🧬 2D Confocal Microscopy Image Segmentation & Quantitative Analysis

This repository provides a fully automated pipeline for analyzing multi-channel confocal microscopy images of cells. It uses:

• Cellpose for instance segmentation of cytoplasm and nuclei
• Percentile-based thresholding for mitochondria and 2SC detection
• Quantitative extraction of intensity metrics and colocalization statistics

The pipeline operates on .tiff images with 4 channels and outputs per-image segmentation masks, overlays, and summary CSV reports.

---

🧠 Workflow Summary

1. 🖼️ Preprocessing

   • Recursively loads .tiff images and normalizes channels using global percentile thresholds.

2. 🧫 Segmentation

   • Cytoplasm & nuclei via Cellpose
   • Mitochondria & 2SC via global thresholding (e.g. 90th / 95th percentile in normalized space)

3. 📊 Quantification

   • Calculates 2SC intensity in cytoplasm (excluding nucleus & mitochondria) and in nucleus (excluding mitochondria)
   • Computes mitochondrial colocalization with 2SC
   • Estimates per-cell areas for cytoplasm, nuclei, and mitochondria

4. 🎨 Overlay Generation

   • Produces visual overlays of segmented structures for interpretation

5. 📁 Batch Processing

   • Automatically processes all .tiff images in subdirectories
   • Stores outputs in organized folders
   • Saves CSV summary reports

---

🧬 Input Format

Each input image must be a .tiff file with 4 channels, representing:

Channel Index	Target Structure
0	Nucleus
1	Cytoplasm
2	Mitochondria
3	2SC Staining

Images can be organized into subfolders (e.g., by experimental condition or cell type). The script handles recursive traversal and processing.

---

🧱 Requirements

Install dependencies with:

pip install numpy tifffile scikit-image matplotlib opencv-python cellpose

Cellpose will optionally use a GPU if available.

---

▶️ How to Run

Modify the paths and channel indices in the __main__ section of multi_channel_cell_segmentation.py:

raw_data_path = '/path/to/raw_images/'  # input directory (recursively scanned)
top_level_output_path = '/path/to/output/'

nuclei_channel = 0
cyto_channel = 1
mito_channel = 2
sc_channel = 3

Then, run the script:

python multi_channel_cell_segmentation.py

---

📂 Output Structure

For each image:

Output File	Description
*_Processed_Image.tiff	Normalized and downsampled image
*_Cytoplasm_Mask.tiff	Segmented cytoplasm overlay
*_Nucleus_Mask.tiff	Segmented nucleus overlay
*_Mitochondria_Mask.tiff	Mitochondria mask from thresholding
*_Segmentation_Overlay.tiff	Overlay of cytoplasm and nucleus
*_MitoCyto2SC_Overlay.tiff	Combined overlay of all structures
*_cellwise_areas.csv	Per-cell area stats (cytoplasm, nucleus, mito)
Intensity_Colocalisation_data.csv	Summary stats per image (2SC, colocalisation)

---

🧪 Use Cases

This pipeline is ideal for:

• Quantifying 2SC localization patterns across cell types
• Evaluating mitochondrial-2SC colocalization metrics
• Extracting cell-wise morphometrics for statistical analysis
• Generating overlay visualizations for publication or QC

---

📬 Contact

For questions, bugs, or contributions, feel free to open an issue or reach out!