Dear Ana,
The mean signal (and therefore the SNR) in diffusion imaging depends on the b-value, brain structure and the gradient direction, because the signal attenuation depends on the diffusion properties.
To simplify, one can use the mean b0 signal in the white matter, divided by the background noise standard deviation.
The problem with using the inverse brain mask is that you will still have some non-brain structures with high signal, i.e. skull, neck and face tissues.
I would therefore recommend that you draw a mask manually in the background to compute the standard deviation of the noise.
Once you have the WM mask and the background mask, you can get the SNR for the b0, or any b-value with the commands:
-> mean at b=0: fslstats mean_b0 -k wm_mask -m
-> mean at b=1000: fslstats mean_b1000 -k wm_mask -m
-> standard deviation of the noise: fslstats mean_b0 -k background_mask -s
Hope that helps,
Cheers,
Manu
> On 10 Mar 2017, at 10:04, Ana E. <[log in to unmask]> wrote:
>
> Hello FSLusers,
>
> I would like to calculate the SNR in my diffusion image. I did the mean(signal)/std(noise).
>
> As I understood the noise is calculated with the invert mask ( values out mask). Could someone give me some tips about this field?
>
>
> Best regards,
>
> Ana.
|