Wearable anesthesia depth monitoring device and monitoring method thereof

Abstract

The application discloses a wearable anesthesia depth monitoring device and a monitoring method thereof, and relates to the technical field of anesthesia monitoring.The method comprises the following steps: collecting electroencephalogram signals, blood oxygen saturation, integrated heart rate and skin conductance signals of a user through a head-mounted module and a wristband module, sending data to a cloud server in combination with physiological characteristic information and surgical characteristic information, and configuring an anesthesia monitoring precision coefficient.According to the precision coefficient, anesthesia depth recognition is performed by using multi-modal physiological signals, anesthesia depth monitoring results are generated, and real-time display is performed.The application solves the technical problems that existing anesthesia depth monitoring devices rely on a single signal, have insufficient accuracy and real-time performance, have poor portability, and cannot adapt to the multi-scene requirements of resource-limited environments, and achieves the technical effects of improving the accuracy, real-time performance and portability of anesthesia depth monitoring by means of multi-modal physiological signal collection and individualized monitoring precision configuration, and meeting the multi-scene application requirements.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of anesthesia monitoring, in particular to a wearable anesthesia depth monitoring device and a monitoring method thereof. BACKGROUND

[0002] In clinical anesthesia management, accurate monitoring of anesthesia depth is crucial to ensure patient safety. Existing anesthesia monitoring devices mainly rely on a single electroencephalogram signal (such as bispectral index, BIS), which can assess the patient's consciousness state, but its function is single and depends on fixed installation. The device is large in size, high in cost, complex in operation, and requires special consumables (such as disposable electrode patches). These limitations make it difficult for traditional devices to meet the needs of short-time surgery, local anesthesia, and resource-limited primary medical scenarios. In addition, single signal monitoring cannot fully reflect the patient's physiological state and is easily affected by artifacts, affecting the accuracy of monitoring. SUMMARY

[0003] The present application provides a wearable anesthesia depth monitoring device and a monitoring method thereof, which solves the technical problems of existing anesthesia depth monitoring devices relying on a single signal, insufficient accuracy and real-time performance, and poor portability, which cannot adapt to the needs of multiple scenarios in resource-limited environments.

[0004] In view of the above problems, the present application provides a wearable anesthesia depth monitoring device and a monitoring method thereof.

[0005] In a first aspect, the present application provides a wearable anesthesia depth monitoring device, which comprises a head-mounted module, a wristband module, and an anesthesia depth monitoring system. The system comprises: a physiological signal monitoring unit for monitoring the electroencephalogram, oxygen saturation, integrated heart rate, and skin conductance of a user wearing the device through the head-mounted module and the wristband module, obtaining the electroencephalogram, oxygen saturation, integrated heart rate, and skin conductance, wherein the head-mounted module and the wristband module are respectively configured with electroencephalogram sensors, oxygen saturation sensors, integrated heart rate sensors, and skin conductance sensors; a characteristic information acquisition unit for acquiring physiological characteristic information and surgical characteristic information of the user; an anesthesia monitoring accuracy configuration unit for sending the electroencephalogram, oxygen saturation, integrated heart rate, skin conductance, physiological characteristic information, and surgical characteristic information to a cloud server, configuring an anesthesia monitoring accuracy coefficient for monitoring the anesthesia depth of the user according to the physiological characteristic information and surgical characteristic information; an anesthesia depth recognition unit for recognizing the anesthesia depth according to the electroencephalogram, oxygen saturation, integrated heart rate, and skin conductance according to the anesthesia monitoring accuracy coefficient, obtaining anesthesia depth monitoring results and receiving, and displaying the anesthesia depth monitoring results.

[0006] In a second aspect, the application provides a wearable anesthesia depth monitoring method, comprising: monitoring electroencephalogram signals, blood oxygen saturation, integrated heart rate and skin conductance of a user wearing the device through a head-mounted module and a wristband module, obtaining electroencephalogram signals, blood oxygen saturation, integrated heart rate and skin conductance, wherein the head-mounted module and the wristband module are respectively configured with electroencephalogram sensors, blood oxygen saturation sensors, integrated heart rate sensors and skin conductance sensors; obtaining physiological characteristic information and surgical characteristic information of the user; sending the electroencephalogram signals, blood oxygen saturation, integrated heart rate, skin conductance, physiological characteristic information and surgical characteristic information to a cloud server, configuring an anesthesia monitoring precision coefficient for monitoring the anesthesia depth of the user according to the physiological characteristic information and the surgical characteristic information; identifying the anesthesia depth according to the electroencephalogram signals, blood oxygen saturation, integrated heart rate and skin conductance according to the anesthesia monitoring precision coefficient, obtaining anesthesia depth monitoring results and receiving, and displaying the anesthesia depth monitoring results.

