Quantum-convolution-based hybrid neural network model for arrhythmia detection

This paper proposes a quantum convolutional hybrid neural network (QCHNN) model for cardiac arrhythmia detection that integrates quantum computing (QC) into the convolutional neural network (CNN). By leveraging the automatic feature extraction capability of the CNN and the entanglement property of QC, the robustness against noise is improved. QCHNN ultimately realizes a robust and highly accurate detection model. First, the model converts the electrocardiography (ECG) signal into two-dimensional gray-scale images, thereby mitigating the effects of signal noise. We then use an annular parameterized quantum circuit (PQC) to form a fully connected quantum (FCQ) layer, which enhances robustness effectively. In this quantum layer, the PQC not only provides sufficient entanglement to complete the feature fusion task, but also can control the number of parameters. Finally, we perform experimental simulations to verify the high accuracy and robustness of the proposed model. The results show that the detection accuracy of QCHNN can be improved by about 15.94% in a small sample, and the best accuracy achieved for five types of arrhythmias is 98.58%. Moreover, we prove that the QCHNN model is more tolerant of noise than the CNN model by testing four types of noise.