Classify cells

Classify cells with a user defined criteria

Select Classify cells in the main pipeline step (1)

1. You can also launch it out of the pipeline in Plugins>FishFeats>Classify segmented cells

Cells must have been created/saved before to use this step

The file named imagename_cells2D.tif should have been saved, or the get cells step should have been performed in the main pipeline before.

When you start this step, you get an interface with a table containing all the cells and all the features (classifications) already done. You can click on one cell to see it in the image.

In the second onglet of the interface, you can add a new feature/classification to do, either by writing a new feature name to create a new one, or by selecting an already present one in the list to edit it. Choose a feature name that will be reelvant for you (eg. "PCNA", "DoubleNucleus", "SuperCell"...). The program will add "Feat_" in front of the name of the feature to indicate that it is a feature.

Click on Do feature to open the interface to choose how you want to initialize your classification.

Classification prefilling

When you click on Do feature, it will either reload the feature if it was already present or open a parameter interface to choose how to initialize the feature.

In the first line, the interface shows the feature name that you have entered.

You can choose the number of possible classes for this feature (Nb classes parameter)(1). For example, if the feature can only be positive (=2) or negative (=1), the number of classes will be 2.

1. Note that you can also change it later by clicking Add one class in the edit interface.

Then you can select the method to use to prefill the classification automatically (you will be able to manually edit it afterwards).

You can either have a prefilled classification with all the cells in the same class (Initialize all cells at 1), based on a thresholding of one chanel (from projection+threshold), based on the position of cells on edges or not (Boundary cells.

Click on Create new feature to launch the classification of your new feature. It will add a new layer, called Feat_featurenameCells, prefilled according to the selected method. In that layer, one color corresponds to one class so if nb classes is 7, you can have 7 colors.

You now have the possibility to manually edit the classification, with the Feat_ featurename parameter interface and to see the table of cells with their corresponding classification with the Features table interface. Click on Update/save table button to update the displayed table in the Feature table onglet and save it in the results.csv file. If you do another feature later (running again the Edit/Add feature interface), it will be added to this table so you can accumulate the analysis.

Empty prefilling

The classification can be prefilled automatically if your classification is binary ("yes" or "no") and depends on the intensity in one chanel of the image.

From intensity projection

Choose the method from projection+threshold and the corresponding chanel proj chanel. It will make a 2D projection of this chanel, and classify as "yes" (2) cells that have at least threshold_areaprop% of its pixel brighter than: mean(intensity) * threshold_frommean.

You can then see the table with the automatically calculated feature and manually edit the automated classfication by using specific shortcuts.

Boundary classification

With this option, you can choose to automatically classify the cells according to if they touch the border of the image (and thus might not be complete), or are on the edge of the tissue (no neihbors in one side), or next to a big hole in the tissue. By selecting Boundary cells, you can choose if you want to classify the cells that touch the border of the image Image border and/or on the edges (of tissue or holes) Tissue boundary. The cells will be classified as 1 if they are not a border or a boundarie, 2 if they are a boundary cell and 3 if they are a border cell.

Load classification

If you select a feature name from the proposed list on the feature name parameter and click on Do feature, it will automatically load the previous classification (that is saved and loaded from the file imagename_results.csv. You can directly edit it.

Manual editing

To modify the classification of some cells, you can set the current value to assign with the Class value parameter. This parameter can take value from 1 to the nb classes parameter that you chose previously in the Do feature interface (the number of classes). You can increase the maximum number of classes to add a new one by clicking the button add one class in the right-side interface.

You can change the current value of this Class value parameter by sliding the bar in the interface, or by pressing i to increase its value or d to decrease it.

Also, if you right-click on a cell, you can set it to the cell's class. It will automatically set-up the Class value parameter to the classification of the selected cell.

To change the class of a cell and set it to the current value of Class value parameter, press Control+left click on the cell. It's color will be udpate directly after the click.

When you click on Update/save table, the current classification will be saved (in the filename_results.csv file along with the other results) and the displayed table in the Feature table onglet will be updated. The current step is not stopped, so don't hesitate to save regularly.

You can then click on Feature featurename done to close the interface of this feature and do another feature or finish this step. It will remove the feature layer, but the results of the classification will still be present in the features table.

If you want to export the view of the classified cells, click on export feature image. It will save it in a file called imagename_feat_featurename.png in the results folder.

Features table

All the results of the classified features are summarized in the Feature table. Each row is one segmented cell and each column the features that have been defined. The table will be saved in the results folder in the results file called imagename_results.csv . If the plugin is closed and open again on the same image, the features will be automatically reloaded. You can edit them by loading the corresponding feature.

If you click on Stop and Save, this feature table will be closed, and all opened feature layers and interfaces will be closed.