Sleep quality prediction method and system

Abstract

The invention provides a sleep quality prediction method and system, and the method comprises the following steps: building a sleep quality prediction model based on the HRV state information data before sleep; and performing sleep quality prediction by using the sleep quality prediction model to obtain a sleep prediction result. According to the technical scheme, the sleep quality problem which does not occur can be predicted in advance, and a basis is provided for improving the sleep quality.

Description

Background Art

[0002] In recent years, portable devices have developed rapidly in the field of sleep monitoring. Accelerometer body movement, blood oxygen saturation monitoring, HRV and other indicators are used to identify and analyze sleep-wake cycles, sleep disorders, etc. HRV, as a core indicator reflecting autonomic nervous function, can reveal the sleep-related neural regulation mechanism from the sympathetic-parasympathetic balance pattern. [1] Studies have shown that an increase in HRV, especially an increase in parasympathetic nerve activity (such as an increase in high-frequency components), helps to speed up falling asleep and increase the proportion of deep sleep, while excessive sympathetic nerve excitement is closely related to difficulty falling asleep and an increase in the proportion of light sleep. [2] In addition, HRV biofeedback training or relaxation training (such as breathing regulation and meditation) can improve sleep quality by regulating HRV. [1] This suggests that HRV not only has a direct regulatory effect on sleep state, but may also be an important physiological predictor of sleep quality.

[0003] With the popularity of wearable devices, the collection of HRV data has become simple and easy. HRV analysis is a non-invasive monitoring method that does not interfere with the user's daily activities and sleep process. For example, devices such as smart watches and bracelets can monitor HRV in real time and non-invasively. The convenience of this technology makes it widely used in daily life. It can not only be used to monitor recovery ability, stress level, etc., but also can be easily integrated into personal health management and sleep monitoring. HRV is extremely sensitive to physiological and psychological states such as stress, fatigue and mood changes, which are directly related to sleep quality and insomnia. Therefore, HRV changes can provide early warning signals for potential sleep disorders. High-quality sleep is usually accompanied by an improvement in HRV, while insomnia, shallow sleep and sleep interruption can cause abnormal fluctuations in HRV. By analyzing the HRV before bedtime, the possible sleep state and quality of an individual can be predicted to a certain extent, helping to identify the risk of insomnia. In summary, the use of HRV before bedtime for insomnia prediction and sleep quality assessment has scientific basis and practical value.

[0004] References:

[0005] [1]LIQ,SHI M,STEWARD C J,et al.AComparison Between Pre-Sleep HeartRate Variability Biofeedback and Electroencephalographic Biofeedback Trainingon Sleep in National Level Athletes with Sleep Disturbances[J].AppliedPsychophysiology and Biofeedback,2024,49(1):115-124.doi:10.1007 / s10484-023-09604-3;

[0006] [2]SHAFFER F,GINSBERG J P.An Overview ofHeart Rate VariabilityMetrics and Norms[J].Front Public Health,2017,5:258.doi:10.3389 / fpubh.2017.00258;

[0007] [3]WANG Q Q,YU S C,QI X,et al.[Overview of logistic regression modelanalysis and application][J].Zhonghua Yu Fang Yi Xue Za Zhi.Chinese Journalof Preventive Medicine,2019,53(9):955-960.doi:10.3760 / cma.j.issn.0253-9624.2019.09.018;

[0008] [4]LIQ,STEWARD C J,CULLEN T,et al.Presleep Heart-Rate VariabilityBiofeedback Improves Mood and Sleep Quality in Chinese Winter BobsleighAthletes[J].International Journal of Sports Physiology and Performance,2022,17(10):1516-1526;

[0009] [5]SATEIAM J.International classification of sleep disorders-thirdedition:highlights and modifications[J].Chest,2014,146(5):1387-1394.doi:10.1378 / chest.14-0970;

[0010] [6]GARBARINO S,BRAGAZZI N L.Revolutionizing Sleep Health:TheEmergence and Impact of Personalized Sleep Medicine[J].Journal ofPersonalized Medicine,2024,14(6):598;

[0011] [7]KONTOS A,BAUMERT M,LUSHINGTON K,et al.The Inconsistent Nature of Heart Rate Variability During Sleep in Normal Children and Adolescents[J].Front Cardiovasc Med,2020,7:19.doi:10.3389 / fcvm.2020.00019. Summary of the invention

[0012] The present application provides a sleep quality prediction method and system for predicting sleep quality problems that have not occurred in advance, thereby providing a basis for improving sleep quality.

[0013] In a first aspect, a sleep quality prediction method is provided, comprising the following steps:

[0014] Based on the HRV status information data before going to bed, a sleep quality prediction model is constructed;

[0015] The sleep quality prediction model is used to predict sleep quality and obtain a sleep prediction result.

[0016] In the above technical solution, a sleep quality prediction model is constructed based on the pre-sleep HRV state information data; the sleep quality prediction model is used to predict sleep quality and obtain a sleep prediction result; sleep quality problems that have not occurred can be predicted in advance, providing a basis for improving sleep quality.

[0017] In a specific embodiment, it also includes:

[0018] The pre-sleep HRV status information data is obtained.

[0019] In a specific implementation scheme, the steps of constructing a sleep quality prediction model based on the pre-sleep HRV status information data specifically include:

[0020] Based on the HRV status information data before sleep, an insomnia prediction sub-model was constructed;

[0021] Based on the HRV status information data before sleep, a sleep effect potential sub-model is constructed;

[0022] Based on the HRV status information data before sleep, a sleep-wake time sub-model is constructed.

[0023] In a specific implementation manner, the insomnia prediction submodel adopts a binary logistic regression model.

[0024] In a specific implementation scheme, the sleep efficacy submodel adopts a stepwise linear regression model.

[0025] In a specific implementation scheme, the sleep-wake time sub-model is an MLP neural network model.

[0026] In a specific implementation scheme, the formula of the binary logistic regression model is:

[0027] ln(p / 1-p)

[0028] =62.809-0.129*RMSSD-0.214*PNN50-0.152*MHR-23.062*SampEn;

[0029] Among them, p is the probability of insomnia being 1, 1-p represents the probability of insomnia being 0; RMSSD is the root mean square of the continuous difference between RR intervals; PNN50 is the percentage of adjacent sinus RR interval differences greater than 50ms; MHR is the mean heart rate; SampEn is the sample entropy.

[0030] In a specific implementation scheme, the formula of the stepwise linear regression model is:

[0031] SOL = 0.243-0.270*RMSSD;

[0032] AT=0.892-0.588*PNN50-0.300*MHR-0.267*ApEn;

[0033] SE=0.502+0.235*RMSSD+0.332*PNN50+0.201*MHR;

[0034] DST=0.374+0.201*MRR+0.194*ApEn;

[0035] Among them, SOL is sleep latency; AT is wake time; SE is sleep efficiency; DST is deep sleep; RMSSD is the root mean square of continuous differences in RR intervals; PNN50 is the percentage of adjacent sinus RR interval differences greater than 50ms; MHR is mean heart rate; ApEn, approximate entropy; SampEn is sample entropy.

[0036] In a second aspect, a sleep quality prediction system is provided, comprising:

[0037] The prediction model module is used to build a sleep quality prediction model based on the HRV status information data before going to bed;

[0038] The prediction module is used to use the sleep quality prediction model to predict sleep quality and obtain a sleep prediction result.

[0039] In the above technical solution, a sleep quality prediction model is constructed based on the pre-sleep HRV state information data; the sleep quality prediction model is used to predict sleep quality and obtain a sleep prediction result; sleep quality problems that have not occurred can be predicted in advance, providing a basis for improving sleep quality.

[0040] In a specific implementation scheme, the prediction model module includes:

[0041] An insomnia sub-model unit, used to construct an insomnia prediction sub-model based on the pre-sleep HRV state information data;

[0042] The sleep effect potential sub-model unit is used to construct a sleep effect potential sub-model based on the HRV state information data before sleep;

[0043] The wake-up time sub-model unit is used to construct a sleep-wake-up time sub-model based on the pre-sleep HRV state information data. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flowchart of a sleep quality prediction method provided in an embodiment of the present application;

[0045] Figure 2 A structural block diagram of a sleep quality prediction system provided in an embodiment of the present application;

[0046] Figure 3 A schematic diagram of the HRV before bedtime prediction of sleep quality and its application process provided in the embodiment of the present application;

[0047] Figure 4 A flowchart of model establishment and retraining provided in an embodiment of the present application;

[0048] Figure 5 An example diagram for establishing a multi-algorithm model provided in an embodiment of the present application;

[0049] Figure 6 A logical diagram of the binary logistic regression model provided in the embodiment of the present application for determining whether there is insomnia;

[0050] Figure 7 A schematic diagram of the stepwise linear regression and neural network model and establishment process provided in the embodiments of the present application. DETAILED DESCRIPTION

[0051] The present application is further described in detail below through the accompanying drawings and embodiments. Through these descriptions, the characteristics and advantages of the present application will become clearer and more specific.

[0052] The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise noted.

