🧠 MyBachelorThesis: AI Pipeline for TBI Analysis

An advanced machine learning pipeline for analyzing Traumatic Brain Injury (TBI) data using multi-modal approaches.

🎯 Introduction

This project implements a sophisticated AI pipeline for TBI analysis, leveraging state-of-the-art machine learning architectures including NAIM and TabNet.

✨ Key Features

• 📊 Flexible data processing pipeline for TBI datasets
• 🔄 Multi-modal learning support (clinical + text data)
• ⚙️ Hydra-based configurable model architectures
• 📈 Integrated logging with TensorBoard and W&B
• 📊 Comprehensive evaluation metrics and visualization

🚀 Getting Started

Prerequisites

• Python 3.8 or higher
• CUDA-compatible GPU (recommended)
• Git

🛠️ Installation

1. Clone the repository

git clone https://github.com/khiemducdoan/MyBachelorThesis.git
cd MyBachelorThesis

2. Set up virtual environment

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

3. Install dependencies

pip install -r requirements.txt

📖 Usage Guide

1️⃣ Data Preparation

Step	Description
Raw Data	Place TBI data in dataset/raw/
Text Data	Ensure CSV format with required columns
Preprocessing	Use notebooks in dataset/ for cleaning

2️⃣ Configuration

Configurations are managed through Hydra in the configs/ directory:

configs/
├── main.yaml          # Main configuration
├── model/            # Model-specific configs
├── data/             # Dataset configs
└── default/          # Default parameters

Modifying Settings:

# Via command line
python train.py model=naim_text data.batch_size=32

3️⃣ Training

# Basic training
python train.py

# With specific config
python train.py experiment=tbi_naim

# Hyperparameter sweep
python train.py logging.sweep=true

4️⃣ Monitoring

Tool	Location	Purpose
TensorBoard	outputs/logs/	Training progress
Checkpoints	outputs/<date>/	Model saves
W&B	Online dashboard	Experiment tracking

5️⃣ Project Structure

MyBachelorThesis/
├── 📁 configs/        # Configuration files
├── 📁 dataset/       # Data and preprocessing
├── 📁 src/          # Source code
│   ├── models/     # Model implementations
│   └── utils/      # Utility functions
├── 📁 outputs/     # Training outputs
└── 📁 notebooks/   # Analysis notebooks

📫 Contact

For questions about the dataset or project, please email: vinakhiem120@gmail.com

---

Made with ❤️ for Traumatic Brain Injury Research