Fiber Orientation Prediction Project

Overview

Advanced machine learning project for predicting and analyzing fiber orientations using state-of-the-art image processing and neural network techniques.

Project Structure

fiber-orientation/
│
├── src/
│   └── fiber_predictor/
│       ├── apps/           # Application-specific modules
│       ├── neural_networks/# Neural network models
│       ├── svr_hog/        # Support Vector Regression models
│       └── utils/          # Utility functions
│
├── notebooks/      # Exploratory Jupyter notebooks
│   ├── HOG/       # HOG feature analysis notebooks
│   └── NN/        # Neural network analysis notebooks
│
├── models/        # Trained models
│   ├── cnn/       # CNN model weights
│   └── svr_hog/   # SVR-HOG model weights
│
├── images/        # Image data
├── environment.yml # Conda environment configuration
└── README.md      # Project overview

Environment Setup

This project uses Conda for environment management. The environment includes PyTorch for deep learning, scikit-learn for machine learning, OpenCV for image processing, and other scientific computing libraries.

Prerequisites

• Conda (Miniconda or Anaconda)

Creating the Environment

# Clone the repository
git clone https://github.com/yourusername/fiber-orientation.git
cd fiber-orientation

# Create and activate conda environment
conda env create -f environment.yml
conda activate fiber-orientation

Key Dependencies

• Python 3.12
• PyTorch & TorchVision
• scikit-learn & scikit-image
• OpenCV
• NumPy & Pandas
• Jupyter Notebook

Usage

Jupyter Notebooks

The project includes several Jupyter notebooks for analysis:

• notebooks/CNN_Inference.ipynb: CNN model inference and visualization
• notebooks/HOG_Inference.ipynb: HOG-based model inference
• notebooks/EDA.ipynb: Exploratory Data Analysis

The notebooks/HOG/ and notebooks/NN/ directories contain detailed analysis notebooks for each approach.

Models

Pre-trained models are available in the models/ directory:

• CNN models in models/cnn/
• SVR-HOG models in models/svr_hog/

Each model type includes various versions (raw, augmented, balanced) for different use cases.

Citation

If you use this software in your research, please cite the associated publication:

E.N. Saruhan, H. Ozturk, D. Kul, B. Sevgin, C. Pekkan, "Learning-enhanced 3D fiber orientation mapping in thick cardiac tissues," Biomedical Optics Express, vol. 16, no. 8, 2025. DOI: 10.1364/BOE.532799

BibTeX

@article{saruhan2025learning,
  title={Learning-enhanced 3D fiber orientation mapping in thick cardiac tissues},
  author={Saruhan, E.N. and Ozturk, Hakancan and Kul, D. and Sevgin, B. and Pekkan, C.},
  journal={Biomedical Optics Express},
  volume={16},
  number={8},
  year={2025},
  doi={10.1364/BOE.532799},
  url={https://opg.optica.org/abstract.cfm?uri=boe-16-8-3315}
}

Citing the Software

This repository includes a CITATION.cff file with detailed citation metadata. The CITATION.cff format is the standard for citing research software and is supported by GitHub, Zotero, and other research tools.

See: CITATION.cff

---

License

MIT License - see LICENSE file for details.