ZMQ connection between CellProfiler and Knime
The Knime bridge makes a TCP connection to a CellProfiler worker and communicates via a protocol that lets the client (either Knime or an ImageJ plugin) run a CellProfiler pipeline.
The Knime-bridge server protocol is supported on Windows and Linux builds dated 1/28/2015 and later. The protocol will be ready on OS/X shortly.
On Windows and Linux, start the server by running the analysis_worker executable. On Linux, this should be on your path after installing CellProfiler, on Windows, it is in the CellProfiler installation directory - typically, "c:/Program Files/CellProfiler". To start the Knime bridge on the worker, use the --knime-bridge-address command-line switch:
analysis_worker --knime-bridge-address=tcp://127.0.0.1:8086
for instance, to start the Knime bridge so that it is only accessible from the local machine on port 8086.
You should use the same address for the Knime bridge client.
You can access the Knime bridge through Java using the
class, org.cellprofiler.knimebridge.KnimeBridgeFactory.
KnimeBridgeFactory().newKnimeBridge() will give you an
unconnected instance of the bridge. You first connect to the bridge,
supplying the same address as you did to the worker server, then
you can call loadPipeline to load a CellProfiler pipeline into
the bridge. At this point, you can ask the bridge for the image
inputs and measurement outputs and then supply one imgPlus() image
for each of the channels. The bridge will communicate with the
server to run your pipeline on the supplied images, returning
the measurements.
There is an example that runs the Knime bridge in the src/test/java
directory: org.cellprofiler.knimebridge.Main.java
If you start ImageJ 2.0 with the Knime bridge jars and their dependencies on the class path, you should see "CellProfiler" on the plugins menu and "Run pipeline" in the submenu. If you run the "Run pipeline" plugin, it will ask you for the URL of your server (tcp://127.0.0.1:8086 in the example above) and your pipeline file (hit the Browse button to find it on disk). Once you finish this dialog, the bridge will contact the server to get the details of your pipeline. You'll be asked to choose your images for your pipeline and after running, you'll get tables for each of the types of objects you have in the pipeline, plus an image measurements table.
At present, there are some limitations to the pipelines you can run. Grouping is not yet supported. There is no mechanism to set properties, so CellProfiler will use the defaults for any that aren't entered via command-line switches to analysis_worker. SaveImages can only output to fixed file names and any part of the pipeline that extracts metadata from file names will not work. In general, you should not use modules that perform I/O such as SaveImages, ExportToSpreadsheet and ExportToDatabase.
The input modules and the LoadImages module can be used - the bridge finds the names of the channels from these modules. It's possible for you to use LoadData, but only if you read the image list file from a fixed location - it's best to modify your pipeline to use some other image input method.