CN115281651A - Non-inductive integrated sleep respiratory disease diagnosis system - Google Patents

Abstract

The invention provides a non-inductive integrated sleep respiratory disease diagnosis system, which comprises a pre-diagnosis system, a monitoring platform, a data acquisition system and a quantitative evaluation module, wherein the pre-diagnosis system comprises a pre-diagnosis system body and a quantitative evaluation module; the pre-diagnosis system is used for obtaining a pre-inquiry result and feeding the pre-inquiry result back to the monitoring platform; the data acquisition system acquires various electrophysiological signals and wirelessly transmits the electrophysiological signals to the monitoring platform; the monitoring platform is used for receiving and storing data sent by the pre-diagnosis system and the data acquisition system; the quantitative evaluation module is used for constructing a respiratory event automatic diagnosis model based on multi-mode deep learning and automatically outputting a diagnosis result of a respiratory event by using the model. According to the method, a mode of combining software and hardware is adopted, a quantitative evaluation model is finally constructed through non-contact data acquisition, and the quantitative evaluation of the insensitive integrated disease summary exception reminding, diagnosis, classification screening and treatment effectiveness follow-up can be simply, conveniently, efficiently, intelligently and accurately carried out on the sleep respiratory diseases.

Description

Non-inductive integrated sleep respiratory disease diagnosis system

Technical Field

The invention relates to the technical field of disease diagnosis, in particular to a sleep respiratory disease diagnosis system.

Background

Sleep respiratory disease is a key risk factor of cardiovascular and cerebrovascular diseases and metabolic diseases, is an important cause of labor loss and death of middle-aged and elderly people, and generates heavy social and economic burden. The diagnosis and treatment is a problem to be solved urgently in the prevention and treatment of chronic diseases in China. However, the diagnosis and treatment of sleep respiratory diseases require a large amount of manpower and material resources, the average time from the disease symptom to the final diagnosis and treatment in developed countries is more than 4 years, and the number of sleep respiratory disease patients in China exceeds 1.3 hundred million, which is more serious. The sleep respiratory disease gold standard diagnosis relies on a multi-lead and contact type sleep monitoring technology, but the technology has four major bottleneck problems, namely expensive equipment, low popularization rate, high labor cost and poor wearing comfort level, and the diagnosis requirements of a large patient group cannot be met.

The traditional contact type 'gold standard' polysomnography monitor has the following limitations:

firstly, because traditional sleep monitor equipment that leads more is expensive, and the operation is complicated, needs the patient to be in hospital at least and carries out the monitoring night, and the bed is limited, and monitoring cost is higher, and the interpretation of sleep monitor report need consume a large amount of manpowers (every sleep monitor report manual interpretation recheck needs about 1 hour), and artifical chart still has the variation rate that reaches 5-20%.

Secondly, the existing equipment lacks compatibility, and the existing sleep monitoring device cannot automatically interpret sleep monitoring data of different manufacturers and models, so that data of different hospitals and different equipment cannot be shared, which becomes a bottleneck of grading diagnosis and treatment and medical resource and information sharing, and the data acquisition efficiency and the diagnosis timeliness of sleep respiratory disease screening are low. Due to the reasons, a large number of sleep respiratory disease patients cannot be diagnosed in time, and the patients cannot be seen in a hospital until serious complications such as heart disease, diabetes and the like occur to many sleep respiratory disease patients, so that the optimal treatment time is delayed, great harm is caused to the patients, families and society, and huge social and economic burden is caused.

Disclosure of Invention

In order to overcome the technical defects that a sleep monitoring device in the prior art is high in price, low in popularization rate, high in labor cost, poor in wearing comfort level and incapable of meeting the diagnosis requirements of huge patient groups, the invention aims to provide an integrated sleep respiratory disease diagnosis system without induction, which comprises a pre-diagnosis system, a monitoring platform, a data acquisition system and a quantitative evaluation module;

the pre-diagnosis system is used for collecting the illness information of a user based on a disease knowledge base, disease prediction and a multi-turn dialogue module, performing targeted inquiry clarification to obtain available information, converting the available information into a pre-diagnosis report, and feeding back the result of pre-inquiry to the monitoring platform; the multi-turn dialogue module comprises an account management module, a corpus management module, a configuration module, a human-computer interaction management module and a training visualization module;

