CN117409985B

CN117409985B - Sleep monitoring and improving method, equipment and storage medium for heart failure patient

Info

Publication number: CN117409985B
Application number: CN202311711096.8A
Authority: CN
Inventors: 邢晓曼; 董文飞
Original assignee: Suzhou Guoke Medical Technology Development Group Co ltd; Suzhou Institute of Biomedical Engineering and Technology of CAS
Current assignee: Suzhou Guoke Medical Technology Development Group Co ltd; Suzhou Institute of Biomedical Engineering and Technology of CAS
Priority date: 2023-12-13
Filing date: 2023-12-13
Publication date: 2024-03-22
Anticipated expiration: 2043-12-13
Also published as: CN117409985A

Abstract

The invention relates to a sleep monitoring and improving method, equipment and a storage medium for heart failure patients, wherein the method comprises the following steps: acquiring a ballistocardiogram waveform acquired in a non-contact manner; accurately acquiring contraction point information in real time during heartbeat through the ballistocardiogram waveform; the contraction point information comprises L peak position information behind a J peak in the ballistocardiogram; and starting the electric contraction piece to contract according to the preset amplitude according to the contraction point information. Aiming at the problem that heart function is affected by heavy heart load and unsmooth blood pressure backflow in sleep of heart failure patients, the invention realizes the external counterpulsation and heating regulation of the intelligent body by designing the linkage of the piezoelectric monitoring mattress and the regulation socks, so that the heart can rest. In a home environment, when heart failure patients heart beat, the contraction points are accurately obtained in real time, and external counterpulsation is started after extremely short delay, so that the risk of false start is eliminated. The device is simple and comfortable, and can be used for sleeping at home; the heart rehabilitation nursing device is quick in starting and convenient to wear, and can realize heart rehabilitation nursing every night.

Description

Sleep monitoring and improving method, equipment and storage medium for heart failure patient

Technical Field

The invention relates to the technical field of heart failure monitoring, in particular to a sleep monitoring and improving method, equipment and medium for heart failure patients.

Background

At present, although some sleep monitoring and improving technologies aiming at relatively healthy people exist, for heart failure patients, traditional suggestions such as urination and pillow raising still remain before sleep, and dynamic adjustment cannot be performed according to the situation of the patients. If heart failure can not be well nursed, sudden cardiac arrest is easily caused, and the worsening of the condition of a patient is accelerated.

Disclosure of Invention

To achieve the above and other advantages and in accordance with the purpose of the present invention, a first object of the present invention is to provide a sleep monitoring and improving method for heart failure patients, comprising the steps of:

acquiring a ballistocardiogram waveform acquired in a non-contact manner;

accurately acquiring contraction point information in real time during heartbeat through the ballistocardiogram waveform; the contraction point information comprises L peak position information behind a J peak in the ballistocardiogram;

and starting the electric contraction piece to contract according to the preset amplitude according to the contraction point information.

Further, after the step of starting the electric shrinkable member to shrink according to the preset amplitude according to the shrink point information, the method further comprises the following steps:

Calculating blood pressure through a non-contact ballistocardiogram;

judging whether the blood pressure is near the target blood pressure;

if the blood pressure is not near the target blood pressure, starting a heating regulation and control module to directionally focus and heat the tiny blood vessel to regulate and control the vascular resistance.

Further, after the step of activating the heating regulation module, the method further comprises the following steps:

adjusting the adjusting step length of the heating adjusting and controlling module according to the received peripheral perfusion lifting condition until reaching a set maximum value, and waking up the patient if the blood pressure is still not near the target blood pressure, and adjusting the bed or adding the oral medicine.

Further, after the step of acquiring the ballistocardiogram waveform acquired in a non-contact manner, the method further comprises the following steps of:

comparing the maximum amplitude ratio of the ballistocardiogram among the plurality of measurement points with a base line;

if the change exceeds the preset range, the contraction of the electro-contraction member is stopped.

Further, the step of accurately acquiring the contraction point information in real time during the heartbeat through the ballistocardiogram waveform comprises the following steps:

obtaining a waveform template of the ballistocardiogram through a sliding window; the waveform template of the ballistocardiogram comprises the position and the height of a J peak of the ballistocardiogram and the distance between the J peak and an L peak;

And pre-judging the subsequent ballistocardiogram waveform according to the real-time trace, and correcting the ballistocardiogram waveform within a preset range.

Further, the step of pre-judging the subsequent ballistocardiogram waveform according to the real-time trace and correcting within a preset range comprises the following steps:

in the acquired newly acquired online waveform, if the signal-to-noise ratio of the waveform is higher than a set threshold, the covariance exceeds the set threshold, and the peak value reaches the floating range of the height of the J peak of the waveform template central impact diagram of the ballistocardiogram and is marked as the J peak when the peak begins to slide downwards at the highest position;

if the trace falls from the peak value in the floating range of the distance between the J peak and the L peak of the waveform template center impact diagram of the ballistocardiogram, the L peak is judged.

Further, the step of activating the electro-active shrink element to shrink by a preset amplitude according to the shrink point information further comprises activating the electro-active shrink element within a preset time after the L peak.

Further, the method also comprises the following steps:

and if the heart rate variation exceeds the preset range, dynamically adjusting the waveform template of the heart attack graph through a dynamic time normalization algorithm.

Further, the step of calculating the blood pressure by the non-contact ballistocardiogram includes the steps of:

Extracting time and amplitude information of each characteristic peak in the time domain of the ballistocardiogram;

extracting a frequency domain main component, and calculating time complexity or time fractal information of the frequency domain main component;

and (5) combining the characteristics and inputting the combined characteristics into a Bayesian neural network to obtain the blood pressure.

Further, the construction of the bayesian neural network comprises the following steps:

dividing the ballistocardiogram data into preset time windows;

dividing data by adopting a leave-one-out cross-validation method;

excluding data in the training set, wherein the standard deviation is that the I or J peak value is higher than the preset range of the average value or the continuous diastolic blood pressure change is greater than the preset value;

extracting each characteristic peak and the interval between the characteristic peaks;

performing continuous one-dimensional wavelet transformation by using wavelet kernels, performing Fourier transformation on each frequency spectrum component, extracting long-term frequency domain fluctuation, and calculating time complexity or time fractal information of the long-term frequency domain fluctuation;

constructing a BCG-BP model through each characteristic peak, the interval between the characteristic peaks, the long-term frequency domain fluctuation biological characteristic, the heart rate and the respiratory characteristic;

creating a balanced distribution of blood pressure values using an easyseomb procedure, and generating a balanced distribution of relative blood pressure changes compared to a baseline measurement;

training the data set by adopting a Bayesian neural network, and improving BP estimation of the own data set by using network coefficients of the public data set as a starting point;

Measuring inter-subject and intra-subject correlations between the predicted and reference values using the mean absolute error of blood pressure and performance of a fluctuation assessment algorithm of the in-subject BP relative to baseline;

the measured value for each subject is added to the training process as a calibration data point to adjust the coefficients in the model.

