Hi Sharmili,
Do you really have just 2 subjects, or is this only an example? If there are just 2 subjects, then there isn't enough data I'm afraid, and trying to analyse will likely violate assumptions. To see why, let's forget for a minute that you have 7 measurements per subject and consider just 2: baseline and post-intervention. Then to test the effect of intervention, you could subtract the baseline from the post-intervention scan, leaving 1 image (difference) per subject. WIth 2 subjects, you'd have just 2 images to analyse. With just 2 images, there's just one (!) degree of freedom, and nothing serious can be inferred.
Now, you have 7 images per subject, with 14 images overall. It may seem better, but actually, it's the same situation as above. Although your design has more degrees of freedom (14-8 = 6), not only this isn't much, but it also requires that all observations are pooled to estimate the error, which needs same variances (probably a fair assumption), but also requires compound symmetry, something hard to assume with 7 observations along time.
Perhaps an alternative option, for a different analysis, could be if you had a hypothesis about the response to the intervention (say, that the response would follow a certain curve). Then that curve could be modelled in the GLM, akin to a 1st level fMRI analysis. Still, this would produce one result per subject (fixed effects) but a group inference with just 2 won't be really possible.