the account management module: the account management module comprises the following scenes: user registration, user login, user logout (logout), user password modification, user password recovery, user safe mailbox modification, user grouping checking and the like;

the corpus management module: the retrieval modes are various, complex retrieval expressions are supported, the corpus range can be set, retrieval in a specific field, user-defined retrieval/duration retrieval are realized, the expected scale is large, and the types are rich;

a configuration module: relative independence, interchangeability, and versatility;

the man-machine interaction management module: the man-machine interaction management module mainly comprises the following components: multi-modal input/output, visual synthesis, dialog systems, knowledge processing, intelligent interface agents;

training a visualization module: the data information is updated in real time, and the displayed multidimensional and multi-data integration optimization is met;

the data acquisition system comprises but is not limited to an ultra-wideband biological radar wave detector, an RGB-D-camera, a depth sensor, a temperature sensor, a wireless snore detector (for example, a first non-intrusive intelligent snore relieving device-Nora all over the world invented by a company in san Francisco, USA) and a single-chip electroencephalogram sensor, wherein the ultra-wideband biological radar wave detector is used for detecting chest and abdomen movement and heart rate and wirelessly transmitting the chest and abdomen movement and heart rate to a monitoring platform; the RGB-D-camera is used for collecting face video data, estimating the blood oxygen saturation by using a dual-wavelength method and wirelessly transmitting the blood oxygen saturation to the monitoring platform; the depth sensor is used for capturing limb movement signals during sleep and wirelessly transmitting the signals to the monitoring platform; the temperature sensor is used for capturing the temperature change of the inspiratory phase and the expiratory phase of the patient, detecting a breathing signal and wirelessly transmitting the breathing signal to the monitoring platform; the wireless snore detector is used for continuously collecting snore signals all night; the single-chip electroencephalogram sensor comprises a wireless electroencephalogram collector and a signal receiving adapter for collecting unbound electroencephalogram signals and transmitting the unbound electroencephalogram signals to the monitoring platform in a wireless mode;

the monitoring platform is used for receiving and storing data sent by the pre-diagnosis system and the data acquisition system;

the quantitative evaluation module is used for taking all data (such as pre-inquiry information, chest and abdomen movement, heart rate, blood oxygen saturation, respiratory movement, snore, electroencephalogram and other signals) stored by the monitoring platform as input signals of the multi-mode deep learning-based model, constructing the multi-mode deep learning-based respiratory event automatic diagnosis model, and automatically outputting a respiratory event diagnosis result by using the model.

The common type of the sleep respiratory disease is obstructive sleep apnea hypopnea syndrome, and the realization process of the respiratory event automatic diagnosis model based on the multi-modal deep learning comprises the following steps: selecting various combination schemes of the acquired signals to jointly construct a multi-modal-based deep learning model, adjusting parameters, searching for an optimal model, comparing model expression results, and selecting a model with better channel quantity and model expression as a multi-modal-based deep learning respiratory event automatic diagnosis model. And (4) obtaining grading of the light, medium and heavy degrees and corresponding analysis results according to the obtained respiratory event statistics. After a complete respiratory event sequence of the patient is obtained, an Apnea-Hypopnea Index (AHI) of the patient, namely the number of occurrences of Apnea and Hypopnea per hour during night sleep, is counted according to the event sequence. In staging diagnosis, OSAHS severity staging is performed by AHI, generally speaking, AHI <5 is considered no OSAHS; 5-Ap AHI-Ap 15 was considered mild OSAHS; 15-Ap AHI-s 30 were considered to be moderate OSAHS; AHI >30 is considered to be severe OSAHS. Thus, we obtained a hierarchical diagnosis of OSAHS.

Further, the disease knowledge base is constructed by structuring medical texts.

Furthermore, the ultra-wideband biological radar wave detector is used for detecting the motion of the chest and the abdomen and the heart rate in a non-contact mode, can penetrate through clothes and bedding of a patient, and is low in transmitting power, low in distance resolution and high in interference resistance, and the transmitting power is less than 1 milliwatt.

