Classification

The Classification tab is where you define the object populations that your pipelines will detect and measure. Every detected object is assigned to exactly one class.

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.

Adding and Editing Classes

Click the + button to add a new class. Double-click an existing class to open the Class Editor, which lets you set:

Field	Description
Name	Class label, e.g. cy5@spot
Colour	Display colour used for detected objects in the viewer
Notes	Optional free-text description
Default metrics	Which measurement columns to display by default after the analysis

The default metrics can be changed at any time, even after an analysis has been completed, without re-running the analysis.

Auto-populate from Image Metadata

Click the Magic Stick button to have EVAnalyzer automatically create classes based on the channel information read from the current image. This creates one class per image channel as a starting point.

Classification Presets

Presets are shared sets of class definitions for consistent naming across experiments:

• Load preset — choose from the drop-down beside the + button to load a predefined set of classes.
• Save as template — save your current classification settings as a .impt template file for reuse and sharing.

Template files are stored in ~/evanalyzer/templates/ and appear in the preset list on the next launch.

How Classes Relate to Pipelines

Each pipeline step that produces objects must specify a target class. There are two ways a class is assigned:

1. Extract ROIs — the first object-extraction step in a pipeline assigns a segmentation class (an internal intermediate class used to hold the binary mask results).
2. Classify ROIs — converts segmentation-class objects into named user classes, applying optional filters (area, circularity, intensity) to accept or reject each detected region.

Downstream steps — Colocalization, Voronoi, Distance Transform — all reference classes by name to select their input objects.