2.6. List of all startup tips¶
This is a list of possibly useful tips that the AFNI gurus have thought up, which may inform you, Dear User, of cool features within AFNI. By default, one gets displayed every time you start up the AFNI GUI (unless you have set an environment variable in “~/.afnirc” to turn off this delightful feature). Hence the name of this list.
We might/should add more over time.
If you set environment variable AFNI_GLOBAL_SESSION to the name of a directory with datasets, then those datasets will be visible in the UnderLay and OverLay choosers. For example, copy the MNI template MNI152_2009_template_SSW.nii.gz to this directory, and then you'll always be able to use it as an underlay dataset.
If the aspect ratio (width/height) of an image viewer window looks bad, you can fix it by typing the 'a' key into the image, or by clicking the left mouse button in the intensity grayscale bar at the right of the image.
The right-click popup menu on the intensity grayscale bar to the right of an image viewer has several useful controls, including: * choosing the numerical Display Range for the underlay * drawing a coordinate Label over the image * applying an Automask to the overlay (e.g.,hide the non-brain stuff) * choosing the color for Zero values in the overlay (e.g., black or white)
Looking at venography or arteriography datasets? The image viewer 'Disp' control panel 'Project' menu lets you look at projections of the underlay dataset through a slab of slices, including Minimum and Maximum. The slab half-thickness is given by the 'Slab +-' control below 'Project'.
If you crop an image, you can move the crop window around by pressing the Shift key plus one of the keyboard arrow keys.
The 'Disp' button in an image viewer pops up a control panel with many useful buttons, including: * Project = combine multiple slices into one underlay * Tran 0D = transform values of the underlay pixelwise * Tran 2D = transform underlay image globally (e.g., blurring) * Rowgraphs = graph the underlay numerical values in 1-9 pixel rows
The 'BHelp' button lets you click on some other button in the GUI to get more information about what that button does.
The right-click popup menu on the coordinate display in the AFNI controller has several useful functions, including: * controlling the coordinate display order * jumping to x,y,z (mm) or i,j,k (voxel index) coordinates
The right-click popup menu on the label above the threshold slider lets you control the threshold in various ways: * pin the Threshold sub-brick to equal the OLay or OLay+1 sub-brick (OLay+1 is very useful for Coef/t-statistic sub-brick pairs) * set the threshold slider to have a given voxelwise p-value (based on the statistical properties of the current Thr sub-brick) * control Alpha fading for colorization of sub-threshold voxels * see only Positive or Negative values, with respect to the threshold (which will affect the p-value, as being 1- or 2-sided)
The right-click popup menu on the label above the color overlay bar lets you control colorization from the OLay sub-brick in several ways: * you can jump crosshairs to the largest OLay value above threshold * you can write the current color palette out to a file for editing, or to an image for use in a figure * you can apply pixelwise or 2D spatial transformations to the OLay values before they are turned into colors
You can run InstaCorr on several subjects at the same time, using multiple AFNI controllers opened with the 'New' button.
The 'New' button (lower left corner of AFNI controller) lets you open another AFNI controller. The UnderLay and OverLay datasets will be listed in the controller window title bar.
Image viewer keypress: q = close window (works in graph viewer too)
Image viewer keypress: S = save image (works in graph viewer too)
Image viewer keypress: o = turn OLay color on or off
Image viewer keypress: u = make underlay image from the OLay dataset press u again to make underlay image from ULay
Image viewer keypress: 4 or 5 or 6 = meld ULay and OLay images (controlled by a slider on top of the image) * 4 = OLay on left side, ULay on right side, slider moves boundary * 5 = OLay on top side, ULay on bottom side, slider moves boundary * 6 = ULay and OLay intensity mixed, slider controls mixing fraction (slider to left = more ULay; to right = more OLay)
Image viewer keypress: z/Z = zoom out or in Zooming is limited to factors of 1-4
Graph viewer keypress: < or > = move focus time down or up 1 TR
Graph viewer keypress: 1 or L = move focus time to first or last TR
Graph viewer keypress: v/V = video the focus time up or down This is how you can make a video of subject head movement, by looking at the image viewers while the graph viewer is doing 'v'.