A second object of the present invention is to provide a computer-readable storage medium having stored thereon program instructions that when executed implement a sleep monitoring and improvement method for heart failure patients.

The third object of the present invention is to provide a sleep monitoring and improving device for heart failure patients, which implements the above method, including a piezoelectric sensor array, a controller, and an electro-contraction member; wherein,

the piezoelectric sensing array is used for acquiring ballistocardiogram waveforms in a non-contact mode;

the controller is used for accurately acquiring contraction point information in real time during heartbeat through the ballistocardiogram waveform, and starting the electric contraction piece to contract according to preset amplitude according to the contraction point information; wherein the contraction point information includes L peak position information after the J peak in the ballistocardiogram.

Further, the device also comprises a heating regulation and control module, wherein the controller calculates the blood pressure through the non-contact ballistocardiogram, judges whether the blood pressure is near the target blood pressure, and starts the heating regulation and control module to directionally focus and heat the tiny blood vessel to regulate and control the vascular resistance if the blood pressure is not near the target blood pressure.

Further, the system also comprises a photoelectric sensor, wherein the photoelectric sensor is used for sensing the tip perfusion lifting condition in real time and transmitting the tip perfusion lifting condition to the controller;

the controller adjusts the step length according to the peripheral perfusion lifting condition fed back by the photoelectric sensor until the step length reaches a set maximum value, and if the blood pressure is still not near the target blood pressure, the controller wakes the patient, adjusts the bed or increases the oral medicine.

Further, a plurality of piezoelectric sensors in the piezoelectric sensing array form a plurality of measurement points;

the heating regulation and control module adopts a radio frequency coil;

the electro-shrinkable member is an electro-shrinkable tape.

Further, the electric contraction piece, the heating regulation and control module and the photoelectric sensor are arranged on the intelligent wearing equipment for the lower limbs, the heating regulation and control module and the photoelectric sensor are located in the foot area of the intelligent wearing equipment for the lower limbs, the electric contraction piece is located in the leg area of the intelligent wearing equipment for the lower limbs, the intelligent wearing equipment for the lower limbs is further provided with a wireless communication module, and the electric contraction piece, the heating regulation and control module and the photoelectric sensor are communicated with the controller through the wireless communication module.

Compared with the prior art, the invention has the beneficial effects that:

the invention provides a sleep monitoring and improving method, equipment and a storage medium for heart failure patients, and aims at solving the problem that heart functions are affected by heart load and unsmooth blood pressure backflow in sleep of heart failure patients.

The traditional heart contraction time is provided by the patch electrocardio, so that the accurate position can be kept, and the heart contraction peak is extremely easy to distinguish. But the electrocardiograph plaster needs patients to wear the electrocardiograph plaster by themselves every night, is more troublesome and easy to lose, has uncomfortableness, and for patients with light and medium heart failure, the load exceeds the range of home use. The invention can be used for carrying out time resolution of heart contraction by using a convenient non-contact ballistocardiogram, is simple and comfortable, and can be used for home sleep.

Because the external counterpulsation technology requires accurate real-time heart contraction time, the technology belongs to the technology with higher punishment on false start. That is, if limb compression is initiated during cardiac ejection, there is a damaging effect on the heart. Therefore, in the home environment, the invention accurately acquires the contraction point in real time when the heart beat of the heart failure patient occurs, and starts external counterpulsation after extremely short delay, thereby eliminating the risk of false start. Compared with the traditional external counterpulsation technology, the invention is simple and comfortable, and can be used for sleeping at home; unlike traditional external counterpulsation technology, the invention has the advantages of quick starting, convenient wearing and realization of heart rehabilitation nursing every night.

The expected external counterpulsation can reduce blood pressure to a certain extent, and when the blood pressure is not ideal (still higher than a preset value by more than 10 mmHg), the radio-frequency heating is started to regulate vascular resistance. The blood vessel is relaxed in warm state, and the blood flow resistance is inversely proportional to the fourth power of the radius of the blood vessel, so that the blood flow resistance can be greatly improved under the condition of smaller temperature rise. The traditional heating mode has poor locatability, and the heat radiation easily causes the diastole of the peripheral large blood vessels of the heart to disturb the blood flow force. Since the blood flow resistance is mostly constituted by tiny blood vessels, heating only tiny blood vessels can have minimal intervention, maximal effect, and not disturb the overall blood flow force. Compared with the traditional heating mode, the radio frequency heating mode can be accurately positioned, is quickly started and closed, and can realize the reduction of heart load with minimum heating regulation and control according to a set program.

The intelligent sock integrated photoelectric pulse sensor provided by the invention can sense the peripheral perfusion lifting condition in real time, judge whether vasodilation reaches the expectation or not, and then adjust the step length of radio frequency heating according to the peripheral perfusion lifting condition fed back by the photoelectric sensor. Unlike the traditional regulation and control method which cannot be individuated, the intelligent regulation and control method can intelligently regulate and control the peripheral perfusion lifting condition of the heart failure patient according to the real-time monitoring result, and can furthest realize the comfortable sleep of the heart failure patient.

The foregoing description is only an overview of the present invention, and is intended to provide a better understanding of the present invention, as it is embodied in the following description, with reference to the preferred embodiments of the present invention and the accompanying drawings. Specific embodiments of the present invention are given in detail by the following examples and the accompanying drawings.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiments of the invention and together with the description serve to explain the invention and do not constitute a limitation on the invention. In the drawings:

FIG. 1 is a schematic diagram of a hypothetical and model derived cardiac output curve of the heart failure central cycle pressure/volume relationship;

FIG. 2 is a schematic diagram of sleep monitoring and improving equipment for heart failure patients in example 1;

FIG. 3 is a schematic diagram of BCG and ECG;

FIG. 4 is a schematic view of a sensor under the chest;

FIG. 5 is a schematic diagram showing peak contrast and time difference contrast of BCG measured by a chest sensor;

FIG. 6 is a schematic diagram of sliding window signal matching;

FIG. 7 is a schematic view of dynamic scaling;

FIG. 8 is an in vitro counterpulsation control flow chart;

FIG. 9 is a schematic diagram of fitting blood pressure by BCG;

fig. 10 is a schematic diagram of BCG time domain information extraction;

FIG. 11 is a schematic diagram of BCG frequency domain information extraction;

FIG. 12 is a schematic diagram of the principal components of a two-dimensional spectrum of a patient;

FIG. 13 is a schematic diagram of the principal components of a two-dimensional spectrum of a healthy person;

FIG. 14 is a schematic diagram of a Bayesian neural network;

FIG. 15 is a schematic diagram of knowledge transfer of a public dataset to an owned dataset;

FIG. 16 is a schematic diagram of a conventional external counterpulsation system;

FIG. 17 is a flow chart of sleep monitoring and improvement apparatus for a heart failure patient;

FIG. 18 is a flow chart of a method for sleep monitoring and improving of heart failure patients in example 2;

FIG. 19 is a flow chart of the RF heating control in embodiment 2;

FIG. 20 is a flow chart of vasodilation control of example 2;

fig. 21 is a flowchart of ballistocardiogram data determination in embodiment 2;

FIG. 22 is a flowchart for accurately acquiring systolic information in real time during heartbeat by using the ballistocardiogram waveform in example 2;

FIG. 23 is a flowchart of dynamic adjustment of the waveform template of the ballistocardiogram of embodiment 2;

FIG. 24 is a flowchart for prejudging subsequent ballistocardiogram waveforms according to the real-time trace of example 2;

FIG. 25 is a flowchart for calculating blood pressure from a non-contact ballistocardiogram according to example 2;

FIG. 26 is a Bayesian neural network construction flow chart in accordance with example 2;

fig. 27 is a schematic diagram of a storage medium of embodiment 3.

Detailed Description

The present invention will be further described with reference to the accompanying drawings and detailed description, wherein it is to be understood that, on the premise of no conflict, the following embodiments or technical features may be arbitrarily combined to form new embodiments.

Heart failure is mainly divided into two types, heart failure in systole and heart failure in diastole. The main manifestations are:

(i) Concomitant decline in contractile force, ischemic heart disease and hypertension.

(ii) Changes in diastolic compliance are associated with constrictive pericarditis, pericardial tamponade, amyloidosis, and cardiac hypertrophy.

As shown in fig. 1, (a) in fig. 1 is a pressure/volume ring model of a systolic, failing heart, described by Emax reduction; fig. 1 (b) is a cardiac output curve derived directly from the model in fig. 1 (a) and is compressed compared to a normal heart, (c) in fig. 1 is a diastolic heart failure model described by an increase in diastolic hardness Ds, and (d) in fig. 1 is cardiac output derived from the model in fig. 1 (c) and is also suppressed compared to a normal heart.

For heart failure patients with reduced contractility, the main problem is that the blood flow resistance is excessive, and in the case of damaged contractility, the heart pump blood volume is reduced, the blood pressure is increased, and cardiovascular damage is caused. In diastolic heart failure patients, however, blood reflux is affected, and the afterload of the heart increases, impairing heart function. Generally speaking, heart failure patients have the above two problems with different degrees, and in order to make individual adjustment, the invention provides a sleep monitoring and improving method and equipment for heart failure patients.

Example 1

A sleep monitoring and improving device for heart failure patients, as shown in figure 2, comprises a piezoelectric sensing array, a controller and an electro-contraction piece; wherein,

the piezoelectric sensing array is used for acquiring a Ballistocardiogram (BCG) waveform in a non-contact manner; in this embodiment, a plurality of piezoelectric sensors in the piezoelectric sensor array form a plurality of measurement points; as shown in fig. 2, the piezoelectric sensing array can be buried under the mattress to realize sensing of the heart cycle, blood pressure and blood flow resistance of a patient, and the sensing of the heart cycle can guide the heart recovery of external counterpulsation, so that the afterload of the heart is reduced; the perception of blood pressure can assist in blood flow resistance regulation and control and obtain regulation and control effects.

The controller is used for accurately acquiring contraction point information in real time during heartbeat through a ballistocardiogram waveform, and starting the electric contraction piece to contract according to preset amplitude according to the contraction point information; wherein the contraction point information includes L peak position information after the J peak in the ballistocardiogram.

The external counterpulsation needs to obtain accurate real-time heart contraction time, and belongs to a technology with higher punishment for false start. That is, if limb compression is initiated during cardiac ejection, there is a damaging effect on the heart. Therefore, in the home environment, the risk of false start needs to be eliminated through the system design, and when a certain risk exists, the external counterpulsation regulation and control is stopped, so that the small dose at home is realized, but the regulation and control are accurate and effective.

Traditional heart contraction time is provided by the electrocardio paster, so that the accurate position can be kept, and a heart contraction peak is extremely easy to distinguish. But the electrocardio paster needs the patient to wear by oneself every night, and is comparatively troublesome, easily loses, and has the uncomfortable, to light and moderate heart failure patient, the load surpasses the scope of using at home. The embodiment can be used for carrying out time resolution of heart contraction by using a convenient non-contact ballistocardiogram. As shown in fig. 3, the peaks of BCG are equivalent to ECG, enabling accurate systolic and diastolic times to be obtained.

Because the sleeping posture is not fixed, the peak time of the BCG has unpredictable deviation from the electrocardio standard peak time. In the embodiment, a plurality of measuring points are arranged under the non-contact BCG intelligent mattress, when a patient deviates from the original measuring position remarkably, namely, when the ratio of the maximum amplitude of the BCG between the measuring points is more than 15% compared with a base line, the external counterpulsation activity stops, and only the radio frequency heating regulation and control are reserved. Wherein 15% is an empirical value representing a weight change or height change exceeding about 10kg or 10cm, as shown in the experimental data of fig. 4 and 5.

In order to accurately acquire the contraction points in real time at the time of heartbeat from the long-term BCG signal sequence which has been measured, and to initiate external counterpulsation after an extremely short delay. Because the highest peak (J peak) time of the BCG is related to the measurement position, the embodiment selects the neck BCG to calculate the J peak time, the highest peak value can be selected according to the heart rate period through the peak searching algorithm, the second derivative is calculated to be the smallest, the first derivative is obtained as 0, the second peak (L peak) after the peak value starts the compression of the electric contraction member, and the L peak time can be obtained through the peak searching algorithm, the second derivative is calculated to be the smallest, and the first derivative is obtained as 0. The L peak is chosen for safety, and early initiation of external counterpulsation can affect cardiac ejection, causing serious consequences.

