This page describes the FreeSurfer Group Descriptor (FSGD) File Format (Version 1) as of 11/12/02. FSGD is a way to describe a group of subjects and their accompanying data (like a spreadsheet). This can include class membership and other continuous varibles (eg, age). The nice thing about the FSGD is that various FreeSurfer programs can automatically compute the design matrix from an FSGD file. The FSGDF is used by mri_glmfit, mris_preproc, tksurfer, tkmedit. FSGD files are created automatically by qdec, but qdec is currently (12/9/08) limited to two continuous variables and two discrete variables with two levles. The FSGD format uses tags to identify information.
In an FSGD, a Class is generally synonymous with a Group. A Class/Group is a specific level of a specific discrete Factor. Eg, if you have Gender as a Factor, then the Factor will have two levels (Male and Female). If the only Discrete Factor in your design is Gender, then there will only be two groups (Males and Females). If you have two Discrete Factors (eg, Gender and Handedness), then you will have four classes/groups (MaleRight, MaleLeft, FemaleRight, and FemaleLeft).
Continuous Factors are referred to as Variables in the FSGD. These include things like Age, IQ, Weight, etc.
There are numerous FSGD examples worked out here.
Example of a legal file:
GroupDescriptorFile 1 Title MyTitle Class Class1 plus blue CLASS Class2 circle green SomeTag Variables Age Weight IQ Input subjid1 Class1 10 100 1000 Input subjid2 Class2 20 200 2000 #Input subjid3 Class2 20 200 2000 DefaultVariable Age
Notes:
* The first line must be "GroupDescriptorFile 1".
* Title is not necessary. This will be used for display.
* CLASS only needs the class name, the next two items (eg, "plus blue)" will be used in the display.
* The third input is ignored because of the hash (#).
* DefaultVariable is the default variable for display.
* SomeTag is not a valid tag, so it will be ignored.
General Rules:
* Each subject is represented by an Input tag.
* The order of the subjects is not important but you must make sure it is consistent with the input to mri_glmfit (--y).
* Each subject must have values for all variables (no empty cells)
* Tags are NOT case sensitive.
* Labels are case sensitive.
* When multiple items appear on a line, they can be separated by any white space (ie, blank or tab).
* Any line where # appears as the first non-white space character is ignored (ie, it is treated as a comment).
* The Variables line should appear before the first Input line.
* All Class lines should appear before the first Input line.
* Variable label replications are not allowed.
* Class label replications are not allowed.
* If a class label is not used, a warning is printed out.
* The DefaultVariable must be a member of the Variable list.
* No error is generated if a tag does not match an expected keyword.
* Empty lines are OK.
* A class label can optionally be followed by a class marker.
* A class marker can optionally be followed by a class color.