High range resolution range: 4cm-75cm;

the anti-interference capability realization process comprises the following steps: techniques for monitoring vital signs using millimeter wave bio-radar technology are still under investigation, with the main challenge being the variation of the reflected signal from person to person. The reflection depends on the skin type, tissue and its composition, the water content in the body and the chemical composition. The ultra-wideband biological radar vital sign detection parameters are different from the principle of continuous waves, when a pulse-mode microwave beam shines on a human body, due to the existence of human body vital movement (respiration, intestinal peristalsis and the like), the repetition period of an echo pulse sequence reflected by the human body is changed, and the repetition period of an echo pulse signal is related to the movement speed and frequency of the human body vital movement. If the pulse sequence (carrying the information related to the measured human body vital movement) is modulated, integrated, expanded and filtered, and then sent to a computer for data processing and analysis, parameters (such as respiration, heart rate and the like) related to the measured human body vital sign are obtained.

Further, the RGB-D-camera is used for collecting face video data; obtaining a pulse wave signal IR of a red channel and a pulse wave signal IG of a green channel according to the video data, respectively calculating a direct current component and an alternating current component of the pulse wave signal IR and the pulse wave signal IG, and estimating the blood oxygen saturation by using a two-wavelength method according to the Lambert-beer law and a light scattering theory.

Further, the temperature sensor is used for capturing the nasal gas temperature change of the inspiratory phase and the expiratory phase of the patient, so as to accurately classify the apnea or hypopnea and calculate the interval time.

Furthermore, the wireless snore detector is used for collecting the snore related signals of the low frequency band and effectively extracting the snore related signal characteristics with high identification degree from the complex environmental noise.

Further, the single-chip electroencephalograph sensor is a TGAM module, and a single EEG electroencephalograph channel of the TGAM module has 3 contact points: EEG (electroencephalogram collection point) REF (reference point) GND (ground point); the battery power supply time is more than 10 hours; the electroencephalogram data are output at 512Hz, the sampling rate is 512Hz, the frequency range is 3Hz-100Hz, and electroencephalogram signals can process and output electroencephalogram band data of different electroencephalogram bands (such as alpha, beta, gamma and the like); the method comprises the steps of collecting EEG signals of an individual in real time, and realizing non-binding EEG collection through a wireless brain wave collector and a signal receiving adapter.

Further, a respiratory event automatic diagnosis model based on multi-modal deep learning is constructed, and the method comprises the following steps:

exemplarily, feature extraction is performed on a two-minute gray feature map using a ResNet50 pre-training model; processing and predicting the time sequence of the extracted features by using a bidirectional long-short term memory network; and carrying out classification by utilizing a full connection layer of the convolutional neural network.

The ResNet50 is a pre-training model based on a large number of pictures, which is used for training when performing feature extraction on a two-minute gray scale feature map. Since the collected signals are timing signals and have certain timing correlation, the collected signals are not directly classified through full connection after the ResNet50 model, but the timing relationship of the collected signals is further analyzed through a bidirectional long-short term memory network (Bi-LSTM). LSTM (Long Short-Term Memory) is a Long Short-Term Memory network, a time-recurrent neural network, suitable for processing and predicting important events with relatively Long intervals and delays in time series. Bi-LSTM can better consider forward and backward data, thereby better processing and prediction. In this problem, the judgment of the respiratory event is the analysis of the channel information on the time sequence, so we access a Bi-LSTM network after the output of the ResNet50 to predict the sequence, i.e. the corresponding respiratory event sequence.

Further, the sleep respiratory disease quantitative evaluation model can be used for performing disease summary exception reminding, diagnosis, classification screening and treatment effectiveness follow-up quantitative evaluation on the sleep respiratory disease.

After the technical scheme is adopted, compared with the prior art, the method has the following beneficial effects:

the system for diagnosing the sleep respiratory disease in the non-inductive integrated manner adopts a mode of combining software and hardware, and finally constructs a quantitative evaluation model through non-contact data acquisition, and the quantitative evaluation model can be used for carrying out non-inductive integrated disease summary exception reminding, diagnosis, classified screening and quantitative evaluation of treatment effectiveness follow-up on the sleep respiratory disease. Finally, the simple, convenient, efficient, intelligent and accurate screening and diagnosis of the sleep respiratory diseases are realized. In addition, the non-inductive integrated sleep-breathing disease diagnosis system is a structural database obtained by structuring medical texts based on a disease knowledge base, so that the compatibility is good, and the data sharing is convenient to realize.

In the field of standard sleep monitoring, the coverage of imported equipment almost reaches 100%. The successful development of the novel sleep monitoring and diagnosing system can reduce the monopoly occupied by the international medical appliance enterprises for a long time, obviously reduce the product price, and in addition, the intelligent sleep respiratory disease diagnosing environment system has low cost and reduces the hospitalizing cost of patients. In addition, since the core technology of the sleep monitoring system is mastered in an external enterprise, it means that information security such as medical diagnosis, treatment data, and patient medical records face certain risks. The successful development of the sleep breathing diagnostic system can reduce the risk of the medical information. The successful development of the sleep monitoring and diagnosing environmental system adopting the novel technology overcomes the core component and standard problem in the sleep respiration monitoring field, advances the localization of system equipment, reduces the cost of the domestic novel sleep monitoring system, and more importantly breaks the monopoly of foreign brands in the key field of the sleep monitoring system.

Drawings

FIG. 1 is a flow chart for constructing the integrated sensorless sleep-disordered breathing diagnostic system of the present application;

FIG. 2 is a structural diagram of the present application of an integrated system for diagnosing sleep disordered breathing;

fig. 3 is a schematic diagram illustrating a construction principle of a respiratory event automatic diagnosis model based on multi-modal deep learning according to the present application.

Detailed Description

The advantages of the invention are further illustrated by the following detailed description of the preferred embodiments in conjunction with the drawings. It is to be understood by persons skilled in the art that the following detailed description is illustrative and not restrictive, and is not to be taken as limiting the scope of the invention.

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

The terminology used in the disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used in this disclosure and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

The word "if" as used herein may be interpreted as "at" \8230; "or" when 8230; \8230; "or" in response to a determination ", depending on the context. In the description of the present invention, unless otherwise specified and limited, it is to be noted that the terms "mounted," "connected," and "connected" are to be interpreted broadly, and may be, for example, a mechanical connection or an electrical connection, a communication between two elements, a direct connection, or an indirect connection via an intermediate medium, and specific meanings of the terms may be understood by those skilled in the art according to specific situations.

In the following description, suffixes such as "module", "part", or "unit" used to indicate elements are used only for facilitating the description of the present invention, and do not have a specific meaning per se. Thus, "module" and "component" may be used in a mixture.

Example construction of an inductively integrated sleep-respiratory disease diagnostic System

As shown in fig. 1, the construction process of the integrated non-inductive sleep-induced respiratory disease diagnosis system mainly includes the following steps:

1) Constructing a pre-diagnosis system for the interrogation of sleep respiratory diseases

Constructing a pre-interrogation system of the sleep respiratory disease interrogation knowledge map. The method is characterized in that a disease knowledge base is structured by using texts and data, based on a disease prediction strategy, the disease information of a user is acquired in a multi-turn dialogue mode, the information is specifically asked and clarified, the available information is converted into a pre-diagnosis report to be stored, then the intelligent pre-diagnosis and the feedback of the result are integrated into an intelligent pre-diagnosis system, finally, the related research of dialogue robot interface configuration is carried out, and a one-stop auxiliary system for creating, training and testing the functions of a department inquiry assistant is provided. Based on the technologies of medical text structuring, knowledge base construction, natural language analysis, dialogue management, disease prediction analysis, text generation aiming at a pre-diagnosis report, system construction and the like, 1) an intelligent pre-inquiry assistant is designed and realized. 2) Based on the intelligent assistant, an intelligent pre-interrogation system based on the sleep disease knowledge map is designed and realized. 3) The configurable multi-turn dialogue system for providing assistant creation service for the pre-inquiry system is designed and realized and comprises an account management module, a corpus management module, a robot configuration module, a human-computer interaction management module and a training visualization module. Finally, an intelligent pre-inquiry system supporting user guidance, disease prediction and professional vocabulary explanation is realized based on a multi-turn conversation technology, and intelligent pre-inquiry entrance and feedback services are provided for patient users. And a configurable multi-turn dialogue pre-diagnosis system provides an interfacing one-stop intelligent assistant creating service for an intelligent pre-inquiry system or other field projects. The results of the pre-interrogation are then fed back to the monitoring platform.

