V Exclude Clusters with STD Cluster Membership Probability 20 Minimum Support 0 95 1 Number of activating experiments Number of activated clusters Patterns of Interest Interesting Patterns Selected Patterns Pattern Information Pattern Number of Clusters Number of Experiments Activating Experiments Figure 1 The PaMiNI graphical user interface immediately after opening For a quick start you can now select a valid input file of the formats txt mat or xls see requirements below by clicking the button labeled with next to the Input File text field Then just press the Start button and PaMiNI will start the pattern mining process After the calculation is finished the results can be inspected in the Patterns of Interest Pattern Viewer and Pattern Information sections of the application 5 Input file requirements PaMiNI accepts three kinds of input files txt files that were exported wi
23. reported peak coordinates in MNI space Furthermore PaMiNI allows to extract the volume data of the clusters and of single pat terns in the NIfTI format and is able to create a text file with the information of selected patterns 2 System requirements PaMiNI is a MATLAB application so a working MATLAB installation is required see http www mathworks com PaMiNI should work on a proper MATLAB installation of version R2008a or later since it is written with MATLAB version R2008a It might also be applicable for older versions but the integrity of the full range of functions cannot be ensured Among others PaMiNI uses functions of the Statistic Toolbox which must be installed An installation of the SPM Toolbox is not required There should be no limitations to the operating system except for the restrictions given by MATLAB However the use of Microsoft Windows is recommended For a proper working of PaMiNI an adequate main memory is required which should be at least 4GB A capable CPU is also recommended 3 Installation For installation only copy the PaMiNI folder to any directory on your harddisk It has to be ensured that the folder contains all m files the MNI nii file and the spm_ methods folder with all files in it No file may be removed from the folder at any time There is no need to make changes to MATLAB set path 4 Getting started To open PaMiNI first start the MATLAB program and change the Cur
24. th BrainMap s software Sleuth xls files of meta analysis data in the format commonly used at the Research Center J lich and a corresponding mat format version of the meta analysis files for further information refer to Simon Eickhoff s eickhoff fz juelich de 5 1 txt files from Sleuth These files can be assessed using BrainMap s Sleuth see http www brainmap org the application to access and search the BrainMap database Make your selections in Sleuth then click on the Export menu and choose Locations GingerALE Text to get a txt file that can be read in by PaMiNI It does not matter which reference space is chosen in the Sleuth preferences if reference space is set to Talairach PaMiNI automatically transforms the coordinates to MNI space 5 2 xls files of meta analysis data To be valid for the use with PaMiNI Microsoft Excel files have to be in the format commonly used for meta analyses at the Research Center J lich Specifically this format requires An exemplary file snippet for this format is given in Figure 2 e Experiments occupy one row per reported coordinate that have to be arranged in consecutive rows e Different experiments are seperated by a free row with at least an empty field in the first column e The x y and z coordinates of the reported foci have to be specified in the columns 3 to 5 in consecutive rows including the first row of an experiment where the further information is
25. that entirely contains the selected pattern If a pattern is already in the Selected Patterns listbox and is related for example the selected pattern itself then its color will be magenta If you try to show related patterns of the standard pattern the first in the listbox an error message dialogue will appear You can remove the markation of related patterns in the scatter plot by simply changing the selection in the Selected Patterns listbox 8 Empty Listbox button This button removes all patterns from the Selected Patterns listbox 8 except for the standard pattern in the first row The color of 13 14 15 16 all scatters in the Scatter distribution scatter plot 6 is turned back to black Before the patterns are actually removed a confirmation dialogue opens to check if you really want to empty the listbox Print Selected Patterns button By pressing this button you can write out all information of the dataset and the settings you used for the current analysis to a txt file Furthermore all patterns contained in the Selected Patterns list box 8 together with their respective information are saved After pressing the button first a file selection dialogue appears to specify a txt file This log file will contain the used inputfile s name and its number of experiments and total number of foci the number of repetitions k Max the selected k Opt the cluster membership probabilit
26. to deliberately interpret the data it provides For example the recommendation for the optimal number of clusters k Opt is pro duced by a statistical measure i e the BIC This gives you a good suggestion which 15 number of clusters could be the best but since BIC values of adjoining number of clusters often show only slight differences the suggested k Opt is not necessarily the neurobio logically most meaningful Therefore the observer should try different selections for the number of clusters around the suggested k Opt value to get a synopsis of the relevant patterns and to be able to make reasonable interpretations The same advice holds for example for the list of interesting patterns The implication that the first pattern in the list indicates the core network in the context of the given dataset would possibly be wrong since the measure of interestingness is chosen quite subjectively to assist the user to find relevant patterns The patterns should always be regarded in the context of the overall picture of the pattern distribution The Show Related Patterns function can provide a good aid to get an overview of the relevant patterns Lastly every interpretation can only be made in the context of the chosen dataset and every observer should be cautious when drawing too generic conclusions out of his data 16
27. y information if clusters were excluded by their standard deviation a maximum standard deviation and the pattern vectors of all patterns in the list together with their number of activated clusters their number of sup porting experiments a list of the centers of gravity of their active components and a list of their supporting experiments Extract Pattern Volume button You can extract the volumes of the clusters activated by a specific pattern into a NIfTI file by clicking this button These vol ume files can be used to further investigate them in other programs like MRIcroN or the SPM Anatomy Toolbox After pressing the button a file selection dialogue will appear to choose nii file where the volume should be stored Then the clusters of the pattern selected in the Selected Pattern listbox 8 are extracted into one volume file The cluster volumes are thresholded on a 0 5 FWHM thresh old level each so that they will appear similar to the cluster representations in the Pattern Viewer 16 when using them in external software Extract All Cluster Volumes button This button facilitates the volume ex traction of all clusters into NIfTI files similar to the Extract Pattern Volume button 14 After pressing the button a file choose dialogue opens for specifying the name trunk of the nii files to be extracted The first volume file extracted is labeled by the name trunk and contains all clusters thresholded on a

PaMiNI

Contents

Download Pdf Manuals

Related Search

Related Contents