Orthanc/TODO
2025-06-23 19:07:37 +05:30

417 lines
19 KiB
Plaintext

=======================
=== Orthanc Roadmap ===
=======================
For higher-level ideas in the roadmap, please first read the
"Contributing to Orthanc" section of the Orthanc Book:
https://orthanc.uclouvain.be/book/contributing.html
Some features are being funded by an OpenCollective one-time donations.
selected features are marked with priorities ((1) - higher, (2) - medium, (3) - nice to have)
=======
General
=======
* Configure an user-defined site UID root if generating DICOM UIDs
("FromDcmtkBridge::GenerateUuid()")
* Improve handling of errors in the command queue:
https://groups.google.com/d/msg/orthanc-users/--njEbqcDDI/rBu8XL-Mm-cJ
* Support partial file retrieval in Orthanc::HttpClient
* Support retry counter in Orthanc::HttpClient
* Option to enable DNS lookups in DICOM:
https://orthanc.uclouvain.be/hg/orthanc/file/Orthanc-1.9.3/OrthancFramework/Sources/OrthancFramework.cpp#l88
* Toolbox::ComputeMD5() fails on files larger than 4GB
* Add an option to run Orthanc in read-only mode both for DICOM and for Rest API.
* Logging: add more specific information to contextualize the logs.
For a DICOM Transfer, that would be nice to include the modality in the context + a study identifier or a job id.
* (1) Accept extra DICOM tags dictionaries in the DCMTK format '.dic' (easier to use than declare
them in the Orthanc configuration file). Even the standard dictionaries could be
overridden by these custom dictionaries.
* Add configurations to enable/disable warnings:
- Modifying an instance while keeping its original SOPInstanceUID: This should be avoided!
- Modifying a study while keeping its original StudyInstanceUID: This should be avoided!
In order to be able to disable/enable warnings in both the server and the framework, we should add a map of
enabled warnings in the logging classes directly and have something like:
LOG_WARNING_IF_ENABLED("Warning_ID") << ...
ENABLE_WARNING("Warning_ID", true)
Warnings from Framework should have a separate range like W999_ or WF001_ ...
* Store the job registry in a dedicatd table in DB ?
https://discourse.orthanc-server.org/t/performance-issue-when-adding-a-lot-of-jobs-in-the-queue/3915/2
Note: that might also be the right time to have a central jobs registry when working
with multiple Orthanc instances on the same DB.
Note: the json serialization of a job "content" can be very large -> compress it before saving it to DB ?
* Right now, some Stable events never occurs (e.g. when Orthanc is restarted before the event is triggered).
Since these events are used to e.g. generate dicom-web cache (or update it !), we should try
to make sure these events always happen.
- Generate the events when setting IsStable=true when starting an Orthanc (ok for SQLite) ?
- Also consider the use case of an Orthanc cluster that is being scaled-down just after one Orthanc instance
has received a few instances -> we can not only check for missing stable events at startup since no Orthanc will start.
We would need to maintain the list of "unstable" resources in DB instead of memory only.
- Also check the PG plugin and its new table InvalidChildCounts, with a timestamp there, we can detect for
how long a study has not been modified !
* In prometheus metrics, implement Histograms or Exponential Histograms to measure durations. Right now, we only provide
"average" durations that are not very relevant
(https://opentelemetry.io/docs/specs/otel/metrics/data-model/#histogram)
- for job durations (+ have one histogram for each job)
- for HTTP request handling
- ...
* Investigate if one could fix KeepAlive race conditions:
https://discourse.orthanc-server.org/t/socket-hangup-with-rest-api/4023/3
* The DICOM file cache shall keep a MD5 of the cached file and compare it with MD5
from the DB. That would allow 2 orthancs in a swarm to realize when the other
Orthanc has updated the file:
https://discourse.orthanc-server.org/t/instances-id-content-api-results-are-different-in-docker-swarm-replicas-of-orthanc/4582
* Allow saving PrivateTags in ExtraMainDicomTags.
Note: they can actually be stored but they then appear as "Unknown Tag & Data" in the responses.