[0053] In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0054] To facilitate understanding of the sleep quality prediction method and system provided in the embodiments of the present application, its application scenario is first explained. The sleep quality prediction method and system provided in the embodiments of the present application are used to predict sleep quality problems that have not occurred in advance, providing a basis for improving sleep quality. In recent years, portable devices have developed rapidly in the field of sleep monitoring. Multiple indicators such as accelerometer body movement, blood oxygen saturation monitoring and HRV are used to identify and analyze sleep-wake cycles, sleep disorders, etc. As a core indicator reflecting autonomic nervous function, HRV can reveal the sleep-related neural regulation mechanism from the sympathetic-parasympathetic balance pattern. [1] Studies have shown that an increase in HRV, especially an increase in parasympathetic nerve activity (such as an increase in high-frequency components), helps to speed up falling asleep and increase the proportion of deep sleep, while excessive sympathetic nerve excitement is closely related to difficulty falling asleep and an increase in the proportion of light sleep. [2] In addition, HRV biofeedback training or relaxation training (such as breathing regulation and meditation) can improve sleep quality by regulating HRV. [1] . This shows that HRV not only has a direct regulatory effect on sleep status, but may also be an important physiological predictor of sleep quality. For this reason, the embodiment of the present application provides a sleep quality prediction method and system to predict sleep quality problems that have not occurred in advance, providing a basis for improving sleep quality. The following is a detailed description of the embodiment in conjunction with specific drawings.

[0055] refer to Figures 1 to 7 , Figure 1 A flowchart of a sleep quality prediction method provided in an embodiment of the present application; Figure 2 A structural block diagram of a sleep quality prediction system provided in an embodiment of the present application; Figure 3 A schematic diagram of the HRV before bedtime prediction of sleep quality and its application process provided in the embodiment of the present application; Figure 4 A flowchart of model establishment and retraining provided in an embodiment of the present application; Figure 5 An example diagram for establishing a multi-algorithm model provided in an embodiment of the present application; Figure 6 A logical diagram of the binary logistic regression model provided in the embodiment of the present application for determining whether there is insomnia; Figure 7 A schematic diagram of the stepwise linear regression and neural network model and establishment process provided in the embodiments of the present application.

[0056] exist Figure 1 In the embodiment of the present application, a sleep quality prediction method is provided, comprising the following steps:

[0057] Based on the HRV status information data before going to bed, a sleep quality prediction model is constructed;

[0058] The sleep quality prediction model is used to predict sleep quality and obtain a sleep prediction result.

[0059] In the above technical solution, a sleep quality prediction model is constructed based on the pre-sleep HRV state information data; the sleep quality prediction model is used to predict sleep quality and obtain a sleep prediction result; sleep quality problems that have not occurred can be predicted in advance, providing a basis for improving sleep quality.

[0060] In a specific embodiment, it also includes:

[0061] The pre-sleep HRV status information data is obtained.

[0062] In a specific implementation scheme, the steps of constructing a sleep quality prediction model based on the pre-sleep HRV status information data specifically include:

[0063] Based on the HRV status information data before sleep, an insomnia prediction sub-model was constructed;

[0064] Based on the HRV status information data before sleep, a sleep effect potential sub-model is constructed;

[0065] Based on the HRV status information data before sleep, a sleep-wake time sub-model is constructed.

[0066] In a specific implementation manner, the insomnia prediction submodel adopts a binary logistic regression model.

[0067] In a specific implementation scheme, the sleep efficacy submodel adopts a stepwise linear regression model.

[0068] In a specific implementation scheme, the sleep-wake time sub-model is an MLP neural network model.

[0069] In a specific implementation scheme, the formula of the binary logistic regression model is:

[0070] ln(p / 1-p)

[0071] =62.809-0.129*RMSSD-0.214*PNN50-0.152*MHR-23.062*SampEn;

[0072] Among them, p is the probability of insomnia being 1, 1-p represents the probability of insomnia being 0; RMSSD is the root mean square of the continuous difference between RR intervals; PNN50 is the percentage of adjacent sinus RR interval differences greater than 50ms; MHR is the mean heart rate; SampEn is the sample entropy.

[0073] In a specific implementation scheme, the formula of the stepwise linear regression model is:

[0074] SOL = 0.243-0.270*RMSSD;

[0075] AT=0.892-0.588*PNN50-0.300*MHR-0.267*ApEn;

[0076] SE=0.502+0.235*RMSSD+0.332*PNN50+0.201*MHR;

[0077] DST=0.374+0.201*MRR+0.194*ApEn;

[0078] Among them, SOL is sleep latency; AT is wake time; SE is sleep efficiency; DST is deep sleep; RMSSD is the root mean square of continuous differences in RR intervals; PNN50 is the percentage of adjacent sinus RR interval differences greater than 50ms; MHR is mean heart rate; ApEn, approximate entropy; SampEn is sample entropy.

[0079] Specifically, refer to Figure 3 The sleep quality prediction method uses HRV and algorithm models to ultimately obtain a comprehensive prediction result about the user's sleep state, which is provided to the user for knowing the sleep status in advance, and targeted measures can also be taken according to the results to improve sleep quality.

[0080] refer to Figure 4 , the basic model includes:

[0081] 1. Model training database

[0082] Input: Contains pre-sleep HRV data and corresponding sleep data.

