Hi Afni Experts,
I have a data with sparse temporal sampling so as to allow presentation of auditory stimuli without background auditory interference from scanner noise. The sequence of trials started with encoding of a single sentence, followed by number trials (zero, one, or two), and finally a retrieval trial. Scan data was only acquired after encoding, number and retrieval trials. The schematic of experimental design is displayed in the attachment.
I’m confused whether the sparse fMRI data can by simply analyzed using the common SPM steps, whether the individual contrast images of interest (encoding vs. number, retrieval vs. number) can be simple created by a SPM first-level analysis.
I’m confused whether the sparse fMRI data can by simply analyzed using the common SPM steps, whether
the individual contrast images of interest (encoding vs. number, retrieval vs. number) can be simple created
by a SPM first-level analysis.
I’m not so sure why you ask an SPM question on the AFNI message board.
It’s possible to analyze such a sparse dataset. However, the intervals between scans seem to be always the same, which makes modeling really difficult in your case.
Thank you very much for your reply.
I now try to use the afni program 3dDeconvolve to create contrasts of interest.
The previous description of the task may be not clear enough.
In this sparse fMRI design, 1-second periods of data acquisition occurred following the encoding, number and recall periods. The task has three trial types.
Trial type 1: listening to a sentence (6s), whole brain scan (1s), spoken free recall of the sentence (6s), whole brain scan (1s).
Trial type 2: listening to a sentence (6s), whole brain scan (1s), listening to single digit numbers and making an immediate spoken odd or even decision for each number (number trials, 6s), whole brain scan (1s), spoken free recall of the sentence (6s), whole brain scan (1s).
Trial type 3: listening to a sentence (6s), whole brain scan (1s), number trials (6s), whole brain scan (1s), number trials (6s), whole brain scan (1s), spoken free recall of the sentence (6s), whole brain scan (1s).
One example for the trial type 2 is displayed in the attachment 1.
The order of the three trial types was randomly determined but fixed within participants.
The first session of the task includes 20 trials for type 1, 14 trials for type 2 and 14 trials for type 3). Thus, there are 139 scans in total: 1 empty scan, 48 scans after encoding processes, 42 scans after number processes, 48 scans after recall scans.
According to the the afni program 3dDeconvolve, I wrote the following code to create individual design matrices, modeling each experimental condition (see attachment2). Movement parameters derived from the realignment stage were incorporated as nuisance variables in all analyses.
I’m not sure whether such modelling is appropriate.
Thank you very much. Looking forward to your reply
Chunjie
The fundamental question is: for which events are you trying to capture their BOLD response? Your current script seems to do that for those three scanning events, not Encoding/Number/Recall?
Hi Gang,
Thank you very much for your reply. In fact, I want to obtain two contrasts images of interest (encoding vs. number, retrieval vs. number). The numbers after “./stim/” correspond to the scan points following the three events (see attachment). Each whole-brain scan was acquired in one second (repetition time, 7 s; delayed acquisition time, 1 s).
I know that in the current analysis, I just obtain three contrasts images corresponding to the three conditions. But I don’t know how to make a subtraction in the
codes. Perhaps I can make a direct subtraction after creating the three contrasts.
But I don’t know whether it is appropriate to analyze it in this way, taking into account the sparse fMRI scanning.
Best wishes.
Chunjie
First, I assume that you want to capture the BOLD response for each of three tasks (Encoding/Number/Recall), not the BOLD response for each event during each of those 1-second scan events. As this is a sparse scanning, you don’t have any data during each of those 6-second tasks. However, the data acquired during those 1-second scan periods should be able to capture the peak of the BOLD response from the tasks.
Secondly, for each task, you should provide the task onset times under option -stim_times, not the time for the scanning events. The timing data are better stored in a text file, which can be fed into 3dDeconvolve. This is a little tricky: how is timing coded in your FMRI data?
Lastly, as for the contrasts among the three tasks, you can directly obtain them in 3dDeconvolve through -gltsym.
Hi Gang,
I’m very thankful for your words. I’ll try it.
Best.
Chunjie
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.