If we try to add the PrivateCreator in the ExtraMainDicomTags, then, the DICOMWeb plugin fails to initialize because the private tags are not known.
* Support hashed passwords in RegisteredUsers. E.g:
"RegisteredUsers": {
"admin": {
"Password": "5baa61e4c9b93f3f0682250b6cf8331b7ee68fd8",
"Hashing": "sha1"} }
============================
Documentation (Orthanc Book)
============================
* Write a getting started guide (step by step) for each platform to replace
https://orthanc.uclouvain.be/book/users/cookbook.html :
- Ubuntu/Debian
- Windows (done)
- OSX
- Docker on Linux
Each step by step guide should contain:
- get binaries
- launch
- open explorer
- edit configuration file
- restart and observe changes
* Explain how log rotation can be achieved (no built-in support in Orthanc)
* Explain how to interface with Mirth/NextGen Connect
https://en.wikipedia.org/wiki/NextGen_Connect
* How to reproduce issues by replacing Orthanc with storescp or wlmscpfs
* How to capture TCP traffic of DICOM protocol using tcpdump and Wireshark
* Add more configurations of viewers (Weasis, Slicer...):
https://orthanc.uclouvain.be/book/integrations.html
* Discuss HL7 in a dedicated page:
https://groups.google.com/d/msg/orthanc-users/4dt4992O0lQ/opTjTFU2BgAJ
https://groups.google.com/g/orthanc-users/c/Spjtcj9vSPo/m/ktUArWxUDQAJ
================
Orthanc Explorer
================
* Option to tune the number of results for a local lookup:
https://groups.google.com/g/orthanc-users/c/LF39musq02Y/
========
REST API
========
----------
Short-term
----------
--------
Mid-term
--------
* (1) Archive jobs: Resume downloads using "range requests":
https://developer.mozilla.org/en-US/docs/Web/HTTP/Range_requests
* (3) Create DICOM from DICOMweb JSON ("application/dicom+json")
with "/tools/create-dicom"
* (2) Create multi-frame images with /tools/create-dicom (by adding a
"MultiFrame" flag to avoid creating a series), or modify PixelData
of a multi-frame image:
https://groups.google.com/g/orthanc-users/c/y3-xa_GcdLM/m/m0Kr5G5UPAAJ
* (1) Specify the transfer syntax in /tools/create-dicom
https://groups.google.com/g/orthanc-users/c/o15Dekecgds/m/xmPE2y3bAwAJ
* Support Palette PNG in /tools/create-dicom:
https://discourse.orthanc-server.org/t/404-on-tools-create-dicom-endpoint-with-specific-png/3562
* Support creation of DICOM files from MP4 in /tools/create-dicom.
Sample python code: https://github.com/salimkanoun/OrthancGif/blob/new-organisation/python/create_dicom_video.py.
We would need to extract frame rate + dimension from the MP4 which would
require ffmpeg or a similar library -> can not be done in the Orthanc core.
-> keep it for a python plugin
-> or require the payload to include rows/columns/cinerate/...
* (1) In the /studies/{id}/anonymize route, add an option to remove
secondary captures. They usually contains Patient info in the
image. The SOPClassUID might be used to identify such secondary
captures.
* Support "/preview" and "/matlab" for LUT color images
* /preview should be able to display a dose report:
https://discourse.orthanc-server.org/t/orthanc-image-preview-shows-empty-image/4459
* Try to transcode files if a simple decoding fails:
https://groups.google.com/g/orthanc-users/c/b8168-NkAhA/m/Df3j-CO9CgAJ
* (2) Ranges of DICOM tags for "Keep" and "Remove" in ".../modify" and ".../anonymize":
https://groups.google.com/g/orthanc-users/c/6dETktKo9v8/m/b0LUvSfwAgAJ
* return error code/reason in HTTP response if resubmit/cancel/pause fails ...
* filter /changes by type e.g: /changes?filter=StablePatient
https://groups.google.com/g/orthanc-users/c/r20kDb0axms/m/2tzbQzYJAgAJ
* save more details in jobs e.g: the resources being sent/exported ...
https://groups.google.com/g/orthanc-users/c/rDDusFG5Lco/m/TzTUjWXLAQAJ
https://discourse.orthanc-server.org/t/some-confusion-about-jobs-function/3887
* allow filtering/ordering on the /jobs route:
https://groups.google.com/g/orthanc-users/c/hsZ1jng5rIg/m/8xZL2C1VBgAJ
* add an "AutoDeleteIfSuccessful": false option when creating jobs
https://discourse.orthanc-server.org/t/job-history-combined-with-auto-forwarding/3729/10
* Allow skipping automatic conversion of color-space in transcoding/decoding.
The patch that was initialy provided was breaking the IngestTranscoding.
This might require a DCMTK decoding plugin ?
https://discourse.orthanc-server.org/t/orthanc-convert-ybr-to-rgb-but-does-not-change-metadata/3533/9
* Implement a 'commit' route to force the Stable status earlier.
https://discourse.orthanc-server.org/t/expediting-stability-of-a-dicom-study-new-api-endpoint/1684
---------
Long-term
---------
* Stick to the JSONapi or JAREST guidelines for a "v2" of the API:
https://groups.google.com/forum/#!msg/orthanc-users/Bag-SwEE9ZI/-w7QXI6p7-oJ
http://www.admiraalit.nl/jarest/
=====
DICOM
=====
--------
Mid-term
--------
* Check how Orthanc shall behave wrt to AcceptedSopClasses in these situations (consider Orthanc
accepts CT but not PT)
- What shall we log/warn if an external modality tries to send a PT/CT
- What shall we log/warn if we try to C-GET a PT/CT
Should the rejected files be logged as Failed, Warning, Refused, ...
Note: some tentative work has been initiated in the get-scu-test branch.
* Support "Retrieve AE Title" (0008,0054) in C-FIND:
- On SCP side: done by https://orthanc.uclouvain.be/hg/orthanc/rev/1ec3e1e18f50
- On SCU side:
https://groups.google.com/d/msg/orthanc-users/wPl0g5mqZco/5X1Z8tEzBgAJ
* Check Big Endian transfer syntax in ParsedDicomFile::EmbedImage and
DicomImageDecoder
* Strict hierarchical C-FIND:
https://groups.google.com/d/msg/orthanc-users/VBHpeGVSNKM/tkaVvjWFBwAJ
* report DIMSE error codes in Rest API and job status for /store /query /move /retrieve
* Log outgoing C-Find queries
* Support other Transfer Syntaxes in the Worklist plugin:
https://discourse.orthanc-server.org/t/could-you-please-create-an-option-to-set-the-transfer-syntax-in-the-worklist-plugin-currently-little-endian-explicit-is-fixed/4871
* Deidentify Trace level logs: https://discourse.orthanc-server.org/t/dicom-log-deidentification-at-trace-log-level/5563
We should implement something like GetDeidentifiedContent(const DcmDataSet& dataset) that would
reproduce DcmDataSet::print but hide individual elements that contain PHI.
---------
Long-term
---------
* Support extended association (e.g. for C-Find relational queries):
https://groups.google.com/d/msg/orthanc-users/xD4d3mpc6ms/srF7E2goAAAJ
* Support C-MOVE-CANCEL:
https://groups.google.com/d/msg/orthanc-users/KnduYBFd06A/o86cl5SeCAAJ
* Combine StudyDate + StudyTime in C-Find matching (also for the worklist plugin with
ScheduledProcedureStartTime & Date).
