Volumetric EPSI Processing - Miami

Last Updated Feb. 2011Volumetric EPSI ProcessingA Quick Introduction to EPSI Processing Using the BATCHProcessing PipelineA. A. MaudsleyContents:1.2.Introduction . 1Importing New Data . 2i) Create a New project . 2ii) Import the Data . 2iii) Data Import for GE and Philips. 43. Copying the Processing Files . 64. Running the Short Processing Pipeline . 65. Deleting the Last Result . 86. Running the Full EPSI Processing Pipeline . 87. Acknowledgments. 101. IntroductionThis document gives an overview of running the automated processing pipeline for a volumetricEPSI dataset and associated MRI. Two examples are provided: the first is a “quick” processingpipeline that can be used for phantom data or for initial evaluation of an in vivo dataset, and thesecond is the complete processing pipeline used for human subject data.The general procedure for the first-time processing of a new data set is as follows:1) Export the MRI and MRSI data from the MR scanner and place it in a temporarydirectory that is visible to the computer running the MIDAS software. (The SI is inDICOM format for Siemens, and proprietary MRS format for GE and Philips)2) Run the MIDAS Importer and:a) Create a new Project, if one is not already available.b) Import the data.3) Copy the processing files to the correct location.4) Start the BATCH program and run the processing pipeline.These steps are described below, with the addition of first running an abbreviated processingprotocol followed by deleting that result and repeating the processing using the completepipeline.1

2. Importing New DataRun MIDAS Tools application, and from the toolbar start the Importer ()i) Create a New projectGo to: File New Project, and the widget shown in the following figure will appear:Now fill in the requested information. An example is given in the following figure:For the “Project Path” the program will create anew subdirectory with the same name as that givenfor the project, but you also have the option ofcreating a new top-level directory to put the dataunder.The “Atlas Path” points to files used in the spatialregistration and data analysis. During the MIDASinstallation data files will have been created underthe \MIDAS\MNI-Atlas directory and the programwill initially display this as the default path. However, you may also wish to copy these files to ashared networked location so that they can be shared by multiple users, in which case the pathused in the Project needs to be changed accordingly.The default location for the “Processing Files” is as a subdirectory under the newly-createdproject location, which is indicated by the path “./ProcessingFiles” as shown in the first figure.However, you can place these files anywhere and set the appropriate path, as in the secondexample. See the following section for additional information on the processing files.The “Reference Path” points to a set of files used in the PRANA data analysis. These are notprovided as part of the MIDAS installation and need to be created by the user. Reference data for3T from the University of Miami can also be downloaded separately. The default location isagain under the MIDAS installation, but again can be placed anywhere. For example, placingthis on a shared drive would ensure multiple users are using the same data.ii) Import the DataAfter the project has been created it will show under the “Subject Browser” on the left side of theImporter. Now go to the “Data Browser” section, select Browse, and from the file selectionnavigate to the directory containing the Dicom MRI and MRSI data files. On making the file2

selection the program will scan all of the data headers, and you will see the “Progress” widgetshown in the following figure:When this completes, the Subject, Study, and all Series corresponding to all data in the selecteddirectory will be shown, as indicated in the next figure:Select the data you want to import. In this case you want all of the data so selection should bedone on either the Study level, as shown in this example, or on the Subject level. If you were toselect on the Series level, only that Series would be imported.Click “Import Files” to start the Import process. The data files are then copied to the MIDASproject location into a subdirectory named “raw”, and all parameter information is placed intothe “subject.xml” file. Since the SI files are can be several Gb in size, this can take a fewminutes.When all files are copied the following widget will appear:3

Here you are asked to fill in some information that is not available in the Dicom headers and toassign a “Series Label” to each image series. In the above example, the first two labels havealready been selected, with the “SI” label corresponding to the EPSI acquisition that has TE 70ms, and the SI Ref to the EPSI data with TE 0 ms. Also shown in the figure is the list of Serieslabels that appears when you click in the “Series Label” column, and in this case the MRI T1label has been selected, corresponding to the MPRAGE series.After all labels are selected click the OK button (at the bottom), and exit the Importer.After data import, the files in the temporary directory can be deleted.iii) Data Import for GE and PhilipsGE and Philips do not support DICOM format for raw spectroscopy data and their proprietarydata formats do not provide some essential information; therefore, for these data it is necessary tofirst import an MRI study before an EPSI study. If all files are located in the same directory thenthe Importer will take care of this automatically and the Study/Series selection is done asdescribed above.Philips EPSI data import requires the single *.raw data file and the *.sin file.GE EPSI data consists of multiple “P” files, e.g. of the form “P06144.7’. For 8-channel datathere are 90 files.When the spectroscopy data is selected in the Importer, the Data Browser window shows “null”in the Subject and Study information, as shown here:4

