Introduction: Prostate cancer remains one of the most prevalent malignancies in men, with accurate histopathological assessment being crucial for diagnosis and treatment planning. Whole-slide imaging (WSI) has enabled computational pathology approaches to automate the analysis of tissue specimens. A major application of AI, especially deep learning models, is to automate the process of precise tissue segmentation. This project aims to develop a deep learning model capable of automated multi-class segmentation of prostate tissue structures.

Goal: The primary objective of this project is to build a deep learning model that can accurately segment prostate biopsy WSIs into different histological classes:

• Background (non-tissue regions)

• Benign epithelium (normal prostate glands)

• Stroma (connective tissue components)

• Malignant Grade 3 (low-grade cancer)

• Malignant Grade 4 (intermediate-grade cancer)

• Malignant Grade 5 (high-grade cancer)

By achieving precise segmentation, the model can facilitate better understanding of tumor architecture, aid in prognostic assessments, and potentially assist pathologists in automating the diagnostic process and identifying aggressive disease features.

Dataset: The PANDA (Prostate cANcer graDe Assessment) dataset, derived from the Karolinska Institute and Radboud University Medical Center, provides a rich collection of annotated prostate biopsy images. For our project, we will specifically utilize the Radboud portion of the dataset, which includes:

• Whole-slide images (WSIs) of prostate biopsy samples.

• Expert-labeled segmentation masks indicating tissue regions.

The segmentation labels from Radboud will serve as ground truth for training our model, allowing us to develop a supervised deep learning approach for semantic segmentation.
The data can be accessed on Kaggle: https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/overview

References:
1. Bulten, W., et al. (2021). “Artificial intelligence-assisted Gleason grading of prostate biopsies: a diagnostic study.” Nature Medicine, 27(1), 89-95. DOI: 10.1038/s41591-021-01620-2

2. Litjens, G., et al. (2017). “A survey on deep learning in medical image analysis.” Medical Image Analysis, 42, 60-88. DOI: 10.1016/j.media.2017.07.005

3. Ronneberger, O., et al. (2015). “U-Net: Convolutional networks for biomedical image segmentation.” MICCAI 2015, 234-241. DOI: 10.1007/978-3-319-24574-4_28

Reading material: The following materials are recommended to get familiar with the basics of Python programming, image analysis, and deep learning:

Python basics:

• "Python for Beginners” (W3Schools) - https://www.w3schools.com/python/
• “Python Programming MOOC 2023” from the Department of Computer Science at the University of Helsinki - https://programming-23.mooc.fi/
• “Python Track” from Exercism - https://exercism.org/tracks/python
   

Deep learning basics:

• “Deep Learning with Python” by François Chollet (Chapter 1-3) - https://www.manning.com/books/deep-learning-with-python
• Coursera: “Deep Learning Specialization” by Andrew Ng - https://www.coursera.org/specializations/deep-learning
• U-Net for segmentation: https://arxiv.org/abs/1505.04597
   

Computer Vision & PyTorch:

• FastAI Course: “Practical Deep Learning for Coders” - https://course.fast.ai/
• PyTorch Tutorials: https://pytorch.org/tutorials/
• Introduction to Convolutional Neural Networks (CNNs) - https://cs231n.github.io/convolutional-networks/

General Introduction to Pathology Image Analysis

• Whole-Slide Imaging (WSI) and Computational Pathology: Omar et al. (2023). “Applications of Digital Pathology in Cancer: A Comprehensive Review” Annual Review of Cancer Biology - https://www.annualreviews.org/content/journals/10.1146/annurev-cancerbio-062822-010523
   

Basic Image Processing & Computer Vision for Pathology

• Image Processing with OpenCV (Beginner-Friendly): “OpenCV-Python Tutorials” - https://docs.opencv.org/4.x/d6/d00/tutorial_py_root.html (Covers basic operations like thresholding, contour detection, and image filtering, which are crucial for pathology image analysis.)
• "Deep Learning for Medical Image Analysis” by Zhou et al. (2017) (Covers CNNs, segmentation networks, and deep learning applications in pathology) - DOI: 10.1016/B978-0-12-810408-8.00001-1
• "Computational pathology definitions, best practices, and recommendations for regulatory guidance: a white paper from the Digital Pathology Association” J Pathol (2019) - https://pmc.ncbi.nlm.nih.gov/articles/PMC6852275/
  
   

Tips for getting started:

1. Set up the environment:

Before working with pathology images, we need to set up the software environment properly. This ensures that all dependencies are installed correctly, and students can work with the necessary tools smoothly.

1.1 Install Python: Python is the programming language we will use for deep learning and pathology image analysis.

Windows:
• Download Python from the official website: https://www.python.org/downloads/

• During installation, check the box “Add Python to PATH”.

• Verify installation by opening a command prompt and typing: 
python --version

Mac:
Python should be pre-installed on macOS. If not, install the latest version via Homebrew:
• Install Homebrew: 
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

• Install Python: 
brew install python

• Verify installation: 
python --version

------------------------
1.2 Install Micromamba
Conda or Micromamba helps manage dependencies and create isolated environments for different projects.

• Install Micromamba (Lightweight alternative to Conda): 
curl -Ls https://micro.mamba.pm/install.sh | bash

• Restart the terminal and run:
micromamba --version

----------------------------
1.3 Create a Micromamba Environment:

• Create the environment (we will call it PCa_segmentation):
 micromamba create -n PCa_segmentation python=3.9

• Activate the environment:
micromamba activate PCa_segmentation

• Install PyTorch (Check https://pytorch.org/get-started/locally/ for latest version):

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

• Install image processing and visualization tools:

pip install numpy opencv-python matplotlib albumentations tifffile openslide-python

• Install deep learning libraries for segmentation:

pip install monai segmentation-models-pytorch

•  Install Jupyter Notebook for running code:

pip install jupyterlab

Project Prepared By:
Mohamed Omar