icometrix / dicom2nifti Goto Github PK

Installed with pip install.

$ dicom2nifti --help
Traceback (most recent call last):
  File "/Users/fedorov/.pyenv/versions/2.7.8/bin/dicom2nifti", line 14, in <module>
    import dicom2nifti.convert_dir as convert_directory
  File "/Users/fedorov/.pyenv/versions/2.7.8/lib/python2.7/site-packages/dicom2nifti/__init__.py", line 20, in <module>
    from .convert_dicom import dicom_series_to_nifti
  File "/Users/fedorov/.pyenv/versions/2.7.8/lib/python2.7/site-packages/dicom2nifti/convert_dicom.py", line 8, in <module>
    import dicom2nifti.patch_pydicom_encodings
  File "/Users/fedorov/.pyenv/versions/2.7.8/lib/python2.7/site-packages/dicom2nifti/patch_pydicom_encodings.py", line 5, in <module>
    import dicom.charset
ImportError: No module named dicom.charset

The assumption that we need to use the private scale and slope flags seems to be wrong as we are encountering some strange values due to this scaling.

It was pointed out to us that we might be lacking support for some GE dti data as can be found here: http://www.cabiatl.com/mricro/mricron/dcm2nii.html.
Looking into the data it seems to be older GE 4D data that does not use the StackID as newer data does. Updating the group dicoms function to handle the older data correctly should fix this.

Input test dataset (it is a time-resolved DCE acquisition): http://slicer.kitware.com/midas3/item/126198

Input dataset as visualized in 3D Slicer (http://slicer.org):

Result of dicom2nifti conversion as visualized in 3D Slicer (relative position preserved):

This warning occur in some dicom directory,while others are ok. Can you tell me what's wrong ?

Hi,

thanks for dicom2nifti. I was wondering if you are considering changes for returning the filename of the created nifti file after calling dicom2nifti.convert_directory. My use-case is converting one directory with one series and I want to pass the nifti file reference to some other program for further processing.

Thanks for your time,
Joshy

PS: I can also create a pull request if you want.

Some Philips Multiframe dicom files do not contain SpacingBetweenSlices on the top level. The affine should be generated without making use of this

Hello sir or madam,
thanks for the dicom2nifti.
I dont understand how the name of the generated nifti file is saved. When I open the data in RadiAnt oder ITK. I get this ONE "t1_mpr_ns_sag_K..RECO" . But when I use the code line with dicom2nifti.convert_directory, I get "21_tfl3d1_ns" which is the saved name of the file. I dont understand how it comes to this name??

Is there a function. if you want to save the nifti files like it was hierarchically defined in the folder
Meaning
Folder A -> Folder A1: Dicom1
-> Folder A2: Dicom2

And then in the code dicom2nifti.convert_directory(Folder A, Folder B), we have

Folder B -> Folder A1: Nifti1
-> Folder A2: Nifti2

Is it possible to keep this hierarchy?

Thanks

They are currently in image(scanner) coordinates:
https://github.com/icometrix/dicom2nifti/blob/master/dicom2nifti/convert_siemens.py#L528

**Hi, I am trying to loop through different folders of dicom data and convert them to nifti files. The code converts the first dicom folder to nifti but then throughs in an error (see below)*. if I try to do one folder at a time it works. Will really appreciate help *

Code:
path=os.chdir('/Users/abhinandanbatra/Desktop/xxxx/xxxxx') newpath='/Users/abhinandanbatra/Desktop/yyyy' subfolders=os.listdir(path) print(subfolders) for f in subfolders: dicom2nifti.dicom_series_to_nifti(f, os.path.join(newpath,f+'_label'), reorient_nifti=True)

Error:

FileNotFoundError Traceback (most recent call last)
/anaconda3/lib/python3.7/site-packages/nibabel/loadsave.py in load(filename, **kwargs)
39 try:
---> 40 stat_result = os.stat(filename)
41 except OSError:

FileNotFoundError: [Errno 2] No such file or directory: '/Users/abhinandanbatra/Desktop/pyradiomics/nifti/C6012011_label'

During handling of the above exception, another exception occurred:

FileNotFoundError Traceback (most recent call last)
in ()
7 for f in subfolders:
8 # dicom_input = dicom2nifti.common.read_dicom_directory(f)
----> 9 dicom2nifti.dicom_series_to_nifti(f, os.path.join(newpath,f+'_label'), reorient_nifti=True)
10 # if f.startswith('C6'):
11 # dicom2nifti.dicom_series_to_nifti(f, os.path.join(newpath,f+'_label'), reorient_nifti=True)

/anaconda3/lib/python3.7/site-packages/dicom2nifti/convert_dicom.py in dicom_series_to_nifti(original_dicom_directory, output_file, reorient_nifti)
84 dicom_input = common.read_dicom_directory(dicom_directory)
85
---> 86 return dicom_array_to_nifti(dicom_input, output_file, reorient_nifti)
87
88 except AttributeError as exception:

/anaconda3/lib/python3.7/site-packages/dicom2nifti/convert_dicom.py in dicom_array_to_nifti(dicom_list, output_file, reorient_nifti)
144 # do image reorientation if needed
145 if reorient_nifti:
--> 146 image_reorientation.reorient_image(results['NII_FILE'], results['NII_FILE'])
147 return results
148

/anaconda3/lib/python3.7/site-packages/dicom2nifti/image_reorientation.py in reorient_image(input_image, output_image)
29 # and get the image data in RAS orientation
30 # print 'Reading nifti'
---> 31 image = load(input_image)
32
33 # 4d have a different conversion to 3d

/anaconda3/lib/python3.7/site-packages/dicom2nifti/image_volume.py in load(nifti_file)
41
42 def load(nifti_file):
---> 43 nifti_image = nibabel.load(nifti_file)
44 return ImageVolume(nifti_image)
45

/anaconda3/lib/python3.7/site-packages/nibabel/loadsave.py in load(filename, **kwargs)
40 stat_result = os.stat(filename)
41 except OSError:
---> 42 raise FileNotFoundError("No such file or no access: '%s'" % filename)
43 if stat_result.st_size <= 0:
44 raise ImageFileError("Empty file: '%s'" % filename)

FileNotFoundError: No such file or no access: '/Users/abhinandanbatra/Desktop/pyradiomics/nifti/C6012011_label'

I only learned about the requirement of gdcmconv (for compressed dicom files) via error messages and looking into the code.

This is generally a bad idea and fails on read only systems:

the dicom2nifti guy is unable to convert the dicom rs fMRI file into Nifty (Philips) and this problem is not there for T-1 volumes. i am not too sure what is the issue .
Can some one advice me please
Joga

Hi;

I was wondering if it was possible to convert Relative blow volume maps (RGB) to nifti using the software you have created

When enabling resampling for gantry tilted data we encounter some crashes.
This is due to trying to run the resampling on non gantry tilted niftis.
Use the general check from the beginning as this seems to perform better than the one in the resampling code

It seems that the pixel spacing is mixed up in the case of non isotropical slices.

We got some data that seems to be "preprocessed" using "PMOD Technologies" software (don't know anything more about this) and has non-isotropical axial slices as input.

Validation using slicer and horos shows that we mix up the pixelspacing direction in the affines.
This is usually not an issue as we never encounter non isotropical slices

Currently ImageType is not optional when converting Siemens data.
This should be optional as it is type 3.

Don't know if it is possible as not all dependencies are available on conda

get the error "AttributeError: 'Dataset' object has no attribute 'fileobj_type'" when trying to work with a dicom data set. I am perfectly able to open the dataset using pydicom and extracting the array and header infor using for loops.

dataset included

edit: using python 3.7 and dicom2nifti 2.0.5

code bellow:

import dicom2nifti a = r'C:\Users\User\Google Drive\Dissertation\Work\Subjects\sub_01' b = r'C:\Users\User\Google Drive\Dissertation\Work' dicom2nifti.convert_directory(a, b)

error message:

"Unable to read: C:\Users\User\Google Drive\Dissertation\Work\Subjects\sub_01\EXP180105_LYR_SCAN14.MR.SYTSAI_BRAIN.0003.0001.2018.01.05.11.29.40.942370.31044194.IMA
Traceback (most recent call last):
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\dicom2nifti\convert_dir.py", line 57, in convert_directory
force=dicom2nifti.settings.pydicom_read_force)
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\dicom2nifti\compressed_dicom.py", line 15, in read_file
if _is_compressed(dicom_file, force):
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\dicom2nifti\compressed_dicom.py", line 107, in _is_compressed
force=force)
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\pydicom\filereader.py", line 880, in dcmread
force=force, specific_tags=specific_tags)
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\pydicom\filereader.py", line 771, in read_partial
is_implicit_VR, is_little_endian)
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\pydicom\dataset.py", line 1364, in init
Dataset.init(self, dataset)
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\pydicom\dataset.py", line 177, in init
dict.init(self, *args)
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\pydicom\dataset.py", line 597, in getitem
self.fileobj_type, self.filename, self.timestamp,
File "c:\users\user\appdata\local\programs\python\python37\lib\site-packages\pydicom\dataset.py", line 520, in getattr
return super(Dataset, self).getattribute(name)
AttributeError: 'Dataset' object has no attribute 'fileobj_type'
Unable to read: C:\Users\User\Google Drive\Dissertation\Work\Subjects\sub_01\EXP180105_LYR_SCAN14.MR.SYTSAI_BRAIN.0003.0002.2018.01.05.11.29.40.942370.31044212.IMA"

sub_01.zip

We sometimes get data that contains duplicate slices (same position/data). Probably due to wrong export form scanner/pacs. This results in an "inconsistent slice increment" error however we could automatically filter aways these slices as they have the same position and data.
This way we can convert the data without manual interactions

numpy.matrix is being deprecated. Getting the following warning:

dicom2nifti/common.py:540: PendingDeprecationWarning: the matrix subclass is not 
the recommended way to represent matrices or deal with linear algebra (see 
https://docs.scipy.org/doc/numpy/user/numpy-for-matlab-users.html). 
Please adjust your code to use regular ndarray.

It's currently being used in:

• dicom2nifti/common.py
• dicom2nifti/convert_philips.py
• dicom2nifti/convert_siemens.py

To avoid pending deprecation warnings, I suggest using numpy.ndarray instead of numpy.matrix.

This is ultimately what is causing #25

Here are the ImagePositionPatient (sorted by InstanceNumber):

array([[-107.980637,  -86.787643,  -64.592575],
       [-107.980637,  -87.787643,  -58.192574],
       [-108.080635,  -88.887642,  -51.792572],
       [-108.080635,  -89.98764 ,  -45.39257 ],
       [-108.180634,  -91.087646,  -38.992569],
       [-108.180634,  -92.187637,  -32.592571],
       [-108.28064 ,  -93.287643,  -26.192572],
       [-108.28064 ,  -94.387642,  -19.792572],
       [-108.380638,  -95.387642,  -13.392571],
       [-108.380638,  -96.48764 ,   -6.992572],
       [-108.480637,  -97.587646,   -0.492571],
       [-108.480637,  -98.687645,    5.907429],
       [-108.580635,  -99.787643,   12.307428],
       [-108.580635, -100.887642,   18.70743 ],
       [-108.680634, -101.887642,   25.107428],
       [-108.680634, -102.98764 ,   31.507429],
       [-108.78064 , -104.087639,   37.907425],
       [-108.78064 , -105.187645,   44.307426],
       [-108.880638, -106.287643,   50.707428],
       [-108.880638, -107.387642,   57.10743 ],
       [-108.980637, -108.48764 ,   63.507423],
       [-109.080635, -109.48764 ,   69.907425]])

and the first difference:

array([[ 0.      , -1.      ,  6.400001],
       [-0.099998, -1.099999,  6.400002],
       [ 0.      , -1.099998,  6.400002],
       [-0.099999, -1.100006,  6.400001],
       [ 0.      , -1.099991,  6.399998],
       [-0.100006, -1.100006,  6.399999],
       [ 0.      , -1.099999,  6.4     ],
       [-0.099998, -1.      ,  6.400001],
       [ 0.      , -1.099998,  6.399999],
       [-0.099999, -1.100006,  6.500001],
       [ 0.      , -1.099999,  6.4     ],
       [-0.099998, -1.099998,  6.399999],
       [ 0.      , -1.099999,  6.400002],
       [-0.099999, -1.      ,  6.399998],
       [ 0.      , -1.099998,  6.400001],
       [-0.100006, -1.099999,  6.399996],
       [ 0.      , -1.100006,  6.400001],
       [-0.099998, -1.099998,  6.400002],
       [ 0.      , -1.099999,  6.400002],
       [-0.099999, -1.099998,  6.399993],
       [-0.099998, -1.      ,  6.400002]])

So it gets broken up into 11 stacks of length 2.