[0007] The one or more technical solutions provided in the application have at least the following technical effects or advantages:

[0008] The wearable anesthesia depth monitoring device and the monitoring method provided by the embodiments of the application relate to the technical field of anesthesia monitoring, collect multi-modal signals such as electroencephalogram, blood oxygen saturation, heart rate and skin conductance through the head-mounted module and the wristband module, configure an anesthesia monitoring precision coefficient in combination with physiological and surgical characteristic information, identify the anesthesia depth by using a cloud server and an intelligent algorithm, generate and display accurate monitoring results in real time, meet the needs of multiple scenes, solve the technical problems that the existing anesthesia depth monitoring devices rely on a single signal, have insufficient accuracy and real-time performance, have poor portability and cannot adapt to the needs of multiple scenes in a resource-limited environment, and achieve the technical effects of improving the accuracy, real-time performance and portability of anesthesia depth monitoring by multi-modal physiological signal collection and personalized monitoring precision configuration, and meeting the needs of multiple scene applications. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 The anesthesia depth monitoring system of the wearable anesthesia depth monitoring device is provided in the application, and a structural schematic diagram is provided.

[0010] Figure 2 The anesthesia depth monitoring method is provided in the application, and a flowchart is provided.

[0011] BRIEF DESCRIPTION OF DRAWINGS

[0012] The physiological signal monitoring unit 11, the characteristic information acquisition unit 12, the anesthesia monitoring precision configuration unit 13 and the anesthesia depth identification unit 14. DETAILED DESCRIPTION

[0013] The wearable anesthesia depth monitoring device and the monitoring method thereof are provided to solve the technical problems of the prior art anesthesia depth monitoring device, i.e., the anesthesia depth monitoring device relies on a single signal, the accuracy and real-time performance are insufficient, the device is not portable, and the device cannot adapt to the multi-scene requirements in a resource-limited environment.

[0014] In an embodiment, as shown in the accompanying drawings, the wearable anesthesia depth monitoring device provided by the present application comprises a head-mounted module, a wristband module and an anesthesia depth monitoring system. Figure 1

[0015] The physiological signal monitoring unit 11 is configured to monitor the electroencephalogram (EEG), the oxygen saturation (SpO2), the integrated heart rate (ECG) and the skin conductance (EDA) of the user wearing the device through the head-mounted module and the wristband module.

[0016] Further, the physiological signal monitoring unit 11 is further configured to perform the following steps:

[0017] P11: The physiological signal monitoring unit 11 is configured to monitor the electroencephalogram (EEG), the oxygen saturation (SpO2), the integrated heart rate (ECG) and the skin conductance (EDA) of the user wearing the device through the head-mounted module and the wristband module.

[0018] It should be understood that the physiological signal monitoring unit 11 of the present application monitors the multi-modal physiological signals of the user wearing the device in real time through the head-mounted module and the wristband module, including monitoring the electroencephalogram (EEG), the oxygen saturation (SpO2), the integrated heart rate (ECG) and the skin conductance (EDA). The head-mounted module integrates the electroencephalogram (EEG) sensor and the oxygen saturation (SpO2) sensor to collect the dynamic changes of the brain neural activity and the oxygen content in the blood of the user. The wristband module is configured with the integrated heart rate (ECG) sensor and the skin conductance (EDA) sensor to reflect the circulatory system state and the skin conductance change of the user, which are closely related to the autonomic nervous system and the emotional state of the user.

[0019] ​Specifically, first, the user is collected by the sensors of the head-mounted module and the wristband module. The integrated electroencephalogram sensor and the blood oxygen saturation sensor in the head-mounted module record the user's electroencephalogram activity and the oxygen saturation in the blood, respectively. The electroencephalogram signal (EEG) can directly reflect the brain nerve activity state, which is an important indicator for assessing the depth of anesthesia; the blood oxygen saturation (SpO2) is used to monitor the oxygen carrying capacity of the blood and reflect the patient's circulatory system condition. At the same time, the heart rate sensor and the skin conductance sensor in the wristband module collect the user's heart rate and skin conductance (EDA) signals, respectively. The former reflects the circulatory system state through the electrical physiological activity of the heart, and the latter reflects the response of the autonomic nervous system to external stimuli. Through these sensor arrays, the device can simultaneously collect electroencephalogram signals, blood oxygen saturation, integrated heart rate and skin conductance raw data. However, these raw data may contain noise, such as muscle artifacts, motion interference or the influence of ambient light on the signal, which needs to be further processed to ensure signal quality.

[0020] Next, the collected raw signals are preprocessed to filter out noise and obtain clean physiological signals. For example, for the electroencephalogram signal (EEG), the device uses a band-pass filtering technique to retain the frequency band of 0.5Hz to 45Hz, and combines an independent component analysis (ICA) algorithm to separate out artifact signals such as eye movement, electromyography and power frequency noise, to ensure that the extracted signal accurately reflects the brain's neural activity. For blood oxygen saturation (SpO2), the photoplethysmography (PPG) technique is used to remove high-frequency noise through low-pass filtering, and the signal amplitude is normalized to eliminate the interference of environmental light changes and user actions on the data. In the processing of the heart rate signal (ECG), first, the baseline drift is removed by high-pass filtering, and then the R-wave detection algorithm is used to identify the characteristic peaks in the electrocardiogram signal and exclude noise peaks, to ensure the accuracy of heart rate calculation. For skin conductance (EDA), background conductance and transient conductance changes are separated by low-frequency filtering method, where the background conductance reflects the long-term physiological state, and the transient conductance change is closely related to the rapid response of the autonomic nervous system. Through preprocessing, the signal quality can be effectively improved to provide accurate input for the intelligent algorithm module of the device, ensuring the reliability and accuracy of the device in multiple scenarios.

[0021] The feature information acquisition unit 12 is configured to acquire physiological feature information and surgical feature information of the user.

[0022] Further, the feature information acquisition unit 12 is further configured to perform the following steps:

[0023] P21: Collecting physiological feature information of the user according to a plurality of physiological feature categories; P22: Collecting the length of the operation time and the difficulty level of the operation as surgical feature information.

[0024] Optionally, the core function of the feature information acquisition unit 12 of the present application is to comprehensively collect the physiological feature information of the user and the feature information related to the operation, as an important input for subsequent anesthesia depth monitoring analysis. The work of this unit is divided into the following two steps:

[0025] First, according to multiple physiological feature categories, the physiological feature information of the user is collected. Physiological feature information refers to basic parameters directly related to the individual state of the user, including but not limited to age, gender, weight, height, and basal metabolic rate (BMR). These information provides individualized reference for algorithm model, ensuring the individualization and accuracy of anesthesia depth monitoring results. For example, age and weight affect the metabolic rate and distribution volume of anesthetic drugs, while gender differences may be related to the sensitivity of physiological response. In addition, the unit also dynamically collects some physiological state indicators of the patient before the operation, such as heart rate variability (HRV) and blood pressure level, which reflect the state of the patient's autonomic nervous system and help to assess their tolerance to anesthesia. By classifying the collection of physiological feature information, the comprehensiveness and applicability of the data can be ensured, providing accurate physiological basic data for intelligent algorithm modules.

[0026] Second, the operation-related feature information is collected, including the operation time length and the operation difficulty level. The operation time length refers to the estimated or actual duration of the operation, which can help predict the maintenance time and dosage adjustment strategy of anesthetic drugs. The operation difficulty level is classified according to the complexity of the operation, the degree of trauma, and the requirement for anesthesia depth (such as mild, moderate, and severe). This information directly determines the control target of anesthesia depth, for example, high-difficulty operations may require deeper anesthesia depth to ensure the patient's complete unconscious state and reduce intraoperative stress response. Accurate collection of these operation feature information not only optimizes the device's assessment of the patient's current state, but also forms the overall input variables of the operation scenario in the intelligent algorithm, making the monitoring results more practical.