We should first filter in SQL by StudyDate only, combine it with StudyTime into a single
DateTime string and filter again in C++.
https://discourse.orthanc-server.org/t/performin-find-within-orthanc-for-time-frames/4704
* Worklist plugin: support MPPS
https://github.com/orthanc-server/orthanc-setup-samples/blob/master/python-samples/worklist-with-mpps.py
--------------------
Internationalization
--------------------
* Support multiple specific character sets (cf. "SCSH32" in orthanc-tests)
- http://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_C.12.html#sect_C.12.1.1.2
- Japanese test: http://dicom.nema.org/MEDICAL/dicom/2017c/output/chtml/part05/sect_H.3.2.html
https://discourse.orthanc-server.org/t/garbled-characters-when-i-uploaded-japanese-patient-name/3204/5
* Support Supplementary Kanji set (ISO 2022 IR 159)
* Create DICOM files with multibyte encodings (Korean, JapaneseKanji, SimplifiedChinese)
===========
Performance
===========
* (2) DicomMap: create a cache to the main DICOM tags index
* (3) Check out rapidjson: https://github.com/miloyip/nativejson-benchmark
* Long-shot & not sure it is even feasible at all: try to reduce memory usage by implementing streaming
when receiving DICOM instances from the Rest API or from DICOM and store files directly to disk as they
are received. Note that this would likely require rewriting large parts of Orthanc. Note sure that
would be compatible with Transcoding.
Use case: receiving 10 1GB instances in parallel can consume up to 20 GB
Alternative option 1: write an "external application/plugin" that would take care of these receptions, write the
file at the right place and send a signal to Orthanc to "adopt" the file.
Alternative option 2: declare a memory resource (X GB) that is available for reception. Every time
Orthanc starts receiving a file, it reserves the memory or twice the memory (through a Semaphore)
if no memory is available, it waits and possibly timeouts returning a 503 or DIMSE A700 (out of resources).
This would at least protect from "out of memory" crashes.
Alternative option 3: Configure DCMTK to "stream" DICOM files on a temporary file on disk. Pass the file handle
to Orthanc and/or the Storage plugin (instead of passing a memory buffer) -> the object-storage plugin could
"stream" the file to the storage. The HTTP server could also "stream" its response from file handles.
Transcoding should be "file based" too.
Alternative option 4: Catch out-of-memory exceptions at quite high level in HTTPHandlers and DICOM receivers
and implement retries. After a few retries, fail for good and return "out-of-resources".
Would be interesting to log these errors and count them in the prometheus metrics.
Note: I'm (maybe naively) thinking that you only need the beginning of the file to get the DICOM tags to be stored
in the database, so maybe the rest of the file could be "streamed" directly to disk and not kept in memory?
* To investigate: usage of mapped_file (not only in the indexer plugin):
https://discourse.orthanc-server.org/t/patch-for-orthanc-indexer-plugin-crashing-on-big-non-dicom-files/3849/7
========
Database
========
* Integration test searching for "\" and "%" in PatientName, PatientID...
=======
Plugins
=======
---
SDK
---
* Implement PluginsJob::GetOutput() to allow user-defined jobs to provide output:
https://groups.google.com/g/orthanc-users/c/GZ7u0bTTVoo/m/Z7iTNTi2BgAJ
* Add plugins for normalized operations (notably so as to support
Print SCU/SCP, cf. "dcmprscp"):
https://web.archive.org/web/20170923150432/https://www.medicalconnections.co.uk/kb/DICOM_Print_Service
* Provide access to the Orthanc::DicomUserConnection class in plugins:
https://groups.google.com/d/msg/orthanc-users/ycDA1xPuTRY/nsT2_GOtEgAJ
* Provide a C++ callback similar to "ReceivedInstanceFilter()" in Lua
https://orthanc.uclouvain.be/book/users/lua.html#filtering-incoming-dicom-instances
https://groups.google.com/d/msg/orthanc-users/BtvLTE5Ni8A/vIMhmMgfBAAJ
* Update the SDK to handle buffer sizes > 4GB (all sizes are currently coded in uint32_t)
* Add a C-Get SCP handler: OrthancPluginRegisterGetCallback
https://groups.google.com/g/orthanc-users/c/NRhPkYX9IXQ/m/mWS11g0jBwAJ
* Add a primitive for user authentication (to generate 401 HTTP status, whereas
the "RegisterIncomingHttpRequestFilter()" can only generate 403 HTTP status)
https://groups.google.com/g/orthanc-users/c/ymtaAmgSs6Q/m/PqVBactQAQAJ
* Add an index on the UUID column in the DelayedDeletion plugin:
https://discourse.orthanc-server.org/t/delayeddeletion-improvement-unique-index-on-pending-uuid-column/4032
* Orthanc shall refuse to start if one registers 2 storage plugins.
