HRF Retinal Image Content-Based Retrieval System

Project Overview

This project implements a Content-Based Image Retrieval (CBIR) system specifically designed for retinal fundus images using the High-Resolution Fundus (HRF) dataset. The system employs advanced computer vision techniques and handcrafted feature engineering to enable efficient retrieval of similar retinal images based on visual content.

Key Objectives

• Develop a robust CBIR system for medical image analysis
• Extract and combine multiple handcrafted features for comprehensive image representation
• Compare different feature engineering techniques for retinal image classification
• Provide insights into feature importance and model performance

Dataset

HRF (High-Resolution Fundus) Dataset

• Total Images: 45 high-resolution retinal fundus images
• Categories:
  • Healthy: 15 images
  • Diabetic Retinopathy: 15 images
  • Glaucoma: 15 images
• Resolution: High-quality fundus photographs
• Format: JPEG/JPG images

Features Implemented

1. Handcrafted Feature Extraction

• HOG (Histogram of Oriented Gradients): Captures edge and gradient information
• LBP (Local Binary Patterns): Analyzes local texture patterns
• Edge Detection: Sobel, Canny, and Laplacian edge features
• Gabor Filters: Texture analysis using frequency domain features
• GLCM (Gray-Level Co-occurrence Matrix): Statistical texture descriptors
• Color Histograms: RGB and HSV color distribution features

2. Feature Combination Strategies

• Individual feature performance analysis
• Feature concatenation and fusion techniques
• Dimensionality reduction using PCA
• Feature selection using statistical methods

3. Machine Learning Models

• Random Forest Classifier
• Support Vector Machine (SVM)
• K-Nearest Neighbors (KNN)
• Logistic Regression

4. Evaluation Metrics

• Classification accuracy
• Precision, Recall, F1-score
• Confusion matrices
• Feature importance visualization
• Cross-validation performance

Installation

Prerequisites

• Python 3.8 or higher
• pip package manager

Setup Instructions

1. Clone the repository

git clone https://github.com/yourusername/HRF-Retinal-Edge-Retrieval.git
cd HRF-Retinal-Edge-Retrieval

2. Create virtual environment

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

3. Install dependencies

pip install -r requirements.txt

Required Libraries

The project dependencies are specified in requirements.txt:

opencv-python>=4.8.0
scikit-image>=0.25.0
scikit-learn>=1.3.0
numpy>=1.24.0
pandas>=2.0.0
matplotlib>=3.7.0
seaborn>=0.12.0
jupyter>=1.0.0
tqdm>=4.65.0
pillow>=10.0.0
scipy>=1.10.0
joblib>=1.3.0
ipykernel>=6.25.0
notebook>=7.0.0

Usage

Running the Jupyter Notebook

jupyter notebook HRF-Retinal-Edge-Retrieval.ipynb

Key Notebook Sections

1. Data Loading and Exploration
2. Feature Extraction Pipeline
3. Model Training and Evaluation
4. Performance Comparison
5. Feature Importance Analysis
6. Results Visualization

Results

Note: Results will be updated after running the complete analysis.

Model Performance Comparison

The notebook provides comprehensive analysis of different feature combinations and their performance on the HRF dataset.

Key Insights

• Feature combination strategies and their effectiveness
• Comparative analysis of different machine learning models
• Feature importance rankings for retinal image classification
• Performance metrics across different classification scenarios

Technical Highlights

Feature Engineering Pipeline

The project implements a comprehensive feature extraction pipeline combining multiple computer vision techniques for robust retinal image analysis.

Model Evaluation Framework

• Stratified K-Fold Cross-Validation
• Statistical significance testing
• Feature importance ranking
• Comprehensive visualization

Project Structure

HRF-Retinal-Edge-Retrieval/
├── HRF/                                          # Dataset directory
│   ├── healthy/                                  # Healthy retinal images (15 images)
│   ├── diabetic_retinopathy/                     # DR images (15 images)
│   └── glaucoma/                                 # Glaucoma images (15 images)
├── venv/                                         # Virtual environment
├── Connection Sensitive Attention U-NET...pdf   # Reference paper
├── HRF-Retinal-Edge-Retrieval.ipynb            # Main notebook
├── README.md                                     # Project documentation
└── requirements.txt                              # Dependencies

Educational Value

This project demonstrates:

• Handcrafted feature engineering in computer vision
• Medical image analysis techniques
• Content-based image retrieval systems
• Feature fusion and selection strategies
• Performance evaluation methodologies

Additional Resources

• Reference Paper: Connection Sensitive Attention U-NET for Accurate Retinal Vessel Segmentation (included in repository)
• Dataset: HRF (High-Resolution Fundus) database for vessel segmentation studies

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

Acknowledgments

• HRF Dataset: Thanks to the creators of the High-Resolution Fundus dataset
• OpenCV Community: For excellent computer vision tools
• scikit-image: For comprehensive image processing capabilities
• scikit-learn: For machine learning implementations

Contact

For questions or collaboration opportunities, please open an issue in this repository.

---

Star this repository if you found it helpful!