Short Lecture: Introduction to Cell Segmentation with Python.

This repository contains a series of Python notebooks for a short lecture on cell segmentation. The notebooks guide the user from the fundamentals of digital images, through "old-school" classical computer vision techniques (like thresholding and watershed), and finally to modern, deep-learning-based methods (Cellpose and StarDist).

The final notebook demonstrates a complete analysis pipeline: segmenting a real-world histology image and extracting quantitative, single-cell data for downstream analysis.

Notebooks

The lecture is broken down into four key notebooks, designed to be run in order:

1. 1.Creating_an_Image.ipynb

• Concept: What is a digital image?
• Covers: The fundamentals of how images are represented as matrices (pixels). It explores color channels (RGB vs. grayscale), bit-depth (8-bit vs. 16-bit), and introduces color deconvolution (rgb2hed) for separating stains in histology images.

2. 2.Segmentation_Thresholding.ipynb

• Concept: The simplest segmentation method and its main problem.
• Covers: An interactive ipywidgets slider to demonstrate basic intensity thresholding on a cells3d nuclei image. This notebook clearly highlights the primary failure of this method: merged, touching objects.

3. 3.Segmentation_Watershed.ipynb

• Concept: The "classic" solution for splitting merged objects.
• Covers: An interactive notebook that implements the full marker-controlled watershed pipeline (Threshold -> Distance Transform -> Find Peaks -> Watershed). It shows how this successfully separates the touching nuclei from the previous notebook.

4. 4.Advanced_Cellpose_and_Stardist.ipynb

• Concept: Modern, deep-learning-based solutions and their applications.
• Covers: This is the capstone notebook, divided into three parts:
  1. Modern Solutions (on cells3d): Introduces Cellpose and StarDist with interactive sliders to adjust their parameters (diameter, flow_threshold, prob_thresh, nms_thresh). This provides a direct comparison to the "classic" methods from notebooks 2 & 3 on the same data.
  2. Real-World Application (on immunohistochemistry): Shows how to apply StarDist's pre-trained '2D_versatile_he' model to a complex H&E (hematoxylin and eosin) stained image, achieving robust segmentation where classic methods fail.
  3. Downstream Analysis: The key "so what?" section. It uses the segmentation mask as a "cookie cutter" to extract quantitative data (area, shape, intensity) for every single cell, converts it to a pandas DataFrame, and generates publication-ready plots (histograms, spatial scatter plots).

---

Setup & Installation

To run these notebooks, you'll need a Python environment with several scientific and deep-learning packages.

1. Create a virtual environment (optional, but highly recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install the required packages:

   pip install jupyterlab matplotlib ipywidgets
   pip install scikit-image scipy scikit-learn
   pip install pandas seaborn
   pip install cellpose
   pip install stardist
   pip install "numpy<2.0"  # StarDist/Numba has issues with NumPy >= 2.0

   Note: Cellpose and StarDist may require PyTorch or TensorFlow. Please follow their official installation guides if you encounter issues.

---

How to Run

1. Clone this repository:

   git clone [https://github.com/mouneem/ShortLecture_Cell_Segmentation.git](https://github.com/mouneem/ShortLecture_Cell_Segmentation.git)
   cd ShortLecture_Cell_Segmentation

2. Activate your virtual environment (if you made one):

   source venv/bin/activate

3. Launch Jupyter Lab:

   jupyter lab

4. Open 1.Creating_an_Image.ipynb to start the lecture.

---

Credits

This lecture and its associated notebooks were created by ESSABBAR AbdelMounm @mouneem.