How to use the YeastMate Fiji plugin
You can download the Fiji plugin from https://github.com/hoerlteam/YeastMateFiji/releases. Just put the single .jar file into the plugins folder of you Fiji installation and YeastMate should appear under Plugins->YeastMate the next time you start Fiji.
The Fiji plugin requires a running instance of the detection backend. You can start the packaged detection backend from within the main GUI (see GUI: Getting Started) or directly start it with the yeastmate_server executable in the YeastMate installation folder in resources->python->YeastMate.
Alternatively, you can start the detection backend from the command line with python yeastmate_server.py after setting up the required Python environment as described in (Python - Get Started).
Usage
The YeastMate Fiji plugin will perform detection in the currently selected image in Fiji. If the current image is a hyperstack, we will use the currently selected channel and z-slice and either the currently selected time point or (optionally) all time points. After clicking Plugins->YeastMate in the Fiji menu, the following main dialog will appear:
- Detection score threshold: the first options allow setting a score threshold for single cells, matings and buddings. Only objects for which our network predicts a score above the threshold will be considered. In practice, we found the values of 0.9 (single cells) and 0.75 (matings and buddings) to work well. You can, however, lower the thresholds to detect more objects, at the risk of more false positives, or raise the threshold to only consider objects for which the network is very sure, at the risk of false negatives.
- Normalization quantiles: Before passing images to the network, YeastMate will normalize the intensities of your images. To mitigate effects of single bright or dark pixels, we normalize not the the minimum and maximum intensity, but a low and high quantile of the intensities in your image. The default values of 0.015 and 0.985 should work fine for most images.
- Output to display: The checkboxes on the bottom half of the dialog allow you to specify what output to produce. You can select to add YeastMate's detections to the Fiji ROI Manger (single cells and mothers/daughters from transition events will be added as outlines and the whole matings/buddings as bounding boxes). You can also display the single cell segmentation mask as a new image (and optionally limit the cells shown in the mask to those participating in transitions which you selected to be added to the ROI Manger, e.g. if you only add buddings to the ROI Manager, only masks for mother and daughter cells involved in budding events will be shown).
- Timeseries processing: In addition to processing the currently selected image, the Fiji plugin can also process all frames in a time series if you select Process every frame in time series. The plugin will track cells and mating/budding events from one frame to the next and assign the same labels if they overlap by more the the value set in Minimum overlap for tracking in time series from frame to frame. If you want to process multiple time points, make sure the input image is a hyperstack with a time axis in
Image->Hyperstack. - IP of detection server: The Fiji plugin requires the detection backend to be running in the background and will communicate with it using HTTP requests. On the bottom of the dialog, you can specify the IP adress and port of your detection backend (if you started the local backend from our all-in-one application, it should be
localhost:11005). Next to the textbox, a status message will indicate if the YeastMate detection server is rechable at the specified address.
After setting the parameters, click OK to start detection. After a few seconds, the results you selected should pop up.
Macro usage
If you activate Fiji's Macro Recorder from the main menu (Plugins->Macros->Record...), the command for YeastMate will be recorded and you can use it in macros, e.g. to process many images sequentially.
Example Output
Example: Segmentation mask:
Example: ROIs for mating events:
