CN111035395A

CN111035395A - Blood oxygen saturation signal analysis method and device, computer equipment and storage medium

Info

Publication number: CN111035395A
Application number: CN202010003990.XA
Authority: CN
Inventors: 关建; 殷善开; 易红良; 盛斌; 李华婷; 刘茹涵; 许华俊
Original assignee: Shanghai Jiaotong University; Shanghai Sixth Peoples Hospital
Current assignee: Shanghai Jiaotong University; Shanghai Sixth Peoples Hospital
Priority date: 2020-01-03
Filing date: 2020-01-03
Publication date: 2020-04-21
Anticipated expiration: 2040-01-03
Also published as: CN111035395B

Abstract

The application relates to a method and a device for analyzing a blood oxygen saturation signal, a computer device and a storage medium. The method comprises the following steps: acquiring a blood oxygen saturation signal to be analyzed; determining an oxygen subtraction signal segment corresponding to an oxygen subtraction event in the blood oxygen saturation signal; extracting local features of the oxygen subtracted signal segments; fusing the local features and the global features of the blood oxygen saturation signals to obtain fused features; and performing oxygen subtraction type identification on the fusion characteristics to obtain an oxygen subtraction type identification result of the oxygen subtraction signal segment. By adopting the method, the analysis efficiency of the blood oxygen saturation degree signal can be improved.

Description

Blood oxygen saturation signal analysis method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for analyzing a blood oxygen saturation signal, a computer device, and a storage medium.

Background

Sleep Apnea Hypopnea Syndrome (SAHS) is a common Sleep-related breathing disorder characterized by multiple reductions and cessation of airflow during Sleep, accounting for about 49.7% of men and 23.4% of women adult suffering from Sleep disordered breathing. The most common standard for diagnosing SAHS is the nocturnal Polysomnography (PSG), in which physiological signals such as oronasal airflow, blood oxygen saturation (SpO2), electrocardiogram and sleep state are recorded, and the SAHS can be identified based on the various physiological signals recorded by the PSG. The oxyhemoglobin saturation signal can be measured by a wearable pulse oxyhemoglobin saturation instrument, the signal is easy to collect, many apnea events are related to obvious oxygen saturation reduction, and the oxyhemoglobin saturation signal gradually becomes a common index for diagnosing SAHS.

In the diagnosis of SAHS, accurate statistical analysis of apnea events is required, and the hypoxemia event, i.e. the decrease in oxygen saturation, recorded by the oximetry signal is not necessarily related to apnea events. At present, a blood oxygen saturation signal analysis method for analyzing whether an apnea event occurs by dividing a blood oxygen saturation signal into intervals of one to two minutes can only determine whether the apnea event occurs in each segmented interval, cannot accurately determine an oxygen reduction signal segment corresponding to each apnea event, and needs a doctor to perform manual judgment, so that the blood oxygen saturation signal analysis efficiency is low.

Disclosure of Invention

In view of the above, it is necessary to provide a blood oxygen saturation signal analysis method, a device, a computer device, and a storage medium capable of improving the efficiency of blood oxygen saturation signal analysis in view of the above technical problems.

A method of oximetry signal analysis, the method comprising:

acquiring a blood oxygen saturation signal to be analyzed;

determining an oxygen subtraction signal segment corresponding to an oxygen subtraction event in the blood oxygen saturation signal;

extracting local features of the oxygen subtracted signal segments;

fusing the local features and the global features of the blood oxygen saturation signals to obtain fused features;

and performing oxygen subtraction type identification on the fusion characteristics to obtain an oxygen subtraction type identification result of the oxygen subtraction signal segment.

In one embodiment, determining the oxygen desaturation signal segment corresponding to the oxygen desaturation event in the blood oxygen saturation signal comprises:

performing wavelet denoising processing on the blood oxygen saturation signal based on a wavelet algorithm to obtain a denoised blood oxygen saturation signal;

smoothing the denoised oxyhemoglobin saturation signal to obtain a preprocessed oxyhemoglobin saturation signal;

determining an oxygen reduction event from the pre-processed blood oxygen saturation signal according to the amplitude change of the pre-processed blood oxygen saturation signal;

an oxygen subtracted signal segment corresponding to the oxygen subtracted event is determined from the blood oxygen saturation signal.

In one embodiment, extracting the local features of the oxygen subtracted signal segment comprises:

extracting an oxygen-subtracted signal segment from the blood oxygen saturation signal;

performing time sequence feature extraction on the oxygen subtraction signal fragment to obtain the time sequence feature of the oxygen subtraction signal fragment;

and carrying out convolution processing on the time sequence characteristics of the oxygen subtraction signal fragment to obtain the local characteristics of the oxygen subtraction signal fragment.

In one embodiment, fusing the local features and the global features of the oximetry signal to obtain fused features comprises:

acquiring global characteristics of a blood oxygen saturation signal; the global feature is obtained by performing global feature analysis on the blood oxygen saturation signal;

and fusing the local features and the global features to obtain fused features.

In one embodiment, the global features include at least one of mean, standard deviation, area under the curve, mean frequency, peak frequency, and center frequency of the blood oxygen saturation signal.

In one embodiment, the blood oxygen saturation signal analysis method is implemented based on a pre-trained oxygen reduction signal analysis model, and the training step of the oxygen reduction signal analysis model comprises the following steps:

acquiring an original blood oxygen saturation signal;

determining an original oxygen subtraction signal segment corresponding to an oxygen subtraction event in the original blood oxygen saturation signal, and obtaining an oxygen subtraction signal training segment carrying an oxygen subtraction type label according to the original oxygen subtraction signal segment;

extracting local training characteristics of an oxygen subtraction signal training fragment through an oxygen subtraction signal analysis model to be trained;

fusing the local training characteristics and the global training characteristics of the oxyhemoglobin saturation training signal through an oxygen reduction signal analysis model to be trained to obtain fused training characteristics;

performing oxygen reduction type recognition on the fusion training characteristics through an oxygen reduction signal analysis model to be trained to obtain an oxygen reduction type training recognition result of the blood oxygen saturation training signal;

and determining model loss according to the oxygen subtraction type training recognition result and the oxygen subtraction type label, adjusting the oxygen subtraction signal analysis model to be trained according to the model loss, and continuing training until the training ending condition is met, so as to obtain the trained oxygen subtraction signal analysis model.

