User Manual - Emory University
Contents
1. which performs estimation and Plot which provides various options for visualizing the results 2 1 Analysis Data File There are three choices to import the data e Import Data loads image format files e g Analyze or Nifty images from the raw preprocessed data e Load Saved Data loads data from a saved mat file This option requires the user has previously loaded the raw image data using Im port Data and saved this input in a mat file e Load Estimation Results loads results from MCMC discussed later 2 1 1 Importing data Import data will allow the user to select the image format and the regional parcellation type either AAL 2 or Brodmann 3 When the user clicks the button it will launch the dialog box in Figure 2 2 Bayesian Spatial Model for brain Activation and Connectivity BSMac v2 0 moas JP Load Saved Data fiestas 20uatave 7 Parietal_Sup_R Supp_Motor_Area_L Supp_Motor_Area_R Figure 2 1 BSMac GUI interface Input Data Image Format Region Type Nitti G AAL Analyze Brodmann Input File List g Output File Name Figure 2 2 Load image data interface Table 2 1 Input File Format Path File Name Group Subject Session Stimulus Others datal data2 data3 data4 data5 data Besides choosing the data format and the region map AAL or Brod mann the user needs to provide a text file which i2. of results v Plot brain activation maps v Plot functional brain connectivity 1 1 System requirements The BSMac Software is a collection of the MATLAB functions than can be run in a GUI Users can run and edit BSMac functions from any computer with MATLAB and a non modifiable executable version of the software can be run from machines without MATLAB To run without MATLAB one needs to install the MATLAB Compiler RunTime MCR package before launching BAMac To install the MCR users must run the MCR installer appropriate for their platform MCRInstaller exe for Windows systems or MCRlnstaller bin for UNIX and Linux systems We provide a link on our web page for installing MCR http www sph emory edu bios CBIS links html BSMac is a small self extraction program around 2MB in size including image templates such as AAL 2 Brodmann 3 and MNI152 templates all are in standard MNI Montreal Neurological Institute space After downloading the software simply extract it to your computer BSMac will automatically create four folders to save the necessary files for running BS Mac v anatomical folder includes the AAL template image and associated region labels the Brodmann template image and associated region labels and the MNI template v data folder for saving extracted image data files such as mat files v results folder for saving MCMC results files vV CBIS folder includes help files BSMac performs parameter esti
3. 10 Figure 3 Regional Posterior Probabilities S vs H 5 vs 1 Elle D6 098330 98 OE Regional Posterior Probabilities 07 06 Probabilities Figure 2 6 Regional posterior probabilities for selected contrast Figure 4 Histogram of voxel level Posterior Probabilities S vs H 5 vs 1 EBR Eile DG6ds 18094 058 Voxel Posterior Probabilities 1800 1600 1400 E v ES o gt 5 co e a de a nN El Figure 2 7 Voxel level posterior probabilities for selected contrast 11 Figure 5 Intra regional Correlations Group Session Stimulus DER File Ai CRESE p614 p 11 2 Correlation Correlation 0 0 12345678 9101112 1234567 8910112 Region Region p 1 13 p 21 1 Correlation Correlation 0 0 12345678 9101112 1234567 89101112 Region Region p 211 2 p 2 1 3 Correlation Correlation a 12345678 9101112 a 1234567 8910112 Region Region Figure 2 8 Intra regional correlation estimates Figure 6 Inter regional Correlations Group Session Stimulus File 8603 3399 9 DB RM R04 cA 12345799012 123456784012 R013 R 2 41 1 cA 123456784012 Al 123456784012 R12 R 2 1 3 El 123456768 012 3 123485784012 Figure 2 9 Inter regional correlation estimates 12 Activity Map 10 x a File Ds vs H 5 vs 1 Hvs S 3 vs 1 IH vs S 5 vs 1 x OEE EN CIO Opacity Or
4. BSMac Bayesian Spatial Model for Brain Activation and Connectivity User s Guide version 2 0 CBIS Emory University November 13 2010 Contents 1 Introduction 1 1 System requirements LL Qu Le 1 2 Preprocessing 2 Softvvare structure 2 1 Analysis Data File 2 1 1 Importing data o LL LL LL LL 21 2 Design Matix lesa podes geal ae aol das 2 2 Region Selection 2 3 Prior parameters 24 Estimation ero a a Busa 25 Misualization i 20 8 a ra ta A 2 5 1 Basic Summary Plots 2 5 2 Activation Maps Qu LL LL LL Le 2 5 3 Functional Bibliography Connectivity o Chapter 1 Introduction BSMac is a statistical and graphical visualization MATLAB toolbox for the analysis of MRI data It simultaneously performs whole brain activation analyses at the voxel and region of interest ROI levels as well as functional connectivity FC analyses using a flexible Bayesian modeling framework 1 Details of the underlying statistical model estimation and interpretations of parameters are provided by Bowman et al 1 In the BSMac toolbox most of the operations can be performed easily within a graphical user interface GUI BSMac runs under both Windows and linux unix platforms BSMac implements the following procedures v Load data Image v Design matrix setup v Parameter setup v Region selection v Model estimation v Plot basic summaries
