This program parses the web version of the DICOM Standard into human and machine-friendly JSON files. The purpose of these JSON files is twofold:
-
To provide a standardized and machine-readable way to access DICOM Standard information for software applications
-
To provide a logical model for the relationships between cross-referenced sections in the DICOM Standard
The finalized JSON output of this program is in the standard directory at the
top level of this project.
A GitHub Actions workflow regenerates the JSON files using the most current web version of the DICOM Standard at the beginning of each month. If there are any changes, the updated files are automatically pushed to master.
The DICOM Standard Parser is useful for modeling and understanding properties of the various abstractions defined by the DICOM Standard (IODs, modules, attributes, etc.) as well as the relationships between them.
The raw HTML or XML represents the DICOM Standard as a document but the data isn't easily machine-readable. We process the data from the HTML format into organized JSON files which follow a set of formatting guidelines and contain natural keys to represent relationships between abstractions.
To find or work with a smaller amount of data, e.g. a single IOD, the DICOM Standard Browser may be appropriate.
Install the latest release with pip install dicom-standard.
To download a specific JSON file, run curl https://raw.githubusercontent.com/innolitics/dicom-standard/master/standard/<filename> -o <destination>.
The generated JSON files conform to these formatting rules:
- JSON files representing objects are lists of dictionaries that each contain information relevant to the object.
- JSON files containing relational data between objects contain "foreign keys" to the relevant objects. These field names end with
Id, e.g.ciodIdandmoduleId.
Occasionally, files may deviate from this format when there is a very compelling reason. For example, references.json should be a list of reference objects where the href link is the id for each object. However, since almost every use case for references.json will use the href as a lookup, it makes more sense for the file to be set up as an object containing href to HTML pairs.
Applications that use the JSON files from this repository may need to re-organize data. A separate script must be written to join data from multiple tables into one file or prune out unnecessary fields.
The JSON generated by this program adheres to the following four rules:
- New fields may be added
- Bugs or incorrect data will be fixed as the standard changes
- No fields are removed, maintaining backwards compatibility
- The shape and organization of the JSON files will remain the same
The JSON files can be viewed here.
Please contact us if you use this software and would like your name or company listed here.
This program currently parses the DICOM Standard sections related to Information Object Definitions, modules, and attributes, as well as cross-referenced sections in other parts of the standard. This translates to the following sections:
Completely processed:
- PS3.3
- PS3.4
- PS3.6
Processed for references:
- PS3.15
- PS3.16
- PS3.17
- PS3.18
The Python scripts used to generate the JSON files are designed to be as extensible as possible. If you want to run the code yourself or configure your own custom parsing stage, you'll need the following system-level dependencies:
- Python 3.7
- Make + Unix tools
You will probably also want to setup a "virtual environment" (e.g. using Conda, or Pyenv + Virtualenv) to install the project dependencies into. Once you are in your "virtual environment", you can run:
$ make
to install and compile everything. Add the -j flag to speed this process up
significantly.
To download and parse the most up-to-date web version of the DICOM Standard, run the following commands:
$ make clean
$ make updatestandard
$ make
To download an older version of the DICOM Standard, run
$ make updatestandard VERSION=<version>
with the year and revision desired, e.g. 2018e, 2019c.
WARNING: Differences between previous versions and the current version may cause bugs when used with the current parser library. We recommend forking this repository if you need to use a specific version of the standard.
Parsing stages are indicated by prefixed names (i.e. extract_xxx.py or
process_xxx.py) and use a variety of utility functions from parse_lib.py
and other *_utils.py modules.
The overall data flow of this program takes the following form:
extract (post)process
Raw HTML ---------> JSON intermediate ---------------> JSON final
During this process, the following invariants are maintained:
- Each step in the parsing process is classified as either an "extract" stage, or a "process" stage.
- Stages are python scripts that take one or more files as inputs, and write their output to standard out.
- "Extract" stages takes one more more HTML input files and print out JSON.
- "Process" stages take one or more JSON files as inputs and print out JSON.
In this way, raw HTML is not touched by any stage other than extract_*.py,
and successive processing steps use increasingly refined JSON.
