Hi Oori,
That's a big change, maybe a factor of 4.
To break it down a bit, let me first verify that your SNR is
computed as the mean signal within some small brain mask,
divided by the standard deviation of background noise.
And note that if that is the case, it doesn't actually matter
whether the physiological data applies to the brain, because
the only value that comes from the brain is a single mean.
The reduction in variance of the background noise would cause
the SNR change.
It might be good to describe how you are computing this, just
to be sure. That would include the question of whether/how
you are avoiding ghosting and eye movement artifacts.
- rick