Spot Count per Cell

This tutorial extends the Spot Count workflow to count spots on a per-cell basis. Three pipelines work together:

Nucleus — segment nuclei from the DAPI channel.
Cell area (optional) — segment the cell body, or use Voronoi approximation.
EV Detection — detect spots, then link each spot to the cell it falls in.

Prerequisites

Multi-channel images with: (1) a nuclear stain (DAPI/Hoechst), and (2) a spot channel (Cy5 etc.).
Optionally: a cell-body stain (cell-fill marker) for more accurate cell boundaries.

Step 1: Define Classes

Class	Purpose
`dapi@nucleus`	Segmented nuclei
`cy5@spot`	Detected spots
`cy5@spot-in-nucleus`	Spots that fall inside a nucleus
`cy5@spot-outside`	Spots outside any nucleus

Step 2: Nucleus Pipeline

Create a pipeline named Nucleus targeting the DAPI channel:

Step	Settings
Rolling Ball	Radius: 20–50 (larger than nuclei)
Gaussian Blur	Kernel: 5
Threshold	Manual or Otsu auto-threshold
Connected Components	—
Watershed	Tolerance: 0.3 (optional)
Extract ROIs	—
Classify ROIs	Target: `dapi@nucleus`; Min area: 500 px²; Min circularity: 0.3

Step 3: EV Detection Pipeline

Follow the Spot Count pipeline for the EV channel to produce cy5@spot objects.

Step 4: Link Spots to Nuclei

After the Classify ROIs step in the EV pipeline, add a second Classify ROIs step:

Setting	Value
Origin class	`cy5@spot`
Target class	`cy5@spot-in-nucleus`
Intersection condition	Intersects with `dapi@nucleus` (min intersection: 10%)

Objects that intersect a nucleus are moved to cy5@spot-in-nucleus. Spots that do not intersect remain as cy5@spot and can optionally be moved to cy5@spot-outside by an additional step.

Step 5: Results

In the results view:

The cy5@spot-in-nucleus Count per image gives total spots inside nuclei.
With the parent object ID relationship, you can query spots per nucleus in the DuckDB results file:

SELECT parent_object_id, COUNT(*) AS spots_per_nucleus
FROM objects
WHERE class = 'cy5@spot-in-nucleus'
GROUP BY parent_object_id;

Voronoi cell approximation (no cell stain)

If no cell-body stain is available, use Voronoi to approximate cell territories from nucleus centres:

Setting	Value
Centers	`dapi@nucleus`
Max radius	30 µm (adjust to expected cell size)
Output class	`cell@voronoi`
Exclude areas at edges	enabled
Exclude areas without centre	enabled

Then use cell@voronoi instead of dapi@nucleus in the intersection step.