This workflow was built with KNIME Analytics Platform 4.0.2, however, similar workflows have been used with KNIME Analytics Platform 3.5 to 4.1 as well.
The sample images and pixel classification model are packaged with this workflow already, but can also be downloaded from https://doi.org/10.6084/m9.figshare.9936287.v2.
Instructions to follow before opening the workflow:
1) Follow instructions at https://knime.com/community/imagej to install the KNIME Image Processing - ImageJ Extension
2) Follow instructions at https://knime.com/wiki/knime-image-processing-nightly-build to add and activate the nightly software site of KNIME Image Processing
3) Go to File > Preferences > KNIME > Image Processing Plugin and add ",F" to the available dimension labels and restart KNIME Analytics Platform
4) Open the workflow and follow instructions to search and install missing extensions
Dear Stefan Helfrich, I failed to practice this workflow. Several reasons here: 1. missing extensions, like Fiji Trainable segmentation Feature 2D and etc. 2. I don’t understand the principle of this workflow. Why the channel of the images are split and then the subtraction is applied. How to interactive annotate the stain? What criteria is based to make the judgement? Thanks!
Could you please follow the instructions below the workflow screenshot at DAB stain_AZ_TN – KNIME Community Hubbefore opening the workflow. This way, KNIME can find the missing extension automatically for you.
You can annotate images by opening the Train Pixel Classification Model metanode (double click) and following the instructions of the workflow annotation.
The decision for each pixel are based on pre-computed features. This is also done in the aforementioned metanode: Fiji Trainable Segmentation Features 2D computes a variety of features for the input image and trains a pixel classifier based on the features and provided annotations.