Quantum ECG Anomaly Detection

Quantum vs Classical anomaly detection for ECG data using One-Class SVM.

Structure

ecg/
├── main.py                 # Main entry point
├── classical_detector.py   # Classical One-Class SVM
├── quantum_detector.py     # Quantum Kernel SVM
├── preprocessor.py         # Data preprocessing and loading
├── metrics_calculator.py   # Performance metrics calculation
├── visualizer.py           # Results visualization
├── requirements.txt        # Dependencies
├── results.png             # Visualization output after execution (root level)
├── data/
│   ├── ECG_Features.npy    # ECG features (112,570 samples × 8 features)
│   └── EEG_labels.npy      # Labels (N=normal, others=anomaly)
└── output_samples/
    ├── results.png         # Visualization output
    ├── ibm_results.png     # Visualization output (IBM backend)
    ├── output.log          # Console output log
    └── ibm_output.log      # Console output log (IBM backend)

Requirements

• Python 3.10
• Qiskit
• Qiskit Machine Learning
• Qiskit IBM Runtime
• NumPy
• Scikit-learn
• Matplotlib

Installation

1. Create and activate a virtual environment (Optional):

python3.10 -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

Usage

python main.py

Notes:

• The quantum detector uses a subset of 100 samples by default due to computational constraints. You can adjust this by modifying the fourth parameter in the run_quantum() call in main.py.
• If you want to run the IBM backend, pelase set your IBM token to the IBM_TOKEN in quantum_detector.py file. Default execution will be on a local simulator.

Dataset

The two files are stored in the data folder. Provided by MIT-BIH Arrhythmia Database:

• 112,570 ECG samples
• 8 features per sample
• Binary classification: Normal ('N') vs. Anomaly

Methods

Classical

• One-Class SVM with RBF kernel
• Trained only on normal samples
• Full dataset evaluation

Quantum

• Angle encoding (RY gates, 8 qubits)
• Quantum kernel via fidelity
• Subset evaluation (computational constraints)

Output

• Accuracy, Precision, Recall, F1-Score
• Confusion matrix
• Visualization saved as PNG