In one embodiment, the oxygen reduction type tags include an apnea oxygen reduction type tag and a non-apnea oxygen reduction type tag; obtaining an oxygen subtraction signal training fragment carrying an oxygen subtraction type label according to the original oxygen subtraction signal fragment:

determining an oxygen subtraction event label corresponding to the original oxygen subtraction signal segment;

determining a positive sample signal segment and a negative sample signal segment from the original oxygen subtraction signal segment according to the oxygen subtraction event label; the positive sample signal segment carries an apnea oxygen reduction type label, and the negative sample signal segment carries a non-apnea oxygen reduction type label;

and obtaining an oxygen subtraction signal training segment according to the positive sample signal segment and the negative sample signal segment.

An oximetry signal analysis device, the device comprising:

the system comprises a to-be-analyzed signal acquisition module, a to-be-analyzed signal acquisition module and a to-be-analyzed signal analysis module, wherein the to-be-analyzed signal acquisition module is used for acquiring a to-be-analyzed oxyhemoglobin saturation signal;

the oxygen reduction signal determining module is used for determining an oxygen reduction signal segment corresponding to an oxygen reduction event in the blood oxygen saturation signal;

the local feature extraction module is used for extracting the local features of the oxygen subtraction signal segments;

the characteristic fusion processing module is used for fusing the local characteristic and the global characteristic of the oxyhemoglobin saturation signal to obtain a fusion characteristic;

and the oxygen subtraction type identification module is used for carrying out oxygen subtraction type identification on the fusion characteristics to obtain an oxygen subtraction type identification result of the oxygen subtraction signal segment.

A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

acquiring a blood oxygen saturation signal to be analyzed;

extracting local features of the oxygen subtracted signal segments;

A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:

acquiring a blood oxygen saturation signal to be analyzed;

extracting local features of the oxygen subtracted signal segments;

The oxyhemoglobin saturation signal analysis method, the oxyhemoglobin saturation signal analysis device, the computer equipment and the storage medium determine an oxyhemoglobin saturation signal segment corresponding to an oxyhemoglobin saturation event in an oxyhemoglobin saturation signal to be analyzed, fuse local features of the oxyhemoglobin saturation signal segment and global features of the oxyhemoglobin saturation signal, and perform oxyhemoglobin saturation type recognition on the obtained fusion features to obtain an oxyhemoglobin saturation type recognition result. In the analysis and processing of the oxyhemoglobin saturation signal, the oxygen reduction type recognition is carried out based on the fusion characteristics obtained by fusing the global characteristics of the oxyhemoglobin saturation signal and the local characteristics of the oxygen reduction signal segments, the global characteristics of the signal are fully considered, the oxygen reduction type recognition result of the oxygen reduction signal segments in the oxyhemoglobin saturation signal can be accurately obtained, further manual judgment by a doctor is not needed, and the analysis efficiency of the oxyhemoglobin saturation signal is improved.

Drawings

FIG. 1 is a diagram illustrating an exemplary embodiment of a method for analyzing a blood oxygen saturation signal;

FIG. 2 is a schematic flow chart diagram illustrating a method for analyzing a blood oxygen saturation signal according to one embodiment;

FIG. 3 is a schematic diagram of a process for determining an oxygen minus signal segment in one embodiment;

FIG. 4 is a waveform diagram illustrating pre-processing of a raw oximetry signal in another embodiment;

FIG. 5 is a diagram illustrating the structure of a convolution module in one embodiment;

FIG. 6 is a schematic diagram of an analysis model of oxygen subtraction signal in one embodiment;

fig. 7 is a block diagram showing the configuration of a blood oxygen saturation signal analyzing apparatus according to an embodiment;

FIG. 8 is a diagram illustrating an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The method for analyzing the blood oxygen saturation signal can be applied to the application environment shown in fig. 1. Wherein the terminal 102 and the server 104 communicate via a network. The terminal 102 collects the oxyhemoglobin saturation signal and sends the oxyhemoglobin saturation signal to the server 104, the server 104 determines an oxyhemoglobin saturation signal segment corresponding to an oxyhemoglobin saturation event in the received oxyhemoglobin saturation signal to be analyzed, and fuses the local feature of the oxyhemoglobin saturation signal segment and the global feature of the oxyhemoglobin saturation signal to perform oxyhemoglobin saturation type recognition on the obtained fusion feature, so as to obtain an oxyhemoglobin saturation type recognition result. Further, the server 104 may acquire the stored blood oxygen saturation signal from the memory alone to perform analysis processing. The terminal 102 may be, but is not limited to, a smart phone, an oximetry device, and a portable wearable device, and the server 104 may be implemented by a separate server or a server cluster composed of a plurality of servers.

In one embodiment, as shown in fig. 2, there is provided a method for analyzing a blood oxygen saturation signal, which is exemplified by the application of the method to the server in fig. 1, and includes the following steps:

in step S202, a blood oxygen saturation signal to be analyzed is acquired.

Here, the blood oxygen saturation is the percentage of the volume of oxygenated hemoglobin (HbO2) bound by oxygen in blood to the volume of total bindable hemoglobin (Hb), i.e., the concentration of blood oxygen in blood, which is an important physiological parameter of the respiratory cycle. The monitoring of the blood oxygen index can well know whether the respiratory system and the immune system of the user are normal or not; the treatment effect is followed by detecting the blood oxygen. Generally, the blood oxygen saturation value of normal people is 94% -100%, and the oxygen supply is insufficient when the blood oxygen saturation value is below 94%. The traditional oxyhemoglobin saturation measuring method is to firstly take blood from a human body, then carry out electrochemical analysis by using a blood gas analyzer, measure the partial pressure of blood oxygen and calculate the oxyhemoglobin saturation, and the method is troublesome and cannot carry out continuous monitoring. When the fingerstall type photoelectric sensor is used for measurement, the sensor is sleeved on a finger of a person, the finger is used as a transparent container for containing hemoglobin, red light with the wavelength of 660nm and near infrared light with the wavelength of 940nm are used as incident light sources, the light transmission intensity passing through a tissue bed is measured, the concentration and the blood oxygen saturation degree are calculated, the blood oxygen saturation degree of the human body can be displayed by the instrument, and the continuous nondestructive blood oxygen measuring instrument is provided for clinic. The blood oxygen saturation signal is the blood oxygen saturation data which is measured by a terminal, such as a blood oxygen saturation tester and needs to be analyzed. The blood oxygen saturation signal can be data measured by the terminal in real time or data stored in a memory.

In step S204, an oxygen-decreasing signal segment corresponding to the oxygen-decreasing event in the blood oxygen saturation signal is determined.

Where an oxygen desaturation event refers to an event in which the blood oxygen saturation is reduced, also called a desaturation event, such as a 3% or 4% reduction in blood oxygen saturation, it is considered that an oxygen desaturation event has occurred. The oxygen decreasing signal segment is the embodiment of the oxygen decreasing event in the blood oxygen saturation signal, and when the oxygen decreasing event occurs, the signal amplitude value in the blood oxygen saturation signal has a correspondingly decreased segment, namely the oxygen decreasing signal segment corresponding to the oxygen decreasing event. In a specific implementation, an oxygen decreasing event may be determined according to a change in the amplitude of the blood oxygen saturation signal, and an oxygen decreasing signal segment corresponding to the oxygen decreasing event may be determined from the blood oxygen saturation signal according to a time point corresponding to the oxygen decreasing event, such as a start and end time point of the oxygen decreasing event. The diagnosis of the SAHS requires analysis and determination of the apnea-related oxygen reduction event, while the oxygen reduction signal segment corresponding to the oxygen reduction event in the embodiment is directly determined from the blood oxygen saturation signal, and oxygen reduction type identification needs to be further performed on the oxygen reduction signal segment to determine whether the oxygen reduction event is the apnea-related oxygen reduction event, where apnea refers to the interruption of respiratory airflow, such as the interruption of adult respiration lasting for at least 10 seconds or the interruption of children lasting for at least 2 breaths, so as to determine each apnea-related oxygen reduction event from the blood oxygen saturation signal, thereby facilitating the subsequent diagnosis and analysis of the SAHS.

In step S206, the local features of the oxygen subtracted signal segment are extracted.

After the oxygen subtraction signal segment corresponding to the oxygen subtraction event is obtained, the local features of the oxygen subtraction signal segment are extracted, specifically, the local feature extraction can be performed through an artificial Neural network, for example, the local feature extraction can be performed through a Convolutional Neural network such as a Bi-directional Long Short-Term Memory Convolutional Neural network (Bi-directional Long Short-Term Memory Convolutional Neural network), and the local features of the oxygen subtraction signal segment are obtained.

And S208, fusing the local features and the global features of the blood oxygen saturation signals to obtain fused features.

And after the local characteristic of the oxygen subtraction signal segment is obtained, fusing the local characteristic of the oxygen subtraction signal segment and the global characteristic of the blood oxygen saturation signal to obtain a fusion characteristic. The global feature of the blood oxygen saturation signal is obtained by performing overall feature analysis on the blood oxygen saturation signal to be analyzed, and may specifically include, but is not limited to, features including an average value, a standard deviation, an Area Under a Curve (AUC), an average frequency, a peak frequency, a center frequency, and the like of the blood oxygen saturation signal. When the local features and the global features of the blood oxygen saturation signal are fused, splicing fusion can be directly performed, or weighted fusion can be performed, for example, according to a first weight corresponding to the preset local features and a second weight corresponding to the global features, the fusion features are obtained through weighted fusion. The fusion characteristics not only comprise the characteristics of the oxygen reduction signal segment, but also comprise the characteristics of the whole segment of the blood oxygen saturation signal, the integral characteristics of the signal can be fully considered, and the accuracy of identifying the oxygen reduction type of the oxygen reduction signal segment is improved.

And step S210, performing oxygen subtraction type identification on the fusion features to obtain an oxygen subtraction type identification result of the oxygen subtraction signal segment.

After the fusion feature corresponding to the oxygen subtraction signal segment is obtained, oxygen subtraction type identification is carried out based on the fusion feature, and the oxygen subtraction type identification result reflects the oxygen subtraction type corresponding to the oxygen subtraction signal segment, and specifically can include an apnea oxygen subtraction type and a non-apnea oxygen subtraction type. The apnea oxygen reduction type is characterized in that an oxygen reduction signal segment is related to apnea and is an apnea-induced oxygen reduction event, and the oxygen reduction signal segment can be used for diagnosis analysis of SAHS; the non-apnea oxygen reduction type represents that the oxygen reduction signal segment is not related to apnea and is not an oxygen reduction event caused by apnea, and the oxygen reduction signal segment is not used for the diagnostic analysis of the SAHS, so that the screening of each oxygen reduction signal segment in the blood oxygen saturation signal is realized, and the accuracy of the subsequent SAHS diagnostic analysis is ensured.

In the method for analyzing the oxyhemoglobin saturation signal, an oxyhemoglobin saturation signal segment corresponding to an oxyhemoglobin saturation event in the oxyhemoglobin saturation signal to be analyzed is determined, local features of the oxyhemoglobin saturation signal segment and global features of the oxyhemoglobin saturation signal are fused, and the obtained fusion features are subjected to oxyhemoglobin saturation type recognition to obtain an oxyhemoglobin saturation type recognition result. In the analysis and processing of the oxyhemoglobin saturation signal, the oxygen reduction type recognition is carried out based on the fusion characteristics obtained by fusing the global characteristics of the oxyhemoglobin saturation signal and the local characteristics of the oxygen reduction signal segments, the global characteristics of the signal are fully considered, the oxygen reduction type recognition result of the oxygen reduction signal segments in the oxyhemoglobin saturation signal can be accurately obtained, further manual judgment by a doctor is not needed, and the analysis efficiency of the oxyhemoglobin saturation signal is improved.

In one embodiment, as shown in fig. 3, determining an oxygen subtracted signal segment corresponding to an oxygen subtracted event in the oximetry signal comprises:

and S302, performing wavelet denoising processing on the blood oxygen saturation signal based on a wavelet algorithm to obtain a denoised blood oxygen saturation signal.

In this embodiment, the oximetry signal is pre-processed to obtain a smoother signal, which helps to make the oxygen reduction event easier to determine, and then the corresponding oxygen reduction signal segment is determined from the oximetry signal according to the determined oxygen reduction event. Specifically, after obtaining the blood oxygen saturation signal, the wavelet denoising processing is performed on the blood oxygen saturation signal based on the wavelet algorithm. During implementation, wavelet decomposition can be carried out on the blood oxygen saturation degree signal through a Daubechies wavelet algorithm, and then threshold filtering and reconstruction are carried out, so that denoising processing is carried out, and the denoised blood oxygen saturation degree signal is obtained.

In one particular application, it is contemplated that the raw SpO2 signal may generate significant noise due to high patient nighttime activity, baseline wander, degradation discontinuities, limited resolution of the blood oxygen sensor, and the like. The approximate effectiveness of Daubechies8 wavelet decomposition is discarded at the 14 th level, and then threshold filtering and reconstruction are carried out to obtain a denoised SpO2 signal, so that the denoising effect is effectively improved.

