Fixed ValueError and IndexError

[charleygros]

This PR aimed at fixing some ValueError or IndexError errors that have been reported to occur when using sct_deepseg_sc or sct_deepseg_lesion.

#2016

To reproduce the error:

cd example_data/mt
sct_image -i mt1.nii.gz -split z
sct_deepseg_sc -i mt1_Z0005.nii.gz -c t2 -centerline cnn

Error:

charley@charley:~/data/example_data/mt$ sct_deepseg_sc -i mt1_Z0005.nii.gz -c t2 -centerline cnn

--
Spinal Cord Toolbox (cg_deepseg_valueError_indexError_i1921_i2015_i2016/9d9b320757fc4a1db5bac3f706a01817c6b351c7)
Running /home/charley/spinalcordtoolbox/scripts/sct_deepseg_sc.py -i mt1_Z0005.nii.gz -c t2 -centerline cnn
/home/charley/spinalcordtoolbox/python/lib/python2.7/site-packages/h5py/__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  from ._conv import register_converters as _register_converters

Method:
	Centerline algorithm: cnn
	Assumes brain section included in the image: True
	Dimension of the segmentation kernel convolutions: 2d

Creating temporary folder...
cp mt1_Z0005.nii.gz /tmp/sct-20181108144431.270174-WGI6gM
Reorient the image to RPI, if necessary...
Resample the image to 0.5 mm isotropic resolution...
/home/charley/spinalcordtoolbox/python/lib/python2.7/site-packages/nipy/io/files.py:145: FutureWarning: Default `strict` currently False; this will change to True in a future version of nipy
  ni_img = nipy2nifti(img, data_dtype = io_dtype)
Finding the spinal cord centerline...
WARNING: To avoid intensity overflow due to convertion to uint8, intensity will be rescaled to the maximum quantization scale.
Traceback (most recent call last):
  File "/home/charley/spinalcordtoolbox/scripts/sct_deepseg_sc.py", line 853, in <module>
    main()
  File "/home/charley/spinalcordtoolbox/scripts/sct_deepseg_sc.py", line 844, in main
    remove_temp_files=remove_temp_files, verbose=verbose)
  File "/home/charley/spinalcordtoolbox/scripts/sct_deepseg_sc.py", line 677, in deep_segmentation_spinalcord
    centerline_fname=file_ctr)
  File "/home/charley/spinalcordtoolbox/scripts/sct_deepseg_sc.py", line 478, in find_centerline
    brain_bool=brain_bool)
  File "/home/charley/spinalcordtoolbox/scripts/sct_deepseg_sc.py", line 391, in heatmap
    z_max = np.max(list(set(np.where(data)[2])))
  File "/home/charley/spinalcordtoolbox/python/lib/python2.7/site-packages/numpy/core/fromnumeric.py", line 2334, in amax
    initial=initial)
  File "/home/charley/spinalcordtoolbox/python/lib/python2.7/site-packages/numpy/core/fromnumeric.py", line 83, in _wrapreduction
    return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
ValueError: zero-size array to reduction operation maximum which has no identity
Sentry is attempting to send 1 pending error messages
Waiting up to 10 seconds
Press Ctrl-C to quit
Total processing time: 0 min 6 s

It should be noted that this error did not occur when using -centerline svm. Also, this is not related to the fact that the input image is a single axial slice, since recent improvements (see PR_2012) allow to handle this kind of input.
For instance, the below command works:

sct_deepseg_sc -i mt1_Z0005.nii.gz -c t2 -centerline cnn

Proposed solution:
The reported error occurs when the spinal cord was not detected (here on the single axial slice). There is no much that we can do about it. Consequently, I proposed in the PR to invite the user to use another -centerline method, such as svm or viewer:

charley@charley:~/data/example_data/mt$ sct_deepseg_sc -i mt1_Z0005.nii.gz -c t2 -centerline cnn

