---
title: Cell Fluorescence Quantification
emoji: 🔬
colorFrom: red
colorTo: blue
sdk: gradio
sdk_version: 5.50.0
app_file: app.py
pinned: false
license: mit
---

# Cell Fluorescence Quantification

Automatic detection and quantification of cytoplasmic fluorescence in
RGB fluorescence microscopy images.

## How it works

Each input image is assumed to be a two-channel fluorescence image stored as RGB:

- **Blue channel** — nuclei stain (e.g. DAPI)
- **Red channel** — cytoplasmic marker whose intensity we want to quantify

The pipeline is:

1. **Segment nuclei** from the blue channel using Gaussian smoothing →
   Otsu thresholding → morphological cleaning → watershed splitting on the
   distance transform (to separate touching nuclei).
2. **Expand each nucleus** outward by a fixed pixel radius to define the
   *whole cell* (nucleus + cytoplasm ring). Background pixels are assigned to
   their nearest nucleus, with a Voronoi-style constraint so cells never
   overlap.
3. **Select representative cells** — non-border, well-formed, spread out, with
   above-median cytoplasm intensity (avoids picking dim background regions).
4. **Measure** in the red channel:

   ```
   Cytoplasm Area    = Cell Area     - Nucleus Area
   Cytoplasm IntDen  = Cell IntDen   - Nucleus IntDen
   Mean Cytoplasm    = Cytoplasm IntDen / Cytoplasm Area
   ```

## Outputs

- Annotated image with two concentric outlines per cell (outer = cell,
  inner = nucleus) and a numbered label.
- A per-cell data table with all measurements.
- A downloadable CSV of the table.

## Local usage

```bash
pip install -r requirements.txt
python app.py
```

Then open the local URL printed in the console.

## Parameters

- **Cells per image** — how many cells to report per image (default 5).
- **Cytoplasm ring thickness** — pixel radius used to expand each nucleus
  into the surrounding cytoplasm (default 12 px). Increase if your cells
  have a thicker cytoplasm ring.