[0083] Processing: Establish basic sleep prediction algorithm through multiple prediction models.

[0084] Output: Basic prediction model.

[0085] 2. Model Validation Database

[0086] Validation: Use an independent test dataset to compare the model predictions with the true test values ​​to evaluate model performance.

[0087] Algorithm logic retraining includes:

[0088] 1. Target user data collection

[0089] Collect HRV data of specific target users as input for algorithm retraining.

[0090] 2. Data Preprocessing

[0091] Process the user's HRV data to adapt it to the current basic model.

[0092] 3. Model retraining

[0093] The processed target user data is added to the model training database and the basic model is retrained to generate a personalized prediction model for the target user.

[0094] Finally, the target model is used to predict the user's sleep quality data.

[0095] refer to Figure 5 The output logic of the algorithm includes: binary logistic regression outputs the probability of insomnia; stepwise linear regression and MLP neural network model output specific indicators of sleep quality prediction value that night. Among them, TST is sleep duration; SOL is sleep latency; AT is awakening time; SE is sleep efficiency; LST is light sleep; DST is deep sleep; REM is rapid eye movement sleep; MRR is the average RR interval; PNN50 is the percentage of adjacent sinus RR intervals with a difference greater than 50ms; RMSSD is the root mean square of the continuous difference of RR intervals; SDNN is the standard deviation of NN intervals; TINN is the triangular interpolation of the NN interval histogram; MHR is the average heart rate; SI is the stress index; HF is the high frequency power peak; LF is the low frequency power peak; SD1 is the standard deviation in the short axis direction; SD2 is the standard deviation in the long axis direction; ApEn is the approximate entropy; SampEn is the sample entropy.

[0096] In this embodiment, the implementation process of the sleep quality prediction method is as follows:

[0097] 1 Research subjects

[0098] Based on the G*power sample size calculation, a total of 98 people are needed to meet the sample size requirements of this study. In order to accurately determine whether there is a risk of insomnia and predict sleep quality, the model should include CI and normal sleep population. CI screening criteria: meet the diagnostic criteria of the International Classification of Sleep Disorders-Third Version (ICSD-3). Including the following core indicators: 1) have obvious insomnia symptoms, such as dissatisfaction with sleep quality and daytime dysfunction; 2) complain of sleep latency>30min or wake time after falling asleep>30min; 3) insomnia symptoms at least 3 nights a week; 4) have at least 2 of the following daytime negative effects: such as fatigue, physical decline, memory, attention loss or emotional involvement; 5) insomnia symptoms last>1 month; 6) exclude insomnia secondary to other diseases. The auxiliary evaluation criteria of the scale include: PSQI score>7 points, while the Hamilton Anxiety Rating Scale and Hamilton Depression Rating Scale scores are both <7 points to exclude sleep disorders caused by serious emotional problems.

[0099] Other inclusion criteria: 1) Age: 18-25 years old; 2) No recent mental stimulation, no psychotropic drugs or supplements; 3) No nap habits (to avoid disturbing nighttime sleep). Through recruitment and screening, a total of 174 people were included in the final data set. After screening and sleep monitoring, it was found that the CI incidence rate in this study was 56.32%, and there was no significant difference in age, height and weight between CI and normal sleepers (P>0.05).

[0100] 2 Experimental Procedure

[0101] In order to avoid the "first night effect" in sleep monitoring, 1 to 2 nights of adaptation were performed in the sleep laboratory before the formal experiment. The sleep adaptation was completed by checking the results of the PSQI scale. People with different sleep quality were subjected to pre-sleep neurological function state test and PSG all-night sleep monitoring that night. After the subjects arrived at the laboratory for the first time, information verification, informed consent form filling and experimental process explanation were carried out. After the adaptation to the sleep laboratory environment and equipment wearing was completed, no naps and no strenuous exercise were performed on the day of the formal test. Dinner was required to be completed 4 hours before bedtime to avoid overeating and stimulating diet. On the night of the formal test, the subjects arrived at the sleep laboratory about 90 minutes in advance according to their daily bedtime, and first performed a pre-sleep neurological function test. Test process: 1) Basic information collection; 2) Neurological function assessment: autonomic nervous state: HRV; 3) Then wear polysomnography (PSG) to monitor sleep all night. During sleep monitoring, there is no mandatory time to fall asleep and wake up, and natural sleep and wake up are required.

[0102] 3 Model establishment and verification

[0103] At present, the relationship between pre-sleep HRV and sleep quality and the most suitable prediction model are still unclear. Therefore, the author intends to use binary logistic regression, stepwise linear regression and regression neural network models for analysis and comparison. Binary logistic regression is used to determine the association between pre-sleep HRV and whether insomnia occurs, and stepwise linear regression and neural network models are used to establish the pre-sleep HRV state to predict the sleep quality index of the night, and compare the prediction performance of each model.