2) Constructing a data acquisition system

The sleep somatosensory detection technology based on visual drive builds a sleep monitoring intelligent environment by using multi-dimensional zero-load intelligent sensor technologies such as ultra-wideband biological radar, RGB-D-cameras, depth sensors, temperature sensors and wearable sensors, and comprises the following steps: 1) The ultra-wideband biological radar is used for detecting the motion and the heart rate of the chest and abdomen, the equipment has non-contact property (no binding feeling and suitability for long-time monitoring), ultra-strong penetrability (no contact with skin and penetrability of clothes and bedding), low power (low transmitting power which is generally less than 1 milliwatt); high resolution (capable of identifying micro motion and accurately positioning) and strong anti-interference capability. 2) RGB-D-camera: collecting face video data; obtaining a red channel pulse wave signal IR and a green channel pulse wave signal IG according to the video data; respectively calculating direct current components and alternating current components of the pulse signal IR and the pulse wave signal IG, and estimating the blood oxygen saturation by using a dual-wavelength method according to the Lambert-beer law and the light scattering theory; 3) The depth sensor captures limb movement during sleep; 4) Temperature sensors are used for apnea/hypopnea detection and classification: the nasal gas temperature change of the inspiratory phase and the expiratory phase of a patient can be sensitively captured, so that the classification of apnea/hypopnea and the calculation of interval time can be accurately carried out; 5) Non-contact wireless transmission snore detection: the snore signal can be collected continuously overnight, the sensitivity is high, the snore related signals of low frequency bands can be collected, the characteristics of the snore related signals with high recognition degree are effectively extracted from complex environmental noises, and a snore signal related sleep disease diagnosis model is effectively established; 6) Wireless electroencephalogram monitoring is used for sleep staging: the most advanced integrated single-chip electroencephalogram sensor TGAM module in the world at present is adopted. There are 3 contact points for a single EEG brain electrical channel: EEG (electroencephalogram acquisition point) REF (reference point) GND (ground point); the advanced noise filtering technology is adopted, the energy consumption is low, and the battery is used for supplying power (the working time is more than 10 hours); the electroencephalogram data are output at 512Hz, the sampling rate is 512Hz, and the frequency range is as follows: 3Hz-100Hz. The brain wave signals can process and output brain wave band data such as alpha, beta, gamma and the like; and collecting the EEG signals of the individual electroencephalograms in real time. The wireless brain wave collector and the signal receiving adapter are used for realizing the non-binding brain wave collection. And then all the acquired signals are wirelessly transmitted to a monitoring platform.

3) Quantitative evaluation module for constructing sleep respiratory diseases by integrating core algorithm of multi-modal intelligent environment signals

The quantitative evaluation module integrates data captured by the multi-channel intelligent sensor group, and performs multi-mode fusion analysis and annotation on the data to develop an accurate sleep respiratory disease diagnosis machine learning model and construct the quantitative evaluation module for the sleep respiratory disease.

As shown in fig. 2, the system for diagnosing an integrated sleep-disordered breathing includes a pre-diagnosis system, a monitoring platform, a data acquisition system and a quantitative evaluation module.

The pre-diagnosis system is used for collecting disease information of a user based on a disease knowledge base, disease prediction and a multi-turn dialogue module, performing targeted inquiry and clarification to obtain available information, converting the available information into a pre-diagnosis report, and feeding back a pre-inquiry result to the monitoring platform; the multi-turn dialogue module comprises an account management module, a corpus management module, a configuration module, a human-computer interaction management module and a training visualization module.

The data acquisition system comprises an ultra-wideband biological radar wave detector, an RGB-D-camera, a depth sensor, a temperature sensor, a wireless snore detector and a single-chip electroencephalogram sensor, wherein the ultra-wideband biological radar wave detector is used for detecting chest and abdomen movement and heart rate and wirelessly transmitting the chest and abdomen movement and heart rate to a monitoring platform; the RGB-D-camera is used for collecting face video data, estimating the blood oxygen saturation degree by using a dual-wavelength method and wirelessly transmitting the data to the monitoring platform; the depth sensor is used for capturing limb movement signals during sleep and wirelessly transmitting the signals to the monitoring platform; the temperature sensor is used for capturing the temperature change of the inspiratory phase and the expiratory phase of the patient, detecting the breathing signal and wirelessly transmitting the breathing signal to the monitoring platform; the wireless snore detector is used for continuously collecting snore signals all night; the single-chip electroencephalogram sensor comprises a wireless electroencephalogram collector and a signal receiving adapter for collecting the unbound electroencephalogram signals and transmitting the unbound electroencephalogram signals to the monitoring platform in a wireless mode.

The monitoring platform is used for receiving and storing data sent by the pre-diagnosis system and the data acquisition system.

The quantitative evaluation module is used for taking all data stored by the monitoring platform as input signals of the multi-mode deep learning-based model, constructing the multi-mode deep learning-based respiratory event automatic diagnosis model, and automatically outputting the diagnosis result of the respiratory event by using the model.

As shown in fig. 3, the construction of the respiratory event automatic diagnosis model based on multi-modal deep learning comprises the following steps: performing feature extraction on the gray level feature map of two minutes by using a ResNet50 pre-training model; processing and predicting the time sequence of the extracted features by using a bidirectional long-short term memory network; and carrying out classification by utilizing a full connection layer of the convolutional neural network. The ResNet50 is a pre-training model based on a large number of pictures, which is used for training when performing feature extraction on a two-minute gray scale feature map. Since the collected signals are timing signals and have certain timing correlation, the collected signals are not directly classified through full connection after the ResNet50 model, but the timing relationship of the collected signals is further analyzed through a bidirectional long-short term memory network (Bi-LSTM). LSTM (Long Short-Term Memory) is a Long Short-Term Memory network, a time-cycle neural network, and is suitable for processing and predicting important events with relatively Long intervals and delays in time sequences. Bi-LSTM can better consider forward and backward data, thereby better processing and prediction. In this problem, the judgment of the respiratory event is the analysis of the channel information on the time sequence, so we access a Bi-LSTM network after the output of the ResNet50 to predict the sequence, i.e. the corresponding respiratory event sequence.

Multi-scene clinical application of effect example non-inductive integrated sleep respiratory disease diagnosis system

The obtained non-inductive integrated sleep respiratory disease diagnosis system can be developed through a three-level hospital sleep center to carry out popularization test research, realize clinical application demonstration, carry out comparative analysis with results of various corresponding lead data of standard sleep monitoring, summarize and analyze data of various sleep monitoring leads such as electroencephalogram, respiration, chest and abdomen movement, snore, blood oxygen and the like, and establish a quantitative evaluation model for sleep respiratory disease diagnosis, classification and screening based on novel zero-load equipment.

Examples of clinical applications are as follows: a suspected sleep disordered breathing patient is treated by ear, nose, throat, head and neck surgery in our hospital, and firstly comes to a nurse station, and the intelligent inquiry and pre-diagnosis system is used for service, and the pre-diagnosis system can sequentially ask the following questions:

1) Is snoring? None; 1 to 5 years (2) 5 to 10 years (3) 10 to 15 years (4) more than 15 years

2) Is there a breath hold? None; is provided with

3) Daily sleep time (last seven days)?