And step S304, smoothing the denoised oxyhemoglobin saturation signal to obtain a preprocessed oxyhemoglobin saturation signal.

After denoising the oxyhemoglobin saturation signal, smoothing the obtained denoised oxyhemoglobin saturation signal, specifically, smoothing the signal by using a Moving Average (MA) filter with a duration of 3 seconds; meanwhile, in view of the high sampling rate of the oxyhemoglobin saturation signal during acquisition and overlarge data volume, which is inconvenient for signal processing efficiency, the oxyhemoglobin saturation signal is resampled to ensure that the sampling frequency of the oxyhemoglobin saturation signal reaches 1 Hz so as to balance the sampling rate and accelerate the calculation speed; and then, linear interpolation is used for replacing the signal detection saturation degree to be lower than 50% so as to further smooth the signal, smooth processing of the denoised oxyhemoglobin saturation degree signal is realized, and the preprocessed oxyhemoglobin saturation degree signal is obtained.

Fig. 4 is a schematic waveform diagram of a pre-processed blood oxygen saturation signal obtained by pre-processing an original blood oxygen saturation signal, i.e. an oximetry signal in the diagram, in an embodiment. As is apparent from the figure, compared with the original blood oxygen signal, the preprocessed blood oxygen signal after the Daubechies8 wavelet denoising processing, the moving average filter smoothing processing, the resampling processing and the linear interpolation processing are sequentially performed on the original blood oxygen signal is smoother, and the determination of the oxygen reduction event existing in the blood oxygen saturation signal is more facilitated.

In step S306, an oxygen reduction event is determined from the preprocessed blood oxygen saturation signal according to the amplitude variation of the preprocessed blood oxygen saturation signal.

And (3) preprocessing the oxyhemoglobin saturation signal to obtain a preprocessed oxyhemoglobin saturation signal, and determining an oxygen reduction event according to the amplitude change of the preprocessed oxyhemoglobin saturation signal. The oxygen reduction event is associated with a change in the amplitude of the oximetry signal, which is reduced to a degree that the oxygen reduction event is deemed to have occurred. In a specific application, according to the amplitude change of the preprocessed blood oxygen saturation signal, the oxygen reduction event can be determined from the preprocessed blood oxygen saturation signal from the first to the last order according to the monitoring time of the blood oxygen saturation signal. The determined events of oxygen depletion are determined solely from the magnitude of the blood oxygen saturation signal, and whether each event of oxygen depletion is due to an apnea, a further determination is required to determine the events of oxygen depletion associated with the apnea for diagnostic analysis thereof to enable diagnostic analysis of the SAHS.

In specific implementation, the extreme point of the preprocessed blood oxygen saturation signal can be found through derivative filtering, and all peaks and troughs in the signal are obtained; connecting based on successive downward segments to create full desaturation unless the detected peak interval exceeds 30 seconds; extracting a start point and a stop point of the saturation by using an extreme point of an SpO2 signal; if the reduction from the start to the end of desaturation is at least 1% and the total length is less than 120 seconds, the desaturation event is retained because it is typically the longest duration of an apnea event and its corresponding desaturation, thereby determining an oxygen desaturation event from the pre-processed blood oxygen saturation signal. As shown in fig. 4 for the oximetry signal during the period of 500 seconds to 600 seconds, there is a significant decrease and rise in the amplitude of the oximetry signal during which the oxygen decrease event occurs.

In step S308, an oxygen-decreasing signal segment corresponding to the oxygen-decreasing event is determined from the blood oxygen saturation signal.

After determining the oxygen reduction event from the preprocessed blood oxygen saturation signal, determining an oxygen reduction signal segment corresponding to the oxygen reduction event from the blood oxygen saturation signal, so as to perform oxygen reduction type identification on the oxygen reduction signal segment and determine whether the oxygen reduction event is related to apnea. In specific implementation, the time starting point and the time stopping point of the oxygen reduction event are determined, and the oxygen reduction signal segment corresponding to the oxygen reduction event is determined from the time starting point and the time stopping point corresponding to the blood oxygen saturation signal.

In this embodiment, the preprocessed blood oxygen saturation signal obtained by preprocessing the blood oxygen saturation signal is only used for detecting desaturation, that is, detecting an oxygen reduction event, and an oxygen reduction signal segment corresponding to the oxygen reduction event is determined from the blood oxygen saturation signal according to the oxygen reduction event, so that the preprocessing of the signal is prevented from affecting the oxygen reduction type identification of the oxygen reduction signal segment, and the accuracy of the oxygen reduction type identification is ensured.

In one embodiment, extracting local features of the oxygen subtracted signal segment comprises: extracting an oxygen-subtracted signal segment from the blood oxygen saturation signal; performing time sequence feature extraction on the oxygen subtraction signal fragment to obtain the time sequence feature of the oxygen subtraction signal fragment; and carrying out convolution processing on the time sequence characteristics of the oxygen subtraction signal fragment to obtain the local characteristics of the oxygen subtraction signal fragment.

In this embodiment, the oxygen subtraction signal segment is subjected to time sequence feature extraction and then subjected to convolution processing, so as to extract local features of the oxygen subtraction signal segment. Specifically, after determining the subtracted signal segment corresponding to the subtracted event in the oximetry signal, the subtracted signal segment is extracted from the oximetry signal, and specifically, the subtracted signal segment may be extracted from the corresponding time start and end point in the oximetry signal according to the time start and end point of the subtracted event. The time sequence feature extraction is carried out on the oxygen reduction signal segment, and specifically, the time sequence feature of the oxygen reduction signal segment can be obtained through the time sequence feature extraction of Bi-directional Long Short-Term Memory (Bi-LSTM). And performing convolution processing on the time sequence characteristics of the oxygen subtraction signal segments, specifically performing convolution processing through a multilayer convolution-pooling layer structure in a convolution neural network to obtain the local characteristics of the oxygen subtraction signal segments.

In a specific application, the local features of the oxygen-subtracted signal segment are extracted based on the Bi-LSTM-CNN, specifically, the time sequence feature of the oxygen-subtracted signal segment is extracted through a time encoder in the Bi-LSTM-CNN to obtain the time sequence feature of the oxygen-subtracted signal segment, and the time sequence feature of the oxygen-subtracted signal segment is subjected to convolution processing through a convolution module in the Bi-LSTM-CNN to obtain the local features of the oxygen-subtracted signal segment. The time encoder is a layer of bidirectional long and short term memory network, and comprehensively extracts time sequence information in the oxygen reduction segment; the convolution module comprises 4 convolution-maximum pooling modules, wherein each convolution-maximum pooling module is composed of a one-dimensional convolution layer and a maximum pooling layer. Fig. 5 is a schematic diagram of a convolution module in an embodiment, where the convolution module is composed of 4 convolution-max pooling module connections. The method comprises the steps that a Bi-LSTM-CNN model is used for identifying apnea events, a bidirectional short-time memory network is used for obtaining time characteristics by constructing a forward network and a backward network, and then significant information is extracted through a convolutional neural network model, so that the model has good model performance; and the model can avoid complicated manual feature extraction steps, save a large amount of feature extraction and selection time and improve the analysis efficiency of the blood oxygen saturation signal.

In one embodiment, fusing the local features and the global features of the oximetry signal to obtain fused features comprises: acquiring global characteristics of a blood oxygen saturation signal; the global feature is obtained by performing global feature analysis on the blood oxygen saturation signal; and fusing the local features and the global features to obtain fused features.

In this embodiment, global feature analysis is performed on the blood oxygen saturation signal, and a fusion feature is obtained by fusing a local feature according to the obtained global feature. Specifically, when the local features and the global features of the blood oxygen saturation signal are fused, the global features of the blood oxygen saturation signal are obtained, and the global features are obtained by performing global feature analysis on the blood oxygen saturation signal. The global characteristics reflect the overall signal characteristics of the blood oxygen saturation signal, and can effectively reflect the overall severity of the SAHS patient. The local features and the global features are fused, splicing and fusion can be directly performed, and weighted fusion can also be performed, for example, the fusion features are obtained through weighted fusion according to a first weight corresponding to the preset local features and a second weight corresponding to the global features.

In one embodiment, the global features include at least one of a mean, a standard deviation, an area under the curve, a mean frequency, a peak frequency, and a center frequency of the blood oxygen saturation signal.

The global characteristic reflects an overall signal characteristic of the oximetry signal. The average value is obtained by summing and averaging the amplitudes of the oxyhemoglobin saturation signals, the standard deviation is the standard deviation of the amplitudes of the oxyhemoglobin saturation signals, the area under the curve is the area enclosed by the signal curve of the oxyhemoglobin saturation signals and the amplitudes and the time coordinate axis, and the average frequency, the peak frequency and the center frequency are obtained by calculation based on the signal frequency of the oxyhemoglobin saturation signals.

In one embodiment, the blood oxygen saturation signal analysis method is implemented based on a pre-trained oxygen subtraction signal analysis model, and the training step of the oxygen subtraction signal analysis model comprises the following steps: acquiring an original blood oxygen saturation signal; determining an original oxygen subtraction signal segment corresponding to an oxygen subtraction event in the original blood oxygen saturation signal, and obtaining an oxygen subtraction signal training segment carrying an oxygen subtraction type label according to the original oxygen subtraction signal segment; extracting local training characteristics of an oxygen subtraction signal training fragment through an oxygen subtraction signal analysis model to be trained; fusing the local training characteristics and the global training characteristics of the oxyhemoglobin saturation training signal through an oxygen reduction signal analysis model to be trained to obtain fused training characteristics; performing oxygen reduction type recognition on the fusion training characteristics through an oxygen reduction signal analysis model to be trained to obtain an oxygen reduction type training recognition result of the blood oxygen saturation training signal; and determining model loss according to the oxygen subtraction type training recognition result and the oxygen subtraction type label, adjusting the oxygen subtraction signal analysis model to be trained according to the model loss, and continuing training until the training ending condition is met, so as to obtain the trained oxygen subtraction signal analysis model.

In this embodiment, the blood oxygen saturation signal analysis method is implemented based on a pre-trained oxygen subtraction signal analysis model. Specifically, after obtaining a blood oxygen saturation signal to be analyzed and determining an oxygen-decreasing signal segment corresponding to an oxygen-decreasing event in the blood oxygen saturation signal, inputting the oxygen-decreasing signal segment into a pre-trained oxygen-decreasing signal analysis model, extracting a local feature of the oxygen-decreasing signal segment by the oxygen-decreasing signal analysis model, fusing the local feature and a global feature of the blood oxygen saturation signal to obtain a fused feature, finally performing oxygen-decreasing type recognition on the fused feature, and obtaining an oxygen-decreasing type recognition result of the oxygen-decreasing signal segment by the output oxygen-decreasing signal analysis model, such as whether the oxygen-decreasing signal segment corresponds to the oxygen-decreasing event related to apnea.

When the oxygen reduction signal analysis model is trained, an original blood oxygen saturation signal is obtained, and the original blood oxygen saturation signal can be extracted from historical medical data. Determining an original oxygen subtraction signal segment corresponding to an oxygen subtraction event in an original blood oxygen saturation signal, and obtaining an oxygen subtraction signal training segment carrying an oxygen subtraction type label according to the original oxygen subtraction signal segment, specifically labeling each original oxygen subtraction signal segment to obtain an oxygen subtraction signal training segment carrying an oxygen subtraction type label. And determining the positive and negative sample types corresponding to the oxygen subtraction signal training fragments according to the oxygen subtraction type labels, and performing oxygen subtraction signal analysis model training based on the oxygen subtraction signal training fragments.

Specifically, extracting local training characteristics of an oxygen subtraction signal training segment through an oxygen subtraction signal analysis model to be trained; fusing the local training characteristics and the global training characteristics of the oxyhemoglobin saturation training signal through an oxygen reduction signal analysis model to be trained to obtain fused training characteristics; and carrying out oxygen reduction type recognition on the fusion training characteristics through an oxygen reduction signal analysis model to be trained to obtain an oxygen reduction type training recognition result of the blood oxygen saturation training signal. In this embodiment, the oxygen subtraction signal analysis model may be a Bi-LSTM-CNN model, as shown in fig. 6, which includes four parts, respectively being a time encoder: bidirectional long-short term memory (Bi-LSTM); a local feature extractor: a convolution module; fusing global features; and fully connected and linear regressors. The time encoder is a layer of bidirectional long and short term memory network, and comprehensively extracts time sequence information in the oxygen reduction segment; the convolution module is a convolution neural network module and comprises 4 convolution-maximum pooling modules, wherein each convolution-maximum pooling module consists of a one-dimensional convolution layer and a maximum pooling layer; in the global feature fusion, the features of the average value, the standard deviation, the area under the curve, the average frequency, the peak frequency, the central frequency and the like of the signal at night are taken as global features, and the global features are spliced with local features extracted from the network to be taken as fusion features; finally, this fused feature will go back and forth through one full link layer to the final result: whether it is an apnea-related oxygen-minus segment.

