It is interesting to note that the DICOMDIR file itself is a valid DICOM file (uses the 'Basic Directory IOD' structure) and information is encoded into it using the 'Explicit VR Little Endian Transfer Syntax (UID=1.2.840.10008.1.2.1)' as no negotiation is permitted as mentioned earlier. Create and burn DICOM CD/DVD disks from a DICOMDIR file Select the menu command 'CD/DVD BurningCreate DICOM CD/DVD From DICOMDIR'. A dialog box appears and it allows you to select the desired DICOMDIR file. A second dialog box allows you to select the desired series from the contents of the DICOMDIR file.

This python module provides a set of utilities for extracting data contained inDICOM files into Numpy ndarrays. It is a higher-level library that builds on the excellent lower-level pydicom library.

The library is quite small at the moment, however, if you have a DICOM-relatedutility function that you think would be appropriate to include, create aGithub Issue!

Dependencies¶

Python 2.7 or Python 3.5+
Numpy
PyDicom

Installation¶

Source Code¶

The source code is hosted on Github.

Combine DICOM Slices¶

The DICOM standard stores MR, CT, and PET scans as a series of images saved ina separate files. A common task is to combine all of the images that make up asingle 3D image into a single scan.

The function that performs this task is combine_slices. Since this librarybuilds on pydicom, combine_slices takes an list of pydicomdatasets.

Example¶

Details¶

dicom_numpy.combine_slices(datasets, rescale=None)¶

Given a list of pydicom datasets for an image series, stitch them together into athree-dimensional numpy array. Also calculate a 4x4 affine transformationmatrix that converts the ijk-pixel-indices into the xyz-coordinates in theDICOM patient’s coordinate system.

Returns a two-tuple containing the 3D-ndarray and the affine matrix.

If rescale is set to None (the default), then the image array dtypewill be preserved, unless any of the DICOM images contain either theRescale Slope or theRescale Interceptattributes. If either of these attributes are present, they will beapplied to each slice individually.

If rescale is True the voxels will be cast to float32, if set toFalse, the original dtype will be preserved even if DICOM rescaling information is present.

The returned array has the column-major byte-order.

Datasets produced by reading DICOMDIR files are ignored.

This function requires that the datasets:

Be in same series (have the sameSeries Instance UID,Modality,and SOP Class UID).
The binary storage of each slice must be the same (have the sameBits Allocated andPixel Representation).
The image slice must approximately form a grid. This means there can notbe any missing internal slices (missing slices on the ends of the datasetare not detected).
It also means that each slice must have the sameRows,Columns,Samples Per Pixel,Pixel Spacing, andImage Orientation (Patient)attribute values.
The direction cosines derived from theImage Orientation (Patient)attribute must, within 1e-4, have a magnitude of 1. The cosines mustalso be approximately perpendicular (their dot-product must be within1e-4 of 0). Warnings are displayed if any of these approximations arebelow 1e-8, however, since we have seen real datasets with values up to1e-4, we let them pass.
The Image Position (Patient)values must approximately form a line.

If any of these conditions are not met, a dicom_numpy.DicomImportException is raised.

dicom_numpy.sort_by_slice_position(slice_datasets)¶

Given a list of pydicom Datasets, return the datasets sorted in the image orientation direction.

This does not require pixel_array to be present, and so may be used to associate instance Datasetswith the voxels returned from `combine_slices.

Change Log¶

Version 0.4.0¶

Ignore DICOMDIR files
Fix bug that was triggered when using from dicom_numpy import *
Make combine_slices work with a single slice
Add support for “channeled slices” (e.g., RGB slices)
Allow HighBit and BitsStored DICOM attributes to be non-uniform
Drop support for Python 3.4; test Python 3.7
Require the SamplesPerPixel DICOM attribute to be invariant among the slices

Version 0.3.0¶

Reverted slice ordering change from v0.2.0, since the DICOM standard definesthe Z-axis direction to be increasing in the direction of the head.
Added support for both PyDicom 0.X and 1.X

Version 0.2.0¶

Changed the behavior of combine_slices to stack slices from head (slice 0)to foot (slice -1). Note that this is the reverse of the behavior in v0.1.*.

Version 0.1.5¶

Added the rescale option to combine_slices
Made combine_slices’s returned ndarray use column-major ordering

Contributing¶

Process¶

Contributions are welcome. Please create a Github issue describing the changeyou would like to make so that you can discuss your approach with themaintainers. Assuming the maintainers like your approach, then create a pullrequest.

Tests¶

Most new functionality will require unit tests.

Create Dicomdir From Dicom Files File

Run all of the tests for each supported python version using:

Run all of the tests for the currently active python version using:

Other Contributors¶

Additional contributions made by:

Jonathan Daniel

Thank you!

About Innolitics¶

Innolitics is a team of talented software developers with medical andengineering backgrounds. We help companies produce top quality medical imagingand workflow applications. If you work with DICOM frequently, our DICOMStandard Browser may be useful to you.

If you could use help with DICOM, let us know! We offer training sessions andcan provide advice or development services.

Licenses¶

Permission is hereby granted, free of charge, to any person obtaining a copy ofthis software and associated documentation files (the “Software”), to deal inthe Software without restriction, including without limitation the rights touse, copy, modify, merge, publish, distribute, sublicense, and/or sell copiesof the Software, and to permit persons to whom the Software is furnished to doso, subject to the following conditions:

The above copyright notice and this permission notice shall be included in allcopies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS ORIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THEAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHERLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THESOFTWARE.

Contents

Transcoding of DICOM files
- Solutions to avoid transcoding

DICOM transcoding refers to the process of changing thetransfer syntax of some DICOM instance.

Transcoding can be used to compress DICOM images, which is usefulto reduce the required storage area or to optimize the networkbandwidth for exchanges between sites. Conversely, transcoding can beused to decompress DICOM images, which is needed as many DICOMviewers or specialized analysis software do not provide support forcompressed transfer syntaxes.

Support for transcoding was introduced in Orthanc 1.7.0.

Let’s consider the following basic workflow, in which some imagingworkstation must access a medical image that originates from a PACSand that is served through an Orthanc proxy:

This is quite a common situation, e.g. in university hospitals whereresearchers must access medical images without having authorization tolog in