[0104] refer to Figure 6 , Binary Logistic Regression Model: It is a statistical model for binary classification problems. The output is a probability value, which indicates the possibility of an event. The model can quantify the impact of independent variables on the probability of an event, and the coefficient reflects the weight of the variable. Its advantages are simple structure, easy to interpret, and suitable for small sample data. The main limitation is that it is only applicable to binary classification problems and has limited modeling capabilities for complex nonlinear relationships. [3] . In this application, the presence of chronic insomnia was determined based on the gold standard PSG for sleep monitoring. Sleep quality (yes = 1, no = 0) was used as the dependent variable, and the pre-sleep HRV index was used as the independent variable. Binary Logistic regression analysis was performed. The independent variable screening method was the stpwise stepwise method. In the model formula, p represented the probability of whether insomnia (0 / 1) was 1, and 1-p represented the probability of whether insomnia (0 / 1) was 0. The Hosmer-Lemeshow (HL) goodness-of-fit test was used to determine the goodness of fit of the model. R 2 =1 indicates perfect fit, 0.8≤R 2 <1 indicates a high degree of fit, 0.5≤R 2 <0.8 indicates a moderate fit, 0.2≤R2 <0.5 indicates a low fit, R 2 <0.2 indicates a poor fit, R 2 =0 means no explanatory power.

[0105] refer to Figure 7 , Stepwise linear regression random model: It is a model used to describe the relationship between a continuous dependent variable and multiple independent variables, assuming that the dependent variable is a linear combination of independent variables. This model is suitable for analyzing the impact of independent variables on dependent variables, has strong explanatory power, and the regression coefficient intuitively reflects the weight of the variable. However, it is limited to modeling linear relationships and is sensitive to outliers and multicollinearity. Before establishing the model in this study, the test data was first normalized (X-Min) / (Max-Min), the HRV index (P<0.05) was set as the independent variable, and various sleep quality indicators were used as dependent variables. The data were randomly divided into a training set (80%) and a test set (20%). The training set used the stepwise method to establish a linear regression random model for predicting the sleep quality of the night. The R2 range is 0 to 1, and the larger the value, the higher the degree of fit. The tolerance of the prediction equation should be greater than 0.1, and the variance inflation factor (VIF) should be less than 10, indicating that there is no multicollinearity. The Durbin-Watson test values ​​are all in the range of 0 to 4, indicating that the variables are independent of each other and there is no autocorrelation.

[0106] Neural network model: It is a type of model that simulates the structure of human brain neurons and is good at learning nonlinear relationships. Through the layer-by-layer abstraction of multi-layer neurons, it is widely used in the recognition and prediction of complex nonlinear patterns. However, compared with linear models, its interpretability is weaker. This study uses the most classic neural network: Multi-Layer Perception (MLP). When constructing the MLP neural network model, the data is first randomly divided into a training set (80%) and a test set (20%). The training set is used to train the model. In order to keep the data dimension uniform, the test data is first normalized (X-Min) / (Max-Min). The activation function uses relu, the weight optimization method is adam, the initial learning rate is 0.001, the L2 regularization coefficient is 0.0001, the maximum number of iterations is 200, the optimization tolerance is 0.0001, the hidden layer is 1, and the number of neurons is 100.

[0107] In the stepwise linear regression and neural network model methods, after the model is established using the training set, the effectiveness of the model is evaluated. The model is verified using the test set data (random 20%) to obtain the predicted value of sleep quality, and then the sleep quality measured value and the predicted value are subjected to consistency test analysis.

[0108] 4 Sleep quality monitoring and evaluation methods

[0109] 1) Subjective sleep quality: The Pittsburgh Sleep Quality Index (PSQI) is a commonly used sleep quality assessment tool. It was used for the initial screening of subjects and the assessment of sleep adaptation in this study.

[0110] 2) Objective sleep quality: Polysomnography (PSG) is the gold standard for sleep monitoring. This study used PSG (Natus Medical Inc, San Car Los, Canada) to record the entire night's sleep process. The signals used by this device to determine sleep stages and structures include EEG (C4), electrooculogram (left eye lower, right eye upper), and chin electrogram (left and right sides of the submandibular muscle), and the data acquisition frequency was 256 Hz.

[0111] 5HRV test method

[0112] Heart rate variability (HRV) is a common indicator for evaluating the tone of cardiac sympathetic and parasympathetic nerves. In this study, the Polar H10 heart rate chest strap (Polar Electro, Finland) was used to record the RR interval (the time between the R waves in two QRS waves) data, and the Kubios Hrv-standard software was used to intercept 5 minutes for processing and calculate HRV to evaluate the autonomic nervous state.

