Back to PLS HelpSignal change refrence point
Posted on 05/07/09 04:19:48
Number of posts: 60
Hello,
I believe that the signal change would be calculated as the change from the reference scan which was specified in the session profile (usually zero).This fact transparently confirm that the signal change for one condition is calculated with respect to time zero for that condition, thereby each condition has its own time zero (reference scan) which means the refrence scan for different conditions would be different. Now, my question is whether there would be an easy way to set this reference scan identical for all three conditions and calculate the signal changes of all conditions with respect to one identical reference?
/Alireza
I believe that the signal change would be calculated as the change from the reference scan which was specified in the session profile (usually zero).This fact transparently confirm that the signal change for one condition is calculated with respect to time zero for that condition, thereby each condition has its own time zero (reference scan) which means the refrence scan for different conditions would be different. Now, my question is whether there would be an easy way to set this reference scan identical for all three conditions and calculate the signal changes of all conditions with respect to one identical reference?
/Alireza
Replies:
Untitled Post
I'm Online
Posted on 05/07/09 11:27:30
Number of posts: 291
There will be no easy way to do that unless you have 3 consecutive conditions across the run.
Here's how the MRI datamat is created: Each time point of data (by TR) is loaded for a size of temporal window (also by TR) that you specified, starting from the first onset time of the first condition of the first run. The Reference Scan Normalization is thus processed temporal window by temporal window. Therefore, at one point, only one temporal window of one condition is in the scope. We do not load all conditions at the same time, and do not load un-used temporal window either. Otherwise, many datamat cannot be created because of the size.
However, if you could send me your data, I will have no problem to look into a customized way to create a special datamat just for that data. When you run PLS analysis, the special datamat can be treated at no difference from the regular datamat. The only thing you need to remember is that the datamat is created based on the reference scan of the average of the 3 conditions.
Finally, as I mentioned on the top, you will have an easy way to do that, if you have 3 consecutive conditions. Here's the example that meet the requirement (assuming Temporal Window Size is 8 TR):
condition 1 onset: 1 25 49
condition 2 onset: 9 33 57
condition 3 onset: 17 41 65
Now, if you put 24 instead of 1 for the Number of Reference Scans for each condition, the reference scan will be ALMOST identical for all three conditions.
Since we only use reference scan normalization to trade off the low frequency drift, the way that you proposed is not necessary. However, if really want to give it a try, I can help you if you could send me the data.
Untitled Post
I'm Online
Posted on 05/07/09 11:50:27
Number of posts: 229
Alireza..
it is not clear why you would want to do this.
- For an event-related design, the jittering should make the onsets essentially "equal" with respect to carry-over from the previous event, so the absolute value of the signal intensity for the reference TR should be irrelevant for calculating %signal change.
- For a block design, if you are concerned that the signal from the previous block has not returned to some equivalent/neutral state for all of the different block types, then you have a spill-over problem no matter how you express the %signal change in the data.
As Jimmy pointed out, for both types of designs, zeroing the data at some point close to event onset will essentially remove any low frequency drift in your signal intensity.
nancy
it is not clear why you would want to do this.
- For an event-related design, the jittering should make the onsets essentially "equal" with respect to carry-over from the previous event, so the absolute value of the signal intensity for the reference TR should be irrelevant for calculating %signal change.
- For a block design, if you are concerned that the signal from the previous block has not returned to some equivalent/neutral state for all of the different block types, then you have a spill-over problem no matter how you express the %signal change in the data.
As Jimmy pointed out, for both types of designs, zeroing the data at some point close to event onset will essentially remove any low frequency drift in your signal intensity.
nancy
Login to reply to this topic.