A map of all extraction and processing pathways is shown below:
+-------+ +----------+ +-------+ +-------+
| PS3.3 | | Other | | PS3.4 | | PS3.6 |
+---+---+ | DICOM | +---+---+ +---+---+
| | Sections | | |
| +-----+----+ | |
| | +----v----+ +-----v------+
+-------------+--------+------+-------------+ | | Extract | | Extract |
| | | | | | SOPs | | Attributes |
+---v-----+ +----v-----+ +------v------+ +---v--v---+ +----+----+ +-----+------+
| Extract | | Extract | | Extract | | Extract | | |
| CIODs/ | | CIODs/FG | | Modules/ | | Sections | v v
| Modules | | Macros | | Macro Attrs + +--------+-+ sops.json attributes.json
+----+----+ +----+-----+ +------+-----+ |
| | | |
+-------------+ | +---------------+ +-----------+
| | | | | |
+-----v-----+ +----v----+ +----v------+ +-----v------+ +-----v------+ |
| Process | | Process | | Process | | Preprocess | | Preprocess | |
| CIOD/ | | CIODs | | CIOD/FG | | Macros | | Modules/ | |
| Module | +----+----+ | Macro | | Attributes | | Attributes | |
| Relations | | | Relations | +-----+------+ +-----+------+ |
+-----+-----+ v +----+------+ | | |
| ciods.json | +-------+ +-------+ |
v | | | | | |
ciod_to_modules.json | +----v----+ | +----v----+ | |
v | Process | | | Process | | |
ciod_to_func_group_macros.json | Macros | | | Modules | | |
| +----+----+ | +----+----+ | |
| | | | | |
| v | v | |
| macros.json | modules.json | |
| | | | |
| | +-------v---+ +-------v---+ |
| | | Process | | Process | |
| | | Macro | | Module | |
| | | Attribute | | Attribute | |
| | | Relations | | Relations | |
| | +-------+---+ +-------+---+ |
| | | | |
| | +-v---------------v------v-+
| | | Postprocess |
| | | Add References +-----------+
| | +-+---------------+--------+ |
| | | | |
+------v-----------v---+ +-v---------------v--------+ +--------v--------+
| Postprocess <-----+ Postprocess | | Save References |
| Integrate Functional | | Merge Duplicate Nodes | +--------+--------+
| Group Macros <--+ +-------------+------------+ |
+-----------+----------+ | | |
| | v v
v +----macros_to_attributes.json references.json
modules_to_attributes.json
To update the parser map, please use ASCIIFlow.
To run the full test suite, install and run tox.
To run a specific test, run tox -e <testenv>. Test environments include:
flake8: check and enforce code style and formatmypy: validate type hintspytest: run a set of unit and end-to-end testsbuild-dist: test building the backend into source and binary distributions
Certain parts of the DICOM Standard site cause errors when running the parser, often due to typos or formatting inconsistencies.
When we find one of these issues, we add a hard-coded fix in the relevant file and add a comment starting with 'Standard workaround' that describes the issue and links to its location in the Standard. To be aware when these fixes are obsolete, we add a unit test that fails once the issue no longer exists.
Current standard workarounds (as of rev.2020b):
| Issue description | Workaround location |
|---|---|
| Table A.39.19-1 ends its title with an upper case "S" | extract_ciod_module_tables.pyparse_lib.py |
| Table A.32.9-2 has "Functional Groups Macros" in its title while other tables use "Functional Group Macros" | extract_ciod_func_group_macro_tables.pyparse_lib.py |
| Table A.52.4.3-1 is missing "Image" from the IOD name portion of the title | extract_ciod_func_group_macro_tables.py |
| Table C.8-125 has an extra "Sequence" in its title (should be "CT X-Ray Details", not "CT X-Ray Details Sequence") | parse_lib.py |
| Table A.84.3.2-1 contains a macro that has an extra "Macro" in its name ("Frame VOI LUT With LUT Macro") | process_ciod_func_group_macro_relationship.py |
| Table C.8.25.16-8 has an include statement with an extra hierarchy marker (two instead of one) | hierarchy_utils.py |
| Table A.32.10-1 is missing values in the "Information Entity" field of two rows | process_ciod_module_relationship.py |
| Table A.35.21-1 contains an upper case "O" in "Frame of Reference", while other tables use the lower case | process_ciod_module_relationship.py |
| Table TID 1004 has a section URL pattern ("sect_TID_1004") that doesn't exist within the HTML version of the standard | parse_lib.py |
Certain subsections are located within the base section rather than having their own section (C.7.16.2.5.1 should be within C.7.16.2.5, but sect_C.7.16.2.5.html is invalid) |
parse_lib.pyextract_modules_macros_with_attributes.py |
| The File-Set Identification Module has no description paragraph | extract_modules_macros_with_attributes.py |
| *The Enhanced MR Color Image IOD references the Enhanced MR Image IOD's functional group macros table instead of having its own (they would be identical tables) | extract_ciod_func_group_macro_tables.py |
| *The "Content Creator's Name" attribute appears twice in Table C.36.8-1 with the same hierarchy without a conditional statement | postprocess_merge_duplicate_nodes.py |
| *Table F.3-3 contains a "Record Selection Keys" attribute with an invalid tag ("See F.5") | preprocess_modules_with_attributes.py |
(*) This issue is not caused by a typo or error in the Standard but rather an exception from the normal format and thus does not have a unit test for a fix.
You can contact us directly through our website.
If you find a bug or have suggestions for improvement, please open a GitHub issue or make a pull request.