Quantum conditional generative adversarial network based on patch method for abnormal electrocardiogram generation.

Area of Science:

• Quantum Computing
• Artificial Intelligence
• Biomedical Signal Processing

Background:

• Scarcity and class imbalance in abnormal electrocardiogram (ECG) databases hinder AI development for cardiovascular disease detection.
• Accurate ECG signal generation is vital for training robust diagnostic algorithms.

Purpose of the Study:

• To propose a novel quantum conditional generative adversarial algorithm (QCGAN-ECG) for generating synthetic abnormal ECG signals.
• To address limitations in existing ECG datasets for AI training.

Main Methods:

• Developed a QCGAN-ECG utilizing a patch-based quantum generator for segment-specific feature generation.
• Incorporated quantum registers for controllable generation based on heartbeat type and probability distributions.
• Simulated experiments using Pennylane for performance evaluation.

Main Results:

• QCGAN-ECG achieved an average accuracy of 88.8% in generating abnormal heartbeats.
• The algorithm accurately fitted the probability distributions of diverse abnormal ECG data.
• Demonstrated significant robustness against various levels of quantum noise interference.

Conclusions:

• QCGAN-ECG effectively generates abnormal ECG signals, addressing data limitations.
• The proposed method is practical for near-term quantum devices and enhances AI-driven cardiac diagnosis.
• The approach shows promise for advancing AI in cardiovascular health monitoring.