Classes & Objects

The Classes panel on the right is where you define the object populations that your pipelines detect and measure. Every detected object is assigned to one or more classes. The Objects panel below it lists the objects of the current image.

What is a Class?

A class represents a distinct object population in your experiment. Examples:

• dapi@nucleus - nuclei stained with DAPI
• cy5@spot - extracellular vesicles in the Cy5 channel
• cy7@spot - vesicles in the Cy7 channel
• coloc@cy5cy7 - vesicles colocalising across both channels

Naming convention

Use <prefix>@<type> where the prefix (typically the fluorophore) is used for sorting in selection drop-downs.

Classes panel

Each class is a row with its color, its name, its number of objects on the current image and an eye that shows or hides its objects. Click a class to select it.

• + (New class) adds a class and opens its settings.
• ⋯ › Edit name, color, … of a class opens its settings.
• The same menu has Copy, Move up / Move down and Delete class….

Class settings:

Field	Description
Name	Class label, e.g. cy5@spot
Notes	Optional free-text description
Class ID	The class’s number - see below
Color	Display color used for the class’s objects in the viewer

Every class has a number, its Class ID - the number a Threshold entry’s segmentation class must match (see Object Classes). Auto gives a new class the number above the highest one, so the number of a deleted class isn’t reused by accident (old pipelines and results may still refer to it). Pick a number from the list to choose a free one yourself.

Changing the ID of an existing class moves every reference with it: pipelines, modules and annotated objects that use the class switch to the new ID. Results of earlier runs keep the old one. The background keeps its ID 0.

You can also create a class where you need it: the class pickers of the steps and the annotate bar’s Paint as have + New class… - the new class is picked at once.

Add classes from the image

The Classes panel’s own ⋯ menu:

• Add classes from image creates one class per channel of the current image, as a starting point, in the channels’ colors.
• Merge classes from image adds only the classes the list doesn’t have yet.
• Delete all classes… empties the list, after a confirmation.
• Hide unclassified objects - see below.

Hiding unclassified objects

⋯ › Hide unclassified objects (in the Classes panel) hides objects that have no class yet - on by default, so a large unfiltered detection doesn’t drown out the objects you’ve actually classified.

Objects panel

The Objects panel lists every object on the current image: the live-preview objects of the pipelines and the objects you drew by hand.

Click a row to select it - the object is highlighted on the image, and selecting an object on the image selects its row. Click it again to unselect it.

Filtering

Above the list are a class filter (All classes or one class) and a search. The search takes an object id, a class name, or a condition on a measurement - area > 50, circ < 0.6 - and several terms separated by commas.

Objects as a table

⋯ › Show as a table moves the objects to a table at the bottom of the window, between the left and right panels: one line per object with its ID, classes, area, circularity and the mean of each channel. ⋯ › Columns picks the columns. The class filter and the search sit in the table’s head. Show as a list brings back the list.

Whether an image opens with the list or the table is set in EVA › Project settings… › Viewport › Objects as a table.

The selected object

The selected object’s details are shown in a card beside the panel:

• its classes,
• Area in pixels² and in the physical unit (nm²) from the image’s pixel size,
• Circularity,
• the mean intensity of each channel,
• Parent - for an object linked by Relate Objects, the classes of the objects it belongs to (“Cell”),
• Children - how many linked objects belong to it, per class (“Spot · 7”),
• Touches - each class it colocalizes with, and how many of that class’s objects.

An object found by a pipeline is read-only (From the pipeline · read-only). An object drawn by hand can be changed.

Managing objects

For the selected object:

• Set class… - pick a class, then Add class adds it to the object’s classes, keeping the ones it has, or Set class replaces them.
• Copy to annotation - copies a pipeline object into the hand-drawn annotations, where it can be edited (for example to train a classifier on it). The copy is selected, so its class can be changed right away.
• Delete object (or Delete) - removes a hand-drawn object, after a confirmation.

The panel’s ⋯ menu, for all objects of the image:

Item	What it does
Show as a table / Show as a list	See Objects as a table.
Copy all to annotations	Copies the image’s pipeline objects of the ticked classes into the hand-drawn annotations - with their own classes or one class for all. The same shape is not copied twice.
Export as GeoJSON NEW	Writes the hand-drawn annotations of the open image to a .geojson file. Tick the classes to export and choose the file name.
Import from GeoJSON NEW	Reads a .geojson file - from EVAnalyzer, QuPath or other tools - into the open image. Each class name in the file is mapped to a project class (by default the one with the same name, ignoring case), a new class, no class, or don’t import, so QuPath’s class names don’t need to match. Shapes without a series go to the series shown.
Re-measure annotations	Measures the hand-drawn objects again on the image file - every channel, full resolution.
Delete annotations…	Removes objects drawn by hand - on this image or all images, of the chosen classes - after a confirmation. Objects found by pipelines are not affected.

Drawn and imported annotations are measured on the image file itself (every channel, full resolution), in the background - so their intensities are right at any zoom, also for whole slide images and through a server.

How Classes Relate to Pipelines

Classes aren’t just labels - they’re what every object-processing pipeline step selects, filters, and reassigns by. See Object Classes for the full lifecycle, from a Threshold entry’s raw segmentation class through to a named class assigned by Classify Objects.