NormalisationTransform has no attribute masking_method

[romainVala]

I get an error with
transfo = Compose(( Rescale(masking_method='mask', verbose=True),))

In the init of NormalizationTransform you use self.masking_method
but it is not assign

[romainVala]

it is not assign if you use a string as argument

[fepegar]

Fixed in v0.11.3. Thanks, @romainVala

[fepegar]

By the way,

1. Are you using Rescale for CT data?
2. You don't need to use Compose if you only have one transform

[romainVala]

Ok pour Compose, (I though it was mandatory)

I do not work with CT data, I use rescale to have number between 0 and 1 (by the way I would change the default for out_min_max 0 1

Currently I do not undersand, why I get value between 1 and 3 instead of value between -1 1 ... (but it woks fine if I ask fo (0,1) ...

[romainVala]

Instead of
array -= self.out_min # [out_min, out_max]
it should be
array += self.out_min # [out_min, out_max]

[fepegar]

Thanks again! I definitely need to write some unit tests...

[fepegar]

Is there a strong reason to choose (0, 1) over (-1, 1)?

[fepegar]

I'd like to make this implementation look more like https://github.com/scikit-image/scikit-image/blob/master/skimage/exposure/exposure.py#L269

[romainVala]

I do not know if it makes a difference from a CNN point of view. ... ?

It is just that mri is usually always positive (see #37 (comment))

Other point may be, after masking you usually get voxel value to zero in the background, so normalizing to -1 1 will break it ...

but not a big deal it is just the default value, easy to change