[0027] Through the above two steps of work, the feature information acquisition unit 12 integrates the physiological feature information and the operation feature information into multi-dimensional input parameters, providing comprehensive support for the subsequent anesthesia depth recognition module of the device. This systematic collection method of feature information not only improves the flexibility and adaptability of the monitoring device, but also significantly improves the accuracy and clinical guidance value of anesthesia monitoring results through personalized and contextual data support.

[0028] The anesthesia monitoring precision configuration unit 13 is configured to send the electroencephalogram signal, blood oxygen saturation, integrated heart rate, skin conductance, physiological feature information, and operation feature information to a cloud server, and configure an anesthesia monitoring precision coefficient for monitoring the anesthesia depth of the user according to the physiological feature information and operation feature information.

[0029] Further, the anesthetic monitoring precision configuration unit 13 is further configured to perform the following steps:

[0030] P31: constructing an anesthetic monitoring precision configurator; P32: inputting the physiological characteristic information and the surgical characteristic information into the anesthetic monitoring precision configurator, and outputting an anesthetic monitoring precision coefficient.

[0031] Specifically, the core task of the anesthetic monitoring precision configuration unit 13 of the present application is to send the collected multi-modal physiological signals, physiological characteristic information and surgical characteristic information to a cloud server, and to construct and calculate the precision coefficient of anesthetic monitoring, thereby ensuring the accuracy and individualization of anesthetic depth monitoring.

[0032] First, an anesthetic monitoring precision configurator is constructed. The anesthetic monitoring precision configurator is a core tool based on cloud computing and algorithm model, which can dynamically analyze the individualized parameters of the user. Specifically, the configurator, through a pre-set machine learning algorithm or neural network model, combines multi-dimensional input parameters (including physiological characteristic information and surgical characteristic information), and quantitatively models the precision requirement of anesthetic monitoring. For example, the configurator will adjust the real-time performance and sensitivity of anesthetic depth monitoring according to the surgical characteristic information (such as the length of the operation time and the difficulty level), and at the same time, it will optimize the individualized parameter weight of the model according to the physiological characteristic information of the user (such as age, weight and basal metabolic rate). In this way, the configurator can generate a dedicated anesthetic monitoring configuration scheme for different users and surgical scenarios, ensuring that the monitoring precision meets the actual needs.

[0033] Next, the physiological characteristic information and the surgical characteristic information are input into the anesthetic monitoring precision configurator, and an anesthetic monitoring precision coefficient is calculated and output. The precision coefficient is a dynamic adjustment parameter used to guide the calculation precision and algorithm weight distribution of the subsequent anesthetic depth recognition module. For example, for patients with stable physiological state, the precision configurator may generate a lower precision coefficient, thereby reducing the computational burden of monitoring; while for scenarios with high surgical difficulty and strict anesthetic requirements, the configurator will output a higher precision coefficient to increase the sensitivity and response speed of the monitoring. The output of the precision configurator not only takes into account the physiological differences of individuals, but also fully considers the complexity of the surgical scenario, thereby making the anesthetic depth monitoring more adaptable and reliable.

[0034] Through the above steps, the anesthetic monitoring precision configuration unit 13 can effectively integrate the individual characteristics of the patient and the surgical requirements into the anesthetic depth monitoring system, generating a personalized monitoring precision configuration scheme. This design not only improves the application efficiency of the device in different scenarios, but also maximizes the demand for precise anesthetic management in clinical practice by dynamically adjusting the monitoring precision, thereby improving the reliability of the monitoring results and the clinical guidance value.

[0035] Further, the anesthetic monitoring accuracy configuration unit 13 is further configured to perform the following steps when constructing the anesthetic monitoring accuracy configurator:

[0036] P31-1: According to the user's surgical anesthesia history data, a sample physiological feature information set and a sample surgical feature information set are collected; P31-2: According to the anesthetic dose corresponding to each sample physiological feature information and sample surgical feature information, a sample anesthetic monitoring accuracy set is labeled; P31-3: The sample physiological feature information set, the sample surgical feature information set, and the sample anesthetic monitoring accuracy set are used as supervised training data, and a machine learning is used to construct and train an anesthetic monitoring accuracy configurator.