The flow of the external counterpulsation control of the sleep monitoring and improving equipment for heart failure patients is shown in fig. 8, the patients wear the intelligent socks provided with the electro-contraction parts, bluetooth and intelligent mattress pairing is started, and BCG signal acquisition is started. When the patient just lies down and is about to enter a sleep state, a sliding window technology is adopted to obtain a waveform template of the BCG, the following waveform is predicted according to real-time tracing, and the waveform is corrected (10 ms) within a certain change range. For example, in the template BCG waveform, the J peak of BCG is at a position of 200ms, the height is a, and the distance between the J peak and the L peak is 100ms; then in the new online waveform acquisition, if the waveform signal-to-noise ratio is higher than a set threshold (for example, snr=10 is set), the covariance exceeds 0.9, the peak reaches a±10% a, and when the peak starts to slide down at the highest place, the peak is marked as a J peak, when the peak starts to fall down at 100ms±10ms, the peak is judged as an L peak, and in-vitro counterpulsation is started within 20ms after the L peak, and the BCG template is updated. If any one of the conditions is not satisfied, locking the external counterpulsation regulation and control, waiting for the stabilization of BCG signals and the reduction of noise. Because, when the signal-to-noise ratio is too low, the peak cannot be accurately determined; when the adjacent waveform covariance is less than 0.9, it cannot be determined whether an abnormal state occurs, in which case no external counterpulsation is performed for safety; similarly, when the peak difference variation is abrupt, body motion or other anomalies cannot be excluded, in which case no external counterpulsation is performed for safety. If there is a large change in heart rate (e.g., falling asleep from 80bpm to 50 bpm), the template is dynamically adjusted so that the template is available throughout the night, as shown in fig. 6 and 7.

In this embodiment, dynamic time warping (Dynamic Time Warping, DTW) is used to dynamically adjust the template, and is a method for comparing and measuring the similarity between two time series. The main idea is to find the best match between two time series by aligning and normalizing them. In alignment, DTW allows the time series to be warped (warp) in the time dimension to better match similar parts. The present embodiment assumes that warp is uniform, i.e., there is only one scaling factor for one heartbeat, which is represented as a straight line moving left or right as a whole in fig. 7, and the calculation load can be greatly reduced without optimizing each pixel.

DTW can handle length differences and time offsets between time series, which allows bending in the time dimension to accommodate different speeds and durations. Thus, DTW is very useful in handling time series of different speeds, shifts or partial similarities.

In this embodiment, the electro-shrinkable member is an electro-shrinkable tape. The Nafion IPMC (ion exchange polymeric metal material) with high-quality metal electrode can realize large deformability and ultra-fast response speed at the same time, has the electro-contraction capability, and can play a role of artificial muscle. In the embodiment, the electrically-induced contraction strip is added to the leg extension part of the intelligent sock, so that the extra-corporeal counterpulsation in the cardiac cycle can be realized at an ultra-fast speed (the electric control speed is extremely fast and the hysteresis is extremely short compared with an air bag), thereby being beneficial to reducing the heart burden, lowering the blood pressure and being beneficial to heart rehabilitation. In the diastole, the external counterpulsation is to apply pressure on the legs to assist the blood to flow back, relieve the heart load, reduce the blood pressure and promote the generation of new blood vessels. As shown in fig. 16, the conventional external counterpulsation technique is expensive and is generally used only in medical places, and the cardiac cycle is monitored by electrocardio, and three pressure air bags are arranged on the legs to assist the periodic reflux of blood in the cardiac cycle. In the embodiment, a ballistocardiogram is used, namely, accurate time of heart contraction is perceived through piezoelectric ceramics or piezoelectric films in the intelligent mattress, a close-fitting electric sensor is not needed to be worn, and the bed sensor is connected with the intelligent regulation socks through Bluetooth, so that non-inductive butt joint can be realized; the electric contraction strip can be used at home and has ultra-fast response speed. While the strip pressure is less uniform than the balloon, its simplicity and ease of use, long term availability (uninterrupted treatment every night) may compensate for its deficiency.

As shown in fig. 2, the device further comprises a heating regulation and control module, the controller calculates the blood pressure through the non-contact ballistocardiogram, judges whether the blood pressure is near the target blood pressure, and if the blood pressure is not near the target blood pressure, starts the heating regulation and control module to directionally focus and heat the tiny blood vessel to regulate and control vascular resistance. In this embodiment, the heating control module employs a radio frequency coil.

As shown in fig. 2, the system also comprises a photoelectric sensor, wherein the photoelectric sensor is used for sensing the tip perfusion lifting condition in real time and transmitting the tip perfusion lifting condition to the controller;

As shown in fig. 2, the electric contraction piece, the heating regulation module and the photoelectric sensor are arranged on the intelligent lower limb wearing equipment (for example, intelligent socks), the heating regulation module and the photoelectric sensor are located in the foot area of the intelligent lower limb wearing equipment, the electric contraction piece is located in the leg area of the intelligent lower limb wearing equipment, and the intelligent lower limb wearing equipment is further provided with a wireless communication module, and the electric contraction piece, the heating regulation module and the photoelectric sensor are communicated with the controller through the wireless communication module.

The external counterpulsation is expected to reduce the blood pressure to a certain extent, and when the blood pressure is not ideal (still higher than the preset value by more than 10 mmHg), the radio frequency heating regulation is started. The present embodiment obtains blood pressure by BCG without contact and load. Specifically, extracting time and amplitude information of each characteristic peak of the time domain of the BCG, extracting a frequency domain main component, calculating time complexity or time fractal information (such as Higuchi fractal dimension) of the time domain main component, and inputting the time complexity or time fractal information into a Bayesian neural network for training after characteristic combination. The balanced public data set can be used for pre-training to adjust the initialization parameters of the Bayesian neural network, so that excellent performance can be obtained by using a new small data set. The overall flow is shown in fig. 9.

For feature extraction of BCG, the data is segmented into 5 second windows in order to reliably estimate respiration rate; adopting data division with a cross validation to avoid pollution of training and testing data; data with standard deviation of 30% higher I or J peak than average or continuous DBP change greater than 10mmHg in the training set were excluded.

As shown in fig. 10, the features of I, J, K and L peaks, and the spacing between them, were used to construct BCG-BP models. The advantage of using morphological features is that the physiological significance of each peak can be explained. For example, the generation of aortic inlet and outlet blood pressure waves corresponds approximately to the onset of the I wave and the peak of the J wave. However, these features are susceptible to environmental interference and the information contained by the individual features is limited.

