|
Size: 37478
Comment:
|
Size: 47573
Comment:
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 1: | Line 1: |
| = Tutorial 17: Time-frequency = ''Authors: Francois Tadel, Dimitrios Pantazis, Sylvain Baillet'' This tutorial introduces how to compute time-frequency decompositions of MEG recordings and cortical currents using complex Morlet wavelets. There are several ways to reach the same result, please read all the sections carefully and then chose the method that is best suited for your own data. |
= Tutorial 24: Time-frequency = ''Authors: Francois Tadel, Dimitrios Pantazis, Elizabeth Bock, Sylvain Baillet'' This tutorial introduces how to compute time-frequency decomposition of MEG/EEG recordings and cortical currents using complex Morlet wavelets and Hilbert transforms. |
| Line 8: | Line 8: |
| == Introduction == Some of the MEG/EEG signal properties are difficult to access in time domain (graphs time/amplitude). A lot of the information of interest is carried by oscillations at certain frequencies, but the amplitude of these oscillations is sometimes a lot lower than the amplitude of the slower components of the signal, making them difficult to observe. The averaging in time domain may also lead to a cancellation of these oscillations when they are not strictly locked in phase across trials. Averaging trials in time-frequency domain allows to extract the power of the oscillation regardless of the phase shifts. For a better understanding of this topic, we recommend the lecture of the following article: Bertrand O, Tallon-Baudry C (2000), [[https://www.ncbi.nlm.nih.gov/pubmed/11102663|Oscillatory gamma activity in humans: a possible role for object representation]]. In Brainstorm we offer two approaches for computing time-frequency decomposition (TF): the first one is based on the convolution of the signal with series of complex Morlet wavelets, the second filters the signal in different frequency bands and extracts the envelope of the filtered signals using the Hilbert transform. |
|
| Line 10: | Line 17: |
| == Introduction to complex Morlet wavelets == Complex Morlet wavelets are very popular in EEG/MEG data analysis for time-frequency decomposition. They have the shape of a sinusoid, weighted by a Gaussian kernel, and they can therefore capture local oscillatory components in the time series. An example of this wavelet is show below, where the blue and red curves represent the real and imaginary part, respectively. Contrary to the standard short-time Fourier transform, wavelets have variable resolution in time and frequency. When designing the wavelet, we basically decide a trade off between temporal and spectral resolution. To design the wavelet, we first need to choose a ''central frequency'', ie the frequency where we will define the mother wavelet. All other wavelets will be scaled and shifted versions of the mother wavelet. Unless interested in designing the wavelet at a particular frequency band, the default 1Hz should be fine. Then, the desirable'' time resolution'' for the central frequency should be defined. For example, we may wish to have a temporal resolution of 3 seconds at frequency 1 Hz (default parameters). These two parameters, uniquely define the temporal and spectral resolution of the wavelet for all other frequencies, as shown in the plots below. Resolution is given in units of Full Width Half Maximum of the Gaussian kernel, both in time and frequency. The relevant plots are given below. {{attachment:waveletOptions.gif||height="291px",width="637px"}} |
== Morlet wavelets == Complex Morlet wavelets are very popular in EEG/MEG data analysis for time-frequency decomposition. They have the shape of a sinusoid, weighted by a Gaussian kernel, and they can therefore capture local oscillatory components in the time series. An example of this wavelet is shown below, where the blue and red curves represent the real and imaginary part, respectively. Contrary to the standard short-time Fourier transform, wavelets have variable resolution in time and frequency. For low frequencies, the frequency resolution is high but the time resolution is low. For high frenqucies, it's the opposite. When designing the wavelet, we basically decide a trade-off between temporal and spectral resolution. . {{attachment:tf_resolution.gif||width="336",height="144"}} To design the wavelet, we first need to choose a ''central frequency'', ie. the frequency where we will define the mother wavelet. All other wavelets will be scaled and shifted versions of the mother wavelet. Unless interested in designing the wavelet at a particular frequency band, the default 1Hz should be fine. Then, the desirable'' time resolution'' for the central frequency should be defined. For example, we may wish to have a temporal resolution of 3 seconds at frequency 1 Hz (default parameters). These two parameters, uniquely define the temporal and spectral resolution of the wavelet for all other frequencies, as shown in the plots below. Resolution is given in units of Full Width Half Maximum of the Gaussian kernel, both in time and frequency. . {{attachment:waveletOptions.gif}} |
| Line 26: | Line 33: |
| From the figure above, which designs the Morlet wavelet, we can see that the default wavelet (central frequency 1Hz, FWHM=3sec) has temporal resolution 0.25sec at 5Hz and 0.1sec at 10Hz. In such case, the edge effects are roughly half these times: 125ms in 5Hz and 50ms in 10Hz. Examples of such edge effects are given in the figures below. {{attachment:edgeEffect5Hz.gif}} {{attachment:edgeEffect10Hz.gif}} == Time-frequency decomposition of MEG Recordings == We are going first to compute the time-frequency decomposition for the two averaged recordings we have in the protocol. Drag and drop the two ERF files from ''Right ''and ''Left ''in ''Process1''. Select "Recordings". {{attachment:fileSelection.gif}} Click on Run. Select the process: '''Frequency > Time-frequency (Morlet wavelets)'''. * The only generic option you have access to in this window is the selection of the sensors on which you want to apply the process. You can use this box to select a mixed list of channel types (ex: "MEG") or channel names (ex: "MRT16"). * For now, you can just leave the default "MEG, EEG". There is no EEG channel but it doesn't matter, the extra information is just ignored. {{attachment:processSelection.gif}} Click on the "Edit" button to get access to the options specific to the time-frequency decomposition process. With this time-frequency option window, you can: 1. define bands in time and frequency or analyze the data in all points, 1. configure the time and frequency resolution of the wavelets, 1. define the output type (complex coefficients or squared coefficients / power). {{attachment:timeFreqOptionsBasic.gif}} == Description of the options == |
From the figure above, which designs the Morlet wavelet, we can see that the default wavelet (central frequency '''fc'''=1Hz, '''FWHM_tc'''=3sec) has temporal resolution 0.6sec at 5Hz and 0.3sec at 10Hz. In such case, the edge effects are roughly half these times: 300ms in 5Hz and 150ms in 10Hz. More precisely, if '''f''' is your frequency of interest, you can expect the edge effects to span over FWHM_t seconds: '''FWHM_t = FWHM_tc * fc / f / 2'''. Examples of such transients are given in the figures below. {{attachment:edgeEffect5Hz.gif||width="299",height="244"}} {{attachment:edgeEffect10Hz.gif||width="302",height="245"}} We also need to consider these edge effects when using the Hilbert transform approach. The band-pass filters used before extracting the signal envelope are relatively narrow and may cause long transients. To evaluate the duration of these edge effects for a given frequency band, use the interface of the process "Pre-process > Band-pass filter" or refer to the filters specifications ([[http://neuroimage.usc.edu/brainstorm/Tutorials/ArtifactsFilter#Filter_specifications:_Low-pass.2C_high-pass.2C_band-pass|tutorial #10]]). == Simulation == We will illustrate the time-frequency decomposition process with a simulated signal. * The following code generates a sum of three sinusoids (2Hz, 20Hz, 50Hz) with random white noise. The 50Hz and noise are present everywhere, the 2Hz and 20Hz start only after two seconds.<<BR>> {{{ f1 = 2; f2 = 20; f3 = 50; i =2000:6000; Data(1,i) = sin(f1*2*pi*t(i)) + 0.4 * cos(f2*2*pi*t(i)); Data = Data + 0.2 * sin(f3*2*pi*t) + 0.4 * rand(1,6000); }}} * Empty the Process1 list (right-click > Clear list) then click on [Run]. * Run process: '''Simulate > Simulate generic signals'''. <<BR>>Ntime='''6000''', Sampling frequency='''1000Hz''' (signal duration = 6000/1000 = 6 seconds).<<BR>>Copy-paste the few lines of code above to generate the sum of three sinusoids.<<BR>><<BR>> {{attachment:simulate_process.gif}} * Double-click on the new file to look at the simulated signal.<<BR>><<BR>> {{attachment:simulate_display.gif}} * In Process1, select the simulated signal. * Run process: '''Frequency > Time-frequency (Morlet wavelets)'''. <<BR>>Select the option '''Spectral flattening''': The normalization will be discussed later.<<BR>><<BR>> {{attachment:example_process.gif}} * Click on the button ['''Edit'''] to see all the process options.<<BR>>Time definition: '''Same as input''', Frequency definition: '''Linear 1:1:60''', Compute measure: '''Power'''. <<BR>><<BR>> {{attachment:example_options.gif}} == Process options == |
| Line 53: | Line 62: |
| '''Time definition''': * This panel describes the time points for which you will get a value at the end of the computation. * '''Same as input file''': If you select this option, you would get the time-frequency decomposition (TF) for each of the time samples in the input file (here: 375 samples between -50ms to 250ms) * '''Group in time bands''': This option adds another step of computation: * It first computes the TF for each of the time points in the initial file, exactly like with the "same as input file" option * Then for each time band: it groups the corresponding time points by averaging the power of their TF values. * Definition of the time bands: two methods * Enter your own time bands in the text area, one line per time band, with following format: '''"name / time definition / function" ''' * Click on the "Generate" button to create automatically a list of time bands with the same length. You will be asked the maximal length of each time band. |
'''Time definition''' * '''Same as input file''': The output file has the same time definition as the input file. <<BR>>In this example, it means: 6000 samples between 0 and 6s. * '''Group in time bands''': This option adds a step of computation. First it computes the TF decomposition for all the input file, then averages the power by time band. To define a time band: * Enter your own time bands in the text area, one line per time band, with the following format: '''"name / time definition / function" ''' * Click on the button [Generate] to automatically create a list of time bands with the same length. You will be asked the maximal length of each time band. |
| Line 65: | Line 70: |
| '''Frequency definition''': * This panel describes the frequencies for which the power will be estimated at each time instant. * '''Linear''': You can specify the frequencies with any Matlab syntax start:step:stop. The default is "1:1:30", which produces 60 values [1, 2, 3, 4, ..., 59, 60] * '''Group in frequency bands''': As for the time definition, this option leads to a two-step process: * It computes the TF for several values in the frequency band (see below ho to define a frequency band) * Then for each frequency band: it averages the power of all the frequencies in the band, so that you get only one value for each frequency band (and for each time point and each sensor). * Definition of the frequency bands: |
'''Frequency definition''': Frequencies for which the power will be estimated at each time instant. * '''Linear''': You can specify the frequencies with the Matlab syntax start:step:stop. <<BR>>The default is "1:1:60", which produces 60 values [1, 2, 3, 4, ..., 59, 60]. * '''Log''': With the option start:N:stop, produces a list of N frequencies logarithmically scaled between "start" and "stop". For example "1:40:80" is converted to [1, 1.5, 2.1, 2.7, ..., 61.5, 65.8, 75, 80] * '''Group in frequency bands''': As for the time definition, this option leads to a two-step process. First it computes the TF decomposition for several values in the frequency band, then it averages the power of TF coefficients per frequency band. To define a frequency band: |
| Line 74: | Line 76: |
| * The name can be any string you want * The frequency definition is a Matlab expression evaluated with an eval() call. If the frequency definition contains only two values, Brainstorm adds two extra values in the middle so that the final averaged value is a bit more robust. * Examples of valid Matlab expressions: * "2,4": evaluates to [2,4], and as there are only two values, it uses the frequency vector [2, 2.66, 3.33, 4] instead * "2, 2.5, 3, 3.5, 4": evaluates to [2 2.5 3 3.5 4] * "2:4": evalues to [2 3 4] * "2:0.1:4": evaluates to a vector of 21 values [2 2.1 2.2 2.3 .... 3.8 3.9 4] * The function is the measure we take to combine the values for all the individual frequencies into one for the frequency band. Possible values are: '''mean, max, std, median'''. * Click on the "Reset" button to go back to the default settings. '''Morlet wavelet options''': * Central frequency: frequency where the mother wavelet is designed. All other wavelets will be shifted and scaled versions of the mother wavelet * Time resolution (FWHM): desirable temporal resolution of the wavelet at the central frequency (in units of Full Width Half Maximum). Click 'display' to see the resolution of the wavelet in other frequencies * See the first section of this tutorial for a description of those parameters. '''Processing options''': * '''Compute the following measure''': * The TF returns the wavelet coefficients for each frequency/time/sensor, which are complex values. Typically, what you want to display is the power of the coefficients, which is the square of the amplitude: abs(TF).