your original membrane images. Open the image that ends on
“_scaled” in Fiji and duplicate the first channel using the
“Duplicate” function. This channel contains a mask of the
segmented cells (“Cell Masks”, Fig.1b 4 ). In addition, open
the original membrane image that you used to run the segmen-
tation (seestep 2).- Start the Fiji plugin “Cell annotation” that is part of the MS-
ECS-2D update site (Plugins>Cell transcript>Cell annota-
tion) (seeNote 40) (Fig.5). Select the Original image and its
matching “Cell Masks” image. You can ignore the Nuclei
channel and Membrane channel fields. - In the Cell annotation tool, set the mode to “Correction”.
Correct any under-segmentations by drawing missing lines
pressing the left mouse button. Correct any over-segmentations
Fig. 4KNIME cascaded random forest pipeline. The KNIME cascaded random forest pipeline consists of two
steps. Instep 1the pipeline is trained to recognize membrane, membrane intersection points (vertex points)
and background on a small set of representative membrane images (seeNote 36). Details on training can be
found at “tinyurl.com/KNIME-MS-ECS”. Once the pipeline has been trained it can be used repeatedly to
predict membrane and vertex points instep 2.Instep 2.1, membrane images are loaded. Instep 2.2the
cascaded random forest pipeline is run. Before running this step, the downsampling factor needs to be
adjusted to the pixel size of the images (seeNote 38). Instep 2.3the results can be viewed and written to the
computer for further processing
smFISH and Automated Analysis in Zebrafish 153