The frequency-dependent transmission of vibrations throughout the body. For BCG signals, a continuous one-dimensional wavelet transform is performed using a Morlet wavelet kernel, and then fourier transform is performed on each spectral component (amplitude) to extract long-term frequency domain fluctuations. Thus, the frequency-time representation of the signal is replaced by the frequency-frequency representation. Fig. 11 provides a detailed illustration of this two-step spectral analysis, resulting in a two-dimensional spectrum with high frequency detail and long-term periodicity.

Central frequency of Fourier transformf ₀ ) Different for each measurement, should be near the heart rate. By selectingf ₀ -0.25 to 0.25 Hzf ₀ The fourier frequency range of +0.25 Hz and the wavelet frequency of 2 to 14 Hz.

To reduce redundant information, the spectrum shown in fig. 11 is compressed by reshaping the two-dimensional matrix into a one-dimensional array and applying Singular Value Decomposition (SVD). Each dataset calculates the first 6 most significant components, which are then remodelled back into the 2D spectrum. The most important harmonics may contain most of the information. The healthy BCG dataset variation and the patient dataset variation are different. A possible explanation is the increased heterogeneity of cardiovascular activity in hospitalized patients. The basis functions in the frequency domain are shown in fig. 12 and 13.

The human body can be mechanically modeled as a combination of springs and dampers. Each heartbeat occurs at a different intensity and time, which results in body and limb vibrations having different time scales. The fractal dimension may be used to estimate the chaotic pattern caused by the propagation of vibrations in the cardiovascular tree. This chaos phenomenon is described using Higuchi fractal dimension in this example.

This embodiment adds biometric features, heart rate and respiration to ensure the integrity of the feature information.

To build the BCG-BP model, input individual information such as age, weight and height is used. Age affects mainly vascular elasticity, with non-negligible impact on aortic blood pressure and blood flow velocity. If the human body is modeled as a distributed weight with connected springs and dampers, the transfer function from the central blood pressure to the body movements will become highly personalized. Weight and height are required to estimate the transfer function.

The study found that age was significantly correlated with systolic blood pressure in healthy populations as shown in table 1. In contrast, age has no correlation with patient systolic or diastolic blood pressure. The correlation between patient weight and blood pressure was small, indicating irregular vibration signaling.

TABLE 1 Pearson correlation coefficient (r) of physiological information and BP of subject

* Indicating that there is a significant correlation between physiological parameters and SBP/DBP.

Respiration regulates the lower frequency (0.2-0.6 Hz) in BCG signals and blood pressure fluctuations. The BCG signal is passed through a butterworth band-pass filter with cut-off frequencies of 0.1Hz and 0.5 Hz. Advanced peak detection algorithms are employed to reduce the impact of environmental interference. Since beat-by-beat heart beat frequency is not required, the main frequency is usedf ₀ To calculate heart rate.

If the measured data is severely unbalanced, the learned algorithm will focus on reducing errors around the median blood pressure, thereby reducing discrimination. To solve this problem and improve the accuracy of blood pressure estimation, two easy Ensemble flows are employed. The first procedure aims at creating a balanced distribution of blood pressure values. SBP readings between 85mmHg and 200mmHg are divided into intervals of 5mmHg. For DBP, the range is 55mmHg to 100mmHg. In each interval, the same number of samples is randomly selected to ensure balance. The second procedure attempts to produce an equilibrium distribution of relative blood pressure changes compared to the baseline measurement (Δbp). A baseline was established by averaging the first 5 measurements of the dataset. Subsequent blood pressure readings were resampled to create a uniformly distributed ΔBP range ranging from-50 mmHg to 50mmHg in steps of 5mmHg. If the number of samples in the interval is insufficient, then all available data is included. These balanced data sets are then used to generate training regression algorithms and evaluation test data. Notably, the test data need not be balanced. Multiple resampling iterations are performed and the resulting algorithm is averaged to enhance. The combination of multiple Δbps will produce a new Δblood pressure based on the selected starting baseline and the time interval between the two. The impact of these factors on the performance of the algorithm depends on the distribution of the data.

The choice of features is also important for blood pressure estimation. In this study, the emphasis was not on a single feature. Instead, several feature combining schemes select the frequency domain + time domain approach to generate the final algorithm.

To calculate blood pressure using the selected features as input, the present embodiment chooses to use a bayesian neural network, as shown in fig. 14. The contribution of features to a target varies from query point to query point. Bayesian optimization can apply probability distributions over a possible neural network, alleviating the over-fitting problem, making it more suitable for algorithms with input correlation coefficients. The independent BNN (Beyesian neural network, bayesian neural network) was initially trained on both data sets. The BP estimation for the owned dataset is then improved using the network coefficients of the public dataset as a starting point.

Inter-and intra-subject correlations between predicted and reference values were measured using mean absolute error of blood pressure (MAE) and intra-subject BP fluctuations from baseline (Δsbp and Δdbp) to assess the performance of the algorithm.

The measured value for each subject is added to the training process as a calibration data point to adjust the coefficients in the model. Furthermore, the calibration factor is subtracted from the subsequent blood pressure estimate to improve accuracy.

Transfer learning involves transferring knowledge learned from a pre-trained model (source domain) to a new, similar but different task (target domain). Such as PPG-BP model. In this embodiment, there are two reasons for adopting the transfer learning technique. First, some studies have also shown that BCG measured by different techniques can be converted to each other. Second, a large public database with comparable sensing technology can be accessed, which makes the transfer learning less complex, especially if similar preprocessing steps are employed. This process is shown in fig. 15.

The analytical features obtained from the ceramic sensor were scaled by amplitude to match the sensitivity of the EMFi sensor, while the frequency domain features were resampled at 100 Hz. The present embodiment does not use default coefficients to initiate BNN training, but rather utilizes pre-trained BNN in the dataset-KSU. Fine tuning is performed by progressively thawing the layers and adapting the weights to the new dataset. This approach can take advantage of the learned features of the pre-trained model to achieve better performance using less labeled data in the target domain. The data processed by the method and the built model can obtain blood pressure evaluation results, as shown in table 2.

Table 2 comparison of the new protocol with published protocols

TD is time domain, FD is frequency domain, HFD is fractal index, and Bios is sign information (age, height, weight).