^2. You can choose if you want to apply this transformation or not. * '''Power''': Applies the "power" transformation right after the computation. This causes a loss of the phase information. * '''None''': Save the TF coefficients as they are computed. This can be useful if you plan to use those decompositions for other purposes that require the phase, but the files are going to be twice as big. * If you select "None" here, the power will be computed on the fly for the various displays. * Some combinations of options may disable this choice. Example: if you ask for time bands or frequency bands, the program will have to compute the power before averaging the values; so in those cases the "None" option is disabled. * '''Output''': * '''Save individual TF maps''': This option stops the computation here, and saves in the database one time-frequency file for each input file. * Be very careful using this option because if you have hundreds of trials, it will fill your hard drive very quickly. * One full TF file in this case contains [nSensors x nTime x nFreq] = 3.397.500 values = 25 Mb. * You can however limit the amount of data produced by using time and/or frequency bands. For 15 time bands and 6 frequency bands, the size of each file drops to: 110 Kb. * '''Save average TF maps''': This option is available only when there are more than one input files (number of files you dropped in the Processes tab). * Instead of saving the TF for each file separately, it automatically computes the average of the power of all the TF. * This is a good choice if you do not plan to use independently all the TF files, because it saves a lot of time and disk space. === Computation === Back to our evoked fields. Let's select the following options: * Time: Same as input * Frequency: Linear 1:1:60 * Morlet wavelet: 1Hz / 3s * Output: Power / Individual TF maps (we want one map for the left, and one for the right) Start the computation: * Click on "Ok" * Back to the Pipeline editor, click on "Run". * After a while, two new files will appear in the tree, representing the TF for the two averaged recordings of the two conditions. == Display time-frequency maps == Right-click on the TF file of the ''Left ''condition to see what are the possible display options . {{attachment:treePopupRecordings.gif}} === Time-freq: One channel === * This is the default view for the TF decompositions of full recordings, ie. what would be displayed if you double-click on this file. It shows for one sensor (or one scout/source) the power of each frequency at each time. * When you load time-frequency related files, the "Display" tab automatically appears in the main Brainstorm window to offer some options for the displays, together with the "current frequency" slider. * '''Current frequency''': Slider that shows the current frequency selected in ALL the figures. <<BR>>Exactly as the time is Brainstorm, the frequency selection/display is centralized and managed by one control only for all the figures. As a consequence, it is impossible to display TF files with different frequency definitions at the same time. This can be perceived as an annoying limitation, but it allows all the simultaneous displays to be consistent at anytime, and make the interface much more intuitive to manipulate, with lower risks of mistakes in the interpretation of the different figures. <<BR>><<BR>> {{attachment:panelTimefreq.gif}} === Display tab === * '''List of sensors''': This drop-down list shows the current entry displayed in the selected figure. In this case, it is by default the MEG sensor "MLC11". Try to change it, it will update the display of the figure, showing the frequency power for another sensor.<<BR>>This selection is relative only to the selected figure, and gets updated when you selected another TF figure. <<BR>><<BR>> {{attachment:timeFreqSensor2.gif}} - {{attachment:timeFreqSensor3.gif}} * '''Hide edge effects''': When this option is selected, the time-frequency coefficients that could not be properly estimated because of a lack of time samples are hidden. It allows you to see only the information that is really reliable in the selected figure. In this example, you can see that it masks a large portion of the window: 375 time samples are not enough to perform a correct time-frequency decomposition. Try to work on larger time windows when using those tools.<<BR>><<BR>> {{attachment:timeFreqHide.gif}} * '''Measure''': The selection reflects the type of measure that is currently represented in the selected figure. The entries that are enabled depend on the type of data that is saved in the file. In this case, we saved directly the power of the wavelet coefficients in the file, we discarded the angle/phase information, hence the "phase" option is disabled. If we name "power" the values that are saved in the file in the database, the other options would perform the following operations: * '''Power''': Shows the power values unchanged - (left figure) * '''Magnitude''': Shows sqrt(power) - (right figure) * '''Log''': Shows 10 * log10(power) * {{attachment:timeFreqPower.gif}} - {{attachment:timeFreqSqrt.gif}} === Mouse and keyboard === |
* The frequency definition is a Matlab expression evaluated with an eval() call. If the frequency definition contains only two values, Brainstorm adds two extra values in the middle so that the final averaged value is a bit more robust. Example of valid expressions:<<BR>>"'''2,4'''": Evaluates to [2,4], and then expands to the frequency vector [2, 2.66, 3.33, 4]<<BR>>"'''2:0.5:4'''": Evaluates to [2 2.5 3 3.5 4]<<BR>>"'''2, 2.5, 3, 3.5, 4'''": Evaluates to [2 2.5 3 3.5 4] * The function is the measure we take to combine the values for all the individual frequencies into one for the frequency band. Possible values are: '''mean, max, std, median'''. '''Morlet wavelet options''' * '''Central frequency''': Frequency where the mother wavelet is designed. All other wavelets will be shifted and scaled versions of the mother wavelet * '''Time resolution (FWHM)''': Temporal resolution of the wavelet at the central frequency (in units of Full Width Half Maximum). Click [Display] to see the resolution of the wavelet for other frequencies. '''Compute the following measure''': * The convolution of the signal with complex Morlet wavelets returns the complex coefficients for each frequency/time/sensor. Typically, what we display is the power of the coefficients (square of the amplitude: abs(TF)^2^). You can choose if you want to apply this transformation or not. * '''Power''': Computes the "power" transformation immediately after the TF decomposition. This discards the phase information, but produces files that are twice smaller and a lot easier to process. * '''Magnitude''': Save the magnitude of the complex values instead of the power: abs(TF). * '''None''': Save the TF coefficients as they are computed (complex values). This can be useful if you plan to use these decompositions for other purposes that require the phase. * Some combinations of options may disable this choice. If you select frequency bands, the program will have to compute the power before averaging the values, therefore "none" is not an option. == Display: Time-frequency map == * Right on the new time-frequency file > '''Time-freq: One matrix''' (same as double-clicking). <<BR>>This menu displays the time-frequency decomposition of the first (and only) signal. The Brainstorm window shows two new elements: the tab "Display" and the frequency slider. <<BR>><<BR>> {{attachment:example_file.gif}} * We can easily identify the three horizontal bars as the three sinusoids in the simulated signals, and the trade-off between accuracy in time and accuracy in frequency. Set the measure to "magnitude" in the Display tab if you don't. Click on the figure to move the time-frequency cursor and explore the two axes of the plane. * '''2Hz''': High frequency resolution but poor time resolution (supposed to start sharply at 2s) * '''20Hz''': Better time resolution but poorer frequency resolution (17-24Hz) * '''50Hz''': Continuous over the 6s - Frequency resolution gets even worse (40-60Hz). It looks discontinuous because this oscillation has the same amplitude as the white noise we added in the signal (weight 0.4, relatively to the 2Hz oscillation). * '''Current frequency''': Slider that shows the current frequency selected in all the figures. <<BR>>Just like the time, the frequency selection is centralized and managed by one control only for all the figures. As a consequence, it is impossible to display TF files with different frequency definitions at the same time. This can be perceived as an annoying limitation, but it allows all the simultaneous displays to be consistent at anytime and makes the interface more intuitive to manipulate, with lower risks of mistakes in the interpretation of the different figures. * '''List of signals''': This drop-down list shows the signal currently displayed in the selected figure. In this case, there is only one channel of data called "s1". It will be more useful later. * '''Hide edge effects''': When this option is selected, the time-frequency coefficients that could not be properly estimated because of a lack of time samples are hidden. It allows you to see only the information that is really reliable. The lower the frequency, the longer the edge effects. In the screen capture below, the colormap has been changed to "jet" and the maximum set manually to 0.2 (measure=power).<<BR>><<BR>> {{attachment:example_hide.gif}} * '''Smooth display''': Re-interpolates the time-frequency maps on a finer grid to produce nicer plots. * '''Measure''': Type of measure that is currently represented in the selected figure. The entries that are enabled depend on the type of data that is saved in the file. In this case, we saved directly the power of the wavelet coefficients in the file, we discarded the angle/phase information, so the "phase" option is disabled. The other options are: '''Magnitude''' = sqrt(power), '''Log''' = 10*log10(power) * '''Colormap''': As explained in the previous tutorials, you can change the colormap by clicking+moving on the colorbar on the right of the figure. Double-click on the colorbar to restore the defaults. == Display: Mouse and keyboard shortcuts == ==== Mouse shortcuts ==== |
| Line 142: | Line 110: |
| * If you click anywhere on a TF figure, it will move the current time/frequency cursor to the clicked point, update the time and the frequency sliders in the Brainstorm window, and also update all other visualization figures. * If you change the current time or the current frequency with the sliders or with another figure, you would observe that the time/frequency cursor position is updated. * '''Left-click + move''': Select a time/frequency range. * Hold the left mouse button and move the mouse to mark a selection in both time and frequency. The legends of the X and Y axis are updated to display the selected time and frequency ranges selected.<<BR>><<BR>> {{attachment:timefreqSelection.gif}} * This selection is used by some functions accessible in the figure's popup menu. * '''Zoom''': * '''Mouse wheel''': Zoom/unzoom in time, centered on the current time cursor. * '''Control + mouse wheel''': Zoom/unzoom in frequencies, centered on the current frequency cursor. * '''Right-click + move''', or Control + left-click + move: Move in the zoomed image. * '''Double-click''': Restore initial view. * __Note for MacBook users__: all the actions involving the mouse wheel can be performed on MacBook laptops by moving up and down two fingers on the pad. <<BR>><<BR>> {{attachment:timeFreqZoom.gif}} * '''Colormap''': * As explained in the previous tutorials, you can change the colormap by left-clicking on the colorbar on the right of the figure, holding the button, and then by moving the mouse. * Up/down: Changes the brightness of the colormap. * Left/Right: Changes the contrast. * This has the same effect than changing the brightness/contrast using the popup menu.<<BR>><<BR>> {{attachment:timeFreqColormap1.gif}} - {{attachment:timeFreqColormap2.gif}} * '''Popup menu''': Right-click on the figure. * {{attachment:figurePopup.gif}} * All the menus have been described in the [[Tutorials/TutExploreRecodings|tutorial #4 "Exploring the recordings"]], except for the "Time-frequency selection menu":<<BR>>(You can see this full menu only if there is a selection defined in the figure) * '''Set selection manually''': Does the same thing as drawing a time/freq selection square on a figure, but by typing the values for time and frequency manually. * '''Export to database''': Save the selection for the displayed sensor in a new TF file in the database. However, it is necessary to keep the same time definition for this new file because as it has to stay compatible with the other files it is linked with (here: the recordings file). So in this case, only the selected frequencies are saved, but the time selection is ignored, all the time points are saved. <<BR>>Remember that only one sensor is exported using this menu, all the others are ignored. * '''Export to file''': Same as "Export to database", but the saved file is not registered in the database. The user has to select where to save the file. Unlike the previous menu, only the selected times AND selected frequencies are exported. * '''Export to Matlab''': Same as "Export to file", but the output structure is sent to a variable in the Matlab base workspace instead of being saved to a file. * Keyboard shortcuts: * '''Left and right arrows''': Change the current time * '''Page-up and page-down''': Also change the current time, but faster (10 time samples instead of one) * '''Up and down arrows''': Change the the sensor displayed in this figure (same as using the drop-list in the Display tab) * '''Control + up and down arrows''': Change the current frequency * '''Enter''': View the original time series for this sensor. * '''Control + R''': View the original MEG recordings (this shortcut and the following are visible in the popup menu) * '''Control + T''': View the time-frequency 2D topography (explained in the next sections) * '''Control + I''': Save as image * '''Control + D''': Dock figure in the Matlab environment. == Display multiple channels == Three menus display the same information (the TF maps of all the channels) with different spatial organizations. * '''Time-freq: All channels''': All the maps are displayed one after the other, in the order they are saved in the file. <<BR>><<BR>> {{attachment:timeFreqAllSensors.gif}} * '''Time-freq: 2D Layout''': Similarly to the "2D Layout" for the recordings, it shows each the position of each TF map where the sensor is located on a flattened 2D map. <<BR>><<BR>> {{attachment:timeFreqAllSensors2.gif}} * '''Time-freq: 2D Layout (no overlap)''': Similar to the "2D Layout", but reorganizing the positions of the small images so that they do not overlap. Overlapping in the "2D Layout" views is not a problem with this old CTF system (151 sensors); it gets much worse with the newer CTF systems (275 sensors), or the Elekta-Neuromag systems (102 chips with 3 sensors at each position). This display is a good intermediate between seeing all the information, and having it spatially organized. <<BR>><<BR>> {{attachment:timeFreqAllSensors3.gif}} What to do with these figures: * Maximize the window to be able to see something * '''Click''': clicking on any small TF image opens a new figure to display only the selected sensor * '''Shift + click''': Same, but opens the original recordings time series of the sensor instead