THIS READ ME IS NOT UPDATED FOR VERSION 2.0 YET

InferenceServer Dicom Relay (ISDR)

This InferenceServer Dicom Relay (ISDR) is setup to serve InferenceServer through regular dicom networking protocols. The supported workflow is as follows:

Flow through ISDN

The Store SCP (`src/scp/scp.py`)

A SCP receives scans with regular dicom networking C_STORE events.
Incoming dicom files are sorted out in a folder structure of /INCOMING/PatientID/StudyDescription/SeriesDescription/Modality/SOPClassUID

The InferenceServer post daemon (`src/daemons/inference_server_daemon.py`)

The post daemon runs with a fixed interval (default is 1 sec) and checks timestamp of all incoming folders (/INCOMING/PatientID/StudyDescription/SeriesDescription/Modality/SOPClassUID).
If a folder is untouched for a specified interval of time (default is 10 sec.), the folder checked via the DB for matching fingerprints. A fingerprint is a specification file which contains information on which modalities and keywords in series descriptions that should trigger which model on which InferenceServer endpoint - and subsequently which dicom node the predictions should be send to (usually PACS or TPS, by can be any Store SCP)

The InferenceServer `get_thread` (`src/daemons/get_thread.py`)

This is a thread that is initialized when a job is posted with the post daemon.

It polls InferenceServer with a fixed interval of time to check whether the job is finished. If a task fails on the server, the get_thread will terminate.
When the InferenceServer has the predictions ready, they are send to the get_thread, which in turn ships the result on to the Store SCP specified in the fingerprint file.

How-to's

Set up configurations

ISDR is meant to run in a docker container. You must bind the following folder structure into the container:

CONFIGURATION
  |-CERTS/cert.crt
  |-FINGERPRINTS
  |  |-fingerprint1.json
  |  |-fingerprint2.json

CONFIGURATION/CERTS/cert.crt is the public key to the TLS of InferenceServers. Multiple certs can be merged into this file. See CONFIGURATION/FINGERPRINTS/fingerprint.json.example for an example on how to configure a fingerprint.

You can set the following environment variables. Defaults shown here - just overwrite the variable when running the docker container

- SCP_HOSTNAME=localhost
- SCP_PORT=104
- SCP_AE_TITLE=DICOM_NODE
- INCOMING_DIR=/INCOMING
- FINGERPRINT_DIR=/CONFIGURATION/FINGERPRINTS
- CERT_FILE=/CONFIGURATION/CERTS/cert.crt
- PYNETDICOM_LOG_LEVEL=none ## other option is ["standard"](https://pydicom.github.io/pynetdicom/dev/reference/generated/pynetdicom._config.LOG_HANDLER_LEVEL.html#pynetdicom._config.LOG_HANDLER_LEVEL)
- POST_INTERVAL=1
- POST_AFTER=10
- POST_TIMEOUT=60
- DELETE_ON_POST=true
- GET_INTERVAL=15
- GET_TIMEOUT=86400
- ASK_INFERENCE_SERVER_TO_DELETE_TASK=true # set to "" if you don't want to delete

Run in docker

Assuming that the project is build (see /build.sh as example on howto), you can run ISDR with

docker run \
  -d \
  -it \
  --restart=always \
  --network=host \
  --volume /absolute/path/to/CONFIGURATION:/CONFIGURATION \
  --name dicom_node \
  mathiser/inference_server_dicom_node:v0.2.1

... or you can adapt and run run.sh.example

Attach to the logs

You can attach to the docker logs with:
docker logs -f dicom_node

Restart ISDR

Can you have updated something in CONFIGURATION, you need to restart ISDR:
docker restart dicom_node

Hard restart ISDR

If everything is wrong or you want to upgrade to a newer version:
docker stop dicom_node && docker rm dicom_node

Update to another build of ISDR

docker stop dicom_node && docker rm dicom_node
docker pull mathiser/inference_server_dicom_node:some_other_tag
docker run \
  -d \
  -it \
  --restart=always \
  --network=host \
  --volume /absolute/path/to/CONFIGURATION:/CONFIGURATION \
  --name dicom_node \
  mathiser/inference_server_dicom_node:some_other_tag

mathiser/inference_server_dicom_node