--
Spinal Cord Toolbox (cg_deepseg_valueError_indexError_i1921_i2015_i2016/9d9b320757fc4a1db5bac3f706a01817c6b351c7)
Running /home/charley/spinalcordtoolbox/scripts/sct_deepseg_sc.py -i mt1_Z0005.nii.gz -c t2 -centerline cnn
/home/charley/spinalcordtoolbox/python/lib/python2.7/site-packages/h5py/__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  from ._conv import register_converters as _register_converters

Method:
	Centerline algorithm: cnn
	Assumes brain section included in the image: True
	Dimension of the segmentation kernel convolutions: 2d

Creating temporary folder...
cp mt1_Z0005.nii.gz /tmp/sct-20181108163141.602759-T16TNs
Reorient the image to RPI, if necessary...
Resample the image to 0.5 mm isotropic resolution...
/home/charley/spinalcordtoolbox/python/lib/python2.7/site-packages/nipy/io/files.py:145: FutureWarning: Default `strict` currently False; this will change to True in a future version of nipy
  ni_img = nipy2nifti(img, data_dtype = io_dtype)
Finding the spinal cord centerline...
WARNING: To avoid intensity overflow due to convertion to uint8, intensity will be rescaled to the maximum quantization scale.

Spinal cord was not detected using "-centerline cnn". Please try another "-centerline" method.

Sentry is attempting to send 1 pending error messages
Waiting up to 10 seconds
Press Ctrl-C to quit
Total processing time: 0 min 6 s

#2015 & #1921

To reproduce:

cd example_data/t2
sct_deepseg_sc -i t2.nii.gz -c t2 -kernel 3d

Error:
It raised an IndexError when an empty slice segmentation is encountered by the function post_processing_slice_wise.

Proposed solution:
Use the center of mass of the neighbour slices to perform the slice wise post processing. If the neighbours are also empty (ie all the slices of a 3D patches are not segmented), then no post-processing is performed.

--> Tested on the example_data and it fixes the present error without impacting the output segmentation.

#2001

This issue suggests to correct the Usage of sct_deepseg_lesion since -c t2_ax is not a type of contrast but a contrast_orientation.

Proposed solution:
Clarify the usage:

 -c {t2,t2_ax,t2s}            type of image contrast. For 't2_like' contrast images, a specific 
                              model for images with axial orientation is available: 't2_ax'.

#2057

svm is now the -centerline default value for all contrasts , as indicated in the Usage.

charley@charley:~/data/example_data/t1$ sct_deepseg_sc -i t1.nii.gz -c t1

--
Spinal Cord Toolbox (cg_deepseg_valueError_indexError_i1921_i2015_i2016/d2f524f877add0fa59cbfc0d78e27db69afc390b)
Running /home/charley/spinalcordtoolbox/scripts/sct_deepseg_sc.py -i t1.nii.gz -c t1
/home/charley/spinalcordtoolbox/python/lib/python2.7/site-packages/h5py/__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  from ._conv import register_converters as _register_converters

Method:
	Centerline algorithm: svm
	Assumes brain section included in the image: True
	Dimension of the segmentation kernel convolutions: 2d

#1825

To reproduce:

cd example_data/t2s
sct_deepseg_sc -i t2s.nii.gz -c t2s -ofolder test_folder

Error:
Duplication of output_folder in the display_syntax_viewer:

fsleyes t2s.nii.gz -cm greyscale test_folder/test_folder/t2s_seg.nii.gz -cm red -a 70.0 &

This is now fixed:

fsleyes t2s.nii.gz -cm greyscale test_folder/t2s_seg.nii.gz -cm red -a 70.0 &

[charleygros]

PR ready for review.

[jcohenadad]

I would be more exhaustive in the description. E.g.: "type of image contrast. t2: T2w scan with isotropic resolution. t2_ax: T2w scan with axial orientation and thick slices. t2s: T2*w scan with axial orientation and thick slices."

[charleygros]

Thank you for your feedback. It has been corrected.

[jcohenadad]

nice-- just a minor change in the description. Thanks for this fix!