Graph viewer keypress: m/M = decrease/increase matrix size of graphs Also can do this from the 'Opt->Matrix' menu.
Graph viewer keypress: w = write time series from central sub-graph to a file Set prefix for file from the 'Opt' menu.
The image viewer 'Mont' button (along bottom) will let you make a montage from multiple slices, which can be Saved to a .jpg or .png file. NOTE: you might want to turn the crosshairs off from the 'Xhairs' menu in the main AFNI controller.
If the image editing program 'gimp' is in your path, then the image viewer Save control panel will include an option to start gimp on your image, so you can further edit it immediately. See https://www.gimp.org/
The graph viewer 'Tran 1D' function Dataset#N (from the 'Opt' main menu) lets you plot extra dataset time series on top of the UnderLay dataset's time series graphs.
You can change the way the graph viewer shows its plots by using the 'Colors, Etc.' sub-menu from the main 'Opt' menu (lower right corner): * Boxes = color of the boxes around each sub-graph * BackG = color of background * Grid = color of vertical grid lines * Text = color of text * Data = color of data graph points only, or points+lines, or boxes * Graph Gap = how many pixels spacing between sub-graphs * Thick = how many pixels wide for 'Thick' lines Most of these settings can also be selected by AFNI environment settings in your .afnirc file; with some work, you can setup the graph viewer to look the way you want it to be permanently.
In the graph viewer, the keypress Ctrl-B will cycle the Data plotting between the available modes, which are Lines = the default graphing mode Points = points only, plotted at each data value Points+Lines = points plotted with lines between them Boxes = a bar graph Box+LabelUp = with sub-brick labels at top of mini-graph Box+LabelTop = labels on top of each box Box+LabelDown = labels at bottom of mini-graph Labels for the boxes are taken from the sub-brick labels of the underlay dataset. Thus, you might want to re-label the dataset to have more meaningful labels on your graphs. This can be done with the '3drefit -relabel_all' command. The B key alone will toggle between 'Lines' and 'Boxes' (no labels).
The graph viewer 'Opt->Detrend' menu item lets you choose a polynomial degree for detrending the graph data. This can help you visualize the features of the data you want to see without be distracted by long term trends up or down. -1 = no detrending ; 0 = remove mean ; 1 = remove linear trend ; et cetera
The graph viewer 'Opt->Tran 1D->Despike' function will despike the time series graphs, which can be useful when you trying to figure out what's going on in a dataset.
Right-clicking in a graph viewer plot will popup a window with some statistics about the data being shown.
The README.environment text file lists many Unix 'environment' variables that can be used to control the way AFNI appears and operates.
The Define Datamode control panel lets you control how the OLay dataset is resampled to fit the ULay dataset (that defines the basis for the pixel grid on which the images are displayed). The options are: * NN = Nearest Neighbor * Li = Linear * Cu = Cubic * Bk = Blocky (between NN and Li) When the OverLay is at a coarser resolution than the UnderLay (common in FMRI), Li will produce 'nicer' looking maps, but NN will be more 'honest' looking.
'Define Datamode->Lock' lets you turn the xyz coordinate lock between AFNI controllers off, if you want. Or, you can turn on 'Time Lock', so that the TR index is locked between controllers, as well as the crosshair location.
* Normally, the grid size of the pixel image created for display in an AFNI image viewer is take from the grid size of the Underlay dataset. * But you can change that using the 'Datamode' control panel, by choosing 'Warp ULay on Demand', then setting the grid resampling mode below (e.g., to Li=Linear or Cu=Cubic interpolation). * Sometimes using this to make the display grid more fine is useful for creating nicer looking functional images, especially when 'Alpha' is turned on (to outline above-threshold clusters and at the same time show below-threshold in faded-out translucent colors).