5. ick Functional Connectivity The user can then select a subset or all of the regions included in the analysis and specify the desired connectivity map for display e g corresponding to 15 Functional Connectivity i Parietal Sup L Parietal Sup R Supp Motor Area L Supp Motor Area R Par ta Sip R Spp Hobr Arez L a PrtmenL Futames R Temporaksip Figure 2 13 Display of functional connectivity results a subgroup session and task The functional connectivity map displays all of the selected regions in a dynamically rotating brain with bars connecting regions that have posterior median correlations exceeding a user specified threshold The connected regions have different line thicknesses depending on the strengths of the functional connections between regions Figure 2 13 shows the result of the functional connectivity with selected regions User can also plot the connectivity with special angle view For example Figure 2 13 illustrates the connectivity with all selected regions for group session stimulus 2 1 1 corresponding to healthy control subjects during session 1 with the specified threshold of 0 1 16 Bibliography 1 F D Bowman B Caffo S S Bassett C Kilts A bayesian hierarchical framework for spatial modeling of fmri data Neurolmage 39 2008 146 156 2 N Tzourio Mazoyer B Landeau D Papathanassiou F Crivello O Etard N Delcroix B Mazoyer M Joliot Automated anatomical
6. ientation Jnit Query Talariach Get Active Hide Crosshairs Crosshair Color Voxel Activity Reset All Figure 2 10 Region level brain activation map 13 Activity Map a lol xj x File I E ai ie Prob Thres B El Query Talariach GetActive Hide Crosshairs Crosshair Color Regional Activity Reset All Figure 2 11 Voxel level brain activation map 14 Figure 101 Eile DS AMSI RR US EI OE Axial slice33 Axial slice35 Figure 2 12 Regional activity plot in axial views for every other slice ranging from 33 to 41 The probability threshold is set to 0 2 coronal views The lower right part of the figure provides some basic in formation such as crosshair cursor position and the posterior probability value background brightness and contrast The user can set the posterior probability threshold and opacity for visualization By clicking the Re gional Activity button the user can switch the activity map between the regional and the voxel view see Figure 2 11 There are also options that allow the user to plot particular slices in selected orientations For example the user may generate plots in an axial view slices between 33 to 41 and with the threshold 0 2 see Figure 2 12 2 5 3 Functional Connectivity This toolbox also allows the user to plot inter regional functional connec tivity To display functional connectivity results the user can click the Display menu at the top and then cl
7. labelling of activations in SPM using a macroscopic anatomical parcella tion of the MNI MRI single subject brain NeuroImage 15 2002 273 289 3 L Garey Brodmans localisation in the cerebral cortex the principles of comparative localisation based on cytoarchitectonics Springer Lon don 1994 english translation of Vergleichende Lokalisationslehre der Grosshirnrinde by Korbinian Brodman 17
8. lities of activation specifically estimating the probability that the contrast is greater than zero Figure 2 6 shows the posterior probabilities for each region We also generate a histogram of the voxel level posterior probabilities of activation Figure 2 7 e g giving an indication of the num ber of highly probable alterations in neural activity across the entire brain or in the included regions Figure 2 8 shows intra regional correlations separately for each subgroup say controls and at risk subjects and Figure 2 12 shows the group specific inter regional correlation matrices thresholded at 0 1 for ease of visualization 2 5 2 Activation Maps After generating the basic summary plots in the above step the user can plot more detailed maps of distributed patterns of changes in neural ac tivity or what we simply call activation maps from the menu Display gt Activation Map We provide both region level and voxel level activation maps which are technically posterior probability maps for our analysis Figure 2 10 shows a snapshot with the region level activation map corre sponding to a selected contrast The activation map displays the posterior probabilities associated with the selected contrast in axial sagittal and Figure 1 Region Legend Figure 2 4 Selected regions labels Figure 2 5 Plot of the regional contrast estimates showing the posterior mean for each region
9. mation based on Markov chain Monte Carlo MCMC methods For some analyses the toolbox may consume a lot of RAM for storing estimation results some of which are temporary More than 1GB RAM is recommended for running BSMac To reduce the required RAM you may change the Thin Factor to a larger number You may also decrease the number of iterations or burn in runs although this should be done with caution as it may affect the performance of the MCMC estimation 1 2 Preprocessing We assume that the user has already performed some initial preprocessing prior to using our software and that the data are registered to MNI space 91 x 109 x 91 for example using FSL http www fmrib ox ac uk fsl SPM http www fil ion ucl ac uk or other software packages Chapter 2 Software structure BSMac provides an easy to navigate GUI environment based on MATLAB Figure 2 1 shows a screen capture of the BSMac GUI interface running under Windows There are five basic steps to implementing the spatio temporal model in BSMac namely Analysis Data File s which allows the user to read in the image data or saved data results from previous runs Region Selection which enables the user to select particular brain regions for inclusion in the analysis Parameters giving the user to specify hyper prior parameters in the Bayesian model select starting values and MCMC information such as the number of iterations Estimate
