Sleep apnea hypopnea syndrome evaluation method and device based on target detection framework

Abstract

The invention provides a sleep apnea hypopnea syndrome (SAHS) evaluation method and device based on a target detection framework, wherein mouth-nose airflow and chest pressure data in sleep monitoring data are selected as data basis of SAHS evaluation, a target detection model fusing the mouth-nose airflow and the chest pressure data is designed, and the model comprises three main parts: respectively processing mouth-nose airflow and chest pressure data in a time sequence form, and fusing extracted features; carrying out adaptive regression on the basis of a region proposal network (RPN) to generate a candidate frame for SAHS fragment detection; and generating scale-invariant features based on the candidate segments and classifying the scale-invariant features. According to the invention, model training adopts an alternate training mode, and a RPN network and a classification network are trained respectively; and a focal loss function is used to alleviate a deviation problem possibly caused by model training under an unbalanced sample.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a method and device for evaluating sleep apnea hypopnea syndrome based on a target detection framework. Background technique [0002] Sleep apnea-hypopnea syndrome (SAHS) refers to the clinical syndrome of recurrent apnea and / or hypopnea, hypercapnia, and sleep interruption caused by various reasons in the sleep state, resulting in a series of pathophysiological changes in the body. It is a typical symptom of sleep disorders. The sleep apnea-hypopnea syndrome assessment method is to assess the sleep apnea-hypopnea syndrome through the physiological signals (such as polysomnography) recorded during sleep, and the results are used as an important reference for the diagnosis of sleep disorders. [0003] Traditional SAHS evaluation methods often use template matching to find the salient features of SAHS fragments and calibrate the fragments that meet the standards; however, this t...

AI Technical Summary

Problems solved by technology

[0003] Traditional SAHS evaluation methods often use template matching to find the salient features of SAHS fragments and calibrate the fragments that meet the standards; however, this type of evaluation algorithm depends on specific tasks, is highly customizable, and is very difficult to transfer between different signal fragment recognition tasks Big

The classification algorithm based on feature engineering or feature learning can be applied to different segment recognition tasks. This type of algorithm obtains a feature set through manual feature extraction or feature learning, and then determines whether the input segment belongs to a specific signal segment through classification; but based on A key problem in the classification algorithm is that it can only determine the type of input fragments, and cannot solve the problem of positioning specific fragments; and because the length of SAHS fragments is uncertain, the sliding window algorithm based on a single fixed-length time window It will also destroy the integrity of the real fragment and cannot meet the needs of precise positioning of the fragment

Images