Smart bracelet for remotely monitoring and predicting epileptic seizure and computing device

Abstract

The invention discloses an intelligent bracelet for remotely monitoring and predicting epileptic seizure and computing equipment, and the intelligent bracelet comprises a first processing module which is used for analyzing physiological signal data of a patient collected by a first server, and generating preliminary result data of epileptic seizure; the preliminary result data is stored in a first data structure form; the second processing module is used for sending the first data structure to a second server, and performing secondary processing and conversion on the first data structure to generate a second data structure; the prediction analysis module is used for combining the prediction probability, the epileptic seizure risk score and the event triggering time into a third data structure according to the second data structure; and the bracelet display module is used for sending the third data structure to the nurse wearing bracelet, and the bracelet carries out vibration and flickering prompt according to the third data structure. Warning of different degrees is given according to different risk levels, nurses can take measures in time, and epilepsy danger of patients is avoided.

Description

Technical Field

[0001] The present invention relates to the field, and particularly to an intelligent bracelet and a computing device for remotely monitoring and predicting epileptic seizures. Background Art

[0002] Traditional epileptic seizure monitoring mainly relies on clinical observation and electroencephalogram (EEG) recording. However, in practical applications, especially in the application on the nurse side, there are still some problems and defects. Since traditional monitoring methods often rely on manual observation of EEG data, real-time warning of epileptic seizures cannot be achieved. Even if a remote monitoring system is introduced, due to problems such as data transmission delay and insufficient data processing capacity, the warning information cannot be delivered to nurses or relevant personnel in a timely manner. Moreover, when the existing remote monitoring system sends warning information to nurses, it often simply sends an alarm signal, lacking detailed risk assessment information and clear intervention guidance. After receiving the warning, it is difficult for nurses to quickly judge the specific situation of the patient and take effective intervention measures, missing the best treatment time.

[0003] To solve the above problems, the present invention proposes an intelligent bracelet for remotely monitoring and predicting epileptic seizures, which can give different levels of warnings according to different risk levels, enabling nurses to take timely measures to avoid the danger of epileptic seizures for patients. Summary of the Invention

[0004] In view of the above problems, the present invention provides an intelligent bracelet and a computing device for remotely monitoring and predicting epileptic seizures.

[0005] According to one aspect of the present invention, there is provided an intelligent bracelet for remotely monitoring and predicting epileptic seizures, comprising:

[0006] A first processing module, configured to analyze the patient physiological signal data collected by a first server to generate preliminary result data of epileptic seizures. The physiological signal data includes electroencephalogram, heart rate variability, and activity level. The preliminary result data is stored in a first data structure, and the first data structure includes a timestamp, a physiological data feature vector, and a preliminary risk assessment value;

[0007] A second processing module, configured to send the first data structure to a second server, perform secondary processing and conversion on the first data structure to generate a second data structure, and the second data structure includes power spectral density, waveform features, and time series patterns;

[0008] A prediction analysis module, which is used to perform prediction analysis based on the second data structure to obtain the predicted probability of a patient's epileptic seizure; generate an epileptic seizure risk score according to the predicted probability and clinical experience rules; combine the predicted probability, the epileptic seizure risk score, and the event trigger time into a third data structure; wherein, the third data structure includes an epileptic risk score level and an epileptic warning status identifier, and the epileptic warning status identifier includes an icon code, a color code, and curve data;

[0009] A bracelet display module, which is used to send the third data structure to a nurse-worn bracelet, and the bracelet gives vibration and flashing prompts according to the third data structure.

[0010] In an alternative manner, in the second processing module, a wavelet transform feature extraction method is used to perform secondary processing on the first data structure, calculate the energy distribution of the signal at different scales to obtain the power spectral density, and identify the waveform features according to the largest Lyapunov exponent.

[0011] In an alternative manner, the calculation formula of the power spectral density is:

[0012]

[0013] wherein, PSD(f) is the power spectral density at frequency f; x(n) is the time series data; N is the number of data points; w(n) is the window function.

[0014] In an alternative manner, in the prediction analysis module, random forest and gradient boosting tree are used to weight the epileptic seizure risk score, and the weight formula for the weighting is:

[0015]

[0016] wherein, w i is the weight of the i-th model; AUC i is the area under the curve of the i-th model; m is the total number of models; γ is the weight adjustment factor.

[0017] In an alternative manner, in the bracelet display module, the calculation formulas for giving vibration and flashing prompts according to the epileptic risk score level are respectively:

[0018] f vib = a·RiskLevel e + b·sin(RiskLevel·π)

[0019] L flash = c·RiskLevel d + f·log(1 + RiskLevel)

[0020] Among them, f vib is the vibration frequency, and L flash is the flashing brightness; RiskLevel is the risk score level; a, b, c, d, e, and f are preset parameters.

[0021] In an alternative manner, in the second processing module, Morlet wavelet is used as the wavelet basis function for feature extraction, and its expression is:

[0022] ψ(t) = π -1 / 4 exp(-t 2 / 2) exp(i2πf0t)

[0023] Among them, f0t is the center frequency; t is the time.

[0024] In an alternative manner, in the prediction and analysis module, early stopping method is used to fit the random forest and gradient boosting tree, and the formula for determining the number of training rounds to stop training is:

[0025] T = argmin L val (t) | t ∈ 1, 2,..., T max , ΔL val (t) < ε

[0026] Among them, L val (t) is the loss on the validation set after the t-th round of training; ΔL val (t) is the difference in the loss on the validation set after two consecutive rounds of training; ε is the preset threshold; T max is the maximum number of training rounds.

[0027] In an alternative manner, it further includes an adaptive calibration module for iteratively estimating the independent components in the EEG signal according to the fast independent component analysis method.

[0028] In an alternative manner, the iterative formula for the independent components in the EEG signal is:

[0029] w (k+1) = E[xg(w (k)T x)] - diag(E[g ′ (w (k)T x)])w (k)

[0030] Among them, w (k) is the weight vector at the k-th iteration; x is the observed signal vector; g(.) is the tanh function; g ′ (.) is the derivative of g(.); E[.] is the expectation function; diag(.) is the operation of converting a vector into a diagonal matrix.

[0031] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface, and a communication bus, through which the processor, the memory, and the communication interface complete communication with each other;

[0032] The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the above-mentioned smart bracelet for remotely monitoring and predicting epileptic seizures.

[0033] The solution provided by the present invention includes a first processing module, which is used to analyze the physiological signal data of a patient collected by a first server to generate preliminary result data of epileptic seizures. The physiological signal data includes electroencephalogram, heart rate variability, and activity level. The preliminary result data is stored in the form of a first data structure, and the first data structure includes a timestamp, a physiological data feature vector, and a preliminary risk assessment value. A second processing module is used to send the first data structure to a second server, perform secondary processing and conversion on the first data structure to generate a second data structure, and the second data structure includes power spectral density, waveform features, and time series patterns. A prediction analysis module is used to perform prediction analysis according to the second data structure to obtain the prediction probability of the patient having an epileptic seizure. According to the prediction probability and clinical experience rules, an epileptic seizure risk score is generated. The prediction probability, the epileptic seizure risk score, and the event trigger time are combined into a third data structure. The third data structure includes an epileptic risk score level and an epileptic warning status identifier, and the epileptic warning status identifier includes an icon code, a color code, and curve data. A bracelet display module is used to send the third data structure to a bracelet worn by a nurse, and the bracelet gives vibration and flashing prompts according to the third data structure. By integrating the first processing module and the second processing module, the present invention further analyzes the electroencephalogram, heart rate variability, and activity level physiological signal data of the patient, and extracts power spectral density, waveform features, and time series patterns, improving the prediction probability of epileptic seizures and the accuracy of risk score. By providing warning prompts to the nurse through the vibration frequency and flashing brightness, the nurse can quickly respond to avoid danger to the patient, thereby improving medical efficiency and safety.

[0034] The above description is only an overview of the technical solution of the present invention. In order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the description. And in order to make the above and other purposes, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention are hereinafter specifically exemplified. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiments. The drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Also, throughout the drawings, the same reference numerals are used to denote the same components. In the drawings:

[0036] Figure 1 A schematic framework diagram of an intelligent bracelet for remote monitoring and prediction of epileptic seizures according to an embodiment of the present invention is shown;

[0037] Figure 2 A schematic diagram of electroencephalogram signal acquisition according to an embodiment of the present invention is shown;

[0038] Figures 3a to 3b A schematic diagram of the change of physiological signal data according to an embodiment of the present invention is shown;

[0039] Figure 4 A schematic structural diagram of a computing device according to an embodiment of the present invention is shown. Detailed Embodiments

[0040] The exemplary embodiments of the present invention will be described in more detail below with reference to the drawings. Although the exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present invention can be more thoroughly understood and the scope of the present invention can be fully conveyed to those skilled in the art.

[0041] Figure 1 A schematic framework diagram of an intelligent bracelet for remote monitoring and prediction of epileptic seizures according to an embodiment of the present invention is shown. Specifically, as Figure 1 shown, it includes:

[0042] A first processing module, configured to analyze the patient physiological signal data collected by the first server to generate preliminary result data of epileptic seizures, wherein the physiological signal data includes electroencephalogram, heart rate variability, and activity level; the preliminary result data is stored in the form of a first data structure, and the first data structure includes a timestamp, a physiological data feature vector, and a preliminary risk assessment value;

[0043] A second processing module, configured to send the first data structure to a second server, perform secondary processing and conversion on the first data structure to generate a second data structure, and the second data structure includes power spectral density, waveform features, and time series patterns;

[0044] A prediction analysis module for performing prediction analysis based on the second data structure to obtain the predicted probability of a patient's epileptic seizure; generating an epileptic seizure risk score according to the predicted probability and clinical experience rules; combining the predicted probability, the epileptic seizure risk score, and the event trigger time into a third data structure; wherein the third data structure includes an epileptic risk score level and an epileptic warning status identifier, and the epileptic warning status identifier includes an icon code, a color code, and curve data;

[0045] A bracelet display module for sending the third data structure to a nurse-worn bracelet, and the bracelet gives vibration and flashing prompts according to the third data structure.

[0046] In this embodiment, multi-source physiological signals such as electroencephalogram, heart rate variability (HRV), and activity level are fused to more comprehensively capture the physiological changes before an epileptic seizure (as Figures 3a to 3b shown). Among them, heart rate variability and activity level reflect the state of the autonomic nervous system and the overall activity of the body. The data processing is divided into two hierarchical processing methods: preliminary processing (the first processing module) and secondary processing (the second processing module) to avoid a large amount of calculations on the bracelet device. Among them, the preliminary processing uses lightweight algorithms and is performed on resource-constrained devices (such as bracelets); the secondary processing performs more complex calculations on the server side. The first, second, and third data structures are used to store and transmit data, which is convenient for data interaction and information sharing between different modules. Multiple visualization methods such as icon codes, color codes, and curve data are used to convey the warning information to the nursing staff, so that they can more intuitively understand the risk status of the patient. Through the vibration and flashing prompts of the bracelet, the medical staff can be notified in time of the patient's epileptic seizure risk, make a quick response, and take intervention measures in time.

[0047] Specifically, the electroencephalogram, heart rate variability, and activity level data of the patient are obtained through the first processing module (bracelet or edge computing device). As Figure 2 shown, the frequency domain features of the electroencephalogram (such as the energy of different frequency bands), the time domain and frequency domain features of heart rate variability, and the statistical features of activity level are extracted, and a classifier is used to evaluate the patient's preliminary risk value, generating a first data structure including a timestamp, a feature vector, and a preliminary risk assessment value, and sending it to the second server.

[0048] The second processing module (server side) receives the first data structure from the first processing module. The wavelet transform is performed on the electroencephalogram signal to extract the power spectral density features. At the same time, the maximum Lyapunov exponent is used to identify the waveform features. The time series change patterns of the electroencephalogram signal and heart rate variability are analyzed to generate a second data structure including power spectral density, waveform features, and time series patterns.

[0049] The machine learning models such as random forest and gradient boosting tree are trained by the prediction analysis module (server side). The second data structure is input into the trained model to obtain the prediction probability of epileptic seizure. The epileptic seizure risk score is calculated according to the prediction probability and clinical experience rules, and a third data structure including the prediction probability, epileptic seizure risk score, event trigger time, epileptic risk score level (such as low, medium, high), and epileptic warning status identifier (icon code, color code, and curve data) is generated.

[0050] The third data structure from the server side is received by the bracelet display module (the nurse wears the bracelet). According to the risk score level, prompts are sent through vibration and flashing, and the prompt intensity can be proportional to the risk level. The warning icon, color, and curve data are displayed on the bracelet screen.

[0051] In an optional manner, in the second processing module, the wavelet transform feature extraction method is used to perform secondary processing on the first data structure, calculate the energy distribution of the signal at different scales to obtain the power spectral density, and identify the waveform features according to the largest Lyapunov exponent.

[0052] In this embodiment, the wavelet transform analyzes the signal by stretching and translating a mother wavelet function, which can capture the frequency changes at different time scales, so as to better extract the characteristics of the electroencephalogram signal. Especially before and after epileptic seizure, the energy of different frequency bands will change significantly. Epileptic seizure is often accompanied by an increase in the energy of certain specific frequencies. For example, high-frequency spike waves and slow waves will appear during the seizure period. By calculating the energy distribution characterization of the power spectral density (PSD), the energy distribution of the signal can be intuitively displayed, and the specific frequency components of the electroencephalogram activity during epileptic seizure can be reflected. The largest Lyapunov exponent is an index used to describe the dynamic behavior of a chaotic system, which can quantify the sensitivity of the system to the initial conditions. In the analysis of biological signals, especially electroencephalogram signals, the largest Lyapunov exponent can identify the non-linear features in the signal (epileptic seizure is considered a non-linear dynamic process, often a typical manifestation of a chaotic system, and will show a chaotic or unstable state before and after the seizure). Among them, the higher the largest Lyapunov exponent, the higher the chaos degree of the signal, the worse the stability, and the higher the risk of epileptic seizure. Combining the power spectral density (time-frequency domain feature) extracted by wavelet transform and the largest Lyapunov exponent can more comprehensively describe the complex characteristics of the electroencephalogram signal and improve the expression ability of the features.

[0053] Specifically, according to the characteristics of the signal, Morlet or Daubechies wavelet functions are selected to decompose the EEG signal into different scales (such as decomposed into approximation components and detail components), and the detail components in the frequency band related to epileptic seizures are selected. For each decomposed detail component, a time window of a fixed length is intercepted. The intercepted time series is windowed by a Hamming window or a Hanning window to reduce spectral leakage. The windowed time series is subjected to Fourier transform to obtain a frequency-domain signal. The EEG signal is reconstructed in phase space by the delay embedding method to analyze its nonlinear dynamic characteristics. For example, the nearest neighbor points of each point in the phase space are found, the average exponential growth rate of the distance between the nearest neighbor points evolving with time is calculated, and the maximum Lyapunov exponent is determined by fitting an exponential growth curve. The calculated power spectral density and the maximum Lyapunov exponent are integrated to form a new feature vector and transmitted as the second data structure.

[0054] In an alternative manner, the calculation formula for the power spectral density is:

[0055]

[0056] where PSD(f) is the power spectral density at frequency f; x(n) is the time series data; N is the number of data points; w(n) is the window function.

[0057] In this embodiment, the time-domain signal is converted to the frequency domain to reveal the different frequency components and their intensities contained in the signal. Due to the artifacts caused by data truncation, spectral leakage causes the energy of the frequency components to disperse to the surrounding frequencies, and the accuracy of the analysis is reduced by reducing spectral leakage through the window function.

[0058] In an alternative manner, in the prediction analysis module, random forest and gradient boosting tree are used to weight the epileptic seizure risk score, and the weighting weight formula is:

[0059]

[0060] where w i is the weight of the i-th model; AUC i is the area under the curve of the i-th model; m is the total number of models; γ is the weight adjustment factor.

[0061] In this embodiment, the random forest model and the gradient boosting tree model are trained respectively using the training data, the trained models are evaluated on the validation set or the test set, and the AUC value of each model is calculated. The weight adjustment factor is optimized using methods such as grid search. The AUC value and the selected weight adjustment factor are substituted into the formula to calculate the weight of each model. The prediction results of each model are weighted according to the calculated weights to obtain the final prediction result.

[0062] In an alternative approach, in the bracelet display module, the calculation formulas for vibration and flashing prompts based on the epilepsy risk score level are respectively:

[0063] f vib = a·RiskLevel e + b·sin(RiskLevel·π)

[0064] L flash = c·RiskLevel d + f·log(1 + RiskLevel)

[0065] where f vib is the vibration frequency, and L flash is the flashing brightness; RiskLevel is the risk score level; a, b, c, d, e, f are preset parameters.

[0066] In this embodiment, the vibration frequency and flashing brightness are dynamically adjusted according to the risk score level to provide more intuitive and effective prompts. At a high risk level, the prompts are more obvious, and at a low risk level, false alarms and user fatigue are avoided. Nonlinear response is achieved through exponential and logarithmic functions, and the calculation amount is small, which is suitable for implementation on bracelet devices.

[0067] In an alternative approach, in the second processing module, Morlet wavelet is used as the wavelet basis function for feature extraction, and its expression is:

[0068] ψ(t) = π -1 / 4exp(-t 2 / 2)exp(i2πf0t)

[0069] where f0t is the central frequency; t is the time.

[0070] In this embodiment, Morlet wavelet provides a good resolution balance in both the time and frequency dimensions, being able to capture instantaneous features in the time domain and analyze spectral features in the frequency domain.

[0071] In an alternative approach, in the prediction and analysis module, early stopping is used to fit the random forest and gradient boosting tree, and the formula for determining the number of training rounds to stop training is:

[0072] T = argminL val (t)|t∈1,2,...,T max ,ΔL val (t)<ε

[0073] where L val (t) is the loss on the validation set after the t-th round of training; ΔLval (t) is the difference in the loss on the validation set after two consecutive rounds of training; ε is a preset threshold; T max is the maximum number of training rounds.

[0074] In this embodiment, training is stopped when the loss change on the validation set reaches the preset threshold, reducing unnecessary training time.

[0075] In an alternative manner, it further includes an adaptive calibration module for iteratively estimating the independent components in the EEG signal according to the Fast Independent Component Analysis method.

[0076] Among them, the iterative formula for the independent components in the EEG signal is:

[0077] w (k+1) = E[xg(w (k)T x)] - diag(E[g ′ (w (k)T x)])w (k)

[0078] Among them, w (k) is the weight vector at the k-th iteration; x is the observed signal vector; g(.) is the tanh function; g ′ (.) is the derivative of g(.); E[.] is the expectation function; diag(.) is the operation of converting a vector into a diagonal matrix.

[0079] In this embodiment, the separation matrix (the identity matrix) is initialized, and iterative calculations are performed using the Fast Independent Component Analysis method (FastICA algorithm). Specifically, a random weight vector is selected, and the data matrix is projected onto the weight vector. Nonlinear transformation is performed on the projected data, and the nonlinear-transformed data is normalized and the weight vector is updated to maximize the objective function. The above steps are repeated until convergence. The separation matrix is calculated and the EEG signal is projected onto the separation matrix to obtain the separated independent components.

[0080] According to the solution provided by the present invention, it includes a first processing module, which is used to analyze the physiological signal data of patients collected by the first server and generate preliminary result data of epileptic seizures. Among them, the physiological signal data includes electroencephalogram, heart rate variability, and activity level; the preliminary result data is stored in the form of a first data structure, and the first data structure includes a timestamp, a physiological data feature vector, and a preliminary risk assessment value; a second processing module, which is used to send the first data structure to a second server, perform secondary processing and conversion on the first data structure to generate a second data structure, and the second data structure includes power spectral density, waveform features, and time series patterns; a prediction analysis module, which is used to perform prediction analysis according to the second data structure to obtain the prediction probability of a patient's epileptic seizure; generate an epileptic seizure risk score according to the prediction probability and clinical experience rules; combine the prediction probability, the epileptic seizure risk score, and the event trigger time into a third data structure; among them, the third data structure includes an epileptic risk score level and an epileptic warning status identifier, and the epileptic warning status identifier includes an icon code, a color code, and curve data; a bracelet display module, which is used to send the third data structure to a nurse's wearable bracelet, and the bracelet gives vibration and flashing prompts according to the third data structure. The present invention further analyzes the electroencephalogram, heart rate variability, and activity level physiological signal data of patients by integrating the first processing module and the second processing module, and extracts power spectral density, waveform features, and time series patterns, improving the prediction probability of epileptic seizures and the accuracy of risk score. By providing warning prompts to nurses through vibration frequency and flashing brightness, nurses can quickly respond to avoid danger to patients, thereby improving medical efficiency and safety.

[0081] Figure 4 FIG. shows a schematic structural diagram of an embodiment of a computing device according to the present invention. The specific implementation of the present invention does not limit the specific implementation of the computing device.

[0082] As Figure 4 shown, the computing device may include: a processor 402, a communications interface 404, a memory 406, and a communication bus 408.

[0083] Among them: the processor 402, the communications interface 404, and the memory 406 communicate with each other through the communication bus 408. The communications interface 404 is used to communicate with network elements of other devices such as clients or other servers. The processor 402 is used to execute the program 410, and specifically can execute the relevant steps in the above-mentioned embodiment of the intelligent bracelet for remote monitoring and prediction of epileptic seizures.

[0084] Specifically, the program 410 may include program code that includes computer operation instructions.

[0085] The processor 402 may be a central processing unit (CPU), or a specific integrated circuit (ASIC) (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention. One or more processors included in the computing device may be of the same type of processor, such as one or more CPUs; or may be of different types of processors, such as one or more CPUs and one or more ASICs.

[0086] The memory 406 is used to store the program 410. The memory 406 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0087] According to the solution provided by the present invention, it includes a first processing module for analyzing the physiological signal data of a patient collected by a first server to generate preliminary result data of epileptic seizures. The physiological signal data includes electroencephalogram, heart rate variability, and activity level; the preliminary result data is stored in the form of a first data structure, and the first data structure includes a timestamp, a physiological data feature vector, and a preliminary risk assessment value; a second processing module for sending the first data structure to a second server, performing secondary processing and conversion on the first data structure to generate a second data structure, and the second data structure includes power spectral density, waveform features, and time series patterns; a prediction analysis module for performing prediction analysis according to the second data structure to obtain the prediction probability of the patient's epileptic seizure; generating an epileptic seizure risk score according to the prediction probability and clinical experience rules; combining the prediction probability, the epileptic seizure risk score, and the event trigger time into a third data structure; wherein, the third data structure includes an epileptic risk score level and an epileptic warning status identifier, and the epileptic warning status identifier includes an icon code, a color code, and curve data; a bracelet display module for sending the third data structure to a bracelet worn by a nurse, and the bracelet performs vibration and flashing prompts according to the third data structure. By integrating the first processing module and the second processing module, the present invention further analyzes the electroencephalogram, heart rate variability, and activity level physiological signal data of the patient, and extracts power spectral density, waveform features, and time series patterns, improving the prediction probability of epileptic seizures and the accuracy of risk score. By providing warning prompts to the nurse through the vibration frequency and flashing brightness, the nurse can quickly respond to avoid danger to the patient, thereby improving medical efficiency and safety.

[0088] Those skilled in the art can understand that the modules in the devices in the embodiments can be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and in addition, they can be divided into multiple sub-modules or sub-units or sub-components. Except that at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all the features disclosed in this specification (including the accompanying claims, abstract and drawings) and all the processes or units of any method or device so disclosed. Unless otherwise explicitly stated, each feature disclosed in this specification (including the accompanying claims, abstract and drawings) can be replaced by an alternative feature that provides the same, equivalent or similar purpose. In addition, those skilled in the art can understand that although some of the embodiments herein include certain features included in other embodiments rather than other features, the combination of the features of different embodiments means that it is within the scope of the present invention and forms different embodiments. For example, in the following claims, any one of the claimed embodiments can be used in any combination. The present invention can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In the unit claims listing several devices, several of these devices can be embodied by the same hardware item. The steps in the above embodiments, unless otherwise specifically stated, should not be construed as a limitation on the execution order.

Claims

1. An intelligent bracelet for remote monitoring and prediction of epileptic seizures, characterized in that, Including: A first processing module, which is used to analyze the physiological signal data of patients collected by the first server to generate preliminary result data of epileptic seizures. The physiological signal data includes electroencephalogram, heart rate variability, and activity level. The preliminary result data is stored in the form of a first data structure, and the first data structure includes a timestamp, a physiological data feature vector, and a preliminary risk assessment value. A second processing module, which is used to send the first data structure to the second server, perform secondary processing and conversion on the first data structure to generate a second data structure. The second data structure includes power spectral density, waveform features, and time series patterns. A prediction analysis module, which is used to perform prediction analysis according to the second data structure to obtain the prediction probability of epileptic seizures in patients; generate an epileptic seizure risk score according to the prediction probability and clinical experience rules; combine the prediction probability, the epileptic seizure risk score, and the event trigger time into a third data structure. The third data structure includes an epileptic risk score level and an epileptic warning status identifier. The epileptic warning status identifier includes an icon code, a color code, and curve data. A bracelet display module, which is used to send the third data structure to the bracelet worn by the nurse, and the bracelet gives vibration and flashing prompts according to the third data structure.

2. The smart bracelet for remotely monitoring and predicting epileptic seizures according to claim 1, wherein In the second processing module, a wavelet transform feature extraction method is used to perform secondary processing on the first data structure, calculate the energy distribution of the signal at different scales to obtain the power spectral density, and identify the waveform features according to the largest Lyapunov exponent.

3. The intelligent bracelet for remote monitoring and prediction of epileptic seizures according to claim 2, characterized in that, The calculation formula of the power spectral density is: where PSD(f) is the power spectral density at frequency f; x(n) is the time series data; N is the number of data points; w(n) is the window function.

4. The intelligent bracelet for remotely monitoring and predicting epileptic seizures according to claim 1, characterized in that, In the prediction analysis module, a random forest and a gradient boosting tree are used to weight the epileptic seizure risk score, and the weighting formula is: Among them, w i is the weight of the i-th model; AUC i is the area under the curve of the i-th model; m is the total number of models; γ is the weight adjustment factor.

5. The intelligent bracelet for remotely monitoring and predicting epileptic seizures according to claim 1, characterized in that, In the bracelet display module, the calculation formulas for vibration and flashing prompts according to the epileptic risk score level are: f vib = a·RiskLevel e + b·sin(RiskLevel·π) L flash = c·RiskLevel d + f·log(1 + RiskLevel) Among them, f vib is the vibration frequency, L flash is the flashing brightness; RiskLevel is the risk score level; a, b, c, d, e, f are preset parameters.

6. The smart bracelet for remote monitoring and prediction of epileptic seizures according to claim 2, wherein, In the second processing module, Morlet wavelet is used as the wavelet basis function for feature extraction, and its expression is: ψ(t) = π -1 / 4 exp(-t 2 / 2) exp(i2πf0t) where f0t is the center frequency; t is the time.

7. The smart bracelet for remotely monitoring and predicting epileptic seizures according to claim 4, characterized in that, In the prediction analysis module, early stopping is used to fit the random forest and the gradient boosting tree, and the formula for determining the number of training rounds to stop training is: T = argminL val (t) | t ∈ 1, 2,..., T max , ΔL val (t) < ε Among them, L val (t) is the loss on the validation set after the t-th round of training; ΔL val (t) is the difference in the loss on the validation set after two consecutive rounds of training; ε is a preset threshold; T max is the maximum number of training rounds.

8. The intelligent bracelet for remotely monitoring and predicting epileptic seizures according to any one of claims 1-7, characterized in that It also includes an adaptive calibration module, which is used to iteratively estimate the independent components in the electroencephalogram signal according to the fast independent component analysis method.

9. The smart bracelet for remote monitoring and prediction of epileptic seizures according to claim 8, characterized in that, The iterative formula for the independent components in the electroencephalogram signal is: w (k+1) = E[xg(w (k)T x)] - diag(E[g ′ (w (k)T x)])w (k) where, w (k) is the weight vector at the k-th iteration; x is the observed signal vector; g(.) is the tanh function; g ′ (.) is the derivative of g(.); E[.] is the expectation function; diag(.) is the operation of converting a vector into a diagonal matrix.

10. A computing device, comprising: A processor, a memory, a communication interface, and a communication bus. The processor, the memory, and the communication interface complete mutual communication through the communication bus. The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the intelligent bracelet for remote monitoring and prediction of epileptic seizures.

Images