4) Is there palpitations? None; is provided with

5) Is there a nasal obstruction? None; is provided with

6) Is there headache and weakness after morning onset? None; is provided with

7) Is there hypertension? None; comprises the following steps of; is not aware of

8) Is heart disease diagnosed? None; and if yes, finding time:

9) Is there diabetes? None; is provided with

10 Degree of lethargy while sitting (0; 1;2; 3)

11 Degree of sleepiness while watching television (0; 1;2; 3)

12 When sitting still in public places (e.g. theatres and conferences) (0; 1;2; 3)

13 Continuous riding for one hour without rest (0; 1;2; 3)

14 When lying down for rest afternoon (0; 1;2; 3)

15 When sitting talking to others (0; 1;2; 3)

16 After lunch (no alcohol drinking) sitting still (0; 1;2; 3)

17 When a taxi meets traffic jam and stops for several minutes (0; 1;2; 3)

After the inquiry is finished, the pre-diagnosis system can automatically analyze and output a sleep respiratory disorder disease severity report to prompt the patient whether to carry out further sleep monitoring, and if so, the patient needs to enter an intelligent environment system (namely a data acquisition system) for the diagnosis of the insensitive sleep respiratory disorder disease for sleep respiratory monitoring all night.

The suspicious patient of sleep disorder disease begins to carry out the non-inductive sleep monitoring about 20 pm, and the monitoring signals comprise: performing chest and abdomen exercise; electrocardio; heart rate; (ii) a blood oxygen saturation level; physical movement; oronasal airflow; multi-channel physiological signals such as brain electricity and the like.

The suspected patient can get up about 6. And then, inputting the data stored in the monitoring platform into a quantitative evaluation model of a quantitative evaluation module, automatically outputting a diagnosis report result after about 3 minutes, and making a final diagnosis by a doctor according to the diagnosis report result.

Illustratively, as in tables 1-3, the contents of the sleep monitoring diagnostic report are as follows, including: general information, sleep stages,

The contents of three portions of a sleep apnea event.

TABLE 1 general information

Patient name	Honor a certain point	Record number	20201100
				Sex	For male	Type (B)	Adult
Date of birth	1956-1-1	Starting time	20:20:10
				Age of patient	60	End time	6:02:10
Number of hospitalization	123456	Duration of time	9

TABLE 2 sleep staging

Note: english abbreviation explanation

TIB: time of bed rest

SPT: sleep time (with arousal)

TST: sleep time (without arousal)

WK (SPT): SPT phase arousal

WK (TIB): micro-arousal during the TIB phase

REM: rapid eye movement sleep period

S1: non-rapid eye movement sleep stage 1

S2: non-rapid eye movement sleep stage 2

S3: non-rapid eye movement sleep stage 3

S4: non-rapid eye movement sleep stage 4

MVT: period of leg exercise

TABLE 3 sleep apnea events

TABLE 3-1 Central apnea

	In all	With a decrease in heart rate	With a decrease in blood oxygen
				Total number of times	77	7	75
Maximum time (second)	19.0	17.5	19.0

TABLE 3-2 obstructive apnea

	In all	With a decrease in heart rate	With a decrease in blood oxygen
				Total number of times	14	0	12
Maximum time (second)	21.5	0.0	16.0

TABLE 3-3 Combined apneas

	In all	With a decrease in heart rate	With blood oxygen decline
				Total number of times	85	11	81
Maximum time (second)	21.5	19.0	21.5

TABLE 3-4 Low ventilation

	In all	With a decrease in heart rate	With a decrease in blood oxygen
				Total number of times	147	2	126
Maximum time (second)	33.0	15.0	33.0

TABLE 3-5 respiratory disorder indices

Staging of sleep	REM	NREM	In total
				Disorder index (#/h)	47.5	39.5	41.1

In conclusion, the application constructs the non-inductive integrated sleep respiratory disease diagnosis system by combining software and hardware. The utility model provides a no induction formula integration sleep respiratory disease diagnostic system possesses the analytical ability equivalent with clinician, can provide quick intelligent supplementary for the sleep monitor to the omnidirectional monitoring sleep condition, and reduce doctor's work burden.

It should be noted that the embodiments of the present invention have been described in terms of preferred embodiments, and not by way of limitation, and that those skilled in the art can make modifications and variations of the embodiments described above without departing from the spirit of the invention.