To import the spectroscopy data just select the Study node and “Import Files”. Currently(5/2010), no progress indicator appears for this import, but it just takes a few minutes before theParameters widget appears.For GE (and for some DICOM servers from other instruments) there is an additional differencein that the MRI DICOM files are organized with a separate directory for each image Series, andusing the same filenames in each directory. This means it is necessary to select each MRIdirectory and import each Series in turn. The following example illustrates the procedure for GEdata:Step 1: Import the T1 MRI data. On hitting the Browse button, navigate to the correct directory(directory 005 in this case):Select the directory and import the MRI.Step 2: Browse navigate to the directory containing the p-files for the SI data. The datainformation will then appear in the Data Browser window, and if you expand the data tree youwill see 2 Series listed, i.e.:5

Select at the Study level, hit “Import Files”, and the p-files files will be copied to the project(which can take several minutes).3. Copying the Processing FilesExample processing files are provided with the MIDAS distribution and are available from theMIDAS web site. These files must be copied to the location specified for the Processing Filespath when the project was created.For example, the MIDAS distribution should contain the directory:“C:\Midas\ProcessingFiles\3T PACoil TE70”which are the files for 3T with phased-array detection. In the above example, these would getcopied to N:\TestProject\ProcessingFilesFor the situation where multiple projects use the same set of processing files it is recommendedthat a single common ProcessingFiles directory be shared among all projects, so that only asingle directory need be maintained. This can be done by setting the corresponding path whencreating each project.4. Running the Short Processing PipelineThe full processing pipeline includes operations such as tissue segmentation and lipidextrapolation, which cannot be applied to data obtained from phantom objects, and spectralfitting, which can take several hours. Therefore for phantom data or for quick evaluation of invivo data you will want to use an abbreviated processing pipeline. This processing is describedfirst.Start the BATCH program:You must first select the Study to be processed. Click the Browse button next to the “StudySelection” section of the widget to bring up the MIDAS Browser, and select the Subject orStudy corresponding to the data just imported, e.g. as shown in the figure below:6

Note that if you just created a new project you may have to hit the “refresh” button, which is theblue icon in the upper right corner of the Browser, to make that project visible in the Project list.It doesn’t matter at which level in thelist you make the selection, and afterhitting Done the Subject ID willappear in the Study Selection list.You must now load the file whichdefines the processing steps. Click theBrowse button on the “PipelineDefinition File” line. The fileselection widget will appear and showall files with a .txt extension, and inthis case you want to select the onecorresponding to the short processing,e.g. named something like:“QuickView Processing 3T *.txt”,as shown in the example. Note thatthere are no fixed naming conventionsand the filename in your distributionmay differ from that shown here.At this point the list of processing steps will appear (below):7

Click Start to begin the processing, which may take from 10 minutes to an hour depending onthe number of channels.After processing is complete the Viewer or SID programs can be used to view the result.5. Deleting the Last ResultIf your data is from a human study you will want to reprocess it using the complete processingpipeline, but you must first clear out the last result. There are two options to do this:a) Enabling the “Delete results of previous processing” option in the BATCH program.See the following section.b) Run the “Delete MIDAS Data and Nodes” utility (see the Tutorial document for a moredetailed description). All nodes should be deleted, so just uncheck the “Test Only” optionand hit “Start”.6. Running the Full EPSI Processing PipelineFor the complete EPSI processing the same procedure described in Section 4 is used, but withselection of the processing definition file, which will be named something like:“Batch Processing 3T *.txt”.The figure below shows the list of processing steps, and if the data has not been processed beforeit is only required to click Start to begin the processing, which will take several hours.8

If the data was processed previously then the “Delete results of previous processing” optionshould be set (alternatively you can run the “Delete MIDAS Data and Nodes” utility). Becausethis option doesn’t delete all the SI data the Volumizer and EPSI3D steps for the SI and SI Refdata do not need to be repeated. These can be turned off by clicking on the button to the left ofeach of these processing steps, i.e. that section of the widget should look as shown in thefollowing figure:9