12.12.1. Overview

Auto-image making in AFNI

Here we present examples of automatic image-making with AFNI (in the sense of being command line callable and scriptable) using several tools, such as:

These programs do many of the kinds of things you would do to look at datasets in the AFNI GUI (click buttons, turn on transparency, change colorbar, set thresholds, change overlay/underlay, make montages, etc.). They can be quite useful for making figures, since this process often involves tweaking small things and regenerating images (thanks, Reviewer #2!)– it can save a lot of time to have an adjustable script for making images.

While it is great to look at the datasets interactively, using commands like these one can have images waiting for you to look at various processing steps (like, alignment, ROI placement, distortion correction, artifact detection, etc.). In the era of big data (ooooh, that buzzword!), you are still obligated to know about your analysis steps if you don’t want nasty (and hidden) surprises in your results. These tools allow you to look at your data systematically across a group, and that can be quite useful and powerful.

Each of these “star” programs either “drives” AFNI in a virtual environment or just runs in normal memory, so they don’t actually open up AFNI separately and can be run on remote systems without any special considerations.

The examples in this tutorial section use data sets that are publicly available as part of the AFNI Bootcamp Demo set, freely downloadable as described here. Unless otherwise stated, the data sets in the AFNI_data6/afni/ directory are used. The code snippets are all in tcsh syntax.

Some definitions/abbreviations

ulay

Underlay (dset). Typically grayscale.

olay

Overlay (dset). Typically mapped to a colorbar to present results

thr

Threshold (dset). For voxelwise thresholding– used to determine whether an overlay voxel is seen or not. Need not be the same dataset as “olay”: in FMRI, thr and olay are often different; in DTI, one might set the same dset as olay and thr (e.g., FA map).

coef

The “coefficient” part of a statistical model regressor: the effect estimate. If you have scaled your data during processing to have meaningful units (like “BOLD % signal change” for FMRI), as well as any response model (like a BLOCK, again in FMRI), then this should have physical, meaningful, interpretable and comparable units. This is typically used to set the olay colors.

stat

The “statistic” part of a statistical model regressor. When present, this kind of volume is typically used to threshold the olay data. In both the AFNI GUI and on the commandline, there are functions/programs to convert a p-value to a statistic value, for convenience in thresholding.