Turbo-BrainVoyager 3.2 provides the following new features and enhancements:
- Neurofeedback based on Pattern Classifier. The real-time pattern classification tool introduced in the previous version can now be used as a source for generating classification-based neurofeedback signals. More specifically, it is now possible to use the gradual output value(s) of a trained (multi-class) classifier to generate delayed or moment-to-moment feedback signals. The Neurofeedback dialog contains new options to choose the output from a running SVM classifier as input for the thermometer display. For moment-to-moment feedback, the Real-Time Classification dialog now supports not only the possibility to generate classifier output at the end of a trial but also at each time point during (shifted) protocol conditions.
- SVM Access Plugin. The plugin interface has new commands to access raw classifier output signals for custom processing, e.g. for advanced brain-computer interface (BCI) applications, or for specific calculations for custom neurofeedback visualizations. The new commands are described in the plugin interface documentation and demonstrated in the source code of the provided "Example Plugin - SVM Access" plugin; a compiled version of the plugin is also placed in the TBV plugins folder during program installation and accessible from the Plugins menu.
- BOLD Decoder for Communication BCI. The BOLD Decoder tool can be used for BCI applications such as multiple choice tasks or letter decoding. The tool provides the features described in the publication "Sorger, B., Reithler, J., Dahmen, B., Goebel, R. (2012). A Real-time fMRI-based Spelling Device Immediately Enabling Robust Motor-independent Communication. Current Biology, 22, 1333-1338". The tool can be invoked by selecting the Bold Decoder item in the BCI menu.
- Time Courses Container. The new Time Courses Container allows to visualize time courses and beta plots within a grid layout, i.e. in a number of rows and columns. The zoomable visualization of time course data and estimated betas from all available ROIs provides a manageable overview and is helpful when comparing ROI data in detail, e.g. when selecting regions for neurofeedback applications. The new window can be invoked by selecting the Show Time Courses Container item in the View menu.
- Display of detrended ROI time courses. Time courses of selected ROIs in Time Course Windows of the main window and in the Time Courses Container are now displayed using linear detrending in case that a linear trend is added as a confound in the GLM design matrix. This option can be turned on and off using the D key toggle or the Display Detrended ROI Time Courses item in the View menu.
- Visualization of Maps on Current Volume. It is now possible to visualize calculated maps on the currently processed functional (usually EPI) volume during real-time processing. This is important for quality assurance, i.e. to detect potential scanner problems or problems with reading raw data files. To turn on this display option, click on the Overlay on Current Func Volume item in the View menu. Alternatively, press the C key that acts as a toggle, i.e. pressing it repeatedly will switch back and force between overlay-on-current-volume and overlay-on-first-volume mode. The Overlay on First Func Volume item in the View menu can also be used to switch to overlay-on-first-volume mode. Furthermore, press the B key in overlay-on-current-volume mode to toggle between displaying the current volume either before (default) or after spatial preprocessing has been performed. This possibility can be used to inspect the effect of online preprocessing, including motion correction and spatial smoothing. The currently specified display mode is shown in the standard Log pane in the right lower part of the TBV screen.
- New Neurofeedback Options. Several new options to control ROI activation feedback have been introduced. Feedback can now be turned off for a specified condition, e.g. to show no feedback during a baseline condition. Furthermore, the selection of data points for calculation of a baseline can now be explicitly controlled. In case that a linear trend predictor has been included in (ROI) GLM calculation, detrended ROI time courses (see above) will be used to calculate baseline and feedback values instead of raw signal values. Furthermore, the possibility to preload ROI or VOI files for neurofeedback is now made more easy to access with the new Pre-Load ROI and Pre-Load VOI buttons in the top section of the main window of TBV.
- Bug Fixes, Tips. In very rare cases, Siemens DICOM mosaic header files were not parsed correctly (in case that the header contained more than one DICOM "begin data" key) - this issue has been fixed. The "MultiSliceROIs" value specified in the TBV settings file was not used (erroneously set to 0) - this issue has been fixed. At the begin of a run, beta estimates of ROI time courses could be unstable - this has been fixed; for best beta estimation results at the begin of scans, it is strongly recommended to use linear trend removal (select "1" (default) in the DriftConfoundPredictors spin box of the Statistics tab in the TBV Settings dialog) for temporal high-pass filtering and not higher values for non-linear drift removal.
Turbo-BrainVoyager 3.0 provides the following new features and enhancements:
- Open Architecture via Plugin Interface. The new plugin interface allows full access to all relevant internal data structures supporting any kind of custom processing. In order to quickly learn how to write plugins, the source code of two plugins are provided as examples; the compiled executable of these plugins can also directly be used by selecting them in the Plugins menu. Another includede plugin implements real-time ICA (see below) that uses advanced API functions allowing to visualize custom maps in real-time.
The plugin "Example Plugin - ROI Processing" demonstrates how one may access the content of all currently defined ROIs, both at the level of mean signal values as well as at the voxel level. This feature allows, for example, to combine signal time courses such as subtracting the values of a "background ROI" from the values in functionally specific regions; the result of such processing can be exported for custom real-time visualization software.
The plugin "Example Plugin - Export Volume Data" demonstrates how at each incremental time point preprocessed raw data, beta maps and t maps can be accessed enabling any kind of custom real-time processing as well as export of the full data to external software such as Matlab, e.g. to perform additional calculations or to create custom (feedback) displays.
- Real-Time Multi-Voxel Pattern Classification. This version introduces real-time multi-voxel pattern classification based on support vector machines (SVMs). In a training phase, data from one or more runs can be specified. At any moment in a test run, incremental classification can be turned on to predict the class of evoked activity. This allows to conduct online "brain reading" applications.
- Real-time Independent Component Analysis (rt-ICA). Real-time ICA is now available as a plugin that has been implemented by Fabrizio Esposito (Brain Innovation). When launching it from the Plugins menu, it will ask for a range of slices and a temporal window that will be used for the data-driven multivariate detection of functional networks. If a region-of-interest has been defined, the discovered networks (components) will be ranked with respect to the overlap of the components with the ROI, i.e. the component(s) are visualized that include the selected region. Using the Show plugin maps option, it is possible to show either the calculated plugin maps (components) or the standard GLM-based contrast maps.
- Support for GE Signa HDxt Scanners. The recent generation of GE scanners exports DICOM data incrementally that can now be processed by Turbo-BrainVoyager.
- Advanced 3D Visualization for Philips Data. It is now also possible for exported Philips (DRIN) data to visualize real-time maps in intra-session and extra-session (ACPC, TAL space) 3D anatomical data sets. The new alignment options also allow using defined VOIs across scanning sessions. In the current release this feature requires the use of BrainVoyager QX.
- Improved Support for Mesh Visualization. Beyond folded surfaces supported previously, It is now possible to visualize incrementally analyzed data on inflated and flattened cortex meshes; the inflated mesh(es) need to be specified in the VMR-SRF tab of the TBV Settings dialog; note that the folded meshes that were used to create the inflated/flattened meshes need to be stored in the same folder since the morphed mesh look up the coordinates of vertices in the folded version to perform correct visualization.
Besides bug fixes, Turbo-BrainVoyager 2.8 provides the following enhancements:
- Improved Support for Siemens real-time export. The new Siemens export function introduced with scanner software VB15A is now better supported and documented.
- Delayed feedback. The "Neurofeedback" dialog now allows to provide not only direct (ongoing) feedback but also feedback at the end of blocks. Such "delayed feedback" of the mean level of activity of a whole block may be helpful in the context of some clinical neurofeedback applications. This option can be turned on by enabling the "Delayed feedback" option in the "Feedback type" field of the "Neurofeedback" dialog. Click the "Options" button to invoke the "Delayed Feedback Settings" dialog that can be used to select the protocol condition that should be used for calculating delayed feedback values. You may also specify the data points used for calculating the baseline value ("Baseline interval") as well as the time points used to calculate the average activation value with respect to the baseline ("Modulation interval"). You may also specify a "Feedback file" that will contain the stored feedback values that can be used for providing feedback to the subject using custom software. The "Update feedback file" value specifies when the feedback file should be written (value must be larger than end point of modulation interval) and the "Scale beta value" spin box can be used to specify a value for scaling the estimated percent signal change beta values.
- Improved feedback output for custom applications. In previous versions, volume-by-volume files containing mean ROI signal values (.RTP) files was only produced when the "Neurofeedback" dialog was open and when the time course feedback option was selected. The .RTP files are especially useful for custom processing of real-time ROI data. In this version, the "Log ROI Time Point Files (RTPs)" output option can now be turned on/off from the "Analysis" menu, i.e. opening the "Neurofeedback" dialog is no longer necessary. This option is turned on as default. Furthermore, the mean ROI signal values are now stored as float values for highest precision (also in ".ERT" files that also store data of individual voxels).
Turbo-BrainVoyager 2.6 provides the following enhancements:
- Support for new Siemens real-time export. With recent software releases, Siemens supports a new real-time export functionality saving each volume of data in a mosaic Dicom file. This feature must be turned on requiring an IDEA license. For details, consult the Siemens documentation. Turbo-BrainVoyager 2.6 is able to use the exported data. This requires specification of the "SIEMENS_DICOM_MOSAIC" data type and a specification of the first file, e.g. as "001_000003_000001.dcm" (if this is the scan (series) 3). Subsequent files would be written as "001_000003_000002.dcm", "001_000003_000003.dcm" and so on.
- Conjunction contrasts. A new option is now available allowing to visualize a map as the conjunction of all currently selected contrasts.
- Compatibility with BrainVoyager QX. This version increases the compatibility of the produced output files with BrainVoyager QX. The order of slices is no longer reversed in TBVas in previous versions, which is visible in the multi slice view but not in the volume views since for visualization the order of slices is adjusted in such a way that the brain appears upright in the respective views. The increased compatibility with BrainVoyager QX allows to load and further process TBV data.
Besides improvements and bug fixes, Turbo-BrainVoyager 2.4 provides the following new features:
- Smoothed Feedback. The bar graph display in the Neurofeedback dialog may now show a smoothed version of the calculated feedback signal.The Avergage feedback values field allows to specify how many recent values are averaged to obtain the current display value. With the default value of "1", no averaging takes place; while this gives feedback about momentary changes of the ROI BOLD data, the feedback signal might fluctuate substantially. A large value (e.g. "10" or more) results in a very smooth feedback signal but at the cost of a long delay with respect to the underlying cognitive events. As a good compromise, we suggest values between "3" and "5" implementing a modest temporal low-pass filter.
- Dynamic ROIs. It is now possible to automatically optimize ROIs by specifying that a percentage of "top" voxels should be determined and used as the efffective sub-ROI. Dynamic ROIs generally lead to better signal extraction from ROIs, especially in case of coarse anatomical ROI specification. But also functional ROIs benefit from dynamic selection: By defining rather broad initial ROIs, dynamic voxel selection ensures that the best sub-ROI will be determined even if there are small alignment errors across successive runs or movement-related slice shifts. Dynamic ROIs can be turned on from the Neurofeedback dialog, for details check the Dynamic ROIs documentation.
- Incremental ROI Beta Export. For neurofeedback studies, the Neurofeedback dialog can be used to create basic feedback displays, which can be presented to subjects, e.g. by using the "Feedback Presenter" software. To support creation of special feedback displays for custom software, Turbo-BrainVoyager exports "extended ROI time course" (.ERT) files if the Log ROI Time Courses option is enabled in the Analysis menu. In this file, the signal values of each ROI (and even for each voxel within ROIs) are stored incrementally to disk: When the data for a time point is processed, ROI signal values are appended to the experiment's .ERT file. For several purposes, it might be desirable to know the estimated beta values for each ROI. While custom software could use the exported ROI time course values to calculate beta values, this version of Turbo-BrainVoyager adds the Log ROI Betas option in the Analysis menu, which is enabled as default. With this option turned on, the estimated condition beta values are incrementally saved in a "beta time course" (.BTC) file.
- Preloading VOI Files. ROIs defined in "native" space (.ROI files) can be loaded at any time. A useful option is that a .ROI file can be specified already before a prepared .TBV file is executed, i.e. before any data has been processed. This is useful during a neurofeedback session when ROIs have been determined in a localizer run, which should be used in subsequent neurofeedback runs. ROIs can be also defined in ACPC or Talairach space when using 3D data sets. In this case the ROIs are stored in .VOI files either from Turbo-BrainVoyager or from BrainVoyager QX. While .VOI files could be loaded already in previous versions, this release of TBV allows to preload a VOI file in a similar way as described above for .ROI files.
- Strings for directories (e.g. watch folder, target folder, but also references to files such as protocols) can now be specified in a relative manner with respect to the folder containing the TBV file. A relative path begins with a "./" substring. When reading a TBV settings file, relative paths are expanded by prepending the path containing the TBV file. When saving a TBV file, paths are made relative if they reside within the folder with the TBV file or in any subfolder. This new mechanism allows it to move data to new places without breaking folder and file location specifications.
- The TBV Settings dialog now shows the value of the TimeOutSeconds value. This value had to be changed directly in the ".tbv" file in previous versions but can now be changed in the General tab of the TBV Settings dialog.
- Philips provides a elegant interface to export data during real-time scanning with the new DRIN system that is now supported in Turbo-BrainVoyager.
- When working with ROIs/VOIs, it was not possible to switch between multiple ROIs since all newly opened ROI (interactively or by loading a file) were always added to the already existing ones. The Analysis menu now has an entry Remove ROI Time Courses deleting all currently available ROI time courses. New time courses can be then added interactively or by loading ROI / VOI files from disk.
- The code routines for statistical and visualization routines has been further optimized with an overall gain of 50-100% processing time.
- The Neurofeedback dialog has been improved calculating more sensitive baseline values and also allows to present feedback for 2 ROIs with 2 thermometer displays.
Copyright © 2014 Rainer Goebel. All rights reserved.