CODEWITHSUNDEEP
python-3.xsimpleitk
S.EB
S.EB

Reputation: 2236

How to save an array with the same spacing and origin into a SimpleITK image?

I want to write the image [that I have loaded into array] after changing the values of array into a SimpleITK Image. How can I set the spacing and origin?

[res_file_dir is the path to the nii.gz file]

segmentation_result=sitk.ReadImage(res_file_dir,sitk.sitkUInt8)   #segmentation_result.GetSpacing() (0.3645833432674408, 0.3645833432674408, 0.699999988079071)
seg=sitk.GetArrayFromImage(segmentation_result)  #shape (105, 242, 176)
seg[seg!=obj_label]=0

#segmentation=sitk.Image(segmentation_result.GetSize(),sitk.sitkUInt8)
 numpyOrigin = np.array(list(reversed(segmentation_result.GetOrigin())))  
 numpySpacing = np.array(list(reversed(segmentation_result.GetSpacing())))  
 segmentation = sitk.GetImageFromArray(seg, sitk.sitkVectorUInt8).SetSpacing(numpySpacing)

once I check the spacing of new image, it shows the following:

segmentation.GetSpacing()
*** AttributeError: 'NoneType' object has no attribute 'GetSpacing'

Upvotes: 1

Views: 3417

Answers (1)

Dave Chen
Dave Chen

Reputation: 2085

There are two problems with your last line. The first problem is that segmentation is getting set to the result of the call to SetSpacing, which has a return type of void.

The second problem is your second parameter to sitk.GetImageFromArray. It is a boolean for the named parameter isVector. In your case you want it to be False, which is the default value.

In your code, you are passing sitk.sitkVectorUInt8, which turns out to be an integer of value 13, to isVector. So isVector is getting True, the opposite of what you really want. With True you are getting a 2-d image with each pixel being a vector of length 176.

To double check, I would print out the results of the call to sitk.GetImageFromArray before doing the call to SetSpacing.

So here's what your code should be:

segmentation = sitk.GetImageFromArray(seg, isVector=False)
segmentation.SetSpacing(numpySpacing)

Note that if you want to mask out values in you image, you can do it in SimpleITK without converting to numpy and back. Here's how you could do that:

# create a binary mask image of 
seg = (segmentation_result==obj_label)

# create an image with pixel values of obj_label
segmentation = seg * obj_label

Upvotes: 2

Related Questions