Looking at the DICOM file (first one):

(0018, 0050) Slice Thickness                     DS: "5.000000"
(0018, 0088) Spacing Between Slices              DS: "6.500000"
(0020, 0032) Image Position (Patient)            DS: ['-107.980637', '-86.787643', '-64.592575']
(0020, 0037) Image Orientation (Patient)         DS: ['0.998308', '0.055482', '0.017426', ' -0.057589', '0.984418', '0.166145']
(0028, 0030) Pixel Spacing                       DS: ['0.8593634367', '0.8593869209']

If gdcmconv cannot be found, subprocess.check_output is then called with an executable of None leading to this not so useful stack:

/usr/local/lib/python3.5/dist-packages/dicom2nifti/compressed_dicom.py in _decompress_dicom(dicom_file)
    116         gdcmconv_executable = _which('gdcmconv.exe')
    117 
--> 118     subprocess.check_output([gdcmconv_executable, '-w', dicom_file, dicom_file])
    119 
    120 

/usr/lib/python3.5/subprocess.py in check_output(timeout, *popenargs, **kwargs)
    624 
    625     return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
--> 626                **kwargs).stdout
    627 
    628 

/usr/lib/python3.5/subprocess.py in run(input, timeout, check, *popenargs, **kwargs)
    691         kwargs['stdin'] = PIPE
    692 
--> 693     with Popen(*popenargs, **kwargs) as process:
    694         try:
    695             stdout, stderr = process.communicate(input, timeout=timeout)

/usr/lib/python3.5/subprocess.py in __init__(self, args, bufsize, executable, stdin, stdout, stderr, preexec_fn, close_fds, shell, cwd, env, universal_newlines, startupinfo, creationflags, restore_signals, start_new_session, pass_fds)
    945                                 c2pread, c2pwrite,
    946                                 errread, errwrite,
--> 947                                 restore_signals, start_new_session)
    948         except:
    949             # Cleanup if the child failed starting.

/usr/lib/python3.5/subprocess.py in _execute_child(self, args, executable, preexec_fn, close_fds, pass_fds, cwd, env, startupinfo, creationflags, shell, p2cread, p2cwrite, c2pread, c2pwrite, errread, errwrite, restore_signals, start_new_session)
   1472                     else:
   1473                         env_list = None  # Use execv instead of execve.
-> 1474                     executable = os.fsencode(executable)
   1475                     if os.path.dirname(executable):
   1476                         executable_list = (executable,)

/usr/lib/python3.5/os.py in fsencode(filename)
    860             return filename.encode(encoding, errors)
    861         else:
--> 862             raise TypeError("expect bytes or str, not %s" % type(filename).__name__)
    863 
    864     def fsdecode(filename):

TypeError: expect bytes or str, not NoneType

I have a script to rename the .nii.gz files after conversion, but these names do not carry over if I unzip the files with 7-zip, which needs to be done for use with Vivoquant/

There doesn't seem to be a way to unzip the .nii.gz files directly with python, which would solve the problem, and stopping the script midway to have the user unzip everything wouldn't be ideal.

With that in mind, I would like to ask please whether it's possible to set dicom2nifti to convert DICOMs straight into .nii files? Or, alternatively, whether such functionality could be added in the future?

Thank you for your time.

Update imagevolume to be more like nibabel and use load to open an image

Logger is changed on importing the module

This means that if I use dicom2nifti rather than running as a script, the logger is affected. Better would be to just set logger from runnable scripts.

Currently it seems that we have a mixture of BVEC in both world and image coordinates. We should make this uniform. It would also be nice that we can switch between them using the configurations.

Add a section about error codes and settings to the documentation:
For the errors explain what they mean and how to solve and work around them.
For the settings explain what they do.

Currently we have a way of removing non imaging dicoms in convert_directory however this is not applied when going directly to convert_dicom.
It would be good to move this to common so it can be applied everywhere.

Problem 1: InstanceNumber is Type 2 attribute - it can be missing. As of current version of dicom2nifti, missing InstanceNumber results in no output produced.

