ABP Estimation Using PPG

A machine learning pipeline for estimating Arterial Blood Pressure (ABP) waveforms from Photoplethysmogram (PPG) signals using deep learning and time series forecasting approaches.

Overview

This repository implements multiple approaches to predict blood pressure from PPG signals:

• PPG2ABP: Fully Convolutional Neural Networks (U-Net, MultiResUNet) for signal-to-signal translation
• Linear Regression & CNN: Traditional ML and neural network baselines
• Time Series Forecasting: PyCaret-based ARIMA and AutoML approaches

Repository Structure

ABP-estimation-using-PPG/
├── PPG2ABP/                        # PPG to ABP signal translation module
│   ├── codes_125hz/                # Implementation for 125Hz sampling rate
│   │   ├── models.py               # U-Net, MultiResUNet, and other architectures
│   │   ├── train_models.py         # Training scripts
│   │   ├── evaluate.py             # Evaluation and metrics
│   │   ├── data_processing.py      # Data preprocessing utilities
│   │   ├── streamlit-app.py        # Web demo application
│   │   └── PPG2ABP.ipynb           # Demo notebook
│   ├── codes_25hz/                 # Implementation for 25Hz sampling rate
│   ├── weights/                    # Pre-trained model weights
│   └── requirements.txt            # PPG2ABP-specific dependencies
├── tanzen_data/                    # Data scraping utilities
│   └── data_scrap.ipynb            # Data collection notebook
├── linear_cnn_pycaret.ipynb        # Linear regression, CNN, and PyCaret regression
├── arima_pycaret.ipynb             # Time series forecasting with PyCaret
├── requirements.txt                # Main project dependencies
└── CODE_OF_CONDUCT.md

Tech Stack

• Languages: Python 3.8+
• Deep Learning: TensorFlow/Keras, PyTorch
• ML/AutoML: PyCaret, scikit-learn, XGBoost, LightGBM, CatBoost
• Data Processing: NumPy, Pandas, SciPy
• Visualization: Matplotlib, Seaborn, Plotly
• Web App: Streamlit, Gradio

Installation

# Clone the repository
git clone git@github.com:lucky-verma/ABP-estimation-using-PPG.git
cd ABP-estimation-using-PPG

# Install dependencies
pip install -r requirements.txt

# For PPG2ABP module specifically
pip install -r PPG2ABP/requirements.txt

Usage

PPG2ABP Signal Translation

cd PPG2ABP/codes_125hz
python train_models.py      # Train models
python evaluate.py          # Evaluate on test set
streamlit run streamlit-app.py  # Launch web demo

Jupyter Notebooks

1. linear_cnn_pycaret.ipynb: Baseline models (Linear Regression, CNN) and PyCaret AutoML
2. arima_pycaret.ipynb: Time series forecasting approaches
3. PPG2ABP/codes_125hz/PPG2ABP.ipynb: Full PPG2ABP pipeline demo

Data

The project uses PPG and ABP signal data from MATLAB (.mat) files. Data should be placed in a kaggle_data/ directory with files named part_1.mat through part_12.mat.

Citation

If you use PPG2ABP in your research, please cite:

@article{ibtehaz2020ppg2abp,
  title={PPG2ABP: Translating Photoplethysmogram (PPG) Signals to Arterial Blood Pressure (ABP) Waveforms using Fully Convolutional Neural Networks},
  author={Ibtehaz, Nabil and Rahman, M Sohel},
  journal={arXiv preprint arXiv:2005.01669},
  year={2020}
}

License

MIT License