Normally, voxels whose threshold value is below the slider setting will not be colorized. 'Alpha' fading allows them to get a faded color, while the above-threshold voxel clusters will get a black outline drawn around them. Alpha can be turned on from the right-click popup menu above the threshold slider, or via the AFNI_FUNC_ALPHA environment variable in your .afnirc file.
The InstaCalc function (from the InstaCorr drop-down menu) lets you calculate the overlay dataset on the fly, from multiple inputs, using the same expression syntax as 3dcalc, 1deval, etc.
You can right-click on the label to the left of a drop-down menu (e.g., 'ULay', 'Xhairs', 'Color') to get a chooser panel that lets you control the menu choice in a different way, with a separate chooser.
The 'Rota' arrows (in Define Overlay) lets you rotate the color bar, one color step per click -- if you use Shift+click, it takes 5 color steps per click. The 'F' button to the right will flip the color bar top-to-bottom.
The image viewer right-click popup menu has several useful functions: * Jumpback = take crosshairs to their previous location * Where Am I? = show atlas information about the current location * Image Display = hide GUI controls * Draw ROI Plugin = activate the Drawing plugin
Right-click on the 'Disp' button (lower left) of an image viewer will raise the corresponding AFNI controller to the top. Right-click on the AFNI logo (lower left) of a graph viewer does the same. These functions are here in case you lose the controller somewhere on the screen, and want to get it back.
Right-click on the 'Save' button in an image viewer will popup the list of possible image save formats, and let you choose one. You can do this from the 'Disp' control panel also, but this right-click method is faster.
The 'Rec' button in an image viewer pops up a menu that lets you choose different options for saving image snapshots to a special 'Record' viewer. Once you have recorded the set of images you like, you can save them from the 'Record' viewer. This is one way to make a video of how the overlay image changes as the threshold slider moves, for example. * Next One = record the next image displayed * Stay On = record each new image displayed (until turned Off)
Left-click in the square right of 'Etc->' in an AFNI controller will popup a copy of the splash screen again. Another left-click there will pop the splash window down again. Clicking in the reincarnated splash screen may give funny results. Right-click in that square will give a menu with some fun choices. Middle-click in that square will popup a random insult.
Set environment variable AFNI_DATASET_BROWSE to YES and then when you click on a dataset name in the OverLay or UnderLay popup chooser, AFNI will switch to viewing that dataset immediately (rather than waiting for you to press 'Set'). You can also browse through datasets in these choosers using the keyboard up/down arrows.
You can adjust the brightness and contrast of the underlay (grayscale) image by using the 'b' and 'c' arrows at the right of an image viewer. A more interactive method is to press and hold down the left mouse button, then drag the cursor around up/down (brightness) or left/right (contrast). With this method, you just wiggle the mouse around while left-click is down, and you can adjust the image grayscale until it looks good. The 'Norm' button will reset the grayscale contrast to the startup setting, in case you make things look terrible.
Set environment variable AFNI_CREEPTO to YES, and then the 'Jump to' button will move the crosshairs to the chosen location incrementally, rather than in one big jump. The reasons for using this feature are (a) to help get a feel for the transit, and (b) just plain fun.
Right-click on the color bar in Define Overlay, and you can change the color scale that is used. You can switch the color bar to a discrete set of solid colors by using the menu labeled '#' just beneath the color bar. You can save an image of the color bar by right-clicking on the label above it, and choosing 'Save to PPM' from the popup menu.
You can crop an image by left-clicking the 'crop' button in an image viewer, then selecting the crop region by clicking+dragging in the image. You can Montage cropped images (all will be cropped the same way). Right-clicking on 'crop' will give a chooser where you can specify the cropping region size exactly.
You can use keyboard shortcuts to precisely adjust the threshold slider. Put the mouse over the slider, and then * down/up arrows for tiny adjustments * page up/page down for larger adjustments
In a graph viewer, you can restrict the plotting to a subset of the time points by using the 'Opt->Grid->Index Pin' menu item. This feature is most useful when viewing very lengthy datasets.