Problem 2: There is also no hard rule that requires InstanceNumber to establish the spatial ordering of the frames. The only reliable way to sort instances and reconstruct volume is to sort them using Image Position (Patient) and Image Orientation (Patient) (also see related discussion in http://dclunie.blogspot.com/2013/10/how-thick-am-i-sad-story-of-lonely-slice.html), and this approach is currently not implemented in the tool.

Speaking for myself, until this issue is resolved, it is hard to seriously consider using dicom2nifti.

I get the following DepracationWarning from convert_dir.py:

.../dicom2nifti/convert_dir.py:147: DeprecationWarning: invalid escape sequence \w
.../dicom2nifti/convert_dir.py:148: DeprecationWarning: invalid escape sequence \s

To avoid these warnings, regex should be a raw string:

cleaned_filename = re.sub(r'[^\w\s-]', '', cleaned_filename.strip().lower())
cleaned_filename = re.sub(r'[-\s]+', '-', cleaned_filename)

Add continuous integration by plugging in for example circle ci to this github repo

During conversion of certain siemens dti datasets it seemed that the pixel data was not properly translated to the raw binary data. Seems like a little/big endian problem.

For some Hitachi mri data it became apparent that there is an overflow issue.
Some investigation lead to a but in pydicom where it does not seem to handle cases where bitsstored and bitsallocated do not match. The overflow occurs because the for example a 12 bit signed integer is stored in a 16 bit space. As the sign bit is the most significant bit (12) we cannot read this data as if it where 16 bits.

We should:

• Provide a workaround
• Report a bug with pydicom

Implement structure to further support hitachi data.
We currently support anatomical data but we can already implement the vendor specific structure, tests and documentation.

tempfile.NamedTemporaryFile used to store uncompressed dicom files (via gdcmconv) have a different behavior on Windows.

As the tmpfile is already open the subprocess trying to open it again fails.

It is an open bug but has not received much attention. As of now it is expected behavior breaking dicom2nifti on Windows.

just as the titles says, how can I get an uncompressed nifti output?

for use in certain scenarios it would be great to just return the nibabel object instead of writing to a file. perhaps the output_file can be turned into a keyword argument and if set to None will return an image rather than a path to a file.

Some dicom data from adni cannot be converted like 130_S_2403.
This is possibly due to wrong anonimisation however we could work around this

I have a large group of DICOM series and due to some acquisition rounding errors two series have an inconsistent slice increment. Unfortunately, it is impossible for me to redo the acquisition so i was hoping i could fix it during the conversion. Is there any way to increase the error tolerance for the conversion?

Because it is probably introduced due to rounding the error is extremely small, so i would say it shouldn't really matter for the conversion.

It would be great to have additional support for gantry tilted CT.
We currently support gantry tilted ct by taking it into account in the affine headers, however for certain situations a resampled orthogonal image is prefered.
Adding an option to resample the data would be great.

I have some GE files for which the first and last sorted dicoms have the same image position. This leads to an invalid affine (a column of all 0s).

The issue is that all steps are 0:

This can perhaps just be a better error rather than trying to construct the nifti file.

Hi;

I am trying to convert a dcm file attained from TCIA. however I get the error AttributeError: Dataset does not have attribute 'SeriesNumber'.
. ANy work around for the same

Can I resample the 3D brain MR data to 1mm1mm1mm? What is command for it?

In certain cases (extremely rare) there seems to be an overflow issue.

Hi! I'm trying to convert a DICOM directory to a nifti file with the function dicom2nifti.dicom_series_to_nifti but I always get this error:
ConversionError: GDCMCONV_NOT_FOUND

Can anyone tell me the reason and how I can solve it? I really appreciate any help you can provide.

As pydicom 1.0 was released it is no longer needed to support the dirty double import.
We will thus remove it.

The resampling code seems to hang in some cases.
The issue is that we assume the image to be oriented correctly however we currently do the image reorientation after the resampling. This should be the other way around.

Right now it looks like the only multiframe DICOMs that are supported are those from Philips. There are a bunch of example (generic) multiframe DICOMs that can be found here: ftp://medical.nema.org/MEDICAL/Dicom/Multiframe/MR/

The DICOM format for multiframe is described here:

• http://www.dclunie.com/papers/SS3DICOMObjectsSyllabus.pdf
• and: ftp://medical.nema.org/medical/dicom/final/sup49_ft.pdf
• fields: http://dicom.nema.org/medical/Dicom/2015b/output/chtml/part03/sect_C.8.13.html