Claims (9)

1. An integrated non-inductive sleep-breathing disease diagnosis system is characterized by comprising a pre-diagnosis system, a monitoring platform, a data acquisition system and a quantitative evaluation module;

the pre-diagnosis system is used for collecting the illness information of a user based on a disease knowledge base, disease prediction and a multi-turn dialogue module, performing targeted inquiry clarification to obtain available information, converting the available information into a pre-diagnosis report, and feeding back the result of pre-inquiry to the monitoring platform; the multi-turn dialogue module comprises an account management module, a corpus management module, a configuration module, a human-computer interaction management module and a training visualization module;

the data acquisition system comprises an ultra-wideband biological radar wave detector, an RGB-D-camera, a depth sensor, a temperature sensor, a wireless snore detector and a single-chip electroencephalogram sensor, wherein the ultra-wideband biological radar wave detector is used for detecting chest and abdomen movement and heart rate and wirelessly transmitting the chest and abdomen movement and heart rate to a monitoring platform; the RGB-D-camera is used for collecting face video data, estimating the blood oxygen saturation degree by using a dual-wavelength method and wirelessly transmitting the data to the monitoring platform; the depth sensor is used for capturing limb movement signals during sleep and wirelessly transmitting the signals to the monitoring platform; the temperature sensor is used for capturing the temperature change of the inspiratory phase and the expiratory phase of the patient, detecting a breathing signal and wirelessly transmitting the breathing signal to the monitoring platform; the wireless snore detector is used for continuously collecting snore signals all night; the single-chip electroencephalogram sensor comprises a wireless electroencephalogram collector and a signal receiving adapter for collecting unbound electroencephalogram signals and transmitting the unbound electroencephalogram signals to the monitoring platform in a wireless mode;

the monitoring platform is used for receiving and storing data sent by the pre-diagnosis system and the data acquisition system;

the quantitative evaluation module is used for taking all data stored by the monitoring platform as input signals of the multi-mode deep learning-based model, constructing the multi-mode deep learning-based respiratory event automatic diagnosis model, and automatically outputting the diagnosis result of the respiratory event by using the model.

2. The system of claim 1, wherein the disease knowledge base is constructed by structuring medical text.

3. The system of claim 1, wherein the UWB biometric radar detector is configured to detect thoracoabdominal movements and heart rate in a non-contact manner, and is capable of penetrating the patient's clothing and bedding, with a transmission power of less than 1 mW, high range resolution, and high interference rejection.

4. The system of claim 1, wherein the RGB-D-camera is configured to capture facial video data; obtaining a pulse wave signal IR of a red channel and a pulse wave signal IG of a green channel according to the video data, respectively calculating a direct current component and an alternating current component of the pulse wave signal IR and the pulse wave signal IG, and estimating the blood oxygen saturation by using a dual-wavelength method according to the Lambert-beer law and a light scattering theory.

5. The system of claim 1, wherein the temperature sensor is configured to capture the temperature changes of the patient's inspiratory and expiratory nasal gases, so as to accurately classify apnea or hypopnea and calculate the time interval.

6. The system of claim 1, wherein the wireless snore detector is configured to collect low-frequency snore-related signals and efficiently extract snore-related signal features with high recognition from complex environmental noise.

7. The system of claim 1, wherein the single-chip EEG sensor is a TGAM module having 3 contacts for a single EEG channel: the electroencephalogram collection point, the reference point and the ground wire point; the battery power supply time is more than 10 hours; the electroencephalogram data are output at 512Hz, the sampling rate is 512Hz, the frequency range is 3Hz-100Hz, and electroencephalogram signals can process and output different electroencephalogram band data; the method is characterized by collecting EEG signals of individuals in real time and realizing the unbound EEG collection through a wireless brain wave collector and a signal receiving adapter.

8. The system of claim 1, wherein the method for constructing the respiratory event automatic diagnosis model based on multi-modal deep learning comprises the following steps: performing feature extraction on the gray level feature map by using a ResNet50 pre-training model; processing and predicting the time sequence of the extracted features by using a bidirectional long-short term memory network; classification of respiratory events using the fully connected layer of the convolutional neural network.

9. The system of any one of claims 1-8, wherein the quantitative assessment model of sleep-disordered breathing can be used for quantitative assessment of sleep-disordered breathing for summary disease exception alert, diagnosis, classification screening, and treatment effectiveness follow-up.

Images