[0113] 6 Data processing and analysis methods

[0114] SPSS26.0 was used for statistical analysis. The experimental results were expressed as mean ± standard deviation (M ± SD), with two decimal places reserved, and the statistical analysis results were retained to three decimal places. The KS test was used for the normal distribution test. G*power was used to calculate the sample size. By comparing the sample sizes of binary logistic regression and linear regression random models, the latter required a larger sample size. The sample size of the linear regression model was calculated as follows. Based on the previous research results of this research group [4], the correlation between the proportion of light sleep and HRV index was 0.74 at the highest and 0.55 at the lowest. H1 was set to 0.74, H2 to 0.55, α to 0.05, 1-β to 0.8, and a two-tailed test was performed. The independent variable was 10, and the required sample size was calculated to be 78 people. The linear regression model requires a larger sample size, so 78 people are needed to establish the model as a training set. According to the 20% test set requirement, 20 people are needed as a test set, and a total of 98 people are needed to meet the modeling conditions. Pearson describes the correlation between the measured value and the predicted value of sleep quality. 0.10≤|r|<0.29 is a low correlation, 0.30≤|r|<0.49 is a moderate correlation, 0.50≤|r|<0.69 is a high correlation, 0.70≤|r|<0.89 is a very high correlation, and 0.90≤|r|≤1.00 is an approximate linear correlation. When Bland-Altman is used for consistency test, the mean and difference between the measured value and the predicted value of sleep quality are calculated, and the consistency limit (95% distribution range) is calculated. P<0.05 represents statistical significance.

[0115] In the above technical solution, the research results include:

[0116] 1.1 Sleep quality and HRV test results before bedtime

[0117] Before the formal test, the total scores of the PSQI scale of the two groups were not significantly different from those at the screening (P>0.05), indicating that the subjects had adapted to the sleep monitoring equipment and laboratory environment. During the formal test, the PSQI score of the CI group was significantly higher than that of the normal sleep group (P<0.01). The specific data are not listed due to space limitations.

[0118] Compared with the normal sleep group, CI sleep showed significant differences, among which SOL was significantly prolonged (P<0.001, F=19.515), AT was significantly increased (P<0.001, F=222.376); at the same time, TST (P<0.001, F=29.879), SE (P<0.001, F=203.78), DST (P<0.001, F=203.78) and REM (P<0.05, F=4.068) were all significantly reduced; while LST was significantly increased (P<0.01, F=7.104). Among the HRV indicators before sleep in CI, MRR (P<0.001, F=120.935), RMSSD (P<0.001, F=172.554), PNN50 (P<0.001, F=350.190), ApEn (P<0.001, F=118.432) and SampEn (P<0.001, F=94.453) were significantly lower than those in the normal sleep group. The above results revealed that the parasympathetic nerve activity before sleep in the CI group was lower, and the sleep structure and efficiency were significantly changed. Among them, SOL: sleep latency; AT: awakening time; SE: sleep efficiency; DST: deep sleep; bias: deviation.

[0119] 1.2 Binary Logistic Regression Model to Determine the Outcome of Insomnia

[0120] A binary Logit regression analysis was performed with HRV as the independent variable and insomnia (0 / 1) as the dependent variable. After automatic model identification, the results showed that the changes in RMSSD, PNN50, MHR, and SampEn were highly consistent with insomnia (R 2 =0.914). The HL fit test passed (P=1.000>0.05), and the model fit was good. The accuracy of using HRV to judge whether insomnia is present is as high as 90.844%.

[0121] The final model formula is:

[0122] ln(p / 1-p)

[0123] =62.809-0.129*RMSSD-0.214*PNN50-0.152*MHR-23.062*SampEn.

[0124] 1.3 Results of stepwise linear regression model predicting sleep quality

[0125] As shown in Table 1, a stepwise linear regression prediction model was established with the pre-sleep HRV signal as the independent variable and sleep quality as the dependent variable. All models had VIP < 10, tolerance > 0.1, DW between 0 and 4, and no collinearity problem. Among them, HRV predicted AT with a moderate degree of fit (R 2=0.526, P<0.001), SE poor fit (R 2 =0.481, P<0.001, and the fitting degree of other indicators was poor (R 2 <0.2).

[0126] As shown in Table 2, this study substituted the HRV of 35 validation sets into the prediction equation for verification. Among them, SOL was moderately correlated with the predicted value of HRV (P<0.01), and the measured values ​​of AT, SE, and DST were highly correlated with the predicted value of HRV (P<0.01). There was no correlation between the predicted value of LST and REM and the measured value (P>0.05). The Bland-Altman scatter plot was established with the mean of the measured value and the predicted value as the horizontal axis and the difference between the measured value and the predicted value as the vertical axis. The consistency limit probability of all predicted values ​​was between 91.429 and 97.143%.