of the TF * '''Mouse wheel''': Zoom / unzoom * '''Left+right mouse click + move''': Move in a zoomed figure. == Power spectrum and time series == Those two menus generate similar figures: they represent as a line the evolution of each sensor for one parameter (time or frequency) when the other parameter is fixed. * Power spectrum: For the given time, show the evolution of all the sensors across the frequencies * Time series: For the given frequency, show the evolution of all the sensors in time Open at the same time three figures for the Left/ERF file: * Time-freq: One channel * Power spectrum * Time series. Then navigate in time and frequencies, and observe how each figure gets updated at each change. . {{attachment:timeFreqSpectrum.gif}} === 2D topography === Right-click on the TF file in condition ''Left ''and select successively the first three topography menus. All those three windows represent the same information, in a slightly different way: a spatial map of the power of the current frequency, for all the sensors at the current time. * 3D Sensor cap * 2D Sensor cap * 2D Disc Try to move the time and frequency sliders and see what happens. You can open at the same time a "Time-freq: One channel" view, to keep track easily of the current time and frequency. . {{attachment:timeFreqTopo.gif}} Keyboard shortcuts: * '''Left and right arrows''': Change the current time * '''Page-up and page-down''': Also change the current time, but faster (10 time samples instead of one) * '''Up and down arrows''': Change the current frequency * '''Control + E''': Display the sensors markers * '''Control + E again''': Display the sensors names * '''Shift + Click on a sensor''': Displays the time-frequency decomposition for that specific sensors. {{attachment:topoSelectSensor.gif}} {{attachment:topoSelectSensorTf.gif}} === 2D Layout === The last display mode available for those TF decomposition of recordings is this "2D Layout" topography menu. Right-click on the TF file for the Left condition and select "2D Layout". This represents spatially the power of the current frequency for all the sensors and all the time points. Try to move the current frequency slider to see how the display changes when increasing the frequency. It is a good example to show that the time resolution increase with the frequency. Below: the "2D Layout" for f=8Hz and f=60Hz. . {{attachment:topo2DLayout1.gif}} {{attachment:topo2DLayout2.gif}} Useful operations for this window: * '''Mouse wheel''': Zoom / unzoom * '''Left+right click + move''': (or middle click+move) Move into the zoomed window * '''Shift+mouse wheel''': Increase/decrease the sharpness of the time series * '''Control + mouse wheel''': Increase/decrease the length of the time window displayed around the current time. * '''Click on a line''': Select a sensor * '''Right click + move''': Select a group of sensors * '''Shift + click on a line''': Display the time-frequency decomposition for the selected sensor * '''Right-click''': Display popup menu. More options in the "2D Layout options" sub-menu. * '''Control + E''': Display the channels names * All the other shortcuts and the popup menu are already described in [[Tutorials/TutExploreRecodings|tutorial #4 "Exploring the recordings"]]. == Contents of the "timefreq" files == Right click one of the TF files, and select the menu File > View file contents, to have a look to what is the actual contents of those structures. . {{attachment:viewMatFile.gif}} * '''TF''': [nSensors x nTime x nFrequency] matrix containing all the values of the time-frequency decomposition (complex wavelet coefficients, or power of amplitudes) * nSensors: Number of sensors for which the TF has been estimated * nTime: * If there are time bands defined, nTime = size(TimeBands,1) = number of time bands * If there are no time bands (linear time sampling): nTime = length(Time) * nFrequency: * If there are frequency bands defined, nFrequency = size(Freqs,1) = number of frequency bands * If there are no frequency bands (linear frequency sampling): nFrequency = length(Freqs) |
* '''Left-click + move''': Select a time/frequency range. The legends of the X/Y axis show the selection. <<BR>><<BR>> {{attachment:example_select.gif}} * '''Mouse wheel''': Zoom in time, centered on the current time cursor. * '''Control + mouse wheel''': Zoom in frequencies, centered on the current frequency cursor. * '''Right-click + move''', or Control + left-click + move: Move in the zoomed image. * '''Double-click''': Restore initial view. ==== Keyboard shortcuts: ==== * '''Left/right arrows''': Change the current time. * '''Page-up/page-down''': Change the current time, 10 time samples at a time. * '''Up/down arrows''': Change the the sensor displayed in this figure. * '''Control + up/down arrows''': Change the current frequency. * '''Enter''': View the original time series for this sensor. * '''Control + R''': View the original MEG recordings. * '''Control + T''': View the time-frequency 2D topography. * '''Control + I''': Save as image. * '''Control + D''': Dock figure in the Matlab environment. ==== Figure popup menu ==== . {{attachment:example_popup.gif||width="480",height="191"}} * '''Set selection manually''': Does the same thing as drawing a time/freq selection square on a figure, but by typing the values for time and frequency manually. * '''Export to database''': Save the selection for the displayed sensor in a new file in the database. * '''Export to file''': Same as "Export to database", but the saved file is not registered in the database. * '''Export to Matlab''': Same as "Export to database", but the output structure is sent to a variable in the Matlab base workspace instead of being saved to a file. == Display: Power spectrum and time series == Right-click on the file in the database or directly on the time-frequency figure to access these menus. * '''Power spectrum''': For the current time, shows the power for all the frequencies. * '''Time series''': For the current frequency, shows the power for all the time samples. * Example: Power spectrum density at '''0.5s''' and power time series at '''2Hz'''. <<BR>>We see the oscillation at the 50Hz in the PSD plot, and the oscillation at 2Hz in the TS plot.<<BR>><<BR>> {{attachment:example_psd1.gif}} * Example: Power spectrum density at '''4s''' and power time series at '''20Hz'''. <<BR>>We see all three oscillations in the PSD plot, and the oscillation at 20Hz in the TS plot.<<BR>><<BR>> {{attachment:example_psd2.gif}} * Note that if you right-click on the file in the '''database explorer''' and then select one of these menus, it will show '''all the signals'''. If you right-click on an existing '''time-frequency figure''', it will show only the '''selected signal'''. It doesn't make any difference here because there is only one signal, but it will with the MEG recordings. == Normalized time-frequency maps == The brain is always active, the MEG/EEG recordings are never flat, some oscillations are always present in the signals. Therefore we are often more interested in the transient changes in the power at certain frequencies than at the actual power. A good way to observe these changes is to compute a '''deviation of the power with respect with a baseline'''. There is another reason for which we are usually interested in standardizing the TF values. The power of the time-frequency coefficients are always lower in the higher frequencies than in the lower frequencies, the signal carries a lot less power in the faster oscillations than in the slow brain responses. This '''1/f decrease in power''' is an observation that we already made with the power spectrum density in the [[Tutorials/ArtifactsFilter|filtering tutorial]]. If we represent the TF maps with a linear color scale, we will always see values close to zero in the higher frequency ranges. Normalizing each frequency separately with respect with a baseline helps obtaining more readable TF maps. ==== No normalization ==== The values we were looking at were already normalized (checkbox "Spectral flattening" in the process options), but not with respect to a baseline. We will now compute the non-normalized power obtained with the Morlet wavelets and try the various options available for normalizing them. * In Process1, keep the simulated signal selected. * Run process: '''Frequency > Time-frequency (Morlet wavelets)''', No spectral flattening.<<BR>><<BR>> {{attachment:example_process_none.gif}} * Double-click on the file. As expected, we only see the lower frequencies in this representation: the power of the 2Hz oscillation is a lot larger than the power at 20Hz or 60Hz. <<BR>><<BR>> {{attachment:example_display_none.gif}} ==== Spectrum normalization ==== * In Process1: Select this new non-normalized file "Power,1-60Hz". * Run process: '''Standardize > Spectrum normalization''', Method='''1/f compensation'''. <<BR>><<BR>> {{attachment:norm_multiply.gif}} * This produces exactly the same results as previously (option "Spectral flattening" in the time-frequency process). It '''multiplies the power''' at each frequency bin '''with the frequency''' value (eg. multiplies by 20 the power at 20Hz), in order to correct for the 1/f shape we observe in the power spectrum. This works well for the lower part of the spectrum and up to 60-80Hz, but past this range it tends to '''overcompensate the higher frequencies'''.<<BR>>Note that it does not do any form of baseline correction: the 50Hz oscillation is visible everywhere. ==== Baseline normalization ==== * The second way to proceed is to normalize the power with respect to its average level during a reference time period. We can consider the oscillations at 2Hz and 20Hz as our events of interest, and the 50Hz as noise we want to get rid of. The segment from '''0 to 2 seconds''' does not contain any of the signals of interest, therefore we can consider it as a '''baseline'''. * However, we will not be able to use the full segment [0,2]s because of the edge effects we described at the beginning of this tutorial. The time-frequency map at '''2Hz''' with the display option "Hide edge effects" (left figure below) shows that the power could not be estimated correctly before 0.75s, therefore we shouldn't use it as a baseline. The power time series at 2Hz (right) shows that the power related with the 2Hz oscillation starts to increase significantly after 1.25s, therefore it's not really a "baseline" anymore. This leaves only the segment '''[0.75,1.25]s''' available. <<BR>><<BR>> {{attachment:norm_baseline_2hz.gif}} * At '''20Hz''', the expected transient effects are only 75ms long, therefore we could use a much longer baseline if we were not interested by the lower frequencies: '''[0.075, 1.925]s'''. <<BR>><<BR>> {{attachment:norm_baseline_20hz.gif}} * In this case, we have a very long "resting" time segment (2s), therefore the edge effects are not a problem for picking a time window for the baseline normalization. We will use the first time window mentioned, [0.75,1.25]s as it is long enough (500ms) to estimate the baseline mean power. In real-life cases, with shorter epochs, it is sometimes difficult to find an acceptable '''trade-off between the baseline duration and the exclusion of the edge effects''', especially for the lower frequencies. * Run process: '''Standardize > Baseline normalization''', Baseline='''[0.75, 1.25]s'''.<<BR>>Method='''Event-related perturbation''': ERS/ERD stands for "event-related synchronization / desynchronization", a widely used normalization measure for time-frequency power maps. It evaluates the deviation from the mean over the baseline, in percents: '''(x-mean)/mean*100'''. <<BR>><<BR>> {{attachment:norm_ersd_process.gif}} * Double-click on the new "ersd" file. The colormap type changed from "'''Timefreq'''" to "'''Stat2'''", which uses by default the "'''rwb'''" color set and shows relative values. Indeed, the ERS/ERD values can be '''positive or negative''': the power at a given time sample can be higher or lower than the average level during the baseline. In the simple simulation we used, there is no power decrease at any frequency after 2s, so the strong values are mostly positive. However, if you look in the file, you would see that there are many small negative values (due to the random noise we added). <<BR>><<BR>> {{attachment:norm_ersd_file.gif}} * Note that the '''50Hz disappeared''' because it was present during the baseline, while the 2Hz and 20Hz oscillations show high positive values (between 100% and 2000% increase from baseline). * Remember to select your baseline very carefully according to the frequency range you are interested in. See below examples obtained with different baselines: '''0.075-1.925s''', '''0-2s''', '''0-6s'''. <<BR>>Change the colormap to "jet" if you prefer, and adjust the colormap contrast as needed. <<BR>><<BR>> {{attachment:norm_ersd_compare.gif}} * This video may help you understand better what are the implications of the baseline selection: http://www.mikexcohen.com/lecturelets/whichbaseline/whichbaseline.html <<TAG(Advanced)>> == Tuning the wavelet parameters == ==== Time resolution ==== You can adjust the relative time and frequency resolution of your wavelet transformation by adjusting the parameters of the mother wavelet in the options of the process. * Increasing the option "time resolution" will produce longer wavelets at a given frequency, hence increase the frequency accuracy (lower <<HTML(Δ)>>f) and decrease the time accuracy (higher <<HTML(Δ)>>t). Expect longer edge effects. * Decrease the time resolution will produce shorter wavelets at a given frequency, hence decrease the frequency accuracy (higher <<HTML(Δ)>>f) and increase the time accuracy (higher <<HTML(Δ)>>t). Expect shorter edge effects. * You can modify one or the other parameter, what is important is the '''product of the two'''. All the following combinations fc/FWHM_t produce the same results because their product is constant: (1Hz,3s), (3Hz,1s), (6Hz,0.5s), (60Hz,0.05s) <<BR>><<BR>> {{attachment:tf_resolution_process.gif}} Examples for a constant central frequency of '''1Hz''' with various time resolutions: '''1.5s''', '''4s''', '''10s'''. . {{attachment:tf_resolution_compare.gif||width="599",height="183"}} ==== Frequency axis ==== You can also obtain very different representations of the data by changing the '''list of frequencies''' for which you estimate the power. You can change this in the options of the process. . {{attachment:tf_freqlist_process.gif}} Examples: Log 1:20:150, Log 1:300:150, Linear 15:0.1:25 . {{attachment:tf_freqlist_compare.gif||width="598",height="183"}} <<TAG(Advanced)>> == Hilbert transform == We can repeat the same analysis with the other approach available for exploring the simulated signal in the time-frequency plane. The process "'''Frequency > Hilbert transform'''" first filters the signals in various frequency bands with a band-pass filter, then computes the Hilbert transform of the filtered signal. The magnitude of the Hilbert transform of a narrow-band signal is a measure of the envelope of this signal, and therefore gives an indication of the activity in this frequency band. . {{attachment:slide_hilbert.gif}} ==== No normalization ==== Let's compute the same three results as before: non-normalized, spectral flattening, baseline normalization. * In Process1, select the simulated signal. * Run process: '''Frequency > Hilbert transform''', No spectral flattening, Do not mirror. <<BR>><<BR>> {{attachment:hilbert_process.gif}} * In the advanced options panel, keep the default options: Default frequency bands and '''Power'''.