Have you any suggestions : if I want to evaluate my segmentation but I have my image results as PNG ! what changes should be make to the original script!!!

[catactg]

Hi sebyo: I suggest you convert your PNG files to mhd format. You can do it using SimpleITK like this:

1. load each PNG file using sitk.ImageFileReader() and append each file to a list

   image_list = []


   for filename in filename_list:
    
    fullpath = os.path.join(input_dir, filename)
    
    reader = sitk.ImageFileReader()
    
    reader.SetImageIO('PNGImageIO')
    
    reader.SetFileName(fullpath)
    
    image = reader.Execute()
    
    image_list.append(image)

2. convert the list to a 3d numpy array (note that the coordinate standard for numpy is [z,y,x] while for itk is [x,y,z]

   slices_size = np.shape(sitk.GetArrayFromImage(itkimages[0]))

   number_of_slices = len(itkimages)

   npa = np.zeros((number_of_slices, slices_size[0], slices_size[1]))


   for slice_index, image in enumerate(itkimages):
   
   # only keep one channel (expected that greyscale image has identical channels)
    
   if image.GetNumberOfComponentsPerPixel() > 1:
        
       npa[slice_index, :, :] = sitk.GetArrayFromImage(image)[:, :, 0]
    
   else:
        
       npa[slice_index, :, :] = sitk.GetArrayFromImage(image)

   itk_image_stack = sitk.GetImageFromArray(npa)

3. save it as mhd using sitk.WriteImage()

   outpath = os.path.join(output_dir, your_filename + '.mhd')
   
   sitk.WriteImage(itk_image_stack, outpath)