mCherry‑H2B Nuclear Segmentation Pipeline

Automated, reproducible pipeline for quantifying nuclear morphology and fluorescence in Arabidopsis root tips. The confocal image set is not uploaded due to file size. Google drive link to the folder used in this project: https://drive.google.com/drive/folders/1lyRpDqsC8ukQ7ZqJe-EhZeCihV2n9mO5?usp=drive_link

Highlights

• One‑command batch processing for every .tif/.tiff under a project folder
• Robust adaptive thresholding that copes with uneven illumination
• Dual outputs: SQLite (per‑nucleus) and CSV (per‑image)
• Publication‑ready bar plots with SEM error bars (z‑slice and stitched)
• Modular—drop in deep‑learning models (U‑Net, SAM) without changing downstream code

---

1️⃣ Quick Start

# 1. Clone the repository
git clone https://github.com/JHYUIHGY/Otegui_segmentation.git
cd Otegui_segmentation

# 2. Set up a virtual environment
python3 -m venv venv
source venv/bin/activate

# 3. Install dependencies
pip install -r requirements.txt

Organize your data

data/
├── istl_control_elongation_zone/
│   ├── mchH2B-istl345_seedling3_z01.tif
│   └── …
├── istl_mutant_meristem_zone/
│   └── *.tiff
└── …

Edit seg_process_batch.py

root_dir     = "data"   # top-level folder with images
overwrite_db = True     # wipe old DB on every run

Run the batch processor

python seg_process_batch.py

Create figures

python plot_nuclear_metrics_dual.py
open bar_zslices.png     # macOS example

---

2️⃣ Dependencies

numpy
pandas
matplotlib
seaborn
scikit-image
tifffile
scipy

Install with:

pip install -r requirements.txt

Optional

Package	Purpose
jupyterlab	Interactive exploration & debugging
ipykernel	Use the virtualenv inside notebooks

---

3️⃣ Pipeline Overview

Stage	Script / Function	What it does
Folder crawl	discover_image_folders()	Picks all subdirs containing .tif or .tiff images
Metadata parsing	segmentation.parse_filename()	Extracts region, genotype, seedling, z-slice, channel
Segmentation	segment_and_measure_spots()	Adaptive local threshold → morphological cleanup
Measurements	(same as above)	Area, mean intensity, integrated intensity
Storage	store_measurements_in_sql()	Writes to results.db and image_summary.csv
QC overlays	visualize_segmentation_detailed()	Saves labeled overlays in results/…
Plotting	plot_nuclear_metrics_dual.py	Generates bar plots for z-slices and stitched images

---

4️⃣ Re-running Cleanly

To reset everything:

rm image_summary.csv results.db -r results/
python seg_process_batch.py

Or just set overwrite_db = True and delete image_summary.csv manually.

---

5️⃣ Extending the Pipeline

Task	Where to modify
Background subtraction	Add logic at the top of segment_and_measure_spots()
Use deep learning models	Replace thresholding block with model inference
Add parallel processing	Wrap process_folder() in ThreadPoolExecutor

---

6️⃣ Troubleshooting

Issue	Fix
Duplicate rows in CSV	Delete image_summary.csv before re-running
KeyError in plotting	Ensure seaborn ≥ 0.13 and use latest script
Blurry / incomplete masks	Lower min_area, tweak threshold kernel

---

7️⃣ License & Citation

Released under the MIT License.

If you use this pipeline in a publication, please cite:

Phillip Duan (2025). Automated Quantification of Nuclear Fluorescence in ISTL Mutant Arabidopsis Roots.

---

Happy segmenting! 🧬 Pull requests & issues are always welcome.