[0037] In a possible embodiment of the present application, the anesthetic monitoring accuracy configuration unit 13, when constructing the anesthetic monitoring accuracy configurator, performs deep analysis and modeling on the user's surgical anesthesia history data to ensure that the configurator can accurately adapt to individual needs.

[0038] Specifically, first, according to the user's surgical anesthesia history data, a sample physiological feature information set and a sample surgical feature information set are collected. These sample data come from actual clinical cases and contain multi-dimensional physiological feature information (such as age, weight, gender, basal metabolic rate) and surgical feature information (such as surgical time, surgical complexity, intraoperative blood loss, etc.) of the user. The sample physiological feature information set is used to reflect the physiological reaction difference of the individual to the anesthetic drug, while the sample surgical feature information set is directly related to the requirement of anesthetic depth, for example, complex surgery often requires a deeper anesthetic level. These data provide comprehensive input features for subsequent model training.

[0039] Further, according to the physiological feature information and surgical feature information of each sample, the anesthetic dose is labeled and a sample anesthetic monitoring accuracy set is obtained. The anesthetic dose is recorded according to the anesthetic scheme given by the doctor in the historical data, which can directly reflect the sensitivity and tolerance of the user to anesthesia under different conditions. Based on the relationship between the anesthetic dose and the actual anesthetic depth monitoring result, a corresponding anesthetic monitoring accuracy value is labeled for each sample. This accuracy value represents the sensitivity and response accuracy that the monitoring system needs to achieve under the current feature condition. For example, complex surgery samples with large anesthetic depth fluctuations will be labeled with higher accuracy requirements, while short-time surgery or local anesthesia may correspond to lower accuracy requirements.

[0040] Next, the sample physiological feature information set, the sample surgical feature information set, and the sample anesthesia monitoring accuracy set are used as supervised training data to construct and train the anesthesia monitoring accuracy configurator using machine learning methods. In this process, a supervised learning algorithm such as support vector machine, random forest, or neural network is used to model the training data. The input of the model is the physiological features and surgical features of the sample, and the output is the corresponding anesthesia monitoring accuracy value. During the training process, the model parameters are optimized by minimizing the error function, so that it can accurately predict the anesthesia monitoring accuracy requirements of new users under different physiological and surgical conditions. In addition, to improve the generalization ability of the model, cross-validation and data augmentation techniques are used to ensure that the configurator has good adaptability in different scenarios.

[0041] Through the above steps, the anesthesia monitoring accuracy configurator can dynamically calculate the anesthesia monitoring accuracy coefficient based on the individualized features and surgical requirements of the user, providing accurate guidance for subsequent anesthesia depth recognition. The construction of this configurator combines user historical data and the technical advantages of machine learning, ensuring the reliability and intelligence level of the device in multiple scenarios, while significantly improving the accuracy and safety of anesthesia management.

[0042] The anesthesia depth recognition unit 14 is configured to recognize anesthesia depth according to the EEG signal, blood oxygen saturation, integrated heart rate, and skin conductance based on the anesthesia monitoring accuracy coefficient, obtain anesthesia depth monitoring results, and display the anesthesia depth monitoring results.

[0043] Further, the anesthesia depth recognition unit 14 is further configured to perform the following steps:

[0044] P41: An anesthesia depth recognizer is constructed using integrated machine learning, wherein the anesthesia depth recognizer includes multiple anesthesia depth recognition paths; P42: The anesthesia monitoring accuracy coefficient is multiplied by the number of anesthesia depth recognition paths and rounded to obtain the number of anesthesia monitoring paths, and at least one anesthesia depth recognition path is randomly selected according to the number of anesthesia monitoring paths; P43: The EEG signal, blood oxygen saturation, integrated heart rate, and skin conductance are input into the at least one anesthesia depth recognition path to identify at least one anesthesia depth, calculate the mean value to obtain anesthesia depth monitoring results; P44: Receive the anesthesia depth monitoring results transmitted by the cloud server and display them.

[0045] It should be understood that the anesthesia depth recognition unit 14 of the present application is responsible for accurately recognizing the anesthesia depth of the user based on the collected multi-modal physiological signals and the calculated anesthesia monitoring accuracy coefficient, and displaying the monitoring results in real time.

