Apnea detection model training method based on time domain and frequency domain generative adversarial network

Abstract

The invention discloses an apnea detection model training method based on a time domain and frequency domain generative adversarial network. The method comprises the following steps: constructing a generative network G and a discrimination network; wherein the discrimination network comprises a time domain discrimination network D1 and a frequency domain discrimination network D2; performing generative adversarial training: training a generative network G through a generative adversarial network based on the time domain discrimination network D1 and the frequency domain discrimination network D2; building a training data set of the apnea detection model; the data set at least comprises an apnea tag vector and a corresponding ventilatory capacity signal; and training the apnea detection model by adopting the established training data set. According to the method, the generation network learning real signal distribution is improved, abundant generation signals with real signal distribution can be generated, the data augmentation capability is greatly improved, high-cost data acquisition and data manual annotation can be greatly saved, accumulated manual errors in the annotation process are avoided, and the annotation efficiency is improved. And accurate, reliable and large-quantity simulation training data can be obtained.

Description

technical field [0001] The invention relates to an apnea detection model training method based on time domain and frequency domain generation confrontation network. Background technique [0002] Currently, Sleep Apnea is a sleep disorder that affects a large population. In medicine, the gold standard for diagnosing apnea is Full-Night Polysomnography. With the deployment of modern medical equipment, the Internet of Things, and medical and health intelligence networks, the huge number of sleep disorder patients has brought about a sharp increase in the amount of data. Compared with the traditional manual standard scoring process that relies on medical experts, real-time, high-precision, Highly automated means of apnea detection and identification are imminent. [0003] In recent decades, researchers have been relentlessly advancing fast, high-accuracy, and automated detection schemes for apnea detection. Benefiting from the vigorous development of artificial intelligence i...

AI Technical Summary

Problems solved by technology

[0007] In general, the existing artificial intelligence sleep apnea detection algorithms rely on manually labeled data with high labor costs, and the internal imbalance of the data weakens the generalization ability and robustness of the detection algorithm.

Images