In a graph viewer, the default plotting method has the bottom of each graph using a separate value (the minimum in that voxel). You can also make them have a common baseline (minimum among all voxels in the graph window) or a global baseline (set by you) by using the 'Opt->Baseline' menu items.
At the bottom of a graph viewer is a bunch of text showing various information about what is being shown.
When looking at FMRI data graphs with a regular stimulus timing, it is helpful to set the graph grid lines to match the stimulus timing spacing. You can do this from the 'Opt->Grid->Choose' menu item.
You can have graphs drawn as box plots rather than as connected line segments, by using the 'Opt->Colors, Etc.->(Data) Boxes' menu item, or by pressing the 'B' key when the mouse cursor is over the graph viewer window.
In the graph viewer 'Opt' and 'FIM' menus, items that have keyboard shortcuts have the key inside square brackets, as in 'Opt->Scale->Down [-]', meaning the '-' key will cause the graph to scaled down (vertically).
Advanced graphing: you can change the x-axis values from being 0,1,2,... to be anything you want, chosen from a 1D text file (applies to all voxels) or from a 3D dataset (per voxel x-coordinates). The x-axis for the central sub-plot will be displayed as a vertical graph at the left of the graph viewer window. See the 'Opt->X-axis' menu items to do strange things.
The 'Define Datamode->Misc' menu has a lot of choices, a few of which are: * Voxel Coords? = show voxel indexes instead of mm coordinates in AFNI GUI * ULay Info = show information from the UnderLay dataset header * Purge Memory = eject datasets from memory, forcing reloads when viewed
When saving an image (or a montage), you might want to turn the crosshairs off. You can do this from the 'Xhairs' menu in the AFNI controller. If you want all the sub-images in a montage to have crosshairs (instead of just the central image), turn the 'X+' button on.
Just below the slider bar in an image viewer is a label, such as 'Axial: left=Left'. This label indicates that you are looking at an axial image and the software thinks that the left side of the image viewer is the subject's Left. Similarly, the sagittal viewer label would normally say 'Sagittal: left=Anterior'. However, these labels will change if you alter the image viewing orientation in the 'Disp' control panel.
When saving from the image viewer, the saved image is on the matrix of the dataset. It is NOT a screen capture; that is, the image saved will not depend on the size of the image viewer window. A montage image will be the full size of all the base images catenated together. You can also choose a 'Blowup' factor to scale the image size upward: factors from 2 to 8 are available.
You can tell the graph viewer to ignore the first few time points when plotting. Menu item 'FIM->Ignore' lets you choose how many to ignore by mouse clicks. Keypress 'I' increases the ignore count by 1, 'i' decreases by 1. Ignored points are plotted with little blue circles which take the value of the first non-ignored point.
If you have a complicated AFNI window layout you want to save, you can use 'Define Datamode->Misc->Save Layout' to save a startup script that will be used when you re-start AFNI in the same directory to restore the AFNI windows to (approximately) the same state they had before.
Did you know that AFNI can display datasets stored with various data types? * byte (unsigned 8 bit integers) * short (signed 16 bit integers) * float (32 bit values) * complex (pairs of floats) * RGB (triples of bytes)
AFNI will read in .jpg and .png image files as 2D 'datasets'. Is this useful? It depends on who you ask! If you don't like this, set Unix environment variable AFNI_IMAGE_DATASETS to NO (in your ~/.afnirc file).
The AFNI program 'aiv' (AFNI Image Viewer) can be used for a quick display of images in various formats (.jpg, .png, plus datasets). The interface is the same as the slicer viewer built into the AFNI GUI.
The AFNI GUI now 'knows' about the BIDS file hierarchy. You can open all the datasets from a given subject in a single session, even though BIDS scatters them over several subdirectories. To do this, use the '-bysub' option. See the output of 'afni -help' for the details.
