🫀 ECG Arrhythmia Classification with Streamlit Dashboard

This project presents an end-to-end ECG Arrhythmia Classification system, combining deep learning with an interactive Streamlit web dashboard.
A 1D Convolutional Neural Network (CNN) is trained on ECG signals to detect abnormal rhythms, and users can interactively upload signals, visualize them, and view predictions in real time.

---

Project Workflow

1. Data Preprocessing

   • ECG signals (from MIT-BIH dataset or CSV files) are normalized.
   • Noise reduction and segmentation are applied to ensure high-quality signals.

2. Model Training (1D CNN)

   • A 1D Convolutional Neural Network is implemented in TensorFlow/Keras.
   • Layers: Conv1D → BatchNorm → MaxPooling → Dropout → Dense → Softmax.
   • Loss: categorical_crossentropy, Optimizer: Adam.
   • Model trained on arrhythmia classes.

3. Model Saving

   • Trained model is stored as cnn_ecg.h5 for deployment.

4. Streamlit Deployment

   • Users upload ECG CSV files.
   • ECG waveform is plotted.
   • The model predicts arrhythmia type and shows class probabilities.
   • Provides an easy-to-use interface for clinical and educational purposes.

---

Arrhythmia Classes

The model can classify the following rhythms:

• Normal Sinus Rhythm (NSR)
• Atrial Fibrillation (AFib)
• Ventricular Tachycardia (VTach)
• Premature Ventricular Contraction (PVC)

---

Repository Structure

├── model\_training/
│   ├── train\_model.py       # Script to train the CNN model
│   ├── ecg\_dataset.csv      # Training data (or link to MIT-BIH)
│   └── cnn\_ecg.h5           # Saved trained model
│
├── app/
│   ├── app.py               # Streamlit application
│   └── sample\_ecg.csv       # Example ECG input
│
├── requirements.txt         # Python dependencies
└── README.md                # Documentation

---

Model Architecture

model = Sequential([
    Conv1D(filters=32, kernel_size=5, activation='relu', input_shape=(input_length, 1)),
    BatchNormalization(),
    MaxPooling1D(pool_size=2),
    Dropout(0.2),

    Conv1D(filters=64, kernel_size=5, activation='relu'),
    BatchNormalization(),
    MaxPooling1D(pool_size=2),
    Dropout(0.3),

    Flatten(),
    Dense(128, activation='relu'),
    Dropout(0.4),
    Dense(num_classes, activation='softmax')
])

• Optimizer: Adam
• Loss: Categorical Crossentropy
• Metrics: Accuracy

---

Installation & Setup

1. Clone the repository:

   git clone https://github.com/your-username/ecg-arrhythmia-dashboard.git
   cd ecg-arrhythmia-dashboard

2. Create a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate   # Linux/Mac
   venv\Scripts\activate      # Windows

3. Install dependencies:

   pip install -r requirements.txt

4. Train the model (if you want to retrain):

   python model_training/train_model.py

5. Run the Streamlit app:

   streamlit run app/app.py

---

Usage

• Upload your ECG CSV file in the sidebar (single column of ECG values).

• The app will:

  1. Plot the ECG waveform.
  2. Predict the arrhythmia type.
  3. Display class probabilities with a bar chart.

---

Example Output

• Input ECG Plot: A waveform visualization of the uploaded ECG signal.

• Prediction:

  Predicted Class: Atrial Fibrillation (AFib)
  Confidence: 92.5%
  

• Class Probabilities (Bar Chart):

  • NSR: 5%
  • AFib: 92.5%
  • VTach: 1.5%
  • PVC: 1%

---

Tools & Technologies

• Python
• TensorFlow / Keras
• NumPy and Pandas
• Matplotlib
• Streamlit

---

Future Improvements

• Integrate Grad-CAM for interpretability (highlighting important ECG regions).
• Support multi-lead ECG inputs.
• Deploy app on Streamlit Cloud / Hugging Face Spaces.
• Extend arrhythmia coverage using MIT-BIH Arrhythmia Database.

---