[0127]

[0128] Table 1 Stepwise linear regression formula for predicting sleep quality using pre-sleep HRV

[0129]

[0130] Table 2 Results of stepwise linear regression of HRV in predicting sleep quality

[0131] Among them: SOL: sleep latency; AT: wake time; SE: sleep efficiency; DST: deep sleep; MRR: mean RR interval; PNN50: percentage of adjacent sinus RR interval differences greater than 50ms; RMSSD, root mean square of consecutive differences in RR intervals; SDNN, standard deviation of NN intervals; TINN, triangular interpolation of NN interval histogram; MHR: mean heart rate; SI: stress index; HF: high-frequency power peak; LF: low-frequency power peak; SD1, standard deviation in the short axis direction; SD2, standard deviation in the long axis direction; ApEn, approximate entropy; SampEn, sample entropy.

[0132] 1.4MLP neural network model predicts sleep quality results

[0133] As shown in Table 3, the MLP neural network prediction model was established with the pre-sleep HRV signal as the independent variable and the sleep quality as the dependent variable. Under this model, the pre-sleep HRV predicted value was highly correlated with the AT measured value (P<0.01); the predicted values ​​of other sleep indicators were not correlated with the measured values ​​(P>0.05). The mean absolute error between the measured value and the predicted value ranged from 0.120 to 0.269, the root mean square error ranged from 0.157 to 0.299, the mean absolute percentage error ranged from 0 to 1.059, and the deviation ranged from -0.174 to 0.028. The consistency test results showed that the limit probability was between 94.286 and 97.143%. The above results show that the overall error of the neural network model is small and the consistency is high.

[0134]

[0135] Table 3 Neural network model for predicting sleep quality using pre-sleep HRV

[0136] Among them, AT: awakening time; MAE: mean absolute error; RMSE: root mean square error; MAPE: mean absolute percentage error; bias: deviation.

[0137] In this embodiment, the results of the study showed that the incidence of CI was 56.32%, and the sleep quality was significantly changed, as shown by a significant decrease in TST, SE, DST and REM, and a significant increase in SOL and LST; at the same time, HRV indicators such as RMSSD, pNN50, ApEn and SampEn were significantly reduced, indicating that the parasympathetic nerve activity was low. Therefore, optimizing the autonomic nervous function and sleep quality of CI people is an important direction for improving health status.

[0138] The core role of sleep in health management and disease prevention has been widely confirmed. At present, EEG combined with electromyography and other indicators in polysomnography is used as the gold standard sleep monitoring method. Although it has the advantage of high precision, it is limited in large-scale application due to the complexity of the equipment, high cost and analysis lag [5]. In recent years, portable devices have developed rapidly in the field of sleep monitoring. Accelerometer body movement, blood oxygen saturation monitoring and HRV and other indicators are used to identify and analyze sleep-wake cycles, sleep disorders, etc. Although portable devices have lowered the technical threshold for sleep monitoring, their results often focus on physiological characterization data, but pay insufficient attention to the prediction and development trend of sleep problems. Relying solely on monitoring data may delay early diagnosis and treatment [6]. Therefore, building a model for judging insomnia before bedtime and predicting sleep quality, and making accurate predictions and timely interventions before risks appear, can more comprehensively optimize health status and sports performance.

[0139] HRV not only has a direct regulatory effect on sleep status, but may also be an important physiological predictor of sleep quality. Studies have shown that as sleep deepens, the sympathetic nerves continue to weaken, while the parasympathetic nerves continue to strengthen and gradually dominate the autonomic nervous system. [7] . The improvement of HRV, especially the enhancement of parasympathetic nerve activity (such as the increase of high-frequency components), helps to accelerate falling asleep, while the overexcitation of sympathetic nerves is closely related to difficulty falling asleep and more awakenings. SOL and AT are the main basis for determining CI. In this study, the accuracy of HRV in determining whether CI occurs under the binary logistic regression model is as high as 90.844%, which shows that the HRV index is consistent with previous studies and has a high sensitivity to sleep continuity indicators such as SOL and AT. In addition, the stepwise linear regression model in this study predicted AT and SE with a good degree of fit. Even if SOL and DST have a poor degree of fit, more characteristic variables (such as psychological factors or environmental parameters) may need to be introduced in the future. However, the correlation analysis still found that the predicted values ​​of AT, SE, SOL and DST are closely related to the measured values; the consistency of the MLP neural network prediction model is relatively high, but only the predicted value of AT is highly correlated with the measured value and is better than the linear regression model. This shows that the three models have a good prediction effect on sleep continuity indicators, but poor prediction ability for sleep stage indicators. Specifically, except for AT, the predictive ability of the stepwise linear regression model of SOL, SE and DST indicators is better than that of the MLP neural network model, and there is a strong linear relationship between HRV and sleep.

[0140] In the above technical solution, when using HRV before bedtime to judge whether there is insomnia, the prediction accuracy of the binary logistic regression model is as high as 90.844%; the stepwise linear regression model is used to predict sleep efficiency, sleep latency and deep sleep with high accuracy and strong correlation; the MLP neural network is used to predict awakening time with high accuracy and strong correlation.

