Hi. In the context of alignment, are the correlations used in the cost functions what I understand as a “spatial correlation”? That is, are you simply correlating the value at a point in space in the source dataset with the value at the same point in space in the base dataset and then repeating for all points in space? Of course you are not really “repeating;” you just have two variables/vectors, but you know what I mean. Speaking of spatial correlations, if I wanted to calculate them for a 3D+time source dataset so that I calculated a spatial correlation with a single functional base dataset or “template” for each time subbrick, which AFNI program should I use?
Which alignment program are you referring to? There is choice of cost function, with several different rules for taking 2 volumes as input and returning a scalar value; the cost function for motion correction (EPI → EPI) tends to be “least squares” (ls); for EPI-> anat tends to be local pearson correlation (lpc, lpc+ZZ), noting that these dsets have opposite tissue contrast; and anat-> template tends to be lpa or lpa+ZZ, because the contrast is similar of these vols. There are many other cost functions (Helinger, normalized mutual information, correlation ratio, etc.). But alignment is always between two, 3D volumes.
Also, correlation couldn’t be calculated between only 2 values, it must be a between a set of >2 values (in practice, between maaany more).
A lot of alignment principles and descriptions are provided in this AFNI Academy video:
… with the subsequent videos in that playlist talking about several details for specific applications.
Re. <<if I wanted to calculate [spatial correlations] for a 3D+time source dataset so that I calculated a spatial correlation with a single functional base dataset or “template” for each time subbrick, which AFNI program should I use? >>
… I am afraid I don’t understand. Could you please clarify this. I don’t understand the inputs… in particular, a “single functional base dataset or “template” for each time subbrick”?
Also, are you really talking about alignment in this part, or about measuring similarity of spatial pattern? And if your input for that is 2 x 4D time series, are those datasets each with the same number of time points, and so you want the i-th subbrick in one dset correlated with the i-th subbrick in another one, for all values of i?
Hi. I worded this question poorly. Let me think about this.