Which function for modeling would be the best one for my experimental design doing 3dDeconvolve (first-level analysis) ?

Hello everyone,

I am just a newbie in AFNI having a question of using function when modeling in first-level analysis.

My experimental protocol looks like this.

Fixation cross - [ {cue - stimulus - response screen} - {cue - stimulus - response} - {cue - stimulus - response} ] - fixation cross - [ … ] - …

TR: 2 secs
[ ] = block
{ } = trial
Fixation cross: 6 secs
cue presentation: 2 secs
stimulus presentation: 4 secs
response screen: 3 secs
(There was no interval between trials)

{cue - stimulus - response screen} indicates 1 trial, and 1 block contains 3 trials.
This is basically block design which assign each of experimental conditions to each block,
so I put each block as a regressor when making GLM.

Even though it is composed of block design, but there are three different kinds of events in each block,
there could be a lot of noise, because cue and response screens were made of texts and the stimulus screens were made of complex pictures.

Therefore, I want to take only the screen showing stimulus as a regressor and figure out which brain areas are involved in those times under the experimental conditions like analysis of event-related design.

I had used ‘BLOCK5’ function traditionally when making GLM of each of block for 27 secs; 2s (cue) + 4s (stim) + 3s (response) + 2s + 4s + 3s + 2s + 4s + 3s = 27s, as the block design analysis.
Now I want to try each of stimulus presenting time in each of trial as a new regressor.
The ‘GAM’ function could be one of option which has been normally used in event-related design, and is it proper to my experimental situation too, or are there
any suitable modeling functions you could recommend?
Any advice is welcome!