Right now, this is not possible because OrthancPluginRegisterStorageArea2 does not return any value
and it can not throw an Exception because that's a core function called from a plugin -> the Exception
can not cross the C/C++ frontier safely -> we need a OrthancPluginRegisterStorageArea3 with a return value.
Ex: install DelayedDeletion + S3 storage. Right now, the second plugin to load is just ignored with an error
message in the logs.
-----------
Housekeeper
-----------
* The Housekeeper should just refuse to start if you are using a lossy transfer syntax because that would generate
new orthanc ids as well and there is a risk of messing up things. tools/reconstruct shall return a 400 as well.
https://discourse.orthanc-server.org/t/crashes-on-housekeeper-transcode-storing-in-s3/5330/2
----------------
Ideas of plugins
----------------
* DICOM-RT primitives (RT-STRUCT, RT-PLAN, RT-DOSE)
* Converter to/from NIfTI
* Decode JPEG2k with grok: https://github.com/GrokImageCompression/grok
* Generate dynamic HTTP content using Lua:
https://groups.google.com/d/msg/orthanc-users/KompazkxRSs/5Rh03mzgDAAJ
* More generally, expose more callbacks of the plugin SDK in Lua:
https://groups.google.com/d/msg/orthanc-users/_FbiRHuXPGM/J-OAv7zaCAAJ
* Configuration interface plugin
===
Lua
===
* Configure HTTP headers from Lua (in RestApiPost(), RestApiPut()
and RestApiDelete().
https://groups.google.com/forum/#!msg/orthanc-users/WNnW187OILM/6XX_bm96BwAJ
* Retrieve HTTP status from calls to HttpGet, HttpPost, ...
https://discourse.orthanc-server.org/t/how-to-get-http-status-code-from-httppost/1263/4
================
Code refactoring
================
* Avoid direct calls to FromDcmtkBridge (make most of its
methods private), go through ParsedDicomFile wherever possible
=================
Platform-specific
=================
---------
Packaging
---------
* RHEL through EPEL (as of 2021, CentOS seems to be over):
http://fedoraproject.org/wiki/EPEL_Package_Maintainers
* Fedora: Are DICOM dictionaries correctly located?
https://groups.google.com/g/orthanc-users/c/YB_LTBuUQNA/m/3H1xsrZJDgAJ
------------------------
Big-endian architectures
------------------------
* Check the generated 16bpp PNG images
-----------------
Microsoft Windows
-----------------
* Add compatibility with non-ASCII paths (Orthanc expresses its paths
as UTF-8 strings, but Windows expects them to be translated to the
system locale)
* Fix error message when stopping the service:
https://groups.google.com/g/orthanc-users/c/NyrwUJ9N6Ec/m/sTZIcWvaAgAJ
====
Misc
====
-----------------
Maintenance tools
-----------------
* Standalone tool to detect corrupted files:
https://groups.google.com/g/orthanc-users/c/KASEJ9Gx3vQ/m/MIJFJdRhAwAJ
* Standalone tool to detect orphan files in OrthancStorage
----------------------------------
Code quality and integration tests
----------------------------------
* Have a look at openQA (cf. GNU Health)
* Add integration tests for LUT
* Add more complex testing scenarios like data-migration, change of
configuration files, multiple orthanc interacting togethers with various
config. This should probably look like the python toolbox tests ...
- add a test to validate Modalities and Peers stored in DB are not lost
while upgrading from one version to the other (Sylvain)
* On Ubuntu 20.04, accesses to unitialized memory are sometimes
reported in libgjpeg by valgrind, if running the following command
(this is probably unrelated to Orthanc):
$ ./Start.sh --force Orthanc.test_bitbucket_issue_141 Orthanc.test_create_pdf Orthanc.test_decode_brainix_as_jpeg
---------------------
External applications
---------------------
* Create REST bindings with Slicer
* Create REST bindings with Horos/OsiriX