After the oxygen reduction type training recognition result is obtained, determining model loss according to the oxygen reduction type training recognition result and the oxygen reduction type label, for example, determining loss based on cross entropy, adjusting the oxygen reduction signal analysis model to be trained according to the model loss, and continuing training until the training ending condition is met, and ending the training if the loss function meets the convergence condition or the training frequency reaches the training frequency threshold value to obtain the trained oxygen reduction signal analysis model. The oxygen reduction signal analysis model can analyze an oxygen reduction signal segment corresponding to an oxygen reduction event in the input blood oxygen saturation signal, and output an oxygen reduction type identification result of the oxygen reduction signal segment, wherein the oxygen reduction type identification result can be used for diagnosis assistance of the SAHS.

In one embodiment, the oxygen reduction type tags include an apnea oxygen reduction type tag and a non-apnea oxygen reduction type tag; obtaining an oxygen subtraction signal training fragment carrying an oxygen subtraction type label according to the original oxygen subtraction signal fragment: determining an oxygen subtraction event label corresponding to the original oxygen subtraction signal segment; determining a positive sample signal segment and a negative sample signal segment from the original oxygen subtraction signal segment according to the oxygen subtraction event label; the positive sample signal segment carries an apnea oxygen reduction type label, and the negative sample signal segment carries a non-apnea oxygen reduction type label; and obtaining an oxygen subtraction signal training segment according to the positive sample signal segment and the negative sample signal segment.

In this embodiment, when the method is applied to the SAHS auxiliary diagnosis and judgment, the oxygen reduction type tag includes an apnea oxygen reduction type tag and a non-apnea oxygen reduction type tag, where the apnea oxygen reduction type tag is associated with apnea and the non-apnea oxygen reduction type tag is not associated with apnea. And when an oxygen subtraction signal training fragment carrying an oxygen subtraction type label is obtained according to the original oxygen subtraction signal fragment, determining an oxygen subtraction event label corresponding to the original oxygen subtraction signal fragment. The oxygen reduction event label can be obtained by interpreting and labeling the original blood oxygen saturation signal, and in a specific application, the oxygen reduction event label can contain each desaturation event, namely, a tuple of the start time and the end time of the oxygen reduction event, which is given by a clinician.

Positive and negative sample signal segments are determined from the original oxygen subtracted signal segments according to the oxygen subtracted event signature. PSG records apnea events based on the respiratory signal, with the beginning annotation of the event indicating the beginning of the flow limitation and the end annotation indicating the return of flow to normal. Generally, each respiratory event is accompanied by a desaturation event of the blood oxygenation channel, delayed by about 20-40 seconds. In the training set, each desaturation event has a manual annotation by the physician. Thus, for each annotated apnea event, a desaturation event is found in the 60 second window after the annotation begins and the two events are correlated to determine a positive sample signal segment and a negative sample signal segment from the original oxygen subtracted signal segment for model training. Specifically, an Intersection score may be determined according to an Intersection-over-Union (IoU) between the start-stop time of the original oxygen subtraction signal segment and the start-stop time of the corresponding oxygen subtraction event label, and the original oxygen subtraction signal segment may be determined to be a positive sample signal segment or a negative sample signal segment according to the Intersection score. The positive sample signal segment carries an apnea oxygen reduction type label, and represents that the original oxygen reduction signal segment is related to apnea; the negative sample signal segment carries a non-apnea oxygen subtraction type tag, which characterizes that the original oxygen subtraction signal segment is not related to apnea. In specific implementation, the original oxygen minus signal segment with the intersection score larger than 0.6 may be determined as a positive sample signal segment, the original oxygen minus signal segment with the intersection score smaller than 0.4 may be determined as a negative sample signal segment, and the original oxygen minus signal segment with the intersection score between 0.4 and 0.6 may be ignored and is not used as model training sample data. After a positive sample signal segment and a negative sample signal segment are determined from the original oxygen subtraction signal segment, an oxygen subtraction signal training segment is obtained according to the positive sample signal segment and the negative sample signal segment, and the oxygen subtraction signal training segment is used as model training sample data to perform subsequent model training.

It should be understood that although the various steps in the flow charts of fig. 2-3 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2-3 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternating with other steps or at least some of the sub-steps or stages of other steps.

In one embodiment, as shown in fig. 7, there is provided an oximetry signal analysis device including: a signal to be analyzed acquisition module 702, an oxygen subtraction signal determination module 704, a local feature extraction module 706, a feature fusion processing module 708, and an oxygen subtraction type identification module 710, wherein:

a to-be-analyzed signal acquisition module 702, configured to acquire a blood oxygen saturation signal to be analyzed;

an oxygen decreasing signal determining module 704, configured to determine an oxygen decreasing signal segment corresponding to an oxygen decreasing event in the blood oxygen saturation signal;

a local feature extraction module 706, configured to extract local features of the oxygen-subtracted signal segment;

a feature fusion processing module 708, configured to fuse the local feature and the global feature of the blood oxygen saturation signal to obtain a fusion feature;

and the oxygen subtraction type identification module 710 is configured to perform oxygen subtraction type identification on the fusion features to obtain an oxygen subtraction type identification result of the oxygen subtraction signal segment.

In one embodiment, the oxygen subtracted signal determination module 704 includes a wavelet denoising module, a smoothing processing module, an oxygen subtracted event determination module, and an oxygen subtracted signal correspondence module; wherein: the wavelet denoising module is used for performing wavelet denoising processing on the blood oxygen saturation signal based on a wavelet algorithm to obtain a denoised blood oxygen saturation signal; the smoothing processing module is used for smoothing the denoised oxyhemoglobin saturation signal to obtain a preprocessed oxyhemoglobin saturation signal; the oxygen reduction event determining module is used for determining an oxygen reduction event from the preprocessed blood oxygen saturation signals according to the amplitude change of the preprocessed blood oxygen saturation signals; and the oxygen reduction signal corresponding module is used for determining an oxygen reduction signal segment corresponding to the oxygen reduction event from the blood oxygen saturation signal.