[0141] exist Figure 2 In the embodiment of the present application, a sleep quality prediction system is provided, comprising:

[0142] The prediction model module is used to build a sleep quality prediction model based on the HRV status information data before going to bed;

[0143] The prediction module is used to use the sleep quality prediction model to predict sleep quality and obtain a sleep prediction result.

[0144] In the above technical solution, a sleep quality prediction model is constructed based on the pre-sleep HRV state information data; the sleep quality prediction model is used to predict sleep quality and obtain a sleep prediction result; sleep quality problems that have not occurred can be predicted in advance, providing a basis for improving sleep quality.

[0145] In a specific implementation scheme, the prediction model module includes:

[0146] An insomnia sub-model unit, used to construct an insomnia prediction sub-model based on the pre-sleep HRV state information data;

[0147] The sleep effect potential sub-model unit is used to construct a sleep effect potential sub-model based on the HRV state information data before sleep;

[0148] The wake-up time sub-model unit is used to construct a sleep-wake-up time sub-model based on the pre-sleep HRV state information data.

[0149] Those skilled in the art will appreciate that the present application may be implemented as a system, method or computer program product.

[0150] Therefore, the present disclosure may be specifically implemented in the following forms, namely: it may be completely hardware, it may be completely software (including firmware, resident software, microcode, etc.), or it may be a combination of hardware and software, generally referred to herein as a "circuit", "module" or "system". In addition, in some embodiments, the present application may also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium may contain computer-readable program code.

[0151] Any combination of one or more computer-readable media can be used. Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this document, computer-readable storage media can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0152] Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art can change, modify, replace and modify the above embodiments within the scope of the present application. On this basis, a variety of replacements and improvements can be made to the present application, all of which fall within the scope of protection of the present application.

Claims

1. A sleep quality prediction method, characterized in that: The following steps are involved: Based on the HRV status information data before going to bed, a sleep quality prediction model is constructed; The sleep quality prediction model is used to predict sleep quality and obtain a sleep prediction result.

2. The sleep quality prediction method according to claim 1, characterized in that: Also includes: The pre-sleep HRV status information data is obtained.

3. The sleep quality prediction method according to claim 2, characterized in that: The steps of constructing a sleep quality prediction model based on the HRV status information data before sleep include: Based on the HRV status information data before sleep, an insomnia prediction sub-model was constructed; Based on the HRV status information data before sleep, a sleep effect potential sub-model is constructed; Based on the HRV status information data before sleep, a sleep-wake time sub-model is constructed.

4. The sleep quality prediction method according to claim 3, characterized in that: The insomnia prediction submodel adopts a binary logistic regression model.

5. The sleep quality prediction method according to claim 4, characterized in that: The sleep efficacy submodel adopts a stepwise linear regression model.

6. The sleep quality prediction method according to claim 5, characterized in that: The sleep-wake time sub-model MLP neural network model.

7. The sleep quality prediction method according to claim 6, characterized in that: The formula of the binary logistic regression model is: ln(p / 1-p) =62.809-0.129*RMSSD-0.214*PNN50-0.152*MHR-23.062*SampEn; Among them, p is the probability of insomnia being 1, 1-p represents the probability of insomnia being 0; RMSSD is the root mean square of the continuous difference between RR intervals; PNN50 is the percentage of adjacent sinus RR interval differences greater than 50ms; MHR is the mean heart rate; SampEn is the sample entropy.

8. The sleep quality prediction method according to claim 7, characterized in that: The formula of the stepwise linear regression model is: SOL = 0.243-0.270*RMSSD; AT=0.892-0.588*PNN50-0.300*MHR-0.267*ApEn; SE=0.502+0.235*RMSSD+0.332*PNN50+0.201*MHR; DST=0.374+0.201*MRR+0.194*ApEn; Among them, SOL is sleep latency; AT is wake time; SE is sleep efficiency; DST is deep sleep; RMSSD is the root mean square of continuous differences in RR intervals; PNN50 is the percentage of adjacent sinus RR interval differences greater than 50ms; MHR is mean heart rate; ApEn, approximate entropy; SampEn is sample entropy.

9. A sleep quality prediction system, characterized in that: include: The prediction model module is used to build a sleep quality prediction model based on the HRV status information data before going to bed; The prediction module is used to use the sleep quality prediction model to predict sleep quality and obtain a sleep prediction result.

10. The sleep quality prediction system according to claim 9, characterized in that: The prediction model module includes: An insomnia sub-model unit, used to construct an insomnia prediction sub-model based on the pre-sleep HRV state information data; The sleep effect potential sub-model unit is used to construct a sleep effect potential sub-model based on the HRV state information data before sleep; The wake-up time sub-model unit is used to construct a sleep-wake-up time sub-model based on the pre-sleep HRV state information data.

Images