Plotting

The plotter widget is the heart of the napari-clusters-plotter. To open it, open a napari viewer and find the Plotter Widget either under

• Layers > Visualize > Plot & select features (napari-clusters-plotter)

• Plugins > napari-clusters-plotter > Plot & select features (napari-clusters-plotter)

Widget overview

The core functionality of the plugin is available to you directly upon opening it:

1. Standard matplotlib toolbar: Reset, Undo/Redo, Pan, Zoom, other plot properties & export

2. Tools to select points on the canvas:

   • Lasso selector

   • Ellipse selector (drag and drop, confirm with rightclick)

   • Rectangle selector (drag & drop, confirm with right-click)

3. Selected cluster index: Depending on which color/number is selected here, regions in the plot will be counted as members of this cluster group and highlighted accordingly

4. Show/Hide: Show or hide the selected points or histogram bins on the canvas.

5. Canvas: Features will be visualized here

6. What feature to plot on the x-axis and y-axis, respectively. The Hue dropdown controls the coloring of the data on the canvas, categorical features are highlighted in orange.

7. Reset button: Resets all drawn clusters

Under the Advanced Options tab, you have access to some more customization options for the visualization:

8. Change the colormap for the chosen overlay color. If a layer is colored by a non-categorical feature, this determines the colormap for the depiction. Only enabled if a non-categorical feature is selected in the Hue dropdown.

9. Apply a log-scale to the chosen feature

10. Switch plot type between SCATTER and HISTOGRAM2D

11. Colorap for 2D histogram (only enabled if HISTOGRAM2D is selected in 10.)

12. Change the size of the bins (only enabled if HISTOGRAM2D is selected in 10.)

Visualizing layer features

In order to visualize features from a dataset, you can load some of the sample data which accompanies the napari-clusters-plotter. You find the sample datasets under File > Open Sample > napari-clusters-plotter > .... In this tutorial, we will use the BBBC1 dataset & segmentations dataset, (Jones et al., Proc. ICCV Workshop on Computer Vision for Biomedical Image Applications, 2005), available from the Broad Bioimage Benchmark Collection [Ljosa et al., Nature Methods, 2012] (Link to data source)

The features of a Labels layer can be loaded into the Plotter Widget simply by selecting it in the layer list on the left:

The dropdown menus for x-axis, y-axis and Hue are then automatically populated with all available features. The features will be drawn as soon as you make the first change to the selected features. By default, the MANUAL_CLUSTER_ID is selected as the Hue value. This features stores all the drawn/selected items.

For Label layers, we can observe that the initial random label colors are retained by default (i.e., the selection overlay is invisible by default):

When we start drawing or untoggle the visibility button, we see the respective items showing up in the color we selected as cluster index (see above). If we want to reset the selection, we can press the Reset button. For Labels data, the color will revert back to the default (random colors) as soon as the layer is unselected. To retain the cluster selection, simply keep the layer selected in the napari layer list.

Cross-layer visualization

A key feature of the napari-clusters-plotter is that you can select multiple layers of the same type and explore their features with the Plotter Widget. To do so, simply select multiple layers in the layer list of napari. Again, the aforementioned sample dataset (BBBC1 dataset & segmentations) is suitable to give this a try.

There are two things to keep in mind here:

• The features available for visualization is the intersection of all features that are common to all selected layers. If a feature feature_x is not present in all selected layers, it will not be available for visualization.

• Only features from the same layer type can be compared. If mixed layer types (e.g., Labels and Points) are selected, the drawing will fail.

Visualizing non-categorical features

So far, the Hue selector was always set to MANUAL_CLUSTER_ID, which is by design a categorical column. However, the napari-clusters-plotter supports visualizing any feature as the Hue parameter. If this is done, the points are colored according to the selected feature and each point’s color will be project on the respective object in the napari viewport. In essence, this creates feature maps for each feature you select: