ECG

❤️ Diagnosis of ECG with a Low-Complexity Deep Learning Model

Author: Soroush Soltanizadeh Google Scholar: Profile

---

📌 Overview

Accurate and early detection of cardiac arrhythmias using Electrocardiogram (ECG) signals is vital in modern healthcare. This project introduces a low-complexity 1D Convolutional Neural Network (CNN) for ECG signal classification using the MIT-BIH dataset. To enhance feature representation, Discrete Wavelet Transform (DWT) is applied to the raw ECG signals prior to training. The proposed model is lightweight and suitable for edge devices and real-time wearable applications.

---

🧪 Research Objective

The key objective is to design a computationally efficient CNN that maintains high classification accuracy for ECG-based heart disease diagnosis, especially under constraints relevant to embedded medical systems.

---

📁 Dataset

• Name: MIT-BIH Arrhythmia Dataset
• Format: CSV (MIT_BIH dataset.csv)
• Target Column: target (0 = Normal, 1 = Abnormal)
• Input Features: Extracted ECG signal features

---

🌀 Signal Preprocessing

To extract meaningful patterns from ECG signals:

• Wavelet Transform is applied using pywt.dwt with 'db2' wavelet
• The approximation and detail coefficients are concatenated to form the final feature vector
• Features are standardized using StandardScaler

def apply_dwt(data, wavelet='db2'):
    coeffs = pywt.dwt(data, wavelet)
    return np.hstack(coeffs)

---

🧠 Model Architecture

1D CNN Summary:

• Conv1D: 8 filters, kernel size = 2, padding = "same", activation = ReLU
• MaxPooling1D: pool size = 2
• Flatten
• Dropout: 50%
• Dense: 1 neuron, sigmoid activation

Input Shape   → (n_features, 1)
Conv1D        → filters=8, kernel=2, activation='relu'
MaxPooling1D  → pool_size=2
Dropout       → rate=0.5
Dense         → units=1, activation='sigmoid'

---

⚙️ Training Details

• Optimizer: Adam (learning rate = 0.01)
• Loss Function: Binary Crossentropy
• Evaluation Method: 5-Fold Cross-Validation
• Batch Size: 64
• Epochs: 20

---

📊 Results

Metric	Value
Mean Accuracy	≈{:.2f}%
Standard Deviation	≈{:.2f}%
Model Complexity (CCNN)	≈ (calculated)

💡 The model demonstrates competitive performance while maintaining low computational complexity, making it ideal for real-time monitoring systems.

---

📈 Model Complexity

The complexity is estimated using the formula:

CCNN = ni * nf * nk * (ns + 2*npad - dilation*(nk-1) - 1/stride + 1)

Where:

• ni = number of input features after DWT
• nf = number of filters
• nk = kernel size
• ns = number of time steps
• npad = padding size

---

🩺 Applications

• Wearable ECG Monitors
• Remote Patient Monitoring
• Smart Clinics
• IoT-Based Heart Health Systems

---

🚀 Getting Started

✅ Requirements

pip install tensorflow numpy pandas pywt scikit-learn

▶️ Run the Model

python ecg_diagnosis_cnn.py

---

📚 Future Work

• Explore multi-class ECG classification (e.g., AFib, PVC)
• Integrate explainability methods (e.g., Grad-CAM, LIME)
• Real-time inference on microcontrollers (e.g., Arduino, Raspberry Pi)
• Enhance feature engineering with advanced signal processing techniques

---

📜 License

This project is licensed under the MIT License — feel free to use, modify, and distribute.