imagedata 1.3.1rc1

Read/write medical image data

imagedata

Imagedata is a python library to read and write medical image data into numpy arrays. Imagedata will handle multi-dimensional data. In particular, imagedata will read and sort DICOM 3D and 4D series based on defined tags. Imagedata will handle geometry information between the formats.

The following formats are included:

• DICOM

• Nifti

• ITK (MetaIO)

• Matlab

• PostScript (input only)

Other formats can be added through a plugin architecture.

Install

pip install imagedata

Documentation

See the Documentation page for info.

Example code

A simple example reading two time series from dirA and dirB, and writing their mean to dirMean:

from imagedata.series import Series
a = Series('dirA', 'time')
b = Series('dirB', 'time')
assert a.shape == b.shape, "Shape of a and b differ"
# Notice how a and b are treated as numpy arrays
c = (a + b) / 2
c.write('dirMean')

Sorting

Sorting of DICOM slices is considered a major task. Imagedata will sort slices into volumes based on slice location. Volumes may be sorted on a number of DICOM tags:

• ‘time’: Dynamic time series, sorted on acquisition time

• ‘b’: Diffusion weighted series, sorted on diffusion b value

• ‘fa’: Flip angle series, sorted on flip angle

• ‘te’: Sort on echo time TE

In addition, volumes can be sorted on user defined tags.

Non-DICOM formats usually don’t specify the labelling of the 4D data. In this case, you can specify the sorting manually.

Viewing

A simple viewer. Scroll through the image stack, step through the tags of a 4D dataset. These operations are possible:

• Window/level adjustment: Move mouse with left key pressed.

• Scroll through slices of an image stack: Mouse scroll wheel, or up/down array keys.

• Step through tags (time, b-values, etc.): Left/right array keys.

• Move through series when many series are displayed: PageUp/PageDown keys.

# View a Series instance
a.view()

# View both a and b Series
a.view(b)

# View several Series
a.view([b, c, d])

Converting data from DICOM and back

In many situations you need to process patient data using a tool that do not accept DICOM data. In order to maintain the coupling to patient data, you may convert your data to e.g. Nifti and back.

Example using the command line utility image_data:

image_data --of nifti niftiDir dicomDir
# Now do your processing on Nifti data in niftiDir/, leaving the result in niftiResult/.

# Convert the niftiResult back to DICOM, using dicomDir as a template
image_data --of dicom --template dicomDir dicomResult niftiResult
# The resulting dicomResult will be a new DICOM series that could be added to a PACS

# Set series number and series description before transmitting to PACS using DICOM transport
image_data --sernum 1004 --serdes 'Processed data' \
  dicom://server:104/AETITLE dicomResult

The same example using python code:

from imagedata.series import Series
a = Series('dicomDir')
a.write('niftiDir', formats=['nifti'])   # Explicitly select nifti as output format

# Now do your processing on Nifti data in niftiDir/, leaving the result in niftiResult/.

b = Series('niftiResult', template=a)    # Or template='dicomDir'
b.write('dicomResult')   # Here, DICOM is default output format

# Set series number and series description before transmitting to PACS using DICOM transport
b.seriesNumber = 1004
b.seriesDescription = 'Processed data'
b.write(' dicom://server:104/AETITLE')

Series fields

The Series object is inherited from numpy.ndarray, adding a number of useful fields:

Axes

a.axes defines the unit and size of each dimension of the matrix

Addressing

4D: a[tags, slices, rows, columns]

3D: a[slices, rows, columns]

2D: a[rows, columns]

RGB: a[…, rgb]

patientID, patientName, patientBirthDate

Identifies patient

accessionNumber

Identifies study

seriesNumber, seriesDescription, imageType

Labels DICOM data

slices

Returns number of slices

spacing

Returns spacing for each dimension. Units depend on dimension, and could e.g. be mm or sec.

tags

Returns tags for each slice

timeline

Returns time steps for when a time series

transformationMatrix

The transformation matrix to calculate physical coordinates from pixel coordinates

Series instancing

From image data file(s):

a = Series('in_dir')

From a list of directories:

a = Series(['1', '2', '3'])

From a numpy array:

e = np.eye(128)
a = Series(e)

Series methods

write()

Write the image data as a Matlab file to out_dir:

a.write('out_dir', formats=['mat'])

slicing

The image data array can be sliced like numpy.ndarray. The axes will be adjusted accordingly. This will give a 3D b image when a is 4D.

b = a[0, ...]

Archives

The Series object can access image data in a number of archives. Some archives are:

Filesystem

Access files in directories on the local file system.

a = Series('in_dir')

Zip

Access files inside zip files.

# Read all files inside file.zip:
a = Series('file.zip')

# Read named directory inside file.zip:
b = Series('file.zip?dir_a')

# Write the image data to DICOM files inside newfile.zip:
b.write('newfile.zip', formats=['dicom'])

Transports

file

Access local files (default):

a = Series('file:in_dir')

dicom

Access files using DICOM Storage protocols. Currently, writing (implies sending) DICOM images only:

a.write('dicom://server:104/AETITLE')

Command line usage

The command line program image_data can be used to convert between various image data formats:

image_data --order time out_dir in_dirs