Events & Conditions

Events and conditions in this program are defined in ways identical to the GLM for fMRI, except that events in SRTM are divided into 3 distinguishable types: R1, k2, and k2a.  This can be see in Equation 2 on this page. An alternate formulation is to use only one value of k2' as a global constant, as in Equation 3, and then "k2" is really just a scaled version of R1 (k2=R1*k2'); i.e., k2 is eliminated from the design matrix in favor of a single fixed value (or one value per scan) that is hard-wired into the analysis. Every time point of every scan must have defined event values for R1, k2a, and either k2 or k2’. If one decides to fit parameter k2 for every voxel, set tau2 for the reference region to zero at the top of the GLM file. Alternatively, set the value of tau2 and make no mention of k2 GLM events in the file.


More than one "event" within a single analysis can exist for R1, k2a, and k2 or k2', because there may be more than one scan in the analysis, or there may be time-changing parameters.  For now, only k2a is allowed to have a time variation, but a future version may extend time variation to R1.


Why would one want multiple values of these parameters in an analysis?

1) A single scan could include a challenge. In this case, different time intervals within the scan could have different values of k2a (time-dependent changes in "binding potential").

2) Multiple scans might be performed in the same session. In a double-bolus PET infusion with a challenge before the 2nd infusion, one can choose to fix one more parameters (e.g., R1, k2) across scans or vary the parameter.

3) Multiple scans might be analyzed across sessions, so that different parameters are needed for each scan due to imperfections in registration or due to biological effects.


Conditions should test the value of a given parameter alone, or in combination with events of the same type. For instance, one might want to view differences in k2a between scans, but it makes no sense to compare values of k2 or R1 with k2a.

Joseph B. Mandeville, Athinoula A. Martinos Center for Biomedical Imaging at MGH/MIT/Harvard