Obscure AFNI GUI buttons: EditEnv = Lets you edit some AFNI environment settings interactively; useful when you need to change something and don't want to quit and re-start AFNI. For example, setting AFNI_LEFT_IS_POSTERIOR will flip the usual Sagittal image and graph viewers so that the display's left corresponds to the subject's posterior, rather than the default anterior. NIML+PO = Starts NIML and Plugout socket listening; useful when you meant to do one (or both) of these one the command line (options '-niml' and '-yesplugouts'), but forgot. For example, NIML is needed for 3dGroupInCorr to connect.
The 'Render Dataset' plugin allows you to do 3D volume rendering in the AFNI GUI, with color overlays, animations, and cutouts. (The SUMA GUI also has a volume rendering mode.)
Want your picture in the AFNI splash window at startup? Send us a JPEG image, formatted to be square 128x128, and we can include it!
Do you want ALL the AFNI plugins to be visible in the Plugins menu? Set environment variable AFNI_ALLOW_ALL_PLUGINS to YES in your .afnirc file.
Setting environment variable AFNI_GRAPH_ALLOW_SHIFTN to YES in your .afnirc file will allow you to set the graph viewer matrix size directly using keyboard presses, as in N7<Enter> which will make the graph window have a matrix of 7x7 sub-graphs. It is important to press the <Enter> (or <Return>) key at the end of the digit(s) after N, otherwise the graph window will not respond to any other key presses.
Set environment variable AFNI_STARTUP_SOUND to YES to hear the AFNI startup sound when the GUI opens. Or use the right click popup menu in the logo square right of the 'done' button and select the 'Play startup sound' item. - But whatever you do, DO NOT use the 'Activate Omega-13' menu item! - Sound playing requires the 'sox' software. - To find out if sox is on your system, type the command 'which sox'.
If the 'sox' software is installed on your system, you can play sounds from the AFNI graph viewer window. - Keypress 'p' will play a sequence of tones based on the central sub-graph. - Keypress 'P' will play based on the average of all sub-graphs. - Filtering and detrending the graphs will affect the notes played. - Sound can only be played if you are displaying locally, not remotely. - Environment variable AFNI_SOUND_NOTE_TYPE is used to set the note type: sine square triangle sawtooth trapezium pluck ('pluck' sounds halfway between guitar and piano notes) - See README.environment for a few more details. - To find out if sox is on your system, type the command 'which sox'.
Want bigger fonts in AFNI, for a high resolution screen? Set environment variable AFNI_FONTSIZE to PLUS or BIG (preferably in your .afnirc setup file).
Some environment settings for the AFNI graphical user interface (GUI) that you might want to change (in your ~/.afnirc file) are listed below. The built-in values in the GUI program are shown here in [brackets]: AFNI_DETACH = detach GUI from terminal window [YES] AFNI_RESCAN_AT_SWITCH = find new data when using UnderLay or OverLay [YES] AFNI_GRAPH_FADE = 'fade out' below thresh voxels in graphs [YES] AFNI_DATASET_BROWSE = switch instantly in UnderLay/OverLay [YES] AFNI_PBAR_FULLRANGE = put value labels next to colorscale [YES] AFNI_COLORSCALE_DEFAULT = name of startup colorscale [Reds_and_Blues_Inv] AFNI_THRESH_INIT_EXPON = initial power-of-ten for threshold  AFNI_OPACITY_LOCK = lock 1-9 opacity arrows among image viewers [YES] AFNI_NOSPLASH = turn off the AFNI splash screen at startup [NO] AFNI_STARTUP_SOUND = turn on the AFNI startup sound! [NO] AFNI_LEFT_IS_LEFT = show subject left on screen left [YES] (for Axial and Coronal viewers) AFNI_NOPLUGINS = don't load any of the AFNI GUI plugins [NO] Setting AFNI_NOPLUGINS to YES can speed up scripts that externally drive the AFNI GUI, since such a script usually doesn't use a plugin. * Some of these values are already set in the .afnirc file that is given out with the AFNI binaries. You can also set environment variables in a script before starting the AFNI GUI, which will take priority over values set in the ~/.afnirc file. * All AFNI environment variables (including many that do not affect the GUI) are described in the README.environment file.