[0046] First, an integrated machine learning technique is used to construct the anesthesia depth recognizer. The anesthesia depth recognizer is a multi-path recognition model that integrates multiple machine learning algorithms (such as support vector machines, random forests, deep neural networks, etc.) to enhance the adaptability to different signal patterns. Each path in the recognizer corresponds to an algorithm model or a specific feature processing strategy, for example, the path based on electroencephalogram signals may focus on analyzing the frequency characteristics of brain waves, while the path based on blood oxygen saturation may focus on the dynamic changes of oxygenation level. This integrated architecture not only improves the robustness of the model, but also effectively reduces the recognition error of a single model through multi-path fusion, ensuring the comprehensive accuracy of anesthesia depth evaluation.

[0047] Next, according to the anesthesia monitoring accuracy coefficient, the number of suitable recognition paths is dynamically selected. Specifically, the anesthesia monitoring accuracy coefficient is multiplied by the total number of paths in the anesthesia depth recognizer, and the result is rounded to obtain the number of anesthesia monitoring paths. Subsequently, the system will randomly select at least one path from the multiple recognition paths according to this number for recognition. The core purpose of this process is to flexibly adjust the allocation of computing resources according to the complexity of monitoring requirements. For example, for high-precision requirements (such as complex surgical scenarios), more paths will be enabled to improve the reliability of recognition; while for low-precision requirements (such as simple surgery or short-time monitoring), the number of paths will be reduced to optimize the operating efficiency of the device.

[0048] Next, the collected electroencephalogram signals, blood oxygen saturation, integrated heart rate, and skin conductance data are input into the selected anesthesia depth recognition path. Each path independently recognizes an anesthesia depth value according to its algorithm characteristics. Subsequently, the average of all path recognition results is calculated to obtain the final anesthesia depth monitoring result. This path fusion strategy effectively reduces the deviation that may occur in a single path by integrating the prediction results of multiple models, improving the robustness and credibility of the monitoring results.

[0049] Finally, the anesthesia depth monitoring results transmitted by the cloud server are received in real time, and the display module of the device is used for visual presentation. The display content includes the current anesthesia depth value, monitoring trend chart, and important warning information (such as anesthesia too deep or too shallow). The visual interface is intuitive and clear, allowing medical personnel to quickly understand the patient's anesthesia state and make appropriate adjustments.

[0050] Through the above steps, the anesthesia depth recognition unit 14 combines multi-modal signal input, integrated machine learning, and dynamic path selection, ensuring the accuracy and real-time performance of anesthesia depth monitoring. Its integrated architecture not only flexibly responds to different surgical scenarios and individual differences of users, but also improves the operating efficiency and reliability of the results through the support of cloud data and path optimization, providing strong technical support for clinical anesthesia management.

[0051] Further, the anesthesia depth recognition unit 14, when constructing the anesthesia depth recognizer by using integrated machine learning, is further configured to perform the following steps:

[0052] P41-1: According to the user anesthesia record data, a sample electroencephalogram signal set, a sample oxygen saturation set, a sample integrated heart rate set, a sample skin conductance set, and a sample anesthesia depth set are collected, and an anesthesia depth monitoring training sample set is obtained by combination; P41-2: The anesthesia depth monitoring training sample set is divided to obtain multiple anesthesia depth monitoring training samples; P41-3: The multiple anesthesia depth monitoring training samples are used respectively to train multiple anesthesia depth recognition paths based on integrated machine learning, and an anesthesia depth recognizer is obtained.

[0053] Optionally, the anesthesia depth recognition unit 14, when constructing the anesthesia depth recognizer, uses an integrated machine learning method to deeply model the multi-modal signals and historical anesthesia depth records, so as to ensure that the recognizer can accurately predict the anesthesia depth.

[0054] Illustratively, first, according to the user anesthesia record data, the sample sets required for training are collected, including a sample electroencephalogram signal set, a sample oxygen saturation set, a sample integrated heart rate set, a sample skin conductance set, and a sample anesthesia depth set, and finally an anesthesia depth monitoring training sample set is generated by combination.