10. ncludes additional in formation about the data In the text file the first line row contains information on the variable names no space allowed in the variable name the following lines contain the data corresponding to the variables The text file should include at least 5 columns separated by spaces as shown in Table 2 1 the first five columns The total number of columns exceeding 5 depends on the number of Groups Sessions Stimuli and other covariates The information in the text file will be used to construct the design matrix for modeling After completing the information in the dialog box click the button Ok the BSMac will begin to import the raw image data When BSMac finishes reading the image data the toolbox saves the information as a MAT data file under the folder named data which can be loaded later for future analyses After importing the raw data or loading the saved mat data the Set Contrasts button will become available the Parameter section of the GUI will also become available and will guide the user to the next step of setting design matrix vs S 3vs1 vs Svs1 3vs1 5 vs 1 atra 141411 Figure 2 3 Design Matrix for the second level analysis based on the fBIRN data 2 1 2 Design Matrix The design matrix figure gives an illustration of the design matrix con structed by BSMac from the user specified text file discussed in section 2 1 1 This dialog box also allow
11. r parameters The toolbox is based on a Bayesian hierarchical parametric model Thus the user must specify values of the hyper prior parameter values at the termi nating level of the model BSMac includes the following default parameter values ag 0 1 bg 0 005 co 0 1 dg 0 01 eg 0 1 and fo 0 01 resulting in vague or weakly informative priors The parameter up is ob tained empirically as the mean across all subjects The values for ag bo co dy eg and fo generate densities that place large probability masses on large variances The rationale for using these vague priors is to ensure that the information in the data primarily governs the results The user may keep the default values if there is not specific information available a priori to guide particular selections for these values However the user may set different values if some additional information about the data is available The number of Markov Chain Monte Carlo MCMC iterations may vary from one analysis to the next We suggest running at least a few thousand iterations and discarding the first one thousand When running a large num ber of iterations the Thin Factor can be used to reduce the amount of information retained by keeping draws from say every 5th iteration Thin ning may also help to ensure that draws from subsequent iterations are not highly correlated Caution If you encounter an Out of memory prob lem consider thinning or reducing the n
12. s the user to define contrasts between groups ses sions and tasks Under Current Contrast enter the desired contrast and user specified label name then click update The user can define multi ple contrasts Click Done when finished The contrasts will be saved into the default folder results 2 2 Region Selection Based on the regional parcellation selected in step1 either AAL or Brod mann a list will appear with all of the region names The user can choose regions to be included in the analysis Regions can be defined as either sin gle regions or as combined regions by clicking Add button The combined regions can include up to 4 individual subregions Each row in the com bined region table constitutes a single defined region for the spatio temporal model The selected regions can be saved for reloading later IMPORTANT NOTE We impose a sample size dependent restriction on the number of regions included in the analysis Specifically the number of regions cannot exceed one less than the smallest subgroup size in the data For example in a study with 21 healthy controls and 25 patients we would limit the number of regions to 20 If there are no subgroups present in the data then the number of regions cannot exceed one less than the total sample size 1 This restriction is to ensure stability of the estimation procedures but all regions can be included with a sufficiently large sample size 2 3 Prio
13. umber of iterations Reducing the number of covariates included in the model may also help 2 4 Estimation Once steps 1 3 have been completed the model is now set up and ready for Bayesian estimation The user can execute the estimation step by clicking the Estimate button The MCMC method used here implements Gibbs sampler Applying MCMC methods in our context is complicated by the massive amount of data the vast number of spatial locations and the large number of parameters However the Gibbs friendly model specification fa cilitates estimation by providing substantial reductions in computing time and memory 1 Performing estimation is the most time consuming step in the analysis The estimation time depends on several factors such as the number of co variates the number of regions and the sample size and the speed of the user s computer Our experiences have ranged from around 15 minutes to greater than 3 hours 2 5 Visualization The last step for this toolbox is devoted to visualization This includes gen erating basic summary plots activity maps and 3D functional connectivity maps 2 5 1 Basic Summary Plots The basic summary plots generate several results First a legend of selected regions Figure 2 4 is shown A plot of the posterior mean contrast for each region is also generated see Figure 2 5 The plot gives the contrast estimate for each selected region BSMac also provides regional posterior probabi
Download Pdf Manuals
Related Search