In a main AFNI controller window, the 'Index' field (left-middle) shows the sub-brick/volume/time index currently being displayed in the image viewers. If you right-click on the 'Index' label, a hidden popup control window opens, with these occasionally useful items: Index Step : Lets you set the stepsize for the Index arrow buttons. SLAVE_FUNCTIME : Lets you turn off or on the 'slaving' of the overlay volume index to the underlay volume index; this feature is useful when the overlay is a time- dependent dataset itself. Thr = Olay?+1? Lets you lock the Define Overlay 'Thr' volume index chooser to the 'OLay' volume index chooser, so that as you change which overlay dataset volume you are colorizing, the threshold volume changes in lockstep. The choices on this item are free : Thr and OLay indexes are not locked == : Thr and OLay indexes are locked to be equal +1 : Thr index is locked to be OLay index + 1 which is useful for the Coef/t-statistic volume pairs output by various AFNI codes. When the Thr and OLay indexes are locked, the 'Thr' label in Define Overlay will change to 'Thr*'.
Questions about AFNI? Problems with a program? Try our Message Board at https://afni.nimh.nih.gov/afni/community/board/ * Please be specific and focused, as generic questions without details are very hard to answer well on a Web forum. * If you have a problem with a particular program, give the exact command line you are using, and the exact WARNING or ERROR message that you are seeing.
If you are doing complicated twisted things with AFNI programs, ASK US (on the message board). Often, there is a much easier way to do a task!
REMEMBER: afni_proc.py is your friend when doing time series analyses! In particular, if you are still using custom hand-written scripts for resting-state preprocessing or time series regression, you need to learn to use afni_proc.py (unless you are doing something unusual).
Skull stripping T1-weighted datasets? Programs and scripts include: * 3dSkullStrip - surface expansion program (many options; multi-species) * @NoisySkullStrip - when the dataset is very noisy * @SSwarper - nonlinear warping to MNI space combined with skull stripping (and then the warp can be used in afni_proc.py for time series analyses)
Program 3dUnifize can make the image intensity of a T1-weighted dataset more uniform in space. As a bonus, can also contrast-invert a dataset prior to the uniform-ization, which might be useful for T2-weighted datasets. It also has an experimental option for processing EPI data. ** Please do NOT use the older program 3dUniformize ** ** since it does not do as good a job as 3dUnifize! **
Program 3dcalc does voxelwise calculations on datasets. Doesn't sound exciting to you? Once you get to know it, you will find that 3dcalc is your FRIEND! (: And then you can get to know the interactive InstaCalc :)
AFNI has a lot of downloadable demonstrations; you can find them in your abin directory (if that's where AFNI is for you) by doing ls ~/abin/@Install_* A few examples: @Install_InstaCorr_Demo = data and instructions for using InstaCorr @Install_ClustScat_Demo = data and instructions for interactively plotting time series extracted from Clusterize ROIs @Install_FATCAT_DEMO = data and instructions for using the AFNI FATCAT programs for DTI tractography (etc.)
Program 3drefit can be used to change parameters in a dataset header (e.g., slice timing). Program 3dinfo can be used to display information from a dataset header.
Are you using 3dcalc to compute the average of a bunch of datasets? You should use program 3dMean instead! It is faster and doesn't have the artificial alphabetic limitation of 26 input datasets.
Want to calculate summary values (e.g., mean, standard deviation) at each voxel in a time series dataset? Program 3dTstat is the tool you need -- and there is an interactive version in the GUI.
Programs for extracting information from spatial ROIs: * 3dmaskave = get average across the ROI, one value per time point * 3dROIstats = like 3dmaskave, but for multiple ROIs * 3dmaskSVD = like 3dmaskave, but gives the principal singular vector (time series) across the ROI instead of the mean * 3dmaskdump = just dumps out ALL the values from the ROI at all time points (presumably you will do something fun with these in your own software?)
Programs for computing some local statistics in a neighborhood around each voxel (e.g., a ball): * 3dLocalstat = various statistics from the neighborhood (e.g., mean, median, variance) * 3dLocalBistat = various 2-sample statistics from the neighborhood, calculated between 2 datasets (e.g., Pearson or Spearman correlation) * 3dLocalPV = compute the principal vector (time series) from all the dataset time series inside the neighborhood (a fancy way to 'smooth' the data)
Are you using nonlinear warping to align your subjects' T1-weighted datasets? If not, you should give it a try. The brain images will line up better than using affine alignment (3dAllineate, @auto_tlrc, etc.) and you can get better FMRI results at the group level when you use the nonlinear warps in afni_proc.py. Programs for this: * 3dQwarp = the foundational nonlinear warping program * @SSwarper = uses 3dQwarp and 3dSkullStrip together to align volumes to the MNI template and skull strip them * auto_warp.py = runs 3dQwarp for you, so you don't have to read that program's lengthy help output
Want to create a 'junk' dataset on the command line, just to test to see if something works? AFNI programs can create a dataset in memory from a string; try this example: afni jRandomDataset:64,64,32,96 to create and view a random dataset with 64x64x32 3D volumes, and 96 time points. If you want to create and SAVE such a dataset, try 3dcalc -a jRandomDataset:64,64,32,96 -expr 'a' -prefix Fred.nii If you want a zero-filled dataset, replace 'a' with '0'.
Did you know that AFNI's time series analysis program 3dREMLfit can include voxelwise regressors (a different time series for each voxel)? We use this capability in our Anaticor model for de-noising datasets during activation or resting state analyses.
AFNI programs for individual dataset time series correlation-ing: * 3dTcorr1D = correlate each voxel with a small set of 1D files * 3dTcorrelate = correlate each voxel between 2 datasets * 3dTcorrMap = make of map of how correlated each voxel is to every other voxel in the same dataset * 3dAutoTcorrelate = correlate each voxel to every other voxel in the same dataset and save everything (HUGE)
Program 3dGroupInCorr can be used for Group Instant Correlation interactively via the AFNI GUI. It can also be used in batch mode. The '-batchGRID' option lets you scan over a grid of seed voxels, compute the individual datasets' correlations with their seeds, then compute the t-tests among these correlation maps, and save the results to a collection of datasets.
Want to convert each statistic in a dataset to a (voxelwise) p-value? Use program 3dPval. More complicated statistical conversions can be done with 3dcalc, using the cdf2stat() and stat2cdf() functions. You can explore those interactively using the ccalc program, to make sure you are giving 3dcalc the correct expression.
Want to test dataset values voxelwise for normality (Gaussianity)? Program 3dNormalityTest will apply the Anderson-Darling test and give you a dataset with the voxelwise measure of non-Gaussianity.
Program 1dCorrelate will compute the pairwise correlation coefficient between 2 columns of numbers, AND give you the bootstrap confidence interval for the result. When you have relatively few samples (say, less than 25), bootstrap confidence intervals are more robust than the standard parametric intervals based on the Gaussian assumption.
Program 1dplot is useful for quick-and-dirty plotting of columns of numbers (.1D files). An example, creating a response model function with 3dDeconvolve and sending the time series directly into 1dplot: 3dDeconvolve -num_stimts 1 -polort -1 -nodata 81 0.5 \ -stim_times 1 '1D: 0' 'TWOGAMpw(3,6,0.2,10,12)' \ -x1D stdout: | 1dplot -stdin -THICK -del 0.5 There is also a more beautiful Python version of this program, cleverly named 1dplot.py
Program 1dNLfit does a nonlinear fit of an expression with free parameters to a column of numbers. For example: 1dNLfit -depdata sc.1D -indvar x '1D: 100%0:0.1' \ -expr 'a*sin(b*x)+c*cos(b*x)' \ -param a=-2:2 -param b=1:3 -param c=-2:2 > ff.1D fits a sine+cosine model with 3 free parameters (a,b,c) to the data in file sc.1D, where the 'time' parameter is x.