[0055] Among them, the sample electroencephalogram signal set records the brain electrical activity characteristics under different anesthesia depths, for example, the proportion of low-frequency δ wave increases under deep anesthesia state, and the change rule of α wave with the decrease of consciousness. The sample oxygen saturation set captures the dynamic changes of blood oxygenation level, especially the change trend related to respiratory regulation during surgery. The sample integrated heart rate set contains the dynamic fluctuation and variability (HRV) of heart rate, which reflects the response of the circulatory system to anesthesia. The sample skin conductance set is used to capture the activity characteristics of the user's autonomic nervous system, such as the baseline level of skin conductance and the response amplitude to environmental stimuli. The sample anesthesia depth set is annotated based on historical records, including the actual anesthesia depth state at different time points, as the target variable for training. By integrating the above feature sets, the generated multi-dimensional training samples provide a comprehensive and accurate data basis for the machine learning model.

[0056] Further, the anesthesia depth monitoring training sample set is divided to generate multiple independent training sample subsets, providing data support for the multi-path model of ensemble machine learning. In the division process, a random division method is adopted, while ensuring that each subset contains samples under different anesthesia depths, ensuring the uniformity and balance of feature distribution. For example, special attention is paid to the uniform coverage of samples under deep anesthesia, light anesthesia and wakeful state during data division, thereby avoiding performance degradation of the model due to sample bias. In addition, this division strategy improves the generalization ability and adaptability of the model to new data by providing independent training data for different model paths.

[0057] Next, multiple anesthesia depth recognition paths are trained based on the ensemble machine learning method using multiple anesthesia depth monitoring training samples, and finally a complete anesthesia depth recognizer is generated. Each recognition path corresponds to an independent machine learning model, for example, a path based on a support vector machine (SVM) model focuses on electroencephalogram spectral feature analysis, while a path based on a random forest (RF) model may be more suitable for processing the nonlinear changes of heart rate and blood oxygen. During training, each path independently extracts features and optimizes parameters for its data subset, such as using cross-validation to optimize the hyperparameters of the model to ensure the robustness of the model under different data conditions. Multiple paths use weighted averaging or voting strategies to integrate the prediction results of each path to reduce the impact of single path errors on the final recognition result. Through this multi-path integration method, the recognizer can achieve comprehensive analysis of complex signal patterns, improving the accuracy and stability of anesthesia depth recognition.

[0058] Through the above steps, the anesthesia depth recognizer uses the advantages of ensemble machine learning, combines multi-modal physiological signals and rich historical data, and achieves highly accurate anesthesia depth recognition capability, laying a solid foundation for reliable application of the device in actual clinical practice, ensuring that the device can provide accurate monitoring results in different users and surgical scenarios.

[0059] In summary, the embodiments of the present application have at least the following technical effects:

[0060] The present application collects the user's electroencephalogram signal, blood oxygen saturation, integrated heart rate and skin conductance signal through the head-mounted module and wristband module, combines physiological feature information and surgical feature information, sends the data to the cloud server, and configures the anesthesia monitoring precision coefficient. According to the precision coefficient, anesthesia depth recognition is performed using multi-modal physiological signals to generate anesthesia depth monitoring results and real-time display, providing an accurate and efficient anesthesia depth monitoring solution.

[0061] The technical effect of improving the accuracy, real-time performance and portability of anesthesia depth monitoring through multi-modal physiological signal acquisition and personalized monitoring precision configuration is achieved, meeting the application requirements of multiple scenarios.

[0062] Embodiment two, based on the same inventive concept as the anesthesia depth monitoring system of the wearable anesthesia depth monitoring device in the preceding embodiment, as shown, the present application provides a wearable anesthesia depth monitoring method, the method and system embodiments in the embodiment of the present application are based on the same inventive concept. Among them, the method comprises: Figure 2

[0063] Through the head-mounted module and the wristband module, the user wearing the device is monitored for electroencephalogram, blood oxygen saturation, integrated heart rate and skin conductance, and the electroencephalogram, blood oxygen saturation, integrated heart rate and skin conductance are obtained, wherein the head-mounted module and the wristband module are respectively configured with electroencephalogram sensors, blood oxygen saturation sensors, integrated heart rate sensors and skin conductance sensors; physiological characteristic information and surgical characteristic information of the user are obtained; the electroencephalogram, blood oxygen saturation, integrated heart rate, skin conductance, physiological characteristic information and surgical characteristic information are sent to a cloud server, and an anesthesia monitoring precision coefficient for monitoring the anesthesia depth of the user is configured according to the physiological characteristic information and surgical characteristic information; anