> We are running opt_vbm in spm2 on SPGR images collected from a 1.5T
> machine. We have a question about the default parameters for bias field
> correction.
>
> Are the values for defaults.bias.reg (0.01) and .cutoff (30) optimal for
> bias field correction, or should we try different default values for our
> images? If we should try different values, what range for these defaults
> should we use?
This question can probably only be answered empirically. If your images have
no bias, then segmentation should work better without bias correction (or
with heavy regularisation or a large cutoff). If there is a lot of bias in
your images, then you should use less regularisation. If your bias field is
very low frequency, then you can get by with a broader cutoff. The default
settings are those that worked well for our data at the FIL.
Best regards,
-John
|