<<BR>><<BR>> {{attachment:hilbert_options.gif}} * Double-click on the new file. The figure now has only 6 rows, one for each frequency band. <<BR>><<BR>> {{attachment:hilbert_file.gif}} * The non-normalized results are already easy to interpret: * '''delta (2-4Hz)''': Includes the 2Hz oscillation, contribution starts at 2s * '''beta (15-29Hz)''': Includes the 20Hz oscillation, contribution starts at 2s * '''gamma1(30-59Hz)''': Includes the 50Hz oscillation, contribution starts at the beginning (0s) * Right-click on the file or figure > Time series. Example for delta and beta.<<BR>><<BR>> {{attachment:hilbert_ts.gif}} ==== Normalization ==== * In Process1, select the non-normalized Hilbert-based decomposition. * Run process: '''Standardize > Spectrum normalization''', Method='''1/f compensation'''. * Run process: '''Standardize > Baseline normalization''', Baseline='''[0.75, 1.25]s''', Method='''ERS/ERD'''. * Display the two normalized files side by side. <<BR>><<BR>> {{attachment:hilbert_normalized.gif||width="609",height="155"}} ==== Method specifications ==== * '''Band-pass filters''': Same filters as in the process "Pre-process > Band-pass filter", with the option "Stop-band attenuation: 60dB". For details, see the tutorial [[http://neuroimage.usc.edu/brainstorm/Tutorials/ArtifactsFilter#Filter_specifications:_Low-pass.2C_high-pass.2C_band-pass|Power spectrum and frequency filters]]. * '''Edge effects''': To estimate the duration of the transient effects for each frequency band, select the process "Band-pass filter", enter the frequency band of interest and click "View filter response". Example for the alpha band: <<BR>><<BR>> {{attachment:hilbert_filter_alpha.gif}} * '''Hilbert transformation''': Using the [[http://www.mathworks.com/help/signal/ref/hilbert.html|Matlab's hilbert()]] function. * '''Extraction of the envelope''': Power of the complex Hilbert transform, abs(hilbert(x))^2^. == MEG recordings: Single trials == Let's go back to our auditory oddball paradigm and apply the concepts to MEG recordings. We will use all the trials available for one condition to estimate the average time-frequency decomposition. ==== Spectrum normalization ==== * In Process1, select all the '''deviant trials''' in '''Run#01'''. * Run process: '''Frequency > Time-frequency (Morlet wavelets)''', No spectral flattening. <<BR>>In the advanced options, select: '''Log 1:40:150''', '''Power''', Save '''average''' time-frequency maps. <<BR>><<BR>> {{attachment:tf_trials_process.gif||width="352",height="214"}} {{attachment:tf_trials_options.gif||width="300",height="318"}} * '''Save individual TF maps''': This option stops the computation here and saves in the database one time-frequency file for each input file (40 files), with one TF map for each scout. * '''Save average TF maps''': Instead of saving the TF for each file separately, it automatically computes the average of the power of all the TF. This is a good choice if you do not plan to use independently all the TF files, because it saves a lot of time and disk space. * '''Remove evoked response from each trial before computing TF''': This option first computes the average of all the trials, then subtracts this average from each trial before computing the time-frequency decomposition. This brings the signals to a slightly more stationary state, which may help for the evaluation of the frequency contents. ==== Baseline normalization ==== * Double-click on the new file '''Avg,Power,1-150Hz (MEG)'''. Select "'''Hide edge effects'''". <<BR>>In the drop-down list, select sensor '''MLP56''' (the one with the strongest respons at 90ms). <<BR>>Right-click on the TF figure > '''Time series'''. <<BR>><<BR>> {{attachment:tf_trials_evaluate.gif||width="559",height="163"}} * Defining a baseline is now a lot trickier than with 6s-long simulated signal. The epochs are only 600ms long, and the power at many frequencies could not be estimated correctly. If we want all the values to be "good" after 0s, '''we cannot use anything below 15Hz'''. If we want to normalize the values, we have to go even higher: '''30Hz''' '''if we want a baseline of 50ms''' before 0. * The epochs we use in this tutorial are '''too short to perform a correct time-frequency analysis'''. We should have imported at least 200ms more on each side, just for controlling the edge effects. You should always think carefully about the length of the epochs you import in your database if you are planning to run any form of frequency or time-frequency analysis. * For the purpose of illustrating the tools available in Brainstorm, we will keep on working with these short epochs. Let's try to do the best we can with what we have here. We could use a baseline of 50ms to get a correct estimation above 30Hz, but this is probably a bit too short. We propose to include some more of the baseline (75ms), hoping there are not major edge effects in this segment. * In Process1, select the average time-frequency file. * Run process: '''Standardize > Baseline normalization''', Baseline='''[-75, 0]ms''', Method='''ERS/ERD'''. <<BR>><<BR>> {{attachment:tf_trials_normalize.gif||width="517",height="285"}} * The new menus available to display this file are described in the next section. <<BR>><<BR>> {{attachment:tf_trials_popup.gif||width="625",height="194"}} ==== Things to avoid ==== * Avoid computing the time-frequency decomposition of the average of the trials, you would miss some of the '''induced response''', the brain activity in higher frequencies that is not strictly time-locked to the stimulus, and not aligned in phase across trials. Always prefer the computation of the average of the time-frequency power maps of each trials, as we did here. <<BR>>This is well documented in: [[http://www.iec-lnc.ens.fr/IMG/Files/Visual/Publications/2000_Review_IntJPsychol.pdf|Bertrand O, Tallon-Baudry C (2000)]]. * Avoid using the Hilbert transform approach on short recordings or averages, always use the wavelet approach in these cases. The band-pass filters used for the lower frequency bands may have very high orders, leading to long transients. The example below shows the expected transients for the default frequency bands using the process "Frequency > Hilbert transform", they can be much more invalidating than with the process "Frequency > Time-frequency (Morlet wavelets)". <<BR>><<BR>> {{attachment:hilbert_trials_transients.gif||width="216",height="128"}} == Display: All channels == Three menus display the time-frequency of all the sensors with different spatial organizations. All the figures below represent the '''ERS/ERD'''-normalized average. They use the "jet" colormap, which is not the default configuration for these files. To get the same displays, change the colormap configuration: <<BR>>right-click on the figure > Colormap: Stat2 > Colormap > '''jet'''. * '''All channels''': All the maps are displayed one after the other, in the order they are saved in the file. <<BR>><<BR>> {{attachment:tf_allchannels.gif||width="345",height="295"}} * '''2D Layout (maps)''': Show each TF map where the sensor is located on a flattened 2D map. Most display options are available, such as the colormap management and the option "Hide edge effects".<<BR>><<BR>> {{attachment:tf_2dlayout_maps.gif||width="329",height="279"}} {{attachment:tf_2dlayout_hide.gif||width="328",height="279"}} * '''2D Layout (no overlap)''': Similar to the the previous display, but the positions of the images are reorganized so that they do not overlap. <<BR>><<BR>> {{attachment:tf_2dlayout_nooverlap.gif||width="345",height="294"}} * '''Image [channel x time]''': Shows the values of all the sensors over time for one frequency. <<BR>><<BR>> {{attachment:tf_image.gif||width="314",height="203"}} Useful shortcuts for the first three figures: * '''Click''': Clicking on any small TF image opens a new figure with only the selected sensor. * '''Shift + click''': Opens the original recordings time series of the selected sensor, when available. Here, we display an average of time-frequency maps, so this menu has no effect. * '''Mouse wheel''': Zoom in/out. * '''Right click + move''': Move in a zoomed figure. == Display: Topography == The menus below show the distribution of TF power over the sensors, for one time point and one frequency bin, very similarly to what was introduced in tutorial [[Tutorials/ExploreRecordings|Visual exploration]]. * '''2D Sensor cap''' / '''2D Disc''' / '''3D Sensor cap''': 175ms, 8Hz<<BR>><<BR>> {{attachment:tf_topo.gif||width="474",height="137"}} * '''2D Layout''': 8Hz (black), 35Hz (white)<<BR>><<BR>> {{attachment:tf_2dlayout8.gif||width="260",height="241"}} {{attachment:tf_2dlayout35.gif||width="259",height="240"}} Useful shortcuts for these figures: * '''Left/right arrows''': Change the current time. * '''Up/down arrows''': Change the current frequency. * '''Control + E''': Display the sensors markers/names. * '''Shift + click on a sensor''': Displays the time-frequency decomposition for that specific sensors. * '''Right click + move''': Select a group of sensors. * '''Shift +'''''' scroll''': Change the gain of the time series (2D Layout). * '''Control +'''''' scroll''': Change the length of the window displayed around the current time (2D Layout). <<TAG(Advanced)>> == Scouts == Similar calculations can be done at the level of the sources, either on the full cortex surface or on a limited number of regions of interests. We will start with the latter as it is usually an easier approach. * Drag and drop all the '''deviant trials''' from '''both runs''', select ['''Process sources''']. * Run process "'''Frequency > Time-frequency (Morlet wavelets)'''". <<BR>>Select the option "Use scouts" and '''select all the scouts''' defined in the previous tutorial. <<BR>><<BR>> {{attachment:scouts_process.gif||width="420",height="381"}} * In the advanced options, select "Scout function: '''After'''" and "Output: '''Save average'''". <<BR>>Run the process (it may take a while).<<BR>><<BR>> {{attachment:scouts_options.gif||width="361",height="598"}} * The '''scout function''' was introduced in the previous tutorial. It is the method we use to group the time series for the 20 dipoles we have in each scout into one unique signal. When computing the TF of one scout, we have the choice between applying this function before or after the time-frequency decomposition itself. * '''Before''': Extract the 20 source signals, apply the scout function to get one signal, run the TF decomposition of this signal. This is faster but may lead to information loss. * '''After''': Extract the 20 source signals, run the TF decomposition of the 20 signals, apply the scout function on the power of the TF maps. Always prefer this option when possible. * Rename the new file to add a tag "Deviant" in it. Then right-click > '''Time-freq: All scouts'''. <<BR>><<BR>> {{attachment:scouts_file.gif||width="505",height="193"}} * In Process1, select the new average TF file. * Run process: '''Standardize > Baseline normalization''', Baseline='''[-75, 0]ms''', Method='''ERS/ERD'''. <<BR>><<BR>> {{attachment:scouts_ersd.gif||width="562",height="186"}} <<TAG(Advanced)>> == Full cortical maps == Computing the time-frequency decomposition for all the sources of the cortex surface is possible but complicated because it can easily generate gigantic files, completely out of the reach of most computers. For instance the full TF matrix for each trial we have here would be [Nsources x Ntimes x Nfrequencies] = [15000 x 361 x 40] double-complex = '''3.2 Gb'''! We have two ways of going around this issue: computing the TF decomposition for '''fewer frequency bins''' or frequency bands at a time, or as we did previously, use only limited number of '''regions of interest'''. * In Process1, keep all the '''deviant''' trials both conditions selected, select '''[Process sources]'''. * Run process "'''Frequency > Hilbert transform'''", No spectral flattening, '''Mirror signal''' before.<<BR>>To process the '''entire brain''', do '''not '''select the option "Use scouts".<<BR>><<BR>> {{attachment:sources_process.gif||width="432",height="398"}} * In the advanced options, select "Optimize storage: '''No'''", this option is not available when computing on the fly the average of multiple trials. Save the '''power''', Save the '''average '''Hilbert maps. <<BR>><<BR>> {{attachment:sources_options.gif||width="355",height="477"}} * '''Optimize the storage of the time frequency file''': Let's describe this option in more details. * When computing the TF decomposition of a source file, we are actually applying sequentially two linear transformations to the original recordings: the TF analysis and the source inversion. These two processes can be permuted: TF(Inverse(Recordings)) = Inverse(TF(Recordings)). * Therefore we can optimize the TF computation time by applying the wavelet transformation only to the sensor recordings, and then multiply the wavelet complex coefficients by the inverse operator (ImagingKernel). This trick is always used in the computation of the Hilbert and Morlet transforms. * When we have the option to the save the complex values (constrained sources and no averaging), this can also be used to optimize the storage of the files. In these cases, we save only the wavelet transformation of the sensor data. Later, when the file is loaded for display, the imaging kernel is applied on the fly. This can be disabled explicitly with this option. * Rename the new Hilbert file to include the tag "Deviant", and select it in Process1. * Run process: '''Standardize > Baseline normalization''', Baseline=[-75, 0]ms, Method='''ERS/ERD'''. * Right-click on the Hilbert file > '''Display on cortex'''. <<BR>>The frequency slider now shows frequency bands ("alpha:8-12Hz") instead of frequencies ("12Hz"). You can explore the source activity in time and frequency dimensions. The screen capture below shows the activity at 175ms: a 60% increase in the alpha band around the auditory cortex and a 20% decrease in the beta band around the motor cortex.