You can use make_random_timing.py to make AFNI-compatible random stimulus timing files. You can impose various constraints on the times generated. You can use timing_tool.py to manipulate stimulus timing files in various ways.
Program 1dTsort lets you sort .1D columns of numbers. Program 3dTsort lets you sort each voxel's time series (separately), and can also randomize them if you want.
Program 'count' will generate lists of numbers, which is surprisingly useful in scripting various things. For example, count -dig 1 -comma 0 99 S6 will produce a list of 9 distinct random numbers from 0..99 (inclusive), separated by commas; for example: '31,18,60,62,7,95'. This list could be used to select a random subset of dataset sub-bricks for analysis 3dttest++ -setA Fred.nii[`count -dig 1 -comma 0 333 S20`] (in the above command, the quotes are the single backquote ` and not the single frontquote ').
Most AFNI command line programs accept a common set of options, such as sub-brick selectors. See this page for the details: https://afni.nimh.nih.gov/pub/dist/doc/program_help/common_options.html
Want to resample a dataset to a different grid spacing? Programs: * 3dresample = older program with NN, Linear, and Cubic interpolation * 3dAllineate = for aligning datasets and then resampling, but with the -1Dparam_apply option can just do the resampling function; has more interpolation options, including quintic polynomials and tapered sinc. * 3dUpsample = resamples a dataset to a finer grid in the time direction (the other programs change spatial grids).
Want to blur/smooth a dataset? Programs: * 3dmerge -1blur_fwhm = Gaussian smoothing across whole volume * 3dBlurInMask = similar smoothing, but only inside a mask * 3dLocalPV = 'smooth' a time series dataset by computing the local principal vector around each voxel, instead of the average vector (slow)
Want to create a 3D dataset from a table of numbers? This can be done with program 3dUndump.
Want to slice up a dataset? Or glue datasets together? * 3dZcutup = cut a section of slices out of a dataset * 3dZcat = glue datasets together in the slice direction * 3dXYZcat = glue datasets together in any (spatial) direction * 3dZeropad = add (or subtract) slices to (or from) a dataset * 3dTcat = glue datasets together in the time direction
Did someone give you a dataset with the spatial orientation 'flipped', so that the subject's left is marked as being the right? * Program 3dLRflip can flip the data rows to rearrange the data so the dataset header information and the actual data match again. * Processing scripts afni_proc.py and align_epi_anat.py can check if the EPI and T1w anatomical datasets appear to be left-right flipped. * See Glen et al. (2020): https://www.frontiersin.org/articles/10.3389/fninf.2020.00018/full
Want to write an AFNI '3d' program? If you know C, the code 3dToyProg.c is a good starting point. It shows how to read, calculate, and write datasets, and is heavily commented.
Almost all AFNI command line programs take '-help' as an option, e.g.: 3dTstat -help This method is how you can get the most up-to-date information about using an AFNI program. All '-help' output are also formatted as Web pages and are available here: https://afni.nimh.nih.gov/afni/doc/program_help/index.html
Script @grayplot will read the errts (regression residuals) time series datasets from an afni_proc.py results directory, and make PNG-formatted grayplots, partitioned into gray matter, white matter, and CSF segments. The plots are good for looking at the structure of the residuals -- in an ideal world (alas, hard to find), there would be little spatial or temporal structure in the errts datasets, which are the 'noise' from which the variance parts of single subject t/F statistics are computed.
Interested in the AFNI source code? You can get it here: git clone https://github.com/afni/afni.git This creates a directory called 'afni' in your current working directory. You will find the source code for AFNI (1 million+ lines) in afni/src. Have fun!