Method for obtaining artificial neural network weighted value matrix for identifying atrial fibrillation

Abstract

The invention discloses a method for obtaining an artificial neural network weighted value matrix for identifying atrial fibrillation. The method includes the steps of calling an MIT-BIH arrhythmia database, an MIT-BIH normal sinus rhythm database and a long-time atrial fibrillation database as trainning samples, introducing an artificial neural network for learning training, randomly setting the weighted value of each layer of artificial neural network, and inputting training data samples to conduct repeated iteration to correct the weighted value of each layer until the training error is smaller than a designated value. In this way, the weighted value matrix for determining occurrence of atrial fibrillation can be found, the weighted value matrix is added into an original artificial neural network to establish a new artificial neural network, and human electrocardiosignals are processed with the acquired target human electrocardiosignals as data to obtain a target human body characteristic vector X. The prediction operation is carried out according to the target human body characteristic vector X and the new artificial neural network.

Description

technical field [0001] The invention relates to atrial fibrillation detection, in particular to a method for obtaining an artificial neural network weight value matrix for atrial fibrillation recognition. Background technique [0002] Atrial fibrillation (abbreviated as atrial fibrillation, Auricular Fibrillation, AF) is the most common sustained cardiac arrhythmia. The incidence of atrial fibrillation continues to increase with age, reaching 10% of people over the age of 75. In atrial fibrillation, the frequency of atrial excitement reaches 300-600 beats / min, and the heartbeat frequency is often fast and irregular, sometimes up to 100-160 beats / min, which is not only much faster than normal heartbeat, but also absolutely irregular. When atrial fibrillation occurs, the atrium loses effective systolic function, and the blood easily stagnates in the atrium to form a thrombus. After the thrombus falls off, it can travel with the blood to all parts of the body, resulting in cer...

AI Technical Summary

Problems solved by technology

[0013] Aiming at the disadvantages of dynamic electrocardiogram equipment and technology in the detection of atrial fibrillation, such as serious time lag effect and low detection accuracy due to technical limitations, the present invention can realize long-term continuous detection by using the obtained weight value matrix, and can realize real-time, high-precision Automatically detect atrial fibrillation, provide timely reference for the diagnosis and treatment of patients with atrial fibrillation, and reduce the possibility of high-risk events such as stroke and heart failure caused by atrial fibrillation. Through real-time, continuous ECG signal monitoring and manual Neural network, making predictions and judgments on atrial fibrillation, creating conditions for timely medical intervention, and possibly saving patients' lives

Images