As shown in fig. 17, when the blood pressure is still high, the vascular resistance can be regulated by adopting a proper radio-frequency heating mode. The blood vessel is relaxed in warm state, and the blood flow resistance is inversely proportional to the fourth power of the radius of the blood vessel, so that the blood flow resistance can be greatly improved under the condition of smaller temperature rise. The traditional heating mode has poor locatability, and the heat radiation easily causes the diastole of the peripheral large blood vessels of the heart to disturb the blood flow force. Since the blood flow resistance is mostly constituted by tiny blood vessels, heating only tiny blood vessels can have minimal intervention, maximal effect, and not disturb the overall blood flow force. Rf heating of skin is a non-invasive, localized, precision heating technique in the medical field. The device heats the tissue through the skin surface by using radio frequency energy, thereby generating a series of physiological effects, and the directional focusing of the energy can be realized through the design of the coil. The principle is that a high frequency current generates an electric resistance in skin tissue, thereby generating heat energy. The radio frequency heating can promote blood circulation, increase the delivery of oxygen and nutrients to skin tissue, promote metabolism and cell regeneration. Another advantage of radio frequency heating is that the heating intensity can be quickly started and adjusted, and flexible and efficient time series coding can be performed in conjunction with the monitoring device. In addition, the intelligent sock is integrated with the photoelectric pulse sensor, so that the peripheral perfusion lifting condition can be sensed in real time, and whether vasodilation reaches the expectation is judged. Wherein the peripheral perfusion is the fluctuating amplitude of the pulse/baseline of the pulse.

Example 2

The sleep monitoring and improving method for heart failure patients corresponding to the sleep monitoring and improving device for heart failure patients provided in embodiment 1 may refer to the corresponding description in the above device embodiments for the detailed description of the device, and will not be repeated here. As shown in fig. 18, the method includes the steps of:

s1, acquiring a ballistocardiogram waveform acquired in a non-contact manner; the collection of the ballistocardiogram is realized through a piezoelectric sensing array, and in the embodiment, a plurality of piezoelectric sensors in the piezoelectric sensing array form a plurality of measuring points; as shown in fig. 2, the piezoelectric sensing array can be buried under the mattress to realize sensing of the heart cycle, blood pressure and blood flow resistance of a patient, and the sensing of the heart cycle can guide the heart recovery of external counterpulsation, so that the afterload of the heart is reduced; the perception of blood pressure can assist in blood flow resistance regulation and control and obtain regulation and control effects.

Because the sleeping posture is not fixed, the peak time of the BCG has unpredictable deviation from the electrocardio standard peak time. As shown in fig. 21, after the step of acquiring the ballistocardiogram waveform acquired in a non-contact manner, the steps of:

s11, comparing the maximum amplitude ratio of the ballistocardiogram among a plurality of measurement points with a base line; in the embodiment, a plurality of measuring points are arranged under the non-contact BCG intelligent mattress, and when a patient deviates from the original measuring position remarkably, namely, the ratio of the maximum amplitude of the BCG among the plurality of measuring points is compared with a baseline;

S12, judging whether the change exceeds a preset range;

and S13, stopping the contraction of the electric contraction member if the change exceeds a preset range. I.e. when the change exceeds 15%, the in vitro counterpulsation activity stops, and only the radio frequency heating regulation and control is reserved. Wherein 15% is an empirical value representing a weight change or height change exceeding about 10kg or 10cm, as shown in the experimental data of fig. 4 and 5.

And S14, if the change does not exceed the preset range, continuing to execute the subsequent steps.

S2, accurately acquiring contraction point information in real time during heartbeat through a ballistocardiogram waveform; the contraction point information comprises L peak position information behind a J peak in the ballistocardiogram;

As shown in fig. 22, the step of accurately acquiring the contraction point information in real time at the time of heartbeat by the ballistocardiogram waveform includes the steps of:

s21, obtaining a waveform template of the ballistocardiogram through a sliding window; the waveform template of the ballistocardiogram comprises the position and the height of a J peak of the ballistocardiogram and the distance between the J peak and an L peak; for example, in the template BCG waveform, the J peak of BCG is at a position of 200ms, the height is a, and the distance between the J peak and the L peak is 100ms;

S22, pre-judging the subsequent ballistocardiogram waveform according to real-time tracing, and correcting (10 ms) within a preset range. As shown in fig. 24, the method specifically comprises the following steps:

s221, if the signal-to-noise ratio of the acquired newly acquired online waveform is higher than a set threshold (for example, SNR=10 is set), the covariance exceeds the set threshold (for example, the set threshold is 0.9), the peak value reaches the floating range of the height of the J peak of the central impact diagram of the waveform template of the ballistocardiogram (for example, the floating range of the height of the J peak of the central impact diagram of the waveform template of the ballistocardiogram is A+/-10 percent A), and the peak value is marked as the J peak when the peak begins to slide downwards;

s222, judging that the L peak is in a floating range (100 ms+/-10 ms) of the distance between the J peak and the L peak of the waveform template center impact diagram of the ballistocardiogram when the trace starts to decline from the peak. And, can start external counterpulsation within 20ms after L peak, update BCG template. If any one of the conditions is not satisfied, locking the external counterpulsation regulation and control, waiting for the stabilization of BCG signals and the reduction of noise. Because, when the signal-to-noise ratio is too low, the peak cannot be accurately determined; when the adjacent waveform covariance is less than 0.9, it cannot be determined whether an abnormal state occurs, in which case no external counterpulsation is performed for safety; similarly, when the peak difference variation is abrupt, body motion or other anomalies cannot be excluded, in which case no external counterpulsation is performed for safety. If there is a large change in heart rate (e.g., falling asleep from 80bpm to 50 bpm), the template is dynamically adjusted so that the template is available throughout the night, as shown in fig. 6 and 7.

As shown in fig. 23, the method further comprises the steps of:

s23, if the heart rate variation exceeds a preset range, dynamically adjusting the waveform template of the heart attack graph through a dynamic time normalization algorithm.

S3, starting the electric contraction piece to contract according to the preset amplitude according to the contraction point information. In this example, the electro-constriction is activated within a preset time (20 ms) after the L peak.

The external counterpulsation is expected to reduce the blood pressure to a certain extent, and when the blood pressure is not ideal (still higher than the preset value by more than 10 mmHg), the radio frequency heating regulation is started. As shown in fig. 2, the electric contraction piece, the heating regulation module and the photoelectric sensor are arranged on the intelligent lower limb wearing equipment (for example, intelligent socks), the heating regulation module and the photoelectric sensor are located in the foot area of the intelligent lower limb wearing equipment, the electric contraction piece is located in the leg area of the intelligent lower limb wearing equipment, and the intelligent lower limb wearing equipment is further provided with a wireless communication module, and the electric contraction piece, the heating regulation module and the photoelectric sensor are communicated with the controller through the wireless communication module.

As shown in fig. 17, when the blood pressure is still high, the vascular resistance can be regulated by adopting a proper radio-frequency heating mode. The blood vessel is relaxed in warm state, and the blood flow resistance is inversely proportional to the fourth power of the radius of the blood vessel, so that the blood flow resistance can be greatly improved under the condition of smaller temperature rise. The traditional heating mode has poor locatability, and the heat radiation easily causes the diastole of the peripheral large blood vessels of the heart to disturb the blood flow force. Since the blood flow resistance is mostly constituted by tiny blood vessels, heating only tiny blood vessels can have minimal intervention, maximal effect, and not disturb the overall blood flow force. Rf heating of skin is a non-invasive, localized, precision heating technique in the medical field. The device heats the tissue through the skin surface by using radio frequency energy, thereby generating a series of physiological effects, and the directional focusing of the energy can be realized through the design of the coil. The principle is that a high frequency current generates an electric resistance in skin tissue, thereby generating heat energy. The radio frequency heating can promote blood circulation, increase the delivery of oxygen and nutrients to skin tissue, promote metabolism and cell regeneration. Another advantage of radio frequency heating is that the heating intensity can be quickly started and adjusted, and flexible and efficient time series coding can be performed in conjunction with the monitoring device.

As shown in fig. 19, after the step of activating the electric shrinkable member to shrink by a preset amplitude according to the shrink point information, the steps of:

s4, calculating blood pressure through a non-contact ballistocardiogram;

s5, judging whether the blood pressure is near the target blood pressure;

and S6, if the blood pressure is not near the target blood pressure, starting the heating regulation and control module to directionally focus and heat the tiny blood vessel to regulate and control the vascular resistance.

In addition, the intelligent sock is integrated with the photoelectric pulse sensor, so that the peripheral perfusion lifting condition can be sensed in real time, and whether vasodilation reaches the expectation is judged. Wherein the peripheral perfusion is the fluctuating amplitude of the pulse/baseline of the pulse.

As shown in fig. 20, the following steps are further included after the step of activating the heating regulation module:

s7, adjusting the regulating step length of the heating regulating module according to the received peripheral perfusion lifting condition until the regulating step length reaches a set maximum value;

s8, if the blood pressure is still not near the target blood pressure, waking up the patient, adjusting the bed or adding oral medicines.

The present embodiment obtains blood pressure by BCG without contact and load. Specifically, as shown in fig. 25, the step of calculating the blood pressure by the non-contact ballistocardiogram includes the steps of:

s41, extracting time and amplitude information of each characteristic peak in the time domain of the ballistocardiogram;

S42, extracting a frequency domain principal component, and calculating time complexity or time fractal information (such as Higuchi fractal dimension) of the frequency domain principal component;

s43, inputting the combined features into a Bayesian neural network to obtain the blood pressure.

To calculate blood pressure using the selected features as input, the present embodiment chooses to use a bayesian neural network, as shown in fig. 14. The contribution of features to a target varies from query point to query point. Bayesian optimization can apply probability distributions over a possible neural network, alleviating the over-fitting problem, making it more suitable for algorithms with input correlation coefficients.

As shown in fig. 26, the construction of the bayesian neural network includes the steps of:

s400, dividing the ballistocardiogram data into preset time windows; in order to extract the characteristics of the BCG, the embodiment divides the data into 5 second windows so as to reliably estimate the respiratory rate;

s401, dividing data by adopting a leave-one-out cross-validation method so as to avoid pollution of training and testing data;

s402, excluding data in which the standard deviation in the training set is that the I or J peak value is higher than the preset range of the average value or the continuous diastolic blood pressure change is greater than the preset value; for example, data with standard deviation of 30% higher I or J peaks than average or continuous DBP change greater than 10mmHg in the training set were excluded.

S403, extracting each characteristic peak and the interval between the characteristic peaks; as shown in fig. 10, the features of I, J, K and L peaks, and the spacing between them, were used to construct BCG-BP models. The advantage of using morphological features is that the physiological significance of each peak can be explained. For example, the generation of aortic inlet and outlet blood pressure waves corresponds approximately to the onset of the I wave and the peak of the J wave. However, these features are susceptible to environmental interference and the information contained by the individual features is limited.

The frequency-dependent transmission of vibrations throughout the body. S404, for BCG signals, performing continuous one-dimensional wavelet transformation by using wavelet kernels, then performing Fourier transformation on each frequency spectrum component (amplitude), extracting long-term frequency domain fluctuation, and calculating time complexity or time fractal information of the long-term frequency domain fluctuation; thus, the frequency-time representation of the signal is replaced by the frequency-frequency representation. Fig. 11 provides a detailed illustration of this two-step spectral analysis, resulting in a two-dimensional spectrum with high frequency detail and long-term periodicity.

Central frequency of Fourier transform) Different for each measurement, should be near the heart rate. By selectingf ₀ -0.25 to 0.25 Hzf ₀ The fourier frequency range of +0.25 Hz and the wavelet frequency of 2 to 14 Hz.

S405, constructing a BCG-BP model through each characteristic peak, the interval between the characteristic peaks, the long-term frequency domain fluctuation biological characteristic, the heart rate and the breathing characteristic;

in this embodiment, to build the BCG-BP model, input individual information such as age, weight, and height is used. Age affects mainly vascular elasticity, with non-negligible impact on aortic blood pressure and blood flow velocity. If the human body is modeled as a distributed weight with connected springs and dampers, the transfer function from the central blood pressure to the body movements will become highly personalized. Weight and height are required to estimate the transfer function.

Respiration regulates the lower frequency (0.2-0.6 Hz) in BCG signals and blood pressure fluctuations. The BCG signal is passed through a butterworth band-pass filter with cut-off frequencies of 0.1Hz and 0.5 Hz. Advanced peak detection algorithms are employed to reduce the impact of environmental interference. Since beat-by-beat heart beat frequency is not required, the main frequency is used To calculate heart rate. />

If the measured data is severely unbalanced, the learned algorithm will focus on reducing errors around the median blood pressure, thereby reducing discrimination. To solve this problem and improve the accuracy of blood pressure estimation, S406, creating a balanced distribution of blood pressure values using the easyseomb procedure, and generating a balanced distribution of relative blood pressure changes compared to the baseline measurement;

this embodiment employs two easy Ensemble flows. The first procedure aims at creating a balanced distribution of blood pressure values. SBP readings between 85mmHg and 200mmHg are divided into intervals of 5mmHg. For DBP, the range is 55mmHg to 100mmHg. In each interval, the same number of samples is randomly selected to ensure balance. The second procedure attempts to produce an equilibrium distribution of relative blood pressure changes compared to the baseline measurement (Δbp). A baseline was established by averaging the first 5 measurements of the dataset. Subsequent blood pressure readings were resampled to create a uniformly distributed ΔBP range ranging from-50 mmHg to 50mmHg in steps of 5mmHg. If the number of samples in the interval is insufficient, then all available data is included. These balanced data sets are then used to generate training regression algorithms and evaluation test data. Notably, the test data need not be balanced. Multiple resampling iterations are performed and the resulting algorithm is averaged to enhance. The combination of multiple Δbps will produce a new Δblood pressure based on the selected starting baseline and the time interval between the two. The impact of these factors on the performance of the algorithm depends on the distribution of the data.

S407, training the data set by adopting a Bayesian neural network, and improving BP estimation of the own data set by using a network coefficient of the public data set as a starting point; in the embodiment, the balanced public data set is used for pre-training to adjust the initialization parameters of the Bayesian neural network, so that excellent performance can be obtained by using a new small data set.

S408, measuring inter-and intra-subject correlations between predicted and reference values using Mean Absolute Error (MAE) of blood pressure and fluctuations in BP relative to baseline (ΔSBP and ΔDBP) in the subject to evaluate the performance of the algorithm;

s409, adding the measured value of each subject as a calibration data point to the training process to adjust the coefficients in the model.

Furthermore, the calibration factor is subtracted from the subsequent blood pressure estimate to improve accuracy.

Table 2 comparison of the new protocol with published protocols

Example 3

A computer readable storage medium, as shown in fig. 27, having stored thereon program instructions that when executed implement a method for sleep monitoring and improvement of a heart failure patient. For detailed description of the method, reference may be made to corresponding descriptions in the above method embodiments, and details are not repeated here.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.

The foregoing is illustrative of the embodiments of the present disclosure and is not to be construed as limiting the scope of the one or more embodiments of the present disclosure. Various modifications and alterations to one or more embodiments of this description will be apparent to those skilled in the art. Any modifications, equivalent substitutions, improvements, or the like, which are within the spirit and principles of one or more embodiments of the present disclosure, are intended to be included within the scope of the claims of one or more embodiments of the present disclosure.

Claims

1. A sleep monitoring and improving method for heart failure patients, which is characterized by comprising the following steps:

acquiring a ballistocardiogram waveform acquired in a non-contact manner;

starting the electric contraction piece to contract according to the preset amplitude according to the contraction point information;

the step of accurately acquiring the contraction point information in real time during the heartbeat through the ballistocardiogram waveform comprises the following steps of:

pre-judging the subsequent ballistocardiogram waveform according to real-time tracing and correcting the subsequent ballistocardiogram waveform in a preset range;

the step of pre-judging the subsequent ballistocardiogram waveform according to the real-time trace and correcting within a preset range comprises the following steps:

2. A method for sleep monitoring and improvement in patients with heart failure as claimed in claim 1, wherein: the step of starting the electric contraction piece to contract according to the preset amplitude according to the contraction point information further comprises the following steps:

calculating blood pressure through a non-contact ballistocardiogram;

Judging whether the blood pressure is near the target blood pressure;

3. A method for sleep monitoring and improvement in patients with heart failure as claimed in claim 2, wherein: the method further comprises the following steps after the step of starting the heating regulation module:

4. A method for sleep monitoring and improvement in a heart failure patient according to claim 2 or 3, further comprising the step, after said step of acquiring a non-contact acquired ballistocardiogram waveform, of:

5. A method for sleep monitoring and improvement in patients with heart failure as claimed in claim 1, wherein: the step of starting the electro-active shrink element to shrink according to the shrink point information and the step of starting the electro-active shrink element according to the preset amplitude further comprises starting the electro-active shrink element within the preset time after the L peak.

6. The method for sleep monitoring and improvement of a heart failure patient according to claim 1, further comprising the steps of:

7. A method for sleep monitoring and improvement in patients with heart failure as claimed in claim 2, wherein: the step of calculating the blood pressure by the non-contact ballistocardiogram comprises the steps of:

8. The method for sleep monitoring and improvement in a heart failure patient according to claim 7, wherein: the construction of the Bayesian neural network comprises the following steps:

dividing the ballistocardiogram data into preset time windows;

dividing data by adopting a leave-one-out cross-validation method;

9. A computer readable storage medium, having stored thereon program instructions which, when executed, implement the method of claim 1.

10. Sleep monitoring and improving equipment for heart failure patients, implementing the method according to any one of claims 1-8, characterized in that: the device comprises a piezoelectric sensing array, a controller and an electrostriction piece; wherein,

11. A sleep monitoring and improvement apparatus for heart failure patients as claimed in claim 10, wherein: the controller calculates the blood pressure through the non-contact ballistocardiogram, judges whether the blood pressure is near the target blood pressure, and starts the heating regulation module to directionally focus and heat the tiny blood vessel to regulate and control vascular resistance if the blood pressure is not near the target blood pressure.

12. A sleep monitoring and improvement apparatus for heart failure patients as claimed in claim 11, wherein: the system also comprises a photoelectric sensor, wherein the photoelectric sensor is used for sensing the tip perfusion lifting condition in real time and transmitting the tip perfusion lifting condition to the controller;

13. A sleep monitoring and improvement apparatus for heart failure patients as claimed in claim 11, wherein: a plurality of piezoelectric sensors in the piezoelectric sensing array form a plurality of measurement points;

the heating regulation and control module adopts a radio frequency coil;

the electro-shrinkable member is an electro-shrinkable tape.

14. A sleep monitoring and improvement apparatus for heart failure patients as claimed in claim 12, wherein: the intelligent lower limb wearing equipment comprises a lower limb intelligent wearing equipment, a heating regulation module and a photoelectric sensor, wherein the heating regulation module is arranged on the intelligent lower limb wearing equipment, the photoelectric sensor is positioned in a foot area of the intelligent lower limb wearing equipment, the heating regulation module is positioned in a leg area of the intelligent lower limb wearing equipment, a wireless communication module is further arranged on the intelligent lower limb wearing equipment, and the heating regulation module is communicated with the controller through the wireless communication module.