search

mhe / pynrrd

Simple pure-python module for reading and writing nrrd files.
https://pynrrd.readthedocs.io/
MIT License
116 stars 51 forks source link

oldmin and oldmax support #83

Open tashrifbillah opened 5 years ago

tashrifbillah commented 5 years ago

Nifti1 has scl_slope and scl_inter. The equivalent attributes in the Nrrd are oldmin and oldmax. However, we don't have support for oldmin and oldmax if they are present in intensity calibrated images (for relative intensity images, they are ignored), do we? If we don't have, then I am developing support for it.

addisonElliott commented 5 years ago

Unless I'm misunderstanding the question, pynrrd should read the oldmin, old min, oldmax, and old max attributes as double's. See the code here: https://github.com/mhe/pynrrd/blob/master/nrrd/reader.py#L89

As far as I can tell, they should be read correctly.

tashrifbillah commented 5 years ago

I see, so you left it up to the user to properly scale their data, right? However, NIBABEL scales it automatically before reporting.

addisonElliott commented 5 years ago

Oh I understand! Are you wanting it scaled between 0.0f and 1.0f?

This is essentially what you need then?

(data - oldmin) / (oldmax - oldmin)

If so, do you think it's worthwhile to include that in the pynrrd library or leave it up to the user?

addisonElliott commented 5 years ago

@tashrifbillah ping. Am I on the right track? What are your thoughts on this?

tashrifbillah commented 5 years ago

Label: Follow Up Required Assign: @tashrifbillah

I am organizing the issue in a better way. Let's keep it open, I shall follow up soon.

tashrifbillah commented 5 years ago

Are you wanting it scaled between 0.0f and 1.0f?

Negative, we don't want to normalize data, rather scale data properly taking into account oldmin and oldmax.

So, how do we do that using math?

Given oldmin, oldmax, datatype:

oldmin=(datatype_min)*scl_slope + scl_inter
oldmax= (datatype_max)*scl_slope + scl_inter

then, we should solve the above equations to find scl_slope and scl_inter. Then, the reported data should be:

scaled_data= raw_data*scl_slope+scl_inter

You can take a look at my issue on ITK where I justified reporting scaled data like NIBABEL and FSLview.

addisonElliott commented 5 years ago

👍 I'm with you, all makes sense.

This should be doable from this library via the custom header fields options. Not sure this is worth being built-in functionality personally. If anything, we could add a helper function that users can utilize.

What are your thoughts?

Chida15 commented 5 years ago

Hello, do you know how to get the coronal plane, why can I just get the cross section?

tashrifbillah commented 5 years ago

@Chida15 , do you mind opening a separate issue on your question? This thread is for other issue we are trying to solve.