In one embodiment, the local feature extraction module 706 includes a signal segment extraction module, a timing feature extraction module, and a convolution processing module; wherein: the signal segment extraction module is used for extracting an oxygen reduction signal segment from the blood oxygen saturation signal; the time sequence feature extraction module is used for extracting time sequence features of the oxygen subtraction signal segments to obtain the time sequence features of the oxygen subtraction signal segments; and the convolution processing module is used for performing convolution processing on the time sequence characteristics of the oxygen subtraction signal segments to obtain the local characteristics of the oxygen subtraction signal segments.

In one embodiment, the feature fusion processing module 708 global feature acquisition module and feature fusion module; wherein: the global feature acquisition module is used for acquiring global features of the blood oxygen saturation degree signals; the global feature is obtained by performing global feature analysis on the blood oxygen saturation signal; and the feature fusion module is used for fusing the local features and the global features to obtain fused features.

In one embodiment, the system further comprises a model training module, wherein the model training module comprises an original signal acquisition module, a training fragment determination module, a local training characteristic determination module, a fusion training characteristic determination module, a training type identification module and a model adjustment module; wherein: the original signal acquisition module is used for acquiring an original blood oxygen saturation signal; the training segment determining module is used for determining an original oxygen subtraction signal segment corresponding to an oxygen subtraction event in the original blood oxygen saturation signal and obtaining an oxygen subtraction signal training segment carrying an oxygen subtraction type label according to the original oxygen subtraction signal segment; the local training characteristic determining module is used for extracting the local training characteristics of the oxygen subtraction signal training fragment through the to-be-trained oxygen subtraction signal analysis model; the fusion training feature determination module is used for fusing the local training features and the global training features of the oxyhemoglobin saturation training signals through an oxygen subtraction signal analysis model to be trained to obtain fusion training features; the training type recognition module is used for carrying out oxygen reduction type recognition on the fusion training characteristics through an oxygen reduction signal analysis model to be trained to obtain an oxygen reduction type training recognition result of the oxyhemoglobin saturation training signal; and the model adjusting module is used for determining model loss according to the oxygen reduction type training recognition result and the oxygen reduction type label, adjusting the oxygen reduction signal analysis model to be trained according to the model loss and then continuing training until the training ending condition is met, and obtaining the trained oxygen reduction signal analysis model.

In one embodiment, the oxygen reduction type tags include an apnea oxygen reduction type tag and a non-apnea oxygen reduction type tag; the training fragment determining module comprises an oxygen minus label determining module, a positive and negative sample dividing module and a training sample obtaining module; wherein: the oxygen subtraction label determining module is used for determining an oxygen subtraction event label corresponding to the original oxygen subtraction signal segment; the positive and negative sample dividing module is used for determining a positive sample signal segment and a negative sample signal segment from the original oxygen subtraction signal segment according to the oxygen subtraction event label; the positive sample signal segment carries an apnea oxygen reduction type label, and the negative sample signal segment carries a non-apnea oxygen reduction type label; and the training sample obtaining module is used for obtaining an oxygen subtraction signal training fragment according to the positive sample signal fragment and the negative sample signal fragment.

Specific limitations on the oximetry signal analysis device can be found in the above limitations on the oximetry signal analysis method, and are not described in detail here. The respective modules in the aforementioned oximetry signal analysis apparatus may be wholly or partially implemented by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 8. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of oximetry signal analysis.

Those skilled in the art will appreciate that the architecture shown in fig. 8 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, there is provided a computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

acquiring a blood oxygen saturation signal to be analyzed;

extracting local features of the oxygen subtracted signal segments;

In one embodiment, the processor, when executing the computer program, further performs the steps of: performing wavelet denoising processing on the blood oxygen saturation signal based on a wavelet algorithm to obtain a denoised blood oxygen saturation signal; smoothing the denoised oxyhemoglobin saturation signal to obtain a preprocessed oxyhemoglobin saturation signal; determining an oxygen reduction event from the pre-processed blood oxygen saturation signal according to the amplitude change of the pre-processed blood oxygen saturation signal; an oxygen subtracted signal segment corresponding to the oxygen subtracted event is determined from the blood oxygen saturation signal.

In one embodiment, the processor, when executing the computer program, further performs the steps of: extracting an oxygen-subtracted signal segment from the blood oxygen saturation signal; performing time sequence feature extraction on the oxygen subtraction signal fragment to obtain the time sequence feature of the oxygen subtraction signal fragment; and carrying out convolution processing on the time sequence characteristics of the oxygen subtraction signal fragment to obtain the local characteristics of the oxygen subtraction signal fragment.

In one embodiment, the processor, when executing the computer program, further performs the steps of: acquiring global characteristics of a blood oxygen saturation signal; the global feature is obtained by performing global feature analysis on the blood oxygen saturation signal; and fusing the local features and the global features to obtain fused features.

In one embodiment, the processor, when executing the computer program, further performs the steps of: acquiring an original blood oxygen saturation signal; determining an original oxygen subtraction signal segment corresponding to an oxygen subtraction event in the original blood oxygen saturation signal, and obtaining an oxygen subtraction signal training segment carrying an oxygen subtraction type label according to the original oxygen subtraction signal segment; extracting local training characteristics of an oxygen subtraction signal training fragment through an oxygen subtraction signal analysis model to be trained; fusing the local training characteristics and the global training characteristics of the oxyhemoglobin saturation training signal through an oxygen reduction signal analysis model to be trained to obtain fused training characteristics; performing oxygen reduction type recognition on the fusion training characteristics through an oxygen reduction signal analysis model to be trained to obtain an oxygen reduction type training recognition result of the blood oxygen saturation training signal; and determining model loss according to the oxygen subtraction type training recognition result and the oxygen subtraction type label, adjusting the oxygen subtraction signal analysis model to be trained according to the model loss, and continuing training until the training ending condition is met, so as to obtain the trained oxygen subtraction signal analysis model.

In one embodiment, the oxygen reduction type tags include an apnea oxygen reduction type tag and a non-apnea oxygen reduction type tag; the processor, when executing the computer program, further performs the steps of: determining an oxygen subtraction event label corresponding to the original oxygen subtraction signal segment; determining a positive sample signal segment and a negative sample signal segment from the original oxygen subtraction signal segment according to the oxygen subtraction event label; the positive sample signal segment carries an apnea oxygen reduction type label, and the negative sample signal segment carries a non-apnea oxygen reduction type label; and obtaining an oxygen subtraction signal training segment according to the positive sample signal segment and the negative sample signal segment.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:

acquiring a blood oxygen saturation signal to be analyzed;

extracting local features of the oxygen subtracted signal segments;

In one embodiment, the computer program when executed by the processor further performs the steps of: performing wavelet denoising processing on the blood oxygen saturation signal based on a wavelet algorithm to obtain a denoised blood oxygen saturation signal; smoothing the denoised oxyhemoglobin saturation signal to obtain a preprocessed oxyhemoglobin saturation signal; determining an oxygen reduction event from the pre-processed blood oxygen saturation signal according to the amplitude change of the pre-processed blood oxygen saturation signal; an oxygen subtracted signal segment corresponding to the oxygen subtracted event is determined from the blood oxygen saturation signal.

In one embodiment, the computer program when executed by the processor further performs the steps of: extracting an oxygen-subtracted signal segment from the blood oxygen saturation signal; performing time sequence feature extraction on the oxygen subtraction signal fragment to obtain the time sequence feature of the oxygen subtraction signal fragment; and carrying out convolution processing on the time sequence characteristics of the oxygen subtraction signal fragment to obtain the local characteristics of the oxygen subtraction signal fragment.

In one embodiment, the computer program when executed by the processor further performs the steps of: acquiring global characteristics of a blood oxygen saturation signal; the global feature is obtained by performing global feature analysis on the blood oxygen saturation signal; and fusing the local features and the global features to obtain fused features.

In one embodiment, the computer program when executed by the processor further performs the steps of: acquiring an original blood oxygen saturation signal; determining an original oxygen subtraction signal segment corresponding to an oxygen subtraction event in the original blood oxygen saturation signal, and obtaining an oxygen subtraction signal training segment carrying an oxygen subtraction type label according to the original oxygen subtraction signal segment; extracting local training characteristics of an oxygen subtraction signal training fragment through an oxygen subtraction signal analysis model to be trained; fusing the local training characteristics and the global training characteristics of the oxyhemoglobin saturation training signal through an oxygen reduction signal analysis model to be trained to obtain fused training characteristics; performing oxygen reduction type recognition on the fusion training characteristics through an oxygen reduction signal analysis model to be trained to obtain an oxygen reduction type training recognition result of the blood oxygen saturation training signal; and determining model loss according to the oxygen subtraction type training recognition result and the oxygen subtraction type label, adjusting the oxygen subtraction signal analysis model to be trained according to the model loss, and continuing training until the training ending condition is met, so as to obtain the trained oxygen subtraction signal analysis model.

In one embodiment, the oxygen reduction type tags include an apnea oxygen reduction type tag and a non-apnea oxygen reduction type tag; the computer program when executed by the processor further realizes the steps of: determining an oxygen subtraction event label corresponding to the original oxygen subtraction signal segment; determining a positive sample signal segment and a negative sample signal segment from the original oxygen subtraction signal segment according to the oxygen subtraction event label; the positive sample signal segment carries an apnea oxygen reduction type label, and the negative sample signal segment carries a non-apnea oxygen reduction type label; and obtaining an oxygen subtraction signal training segment according to the positive sample signal segment and the negative sample signal segment.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. A method of oximetry signal analysis, the method comprising:

acquiring a blood oxygen saturation signal to be analyzed;

determining an oxygen-decreasing signal segment corresponding to an oxygen-decreasing event in the blood oxygen saturation signal;

extracting local features of the oxygen subtracted signal segments;

fusing the local features and the global features of the blood oxygen saturation signal to obtain fused features;

2. The method of claim 1, wherein said determining an oxygen subtracted signal segment corresponding to an oxygen subtracted event in said oximetry signal comprises:

determining an oxygen reduction event from the pre-processed oximetry signal based on the change in amplitude of the pre-processed oximetry signal;

determining an oxygen-subtracted signal segment corresponding to the oxygen-subtracted event from the blood oxygen saturation signal.

3. The method of claim 1, wherein the extracting local features of the oxygen subtracted signal segment comprises:

extracting the oxygen subtracted signal segment from the oximetry signal;

and carrying out convolution processing on the oxygen subtraction signal fragment time sequence characteristics to obtain the local characteristics of the oxygen subtraction signal fragment.

4. The method of claim 1, wherein said fusing the local features and the global features of the oximetry signal to obtain fused features comprises:

acquiring global characteristics of the blood oxygen saturation signal; the global feature is obtained by performing global feature analysis on the blood oxygen saturation signal;

and fusing the local features and the global features to obtain fused features.

5. The method of claim 4, wherein the global features comprise at least one of a mean, a standard deviation, an area under a curve, a mean frequency, a peak frequency, and a center frequency of the blood oxygen saturation signal.

6. The method according to any one of claims 1 to 5, wherein the method is implemented based on a pre-trained oxygen subtracted signal analysis model, the training step of the oxygen subtracted signal analysis model comprising:

acquiring an original blood oxygen saturation signal;

extracting local training characteristics of the oxygen subtraction signal training fragment through an oxygen subtraction signal analysis model to be trained;

fusing the local training characteristics and the global training characteristics of the oxyhemoglobin saturation training signal through the oxygen reduction signal analysis model to be trained to obtain fused training characteristics;

performing oxygen reduction type recognition on the fusion training characteristics through the oxygen reduction signal analysis model to be trained to obtain an oxygen reduction type training recognition result of the blood oxygen saturation training signal;

determining model loss according to the oxygen reduction type training recognition result and the oxygen reduction type label, adjusting the oxygen reduction signal analysis model to be trained according to the model loss, and continuing training until the training ending condition is met, so as to obtain the trained oxygen reduction signal analysis model.

7. The method of claim 6, wherein the oxygen reduction type tags comprise an apnea oxygen reduction type tag and a non-apnea oxygen reduction type tag; obtaining an oxygen subtraction signal training fragment carrying an oxygen subtraction type label according to the original oxygen subtraction signal fragment:

determining an oxygen subtraction event tag corresponding to the original oxygen subtraction signal segment;

determining a positive sample signal segment and a negative sample signal segment from the original oxygen subtracted signal segments according to the oxygen subtracted event label; the positive sample signal segment carries the apnea oxygen reduction type label, and the negative sample signal segment carries the non-apnea oxygen reduction type label;

8. An oximetry signal analysis device, the device comprising:

the oxygen reduction signal determination module is used for determining an oxygen reduction signal segment corresponding to an oxygen reduction event in the blood oxygen saturation signal;

the characteristic fusion processing module is used for fusing the local characteristic and the global characteristic of the blood oxygen saturation signal to obtain a fusion characteristic;

9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.