<<BR>><<BR>> {{attachment:sources_display.gif||width="540",height="161"}} * '''Shift + click '''on the cortex surface: Displays the TF decomposition of the selected source. * '''Right-click on the brain''': Selects the closest vertex and displays the popup menu at the same time. The first three menus are relative to the source that was just clicked. <<BR>><<BR>> {{attachment:sources_vertex.gif||width="480",height="163"}} <<TAG(Advanced)>> == Unconstrained sources == In the current example, we are working with the simple case: sources with constrained orientations. The unconstrained case is more difficult to deal with, because we have to handle correctly the three orientations we have at each vertex. * '''Full cortex''': Computes the TF decompositions for all the sources (3*15000=45000), then sum at each location the power for the three orientations. * '''Scouts''': Option "Scout function" in the process. * '''Before''': Extract the 20*3=60 source signals, apply the scout function to get three signals (one per orientation), run the TF decomposition of the three signals, and finally sum the power of the three TF maps. This is faster but may lose some frequency resolution (especially for constrained sources). * '''After''': Extract the 20*3=60 source signals, run the TF decomposition of the 60 signals, apply the scout function on the power of the TF maps for each orientation separately, and finally sum the power obtained for the three orientations. * The storage optimization option is not available with unconstrained sources. <<TAG(Advanced)>> == Getting rid of the edge effects == To avoid making mistakes in the manipulation of the data and producing more readable figures, we encourage you to cut out the edge effects from your time frequency maps after computation. * In Process1, select the very first computed in this tutorial: Test/Simulation/Power,1-60Hz | multiply * Run the process: "'''Extract > Extract time'''", Time window = '''[0.75, 5.25]s''' * Open the new file, select the option "'''Hide edge effects'''": Almost everything left in this new file is correctly estimated. Brainstorm keeps track of the edge effects in the TFmask field of the file. <<BR>><<BR>> {{attachment:tf_cut.gif||width="463",height="166"}} * We recommend you do the same when epoching your recordings: import trials that are longer than necessary, and after the time-frequency estimation, remove the unnecessary segments. <<TAG(Advanced)>> == On the hard drive == Right click one of the first TF file we computed > File > '''View file contents'''. . {{attachment:tf_contents.gif||width="563",height="450"}} ==== Structure of the time-frequency files: timefreq_*.mat ==== * '''TF''': [Nsignals x Ntime x Nfreq] matrix containing all the values of the time-frequency decomposition (complex wavelet coefficients, or double values for power/magnitude/Z-score). * '''TFmask''': [Nfreq x Ntime] logical mask indicating the edge effects (0=edge, 1=valid value). * '''Std''': [Nsignals x Ntime x Nfreq] standard deviation if this file is an average. |
| Line 262: | Line 363: |
| * '''DataType''': Explains from what kind of data this file was computed. Possible values are: 'data', 'results', 'scout, 'matrix'... * '''Time''': Time vector used to estimate this file. * '''TimeBands''': Description of the time bands. Cell array {nTimeBands x 3}, where each line represents a time band {'band_name', 'time definition', 'function'} * '''Freqs''': * If frequency bands: Cell array {nFreqBands x 3}, where each line represents a frequency band {'band_name', 'frequency definition', 'function'} * In linear frequency sampling: vector containing all the frequencies * '''RowNames''': Cell array of strings that describes each row of the TF matrix. In this specific case, it would be the list of all the MEG sensor names. But it could also be a list of names of scouts or clusters. * '''Measure''': Contains the name of the function that was applied right after the computation of the wavelet coefficients. So it represents the type of data contained in the TF matrix. Possible values: * 'none': No measure applied, TF contains the complex wavelet coefficients. * 'power': TF contains the power for each frequency, ie. the square of the amplitude: abs(coefficients).^2 * 'magnitude': abs(coefficients) * 'log': 10 * log10( abs(coefficients).^2) * 'phase': angle(coefficients) * '''Method''': String that identifies the process that generated the file; possible values: {'morlet', 'fft', 'psd', 'hilbert', 'corr', 'cohere', 'granger', 'plv'} * '''DataFile''': Initial file from which this file was computed. In the database explorer, the TF file will be shown as a child of this DataFile file. * '''nAvg''': Number of trials that were averaged to obtain this file (copied from the recordings file at computation) |
* '''DataType''': From what kind of data this file was computed: {'data', 'results', 'scout, 'matrix'} * '''Time''': [1 x Ntime] Time vector used to estimate this file. * '''TimeBands''': [Ntimebands x 3] Cell array where each line represents a time band: <<BR>>{'band_name', 'time definition', 'function'} * '''Freqs''': For regular frequency binning: vector containing all the frequencies.<<BR>>If using frequency bands: [Nfreqbands x 3] cell array, where each line represents a frequency band {'band_name', 'frequency definition', 'function'} * '''RefRowNames''': Used only for connectivity matrices. * '''RowNames''': [1 x Nsignals] Cell array of strings that describes each row of the TF matrix. In this specific case, it would be the list of all the MEG sensor names. But it could also be a list of names of scouts or clusters. * '''Measure''': Contains the name of the function that was applied right after the computation of the wavelet coefficients. So it represents the type of data contained in the TF matrix. Possible values: * '''none''': No measure applied, TF contains the complex wavelet coefficients. * '''power''': Power for each frequency, ie. the square of the amplitude: abs(coefficients)^2^ * '''magnitude''': abs(coefficients) * '''log''': 10 * log10( abs(coefficients)^2^) * '''phase''': angle(coefficients) * '''Method''': String that identifies the process that generated the file: <<BR>>{'morlet', 'fft', 'psd', 'hilbert', 'corr', 'cohere', 'granger', 'plv'} * '''DataFile''': Initial file from which this file was computed. In the database explorer, the TF file will be shown as a child of this DataFile file. * '''SurfaceFile''' / '''GridLoc''' / '''GridAtlas''': Source space that was used, only for source files. * '''Leff''': Effective number of averages = Number of trials that were averaged to obtain this file. * '''ColormapType''': String, force a specific colormap type to be used when displaying this file. * '''DisplayUnits''': String, force to use specific units when displaying this file. |
| Line 279: | Line 383: |
| * '''History''': List of operations performed on this file (better visualized with popup menu File > View file history) == Time and frequency bands == === Frequency bands === * Drag and drop the averaged file ''Left'' /'' ERF ''in the Process tab, click on Run, select the process "Frequency > Time-frequency (Morlet wavelets)". Click on "Edit" to get the options. * Just change the frequency definition: select "Group in frequency bands", leave the default frequency bands, and click on Ok. Click on Run. * Right-click on the file you've just computed "Power,FreqBands" > "Time-freq: One channel".<<BR>><<BR>> {{attachment:freqBands.gif}} * You can notice that now the frequency selection is discreet, both in the figure and in the frequency slider, it is not possible to select a frequency in between. At each time and for each sensor, there are now just 6 values, one per frequency band. * Now if you try to open at the same time the first TF that you have computed "Power,1:1:60Hz", you would get an error message: This frequency definition is not compatible for display with the linear [1...60] scale. * Display the other possible views (All sensors, 2D Sensor Cap, 2D Layout), and change the current time and the current frequency in different ways (sliders, keyboard, clicks on figures). Play with the open figures until you are completely comfortable with those representations. === Time bands === * Compute TF again for file ''Left / ERF'', but this time select the option "Group in time bands" (with "linear" frequency definition). To generate a list of regular time intervals, click on "Generate" and enter "30" ms as the duration of each time band. * You get 11 time bands: t-2 and t-1 before 0ms, and t1 to t9 after 0ms. Click on Ok to start the computation. * Right-click on this new file "Power,1:1:60Hz" > Time-frequency maps.<<BR>><<BR>> {{attachment:timeBands.gif}} * Observe that the behavior is not the same as with the frequency bands: it is still possible to select a specific time in a time band (using the slider or clicking on the figure). This is because the time definition in the interface is still based on the ERF averaged file time definition, which is "continuous". Only 11 values in time are computed, and when requesting values a specific time point, the interface gets the values associated with the time band that time point belongs to. * As a consequence, it is possible to open at the same time the first TF file that you have computed: "Power,1:1:60Hz". You would not get an error message like with the frequencies mismatching. * Display also other types of figures (All sensors, 2D topographies), change current time and frequency several times in several ways (sliders, keyboard, clicks on figures). === Time bands and frequency bands === * Just for curiosity, compute a file with both time and frequency bands.<<BR>><<BR>> {{attachment:timeFreqTimeBand.gif}} <<HTML(<!--)>> == Clusters time series == * Create two clusters "c1" and "c2" for the Left condition: * Display a time series view for the ''Left / ERF'' average: * Select a few sensors, and click on the "New sel" button in the "Cluster" tab. * Press "Escape" to unselect all the selected sensors. * Select another group of sensors, and click on the "New sel" button again. * Drag and drop the ''Left / ERF'' file in the Processes list. Click on run. * Select the process "Frequency > Time-frequency (Morlet wavelets)". * As some clusters are now available, you can have access to them in this interface.<<BR>><<BR>> {{attachment:processesCluster.gif}} * Check the option "Use clusters time series", and select them both (control or shift button). * Click on the "Edit" button. A new option is displayed in the options window, "Cluster function"<<BR>><<BR>> {{attachment:clustersFunction.gif}} * The selection of the "cluster function" is done in the "Cluster" tab. * '''Before''': The selected function (in this case: "mean"), is applied to get only one time series for each cluster, and the TF is computed only for those processed time series. Faster, but may cause a loss of information, if the phase is slightly different for each sensor. * '''After''': The TF is computed for all the sensors, and then then the selected function (mean) is applied to the power of those TF to get only one TF decomposition for each cluster. Slower, but more representative of the power of each frequency. * When the selected function is "All", this option doesn't appear because there is no function applied, it computes the TF for all the sensors. * Other options: * Time: Same as input file * Frequency: Linear 1:1:60 * Measure: Power (None is disabled, if not, it is not possible to combine the values in clusters) * Click on OK. Click on Run. * A new file "Clusters,Power,1:1:60Hz" appears in the database. The size of TF matrix in this file is [nClusters x nTimes x nFrequencies] = [2 x 375 x 60]. * Right-click on it > "Time-freq: All clusters". The list in the "Display" tab now only contains two items: c1 and c2.<<BR>><<BR>> {{attachment:timeFreqClusters.gif}} <<HTML(-->)>> == Cortical sources == * How to display the TF decompositions for the source time series? * It is possible to estimate the TF for each source of the brain, but it would be unrealistic to save this information to a file. The size of the TF matrix would be [nVertices x nTimes x nFrequencies] = [15010 x 375 x 60] complex-double = 5.2 Gb! * We need then to simplify this problem. Two methods: compute the TF only for a few scouts (next section), or use the linear property of the TF decomposition. * As both the source reconstruction process (ImagingKernel * recordings) and the TF are linear operators, it is possible to exchange them:<<BR>>TF(Inverse(Recordings)) = Inverse(TF(Recordings))<<BR>>=> Power(TF(Inverse(Recordings))) = Power(Inverse(TF(Recordings))) * So the solution is to estimate the TF of the recordings, and then multiply it on the fly by the ImagingKernel only for the required sources and/or frequency and/or time instant. This is done in a completely transparent way from the user point a view. * Start the computation of the TF for a source file: * Drag and drop the source file computed for the ''Left / ERF'' file in ''Process1''. * Select the "Process sources" button. * Click on Run. Select "Time-frequency (Morlet wavelets)". Do NOT check the box "Use scouts time series". * Click on Edit. There is a new option available, that offers between saving only the kernel and the TF of sensors, or the full TF matrix. Select the optimized solution. * Note that this solution is not compatible with any of the operations that requires to compute explicitly the power of all the TF decomposition: time bands and frequency bands.<<BR>><<BR>> {{attachment:timeFreqKernel.gif}} * Click on Ok, then Run. A new file appears in the database explorer, as a child of the source file. Right-click on it, try all the visualization menus.<<BR>><<BR>> {{attachment:computeForSources.gif}} * If you understood well how to use the visualization of both the sources and the time-frequency maps, the manipulation of those figures should be rather intuitive. Change several times the current time and the current frequency in different ways. Change current time and frequency so that you can observe the main response in the somatosensory cortex (eg. t = 46.40ms, f = 45Hz). <<BR>><<BR>> {{attachment:timeFreqSources.gif}} * '''Shift + click '''on the cortex surface: This is a useful shortcut that you should remember. It displays the TF decomposition of the selected source.<<BR>><<BR>> {{attachment:selectionSource.gif}} * '''Right-click on the brain''': This selects the closest vertex and displays the popup menu at the same time. The first three menus are relative to the source that was just clicked. <<BR>><<BR>> {{attachment:popupSource.gif}} == Scouts time series == * The previous section introduces an interesting tool to display the TF for a single source, but it is hard to extract meaningful information that can be analyzed statistically or compared between subjects this way. A better option is to define cortical regions of interest (scouts, see [[Tutorials/TutScouts|tutorial #8]]), and compute the TF for the average time series of these scouts. * Define the scouts:<<BR>> * Double click for the source file in the Left condition. * Make sure that your LeftSS and RightSS scouts are still available in your cortex surface.If not, go back to tutorial 8, and recreate them <<BR>><<BR>> {{attachment:createScouts.gif}} * Compute the TF decomposition for these scouts: * Drag and drop the Left source file (''MN: MEG(Constr)'') into ''Process1''. * Select the "Process sources" button. * Click on Run. Select "Frequency > Time-frequency (Morlet wavelets)". * As some scouts are available for the input, you will see this new box:<<BR>><<BR>> {{attachment:selectScouts.gif}} * Select "Use scouts time series", and select both scouts in list. * Click on Edit. This time, most of the options are available again. We are going to compute the full TF matrix at once for the two scouts, so we can compute the power right away, together with time and/or frequency bands. * Leave all the options to the default (linear time, linear frequencies, scout function: After), then click on Ok, then Run. * A new file appears in the tree, as a child of the sources file: "Scouts,Power,1:1:60Hz". * To compute the TF of each scout, the operations performed by the program are the following: * Get time series for each source in the scout * If the scout function is applied __before__ the TF: * Apply the "scout function", that combines the different sources to get only one time series for each scout. * Compute the power of the TF for each scout. * If the scout function is applied __after__ the TF: * Compute the power of the TF for each source * Apply the "scout function", that combines the different sources TF to get one TF only for each scout. * The selection of the "scout function" is done in the "Scout" tab. * Right-click on this file and try all the visualization menus available (including "Time series" and "Power spectrum").<<BR>><<BR>> {{attachment:scoutsTfFile.gif}} == Processing time-frequency files: Z-score == * Time-frequency files, once computed, can be processed with ''Process1'' and ''Process2'', exactly as recordings or sources file, as introduced in [[Tutorials/TutProcesses|Tutorial #9]]. * Only the TF of sensors, clusters and scouts can be processed this way. * The TF of full source files, as explained above, are saved in the database in an incomplete form, that requires a few more on-the-fly computation steps. The processing engine requires an access to the full TF matrix, which is usually too big to be handled in Matlab for those TF/sources files. * If you need to average or do some group-analysis on time-frequency data at the source level, the problem has to be simplified a few steps before: for instance by doing your analysis only on a few scouts instead of the full cortex. * Drag and drop one of the TF files in Process1, select the "Time-frequency" button, click on Run.<<BR>><<BR>> {{attachment:panelProcess.gif}} * One of the process that is commonly applied on the time-frequency maps is the Z-score, to normalize the values in each frequency band with respect to a baseline. This way, all the frequency bands can be compared directly in terms of amplitude. If we look directly at the power of the Morlet coefficients, the values we observe are always lower in the high frequencies than in the low frequencies. The Z-score fixes this and helps getting figures that are easier to read. * Select the process "Standardize > Compute Z-score (static)". Leave the baseline to the default value (the pre-stimulus interval). Click on run. * Double-click on the new file.<<BR>><<BR>> {{attachment:timeFreqZscore.gif}} * Red values are positive, representing an increase in power with respect with the baseline, blue values are negative, representing a decrease. * The time-frequency map may appear all white or all white depending on the way the colormap is configured. To edit the colormap, right-click on the figure > Colormap. == Additional discussions on the forum == * Time and frequency resolution: http://neuroimage.usc.edu/forums/showthread.php?1848 <<EmbedContent("http://neuroimage.usc.edu/bst/get_prevnext.php?prev=Tutorials/Scouts&next=Tutorials/Statistics")>> |
* '''History''': List of operations performed on this file (menu File > View file history). ==== Useful functions ==== * '''in_bst_timefreq'''(TimefreqFile): Read a time-frequency file. * '''in_bst'''(FileName, TimeWindow): Read any Brainstorm file with the possibility to load only a specific part of the file. "TimeWindow" is an range of time values in seconds: [tStart, tStop]. * '''bst_process'''(''''LoadInputFile'''', FileName, Target, TimeWindow): The most high-level function for reading data files. "Target" is a string with the list of sensor names or types to load. * '''morlet_transform'''(): Applies complex Morlet wavelet transform to the time series in input. <<TAG(Advanced)>> == Additional documentation == ==== Articles ==== * Bertrand O, Tallon-Baudry C (2000)<<BR>> [[http://www.iec-lnc.ens.fr/IMG/Files/Visual/Publications/2000_Review_IntJPsychol.pdf|Oscillatory gamma activity in humans: a possible role for object representation]]<<BR>>Int J Psychophysiol, 38(3):211-23 * Bruns A (2004)<<BR>>[[http://www.sciencedirect.com/science/article/pii/S0165027004001098|Fourier-, Hilbert- and wavelet-based signal analysis: are they really different approaches?]]<<BR>>J Neurosci Methods, 137(2):321-32 * Le Van Quyen M, Foucher J, Lachaux J, Rodriguez E, Lutz A, Martinerie J, Varela FJ (2001) <<BR>>[[http://brainimaging.waisman.wisc.edu/~lutz/LeVanQuyen_et_all_JNM_2001.pdf|Comparison of Hilbert transform and wavelet methods for the analysis of neuronal synchrony]]<<BR>>J Neurosci Methods, 111(2):83-98 * Pantazis D, Weber DL, Dale CL, Nichols TE, Simpson GV, Leahy RM (2005)<<BR>>[[https://www.spiedigitallibrary.org/conference-proceedings-of-spie/5674/0000/Imaging-of-oscillatory-behavior-in-event-related-MEG-studies/10.1117/12.600607.short|Imaging of oscillatory behavior in event-related MEG studies]] ([[http://neuroimage.usc.edu/paperspdf/PantazisLeahySPIE2005.pdf|pdf]])<<BR>>in Proceedings of SPIE, Computational Imaging III, Vol. 5674, eds C. Bouman and E. Miller * Pfurtscheller G (1992)<<BR>>[[http://www.sciencedirect.com/science/article/pii/0013469492901333|Event-related synchronization (ERS): an electrophysiological correlate of cortical areas at rest]]<<BR>>Electroencephalogr Clin Neurophysiol, 83(1):62-9 * Series of lectures by Mike X Cohen: <<BR>>http://www.mikexcohen.com/lectures.html ==== Forum discussions ==== * Forum: Time and frequency resolution:<<BR>> http://neuroimage.usc.edu/forums/showthread.php?1848 <<HTML(<!-- END-PAGE -->)>> <<EmbedContent("http://neuroimage.usc.edu/bst/get_prevnext.php?prev=Tutorials/Scouts&next=Tutorials/Difference")>> |
Tutorial 24: Time-frequency
Authors: Francois Tadel, Dimitrios Pantazis, Elizabeth Bock, Sylvain Baillet
This tutorial introduces how to compute time-frequency decomposition of MEG/EEG recordings and cortical currents using complex Morlet wavelets and Hilbert transforms.
Contents
- Introduction
- Morlet wavelets
- Edge effects
- Simulation
- Process options
- Display: Time-frequency map
- Display: Mouse and keyboard shortcuts
- Display: Power spectrum and time series
- Normalized time-frequency maps
- Tuning the wavelet parameters
- Hilbert transform
- MEG recordings: Single trials
- Display: All channels
- Display: Topography
- Scouts
- Full cortical maps
- Unconstrained sources
- Getting rid of the edge effects
- On the hard drive
- Additional documentation
Introduction
Some of the MEG/EEG signal properties are difficult to access in time domain (graphs time/amplitude). A lot of the information of interest is carried by oscillations at certain frequencies, but the amplitude of these oscillations is sometimes a lot lower than the amplitude of the slower components of the signal, making them difficult to observe.
The averaging in time domain may also lead to a cancellation of these oscillations when they are not strictly locked in phase across trials. Averaging trials in time-frequency domain allows to extract the power of the oscillation regardless of the phase shifts. For a better understanding of this topic, we recommend the lecture of the following article: Bertrand O, Tallon-Baudry C (2000), Oscillatory gamma activity in humans: a possible role for object representation.
In Brainstorm we offer two approaches for computing time-frequency decomposition (TF): the first one is based on the convolution of the signal with series of complex Morlet wavelets, the second filters the signal in different frequency bands and extracts the envelope of the filtered signals using the Hilbert transform.
Morlet wavelets
Complex Morlet wavelets are very popular in EEG/MEG data analysis for time-frequency decomposition. They have the shape of a sinusoid, weighted by a Gaussian kernel, and they can therefore capture local oscillatory components in the time series. An example of this wavelet is shown below, where the blue and red curves represent the real and imaginary part, respectively.
Contrary to the standard short-time Fourier transform, wavelets have variable resolution in time and frequency. For low frequencies, the frequency resolution is high but the time resolution is low. For high frenqucies, it's the opposite. When designing the wavelet, we basically decide a trade-off between temporal and spectral resolution.
To design the wavelet, we first need to choose a central frequency, ie. the frequency where we will define the mother wavelet. All other wavelets will be scaled and shifted versions of the mother wavelet. Unless interested in designing the wavelet at a particular frequency band, the default 1Hz should be fine.
Then, the desirable time resolution for the central frequency should be defined. For example, we may wish to have a temporal resolution of 3 seconds at frequency 1 Hz (default parameters). These two parameters, uniquely define the temporal and spectral resolution of the wavelet for all other frequencies, as shown in the plots below. Resolution is given in units of Full Width Half Maximum of the Gaussian kernel, both in time and frequency.
Edge effects
Users should pay attention to edge effects when applying wavelet analysis. Wavelet coefficients are computed by convolving the wavelet kernel with the time series. Similarly to any convolution of signals, there is zero padding at the edges of the time series and therefore the wavelet coefficients are weaker at the beginning and end of the time series.
From the figure above, which designs the Morlet wavelet, we can see that the default wavelet (central frequency fc=1Hz, FWHM_tc=3sec) has temporal resolution 0.6sec at 5Hz and 0.3sec at 10Hz. In such case, the edge effects are roughly half these times: 300ms in 5Hz and 150ms in 10Hz.
More precisely, if f is your frequency of interest, you can expect the edge effects to span over FWHM_t seconds: FWHM_t = FWHM_tc * fc / f / 2. Examples of such transients are given in the figures below.
We also need to consider these edge effects when using the Hilbert transform approach. The band-pass filters used before extracting the signal envelope are relatively narrow and may cause long transients. To evaluate the duration of these edge effects for a given frequency band, use the interface of the process "Pre-process > Band-pass filter" or refer to the filters specifications (tutorial #10).
Simulation
We will illustrate the time-frequency decomposition process with a simulated signal.
The following code generates a sum of three sinusoids (2Hz, 20Hz, 50Hz) with random white noise. The 50Hz and noise are present everywhere, the 2Hz and 20Hz start only after two seconds.
f1 = 2; f2 = 20; f3 = 50; i =2000:6000; Data(1,i) = sin(f1*2*pi*t(i)) + 0.4 * cos(f2*2*pi*t(i)); Data = Data + 0.2 * sin(f3*2*pi*t) + 0.4 * rand(1,6000);
Empty the Process1 list (right-click > Clear list) then click on [Run].
Run process: Simulate > Simulate generic signals.
Ntime=6000, Sampling frequency=1000Hz (signal duration = 6000/1000 = 6 seconds).
Copy-paste the few lines of code above to generate the sum of three sinusoids.
Double-click on the new file to look at the simulated signal.
- In Process1, select the simulated signal.
Run process: Frequency > Time-frequency (Morlet wavelets).
Select the option Spectral flattening: The normalization will be discussed later.
Click on the button [Edit] to see all the process options.
Time definition: Same as input, Frequency definition: Linear 1:1:60, Compute measure: Power.
Process options
Comment: String that will be displayed in the database explorer to represent the output file.
Time definition
Same as input file: The output file has the same time definition as the input file.
In this example, it means: 6000 samples between 0 and 6s.Group in time bands: This option adds a step of computation. First it computes the TF decomposition for all the input file, then averages the power by time band. To define a time band:
Enter your own time bands in the text area, one line per time band, with the following format: "name / time definition / function"
- Click on the button [Generate] to automatically create a list of time bands with the same length. You will be asked the maximal length of each time band.
The function is the measure we take to combine the values for all the individual frequencies into one for the frequency band. Possible values are: mean, max, std, median.
Frequency definition: Frequencies for which the power will be estimated at each time instant.
Linear: You can specify the frequencies with the Matlab syntax start:step:stop.
The default is "1:1:60", which produces 60 values [1, 2, 3, 4, ..., 59, 60].Log: With the option start:N:stop, produces a list of N frequencies logarithmically scaled between "start" and "stop". For example "1:40:80" is converted to [1, 1.5, 2.1, 2.7, ..., 61.5, 65.8, 75, 80]
Group in frequency bands: As for the time definition, this option leads to a two-step process. First it computes the TF decomposition for several values in the frequency band, then it averages the power of TF coefficients per frequency band. To define a frequency band:
One line per frequency band, with the format "name / frequency definition / function"
The frequency definition is a Matlab expression evaluated with an eval() call. If the frequency definition contains only two values, Brainstorm adds two extra values in the middle so that the final averaged value is a bit more robust. Example of valid expressions:
"2,4": Evaluates to [2,4], and then expands to the frequency vector [2, 2.66, 3.33, 4]
"2:0.5:4": Evaluates to [2 2.5 3 3.5 4]
"2, 2.5, 3, 3.5, 4": Evaluates to [2 2.5 3 3.5 4]The function is the measure we take to combine the values for all the individual frequencies into one for the frequency band. Possible values are: mean, max, std, median.
Morlet wavelet options
Central frequency: Frequency where the mother wavelet is designed. All other wavelets will be shifted and scaled versions of the mother wavelet
Time resolution (FWHM): Temporal resolution of the wavelet at the central frequency (in units of Full Width Half Maximum). Click [Display] to see the resolution of the wavelet for other frequencies.
Compute the following measure:
The convolution of the signal with complex Morlet wavelets returns the complex coefficients for each frequency/time/sensor. Typically, what we display is the power of the coefficients (square of the amplitude: abs(TF)2). You can choose if you want to apply this transformation or not.
Power: Computes the "power" transformation immediately after the TF decomposition. This discards the phase information, but produces files that are twice smaller and a lot easier to process.
Magnitude: Save the magnitude of the complex values instead of the power: abs(TF).
None: Save the TF coefficients as they are computed (complex values). This can be useful if you plan to use these decompositions for other purposes that require the phase.
- Some combinations of options may disable this choice. If you select frequency bands, the program will have to compute the power before averaging the values, therefore "none" is not an option.
Display: Time-frequency map
Right on the new time-frequency file > Time-freq: One matrix (same as double-clicking).
This menu displays the time-frequency decomposition of the first (and only) signal. The Brainstorm window shows two new elements: the tab "Display" and the frequency slider.
- We can easily identify the three horizontal bars as the three sinusoids in the simulated signals, and the trade-off between accuracy in time and accuracy in frequency. Set the measure to "magnitude" in the Display tab if you don't. Click on the figure to move the time-frequency cursor and explore the two axes of the plane.
2Hz: High frequency resolution but poor time resolution (supposed to start sharply at 2s)
20Hz: Better time resolution but poorer frequency resolution (17-24Hz)
50Hz: Continuous over the 6s - Frequency resolution gets even worse (40-60Hz). It looks discontinuous because this oscillation has the same amplitude as the white noise we added in the signal (weight 0.4, relatively to the 2Hz oscillation).
Current frequency: Slider that shows the current frequency selected in all the figures.
Just like the time, the frequency selection is centralized and managed by one control only for all the figures. As a consequence, it is impossible to display TF files with different frequency definitions at the same time. This can be perceived as an annoying limitation, but it allows all the simultaneous displays to be consistent at anytime and makes the interface more intuitive to manipulate, with lower risks of mistakes in the interpretation of the different figures.List of signals: This drop-down list shows the signal currently displayed in the selected figure. In this case, there is only one channel of data called "s1". It will be more useful later.
Hide edge effects: When this option is selected, the time-frequency coefficients that could not be properly estimated because of a lack of time samples are hidden. It allows you to see only the information that is really reliable. The lower the frequency, the longer the edge effects. In the screen capture below, the colormap has been changed to "jet" and the maximum set manually to 0.2 (measure=power).
Smooth display: Re-interpolates the time-frequency maps on a finer grid to produce nicer plots.
Measure: Type of measure that is currently represented in the selected figure. The entries that are enabled depend on the type of data that is saved in the file. In this case, we saved directly the power of the wavelet coefficients in the file, we discarded the angle/phase information, so the "phase" option is disabled. The other options are: Magnitude = sqrt(power), Log = 10*log10(power)
Colormap: As explained in the previous tutorials, you can change the colormap by clicking+moving on the colorbar on the right of the figure. Double-click on the colorbar to restore the defaults.
Display: Mouse and keyboard shortcuts
Mouse shortcuts
Left-click: Selection of current time and frequency.
Left-click + move: Select a time/frequency range. The legends of the X/Y axis show the selection.
Mouse wheel: Zoom in time, centered on the current time cursor.
Control + mouse wheel: Zoom in frequencies, centered on the current frequency cursor.
Right-click + move, or Control + left-click + move: Move in the zoomed image.
Double-click: Restore initial view.
Keyboard shortcuts:
Left/right arrows: Change the current time.
Page-up/page-down: Change the current time, 10 time samples at a time.
Up/down arrows: Change the the sensor displayed in this figure.
Control + up/down arrows: Change the current frequency.
Enter: View the original time series for this sensor.
Control + R: View the original MEG recordings.
Control + T: View the time-frequency 2D topography.
Control + I: Save as image.
Control + D: Dock figure in the Matlab environment.
Figure popup menu
Set selection manually: Does the same thing as drawing a time/freq selection square on a figure, but by typing the values for time and frequency manually.
Export to database: Save the selection for the displayed sensor in a new file in the database.
Export to file: Same as "Export to database", but the saved file is not registered in the database.
Export to Matlab: Same as "Export to database", but the output structure is sent to a variable in the Matlab base workspace instead of being saved to a file.
Display: Power spectrum and time series
Right-click on the file in the database or directly on the time-frequency figure to access these menus.
Power spectrum: For the current time, shows the power for all the frequencies.
Time series: For the current frequency, shows the power for all the time samples.
Example: Power spectrum density at 0.5s and power time series at 2Hz.
We see the oscillation at the 50Hz in the PSD plot, and the oscillation at 2Hz in the TS plot.
Example: Power spectrum density at 4s and power time series at 20Hz.
We see all three oscillations in the PSD plot, and the oscillation at 20Hz in the TS plot.
Note that if you right-click on the file in the database explorer and then select one of these menus, it will show all the signals. If you right-click on an existing time-frequency figure, it will show only the selected signal. It doesn't make any difference here because there is only one signal, but it will with the MEG recordings.
Normalized time-frequency maps
The brain is always active, the MEG/EEG recordings are never flat, some oscillations are always present in the signals. Therefore we are often more interested in the transient changes in the power at certain frequencies than at the actual power. A good way to observe these changes is to compute a deviation of the power with respect with a baseline.
There is another reason for which we are usually interested in standardizing the TF values. The power of the time-frequency coefficients are always lower in the higher frequencies than in the lower frequencies, the signal carries a lot less power in the faster oscillations than in the slow brain responses. This 1/f decrease in power is an observation that we already made with the power spectrum density in the filtering tutorial. If we represent the TF maps with a linear color scale, we will always see values close to zero in the higher frequency ranges. Normalizing each frequency separately with respect with a baseline helps obtaining more readable TF maps.
No normalization
The values we were looking at were already normalized (checkbox "Spectral flattening" in the process options), but not with respect to a baseline. We will now compute the non-normalized power obtained with the Morlet wavelets and try the various options available for normalizing them.
- In Process1, keep the simulated signal selected.
Run process: Frequency > Time-frequency (Morlet wavelets), No spectral flattening.
Double-click on the file. As expected, we only see the lower frequencies in this representation: the power of the 2Hz oscillation is a lot larger than the power at 20Hz or 60Hz.
Spectrum normalization
- In Process1: Select this new non-normalized file "Power,1-60Hz".
Run process: Standardize > Spectrum normalization, Method=1/f compensation.
This produces exactly the same results as previously (option "Spectral flattening" in the time-frequency process). It multiplies the power at each frequency bin with the frequency value (eg. multiplies by 20 the power at 20Hz), in order to correct for the 1/f shape we observe in the power spectrum. This works well for the lower part of the spectrum and up to 60-80Hz, but past this range it tends to overcompensate the higher frequencies.
Note that it does not do any form of baseline correction: the 50Hz oscillation is visible everywhere.
Baseline normalization
The second way to proceed is to normalize the power with respect to its average level during a reference time period. We can consider the oscillations at 2Hz and 20Hz as our events of interest, and the 50Hz as noise we want to get rid of. The segment from 0 to 2 seconds does not contain any of the signals of interest, therefore we can consider it as a baseline.
However, we will not be able to use the full segment [0,2]s because of the edge effects we described at the beginning of this tutorial. The time-frequency map at 2Hz with the display option "Hide edge effects" (left figure below) shows that the power could not be estimated correctly before 0.75s, therefore we shouldn't use it as a baseline. The power time series at 2Hz (right) shows that the power related with the 2Hz oscillation starts to increase significantly after 1.25s, therefore it's not really a "baseline" anymore. This leaves only the segment [0.75,1.25]s available.
At 20Hz, the expected transient effects are only 75ms long, therefore we could use a much longer baseline if we were not interested by the lower frequencies: [0.075, 1.925]s.
In this case, we have a very long "resting" time segment (2s), therefore the edge effects are not a problem for picking a time window for the baseline normalization. We will use the first time window mentioned, [0.75,1.25]s as it is long enough (500ms) to estimate the baseline mean power. In real-life cases, with shorter epochs, it is sometimes difficult to find an acceptable trade-off between the baseline duration and the exclusion of the edge effects, especially for the lower frequencies.
Run process: Standardize > Baseline normalization, Baseline=[0.75, 1.25]s.
Method=Event-related perturbation: ERS/ERD stands for "event-related synchronization / desynchronization", a widely used normalization measure for time-frequency power maps. It evaluates the deviation from the mean over the baseline, in percents: (x-mean)/mean*100.
Double-click on the new "ersd" file. The colormap type changed from "Timefreq" to "Stat2", which uses by default the "rwb" color set and shows relative values. Indeed, the ERS/ERD values can be positive or negative: the power at a given time sample can be higher or lower than the average level during the baseline. In the simple simulation we used, there is no power decrease at any frequency after 2s, so the strong values are mostly positive. However, if you look in the file, you would see that there are many small negative values (due to the random noise we added).
Note that the 50Hz disappeared because it was present during the baseline, while the 2Hz and 20Hz oscillations show high positive values (between 100% and 2000% increase from baseline).
Remember to select your baseline very carefully according to the frequency range you are interested in. See below examples obtained with different baselines: 0.075-1.925s, 0-2s, 0-6s.
Change the colormap to "jet" if you prefer, and adjust the colormap contrast as needed.
This video may help you understand better what are the implications of the baseline selection: http://www.mikexcohen.com/lecturelets/whichbaseline/whichbaseline.html
Tuning the wavelet parameters
Time resolution
You can adjust the relative time and frequency resolution of your wavelet transformation by adjusting the parameters of the mother wavelet in the options of the process.
Increasing the option "time resolution" will produce longer wavelets at a given frequency, hence increase the frequency accuracy (lower Δf) and decrease the time accuracy (higher Δt). Expect longer edge effects.
Decrease the time resolution will produce shorter wavelets at a given frequency, hence decrease the frequency accuracy (higher Δf) and increase the time accuracy (higher Δt). Expect shorter edge effects.
You can modify one or the other parameter, what is important is the product of the two. All the following combinations fc/FWHM_t produce the same results because their product is constant: (1Hz,3s), (3Hz,1s), (6Hz,0.5s), (60Hz,0.05s)
Examples for a constant central frequency of 1Hz with various time resolutions: 1.5s, 4s, 10s.
Frequency axis
You can also obtain very different representations of the data by changing the list of frequencies for which you estimate the power. You can change this in the options of the process.
Examples: Log 1:20:150, Log 1:300:150, Linear 15:0.1:25
Hilbert transform
We can repeat the same analysis with the other approach available for exploring the simulated signal in the time-frequency plane. The process "Frequency > Hilbert transform" first filters the signals in various frequency bands with a band-pass filter, then computes the Hilbert transform of the filtered signal. The magnitude of the Hilbert transform of a narrow-band signal is a measure of the envelope of this signal, and therefore gives an indication of the activity in this frequency band.
No normalization
Let's compute the same three results as before: non-normalized, spectral flattening, baseline normalization.
- In Process1, select the simulated signal.
Run process: Frequency > Hilbert transform, No spectral flattening, Do not mirror.
In the advanced options panel, keep the default options: Default frequency bands and Power.
Double-click on the new file. The figure now has only 6 rows, one for each frequency band.
- The non-normalized results are already easy to interpret:
delta (2-4Hz): Includes the 2Hz oscillation, contribution starts at 2s
beta (15-29Hz): Includes the 20Hz oscillation, contribution starts at 2s
gamma1(30-59Hz): Includes the 50Hz oscillation, contribution starts at the beginning (0s)
Right-click on the file or figure > Time series. Example for delta and beta.
Normalization
- In Process1, select the non-normalized Hilbert-based decomposition.
Run process: Standardize > Spectrum normalization, Method=1/f compensation.
Run process: Standardize > Baseline normalization, Baseline=[0.75, 1.25]s, Method=ERS/ERD.
Display the two normalized files side by side.
Method specifications
Band-pass filters: Same filters as in the process "Pre-process > Band-pass filter", with the option "Stop-band attenuation: 60dB". For details, see the tutorial Power spectrum and frequency filters.
Edge effects: To estimate the duration of the transient effects for each frequency band, select the process "Band-pass filter", enter the frequency band of interest and click "View filter response". Example for the alpha band:
Hilbert transformation: Using the Matlab's hilbert() function.
Extraction of the envelope: Power of the complex Hilbert transform, abs(hilbert(x))2.
MEG recordings: Single trials
Let's go back to our auditory oddball paradigm and apply the concepts to MEG recordings. We will use all the trials available for one condition to estimate the average time-frequency decomposition.
Spectrum normalization
In Process1, select all the deviant trials in Run#01.
Run process: Frequency > Time-frequency (Morlet wavelets), No spectral flattening.
In the advanced options, select: Log 1:40:150, Power, Save average time-frequency maps.
Save individual TF maps: This option stops the computation here and saves in the database one time-frequency file for each input file (40 files), with one TF map for each scout.
Save average TF maps: Instead of saving the TF for each file separately, it automatically computes the average of the power of all the TF. This is a good choice if you do not plan to use independently all the TF files, because it saves a lot of time and disk space.
Remove evoked response from each trial before computing TF: This option first computes the average of all the trials, then subtracts this average from each trial before computing the time-frequency decomposition. This brings the signals to a slightly more stationary state, which may help for the evaluation of the frequency contents.
Baseline normalization
Double-click on the new file Avg,Power,1-150Hz (MEG). Select "Hide edge effects".
In the drop-down list, select sensor MLP56 (the one with the strongest respons at 90ms).
Right-click on the TF figure > Time series.
Defining a baseline is now a lot trickier than with 6s-long simulated signal. The epochs are only 600ms long, and the power at many frequencies could not be estimated correctly. If we want all the values to be "good" after 0s, we cannot use anything below 15Hz. If we want to normalize the values, we have to go even higher: 30Hz if we want a baseline of 50ms before 0.
The epochs we use in this tutorial are too short to perform a correct time-frequency analysis. We should have imported at least 200ms more on each side, just for controlling the edge effects. You should always think carefully about the length of the epochs you import in your database if you are planning to run any form of frequency or time-frequency analysis.
- For the purpose of illustrating the tools available in Brainstorm, we will keep on working with these short epochs. Let's try to do the best we can with what we have here. We could use a baseline of 50ms to get a correct estimation above 30Hz, but this is probably a bit too short. We propose to include some more of the baseline (75ms), hoping there are not major edge effects in this segment.
- In Process1, select the average time-frequency file.
Run process: Standardize > Baseline normalization, Baseline=[-75, 0]ms, Method=ERS/ERD.
The new menus available to display this file are described in the next section.
Things to avoid
Avoid computing the time-frequency decomposition of the average of the trials, you would miss some of the induced response, the brain activity in higher frequencies that is not strictly time-locked to the stimulus, and not aligned in phase across trials. Always prefer the computation of the average of the time-frequency power maps of each trials, as we did here.
This is well documented in: Bertrand O, Tallon-Baudry C (2000).Avoid using the Hilbert transform approach on short recordings or averages, always use the wavelet approach in these cases. The band-pass filters used for the lower frequency bands may have very high orders, leading to long transients. The example below shows the expected transients for the default frequency bands using the process "Frequency > Hilbert transform", they can be much more invalidating than with the process "Frequency > Time-frequency (Morlet wavelets)".
Display: All channels
Three menus display the time-frequency of all the sensors with different spatial organizations. All the figures below represent the ERS/ERD-normalized average. They use the "jet" colormap, which is not the default configuration for these files. To get the same displays, change the colormap configuration:
right-click on the figure > Colormap: Stat2 > Colormap > jet.
All channels: All the maps are displayed one after the other, in the order they are saved in the file.
2D Layout (maps): Show each TF map where the sensor is located on a flattened 2D map. Most display options are available, such as the colormap management and the option "Hide edge effects".
2D Layout (no overlap): Similar to the the previous display, but the positions of the images are reorganized so that they do not overlap.
Image [channel x time]: Shows the values of all the sensors over time for one frequency.
Useful shortcuts for the first three figures:
Click: Clicking on any small TF image opens a new figure with only the selected sensor.
Shift + click: Opens the original recordings time series of the selected sensor, when available. Here, we display an average of time-frequency maps, so this menu has no effect.
Mouse wheel: Zoom in/out.
Right click + move: Move in a zoomed figure.
Display: Topography
The menus below show the distribution of TF power over the sensors, for one time point and one frequency bin, very similarly to what was introduced in tutorial Visual exploration.
2D Sensor cap / 2D Disc / 3D Sensor cap: 175ms, 8Hz
2D Layout: 8Hz (black), 35Hz (white)
Useful shortcuts for these figures:
Left/right arrows: Change the current time.
Up/down arrows: Change the current frequency.
Control + E: Display the sensors markers/names.
Shift + click on a sensor: Displays the time-frequency decomposition for that specific sensors.
Right click + move: Select a group of sensors.
Shift + scroll: Change the gain of the time series (2D Layout).
Control + scroll: Change the length of the window displayed around the current time (2D Layout).
Scouts
Similar calculations can be done at the level of the sources, either on the full cortex surface or on a limited number of regions of interests. We will start with the latter as it is usually an easier approach.
Drag and drop all the deviant trials from both runs, select [Process sources].
Run process "Frequency > Time-frequency (Morlet wavelets)".
Select the option "Use scouts" and select all the scouts defined in the previous tutorial.
In the advanced options, select "Scout function: After" and "Output: Save average".
Run the process (it may take a while).
The scout function was introduced in the previous tutorial. It is the method we use to group the time series for the 20 dipoles we have in each scout into one unique signal. When computing the TF of one scout, we have the choice between applying this function before or after the time-frequency decomposition itself.
Before: Extract the 20 source signals, apply the scout function to get one signal, run the TF decomposition of this signal. This is faster but may lead to information loss.
After: Extract the 20 source signals, run the TF decomposition of the 20 signals, apply the scout function on the power of the TF maps. Always prefer this option when possible.
Rename the new file to add a tag "Deviant" in it. Then right-click > Time-freq: All scouts.
- In Process1, select the new average TF file.
Run process: Standardize > Baseline normalization, Baseline=[-75, 0]ms, Method=ERS/ERD.
Full cortical maps
Computing the time-frequency decomposition for all the sources of the cortex surface is possible but complicated because it can easily generate gigantic files, completely out of the reach of most computers. For instance the full TF matrix for each trial we have here would be [Nsources x Ntimes x Nfrequencies] = [15000 x 361 x 40] double-complex = 3.2 Gb!
We have two ways of going around this issue: computing the TF decomposition for fewer frequency bins or frequency bands at a time, or as we did previously, use only limited number of regions of interest.
In Process1, keep all the deviant trials both conditions selected, select [Process sources].
Run process "Frequency > Hilbert transform", No spectral flattening, Mirror signal before.
To process the entire brain, do not select the option "Use scouts".
In the advanced options, select "Optimize storage: No", this option is not available when computing on the fly the average of multiple trials. Save the power, Save the average Hilbert maps.
Optimize the storage of the time frequency file: Let's describe this option in more details.
- When computing the TF decomposition of a source file, we are actually applying sequentially two linear transformations to the original recordings: the TF analysis and the source inversion. These two processes can be permuted: TF(Inverse(Recordings)) = Inverse(TF(Recordings)).
Therefore we can optimize the TF computation time by applying the wavelet transformation only to the sensor recordings, and then multiply the wavelet complex coefficients by the inverse operator (ImagingKernel). This trick is always used in the computation of the Hilbert and Morlet transforms.
- When we have the option to the save the complex values (constrained sources and no averaging), this can also be used to optimize the storage of the files. In these cases, we save only the wavelet transformation of the sensor data. Later, when the file is loaded for display, the imaging kernel is applied on the fly. This can be disabled explicitly with this option.
- Rename the new Hilbert file to include the tag "Deviant", and select it in Process1.
Run process: Standardize > Baseline normalization, Baseline=[-75, 0]ms, Method=ERS/ERD.
Right-click on the Hilbert file > Display on cortex.
The frequency slider now shows frequency bands ("alpha:8-12Hz") instead of frequencies ("12Hz"). You can explore the source activity in time and frequency dimensions. The screen capture below shows the activity at 175ms: a 60% increase in the alpha band around the auditory cortex and a 20% decrease in the beta band around the motor cortex.
Shift + click on the cortex surface: Displays the TF decomposition of the selected source.
Right-click on the brain: Selects the closest vertex and displays the popup menu at the same time. The first three menus are relative to the source that was just clicked.
Unconstrained sources
In the current example, we are working with the simple case: sources with constrained orientations. The unconstrained case is more difficult to deal with, because we have to handle correctly the three orientations we have at each vertex.
Full cortex: Computes the TF decompositions for all the sources (3*15000=45000), then sum at each location the power for the three orientations.
Scouts: Option "Scout function" in the process.
Before: Extract the 20*3=60 source signals, apply the scout function to get three signals (one per orientation), run the TF decomposition of the three signals, and finally sum the power of the three TF maps. This is faster but may lose some frequency resolution (especially for constrained sources).
After: Extract the 20*3=60 source signals, run the TF decomposition of the 60 signals, apply the scout function on the power of the TF maps for each orientation separately, and finally sum the power obtained for the three orientations.
- The storage optimization option is not available with unconstrained sources.
Getting rid of the edge effects
To avoid making mistakes in the manipulation of the data and producing more readable figures, we encourage you to cut out the edge effects from your time frequency maps after computation.
- In Process1, select the very first computed in this tutorial: Test/Simulation/Power,1-60Hz | multiply
Run the process: "Extract > Extract time", Time window = [0.75, 5.25]s
Open the new file, select the option "Hide edge effects": Almost everything left in this new file is correctly estimated. Brainstorm keeps track of the edge effects in the TFmask field of the file.
- We recommend you do the same when epoching your recordings: import trials that are longer than necessary, and after the time-frequency estimation, remove the unnecessary segments.
On the hard drive
Right click one of the first TF file we computed > File > View file contents.
Structure of the time-frequency files: timefreq_*.mat
TF: [Nsignals x Ntime x Nfreq] matrix containing all the values of the time-frequency decomposition (complex wavelet coefficients, or double values for power/magnitude/Z-score).
TFmask: [Nfreq x Ntime] logical mask indicating the edge effects (0=edge, 1=valid value).
Std: [Nsignals x Ntime x Nfreq] standard deviation if this file is an average.
Comment: String displayed in the database explorer to represent the file.
DataType: From what kind of data this file was computed: {'data', 'results', 'scout, 'matrix'}
Time: [1 x Ntime] Time vector used to estimate this file.
TimeBands: [Ntimebands x 3] Cell array where each line represents a time band:
{'band_name', 'time definition', 'function'}Freqs: For regular frequency binning: vector containing all the frequencies.
If using frequency bands: [Nfreqbands x 3] cell array, where each line represents a frequency band {'band_name', 'frequency definition', 'function'}RefRowNames: Used only for connectivity matrices.
RowNames: [1 x Nsignals] Cell array of strings that describes each row of the TF matrix. In this specific case, it would be the list of all the MEG sensor names. But it could also be a list of names of scouts or clusters.
Measure: Contains the name of the function that was applied right after the computation of the wavelet coefficients. So it represents the type of data contained in the TF matrix. Possible values:
none: No measure applied, TF contains the complex wavelet coefficients.
power: Power for each frequency, ie. the square of the amplitude: abs(coefficients)2
magnitude: abs(coefficients)
log: 10 * log10( abs(coefficients)2)
phase: angle(coefficients)
Method: String that identifies the process that generated the file:
{'morlet', 'fft', 'psd', 'hilbert', 'corr', 'cohere', 'granger', 'plv'}DataFile: Initial file from which this file was computed. In the database explorer, the TF file will be shown as a child of this DataFile file.
SurfaceFile / GridLoc / GridAtlas: Source space that was used, only for source files.
Leff: Effective number of averages = Number of trials that were averaged to obtain this file.
ColormapType: String, force a specific colormap type to be used when displaying this file.
DisplayUnits: String, force to use specific units when displaying this file.
Options: Options that were selected in the time-frequency options window.
History: List of operations performed on this file (menu File > View file history).
Useful functions
in_bst_timefreq(TimefreqFile): Read a time-frequency file.
in_bst(FileName, TimeWindow): Read any Brainstorm file with the possibility to load only a specific part of the file. "TimeWindow" is an range of time values in seconds: [tStart, tStop].
bst_process('LoadInputFile', FileName, Target, TimeWindow): The most high-level function for reading data files. "Target" is a string with the list of sensor names or types to load.
morlet_transform(): Applies complex Morlet wavelet transform to the time series in input.
Additional documentation
Articles
Bertrand O, Tallon-Baudry C (2000)
Oscillatory gamma activity in humans: a possible role for object representation
Int J Psychophysiol, 38(3):211-23Bruns A (2004)
Fourier-, Hilbert- and wavelet-based signal analysis: are they really different approaches?
J Neurosci Methods, 137(2):321-32Le Van Quyen M, Foucher J, Lachaux J, Rodriguez E, Lutz A, Martinerie J, Varela FJ (2001)
Comparison of Hilbert transform and wavelet methods for the analysis of neuronal synchrony
J Neurosci Methods, 111(2):83-98Pantazis D, Weber DL, Dale CL, Nichols TE, Simpson GV, Leahy RM (2005)
Imaging of oscillatory behavior in event-related MEG studies (pdf)
in Proceedings of SPIE, Computational Imaging III, Vol. 5674, eds C. Bouman and E. MillerPfurtscheller G (1992)
Event-related synchronization (ERS): an electrophysiological correlate of cortical areas at rest
Electroencephalogr Clin Neurophysiol, 83(1):62-9Series of lectures by Mike X Cohen:
http://www.mikexcohen.com/lectures.html
Forum discussions
Forum: Time and frequency resolution:
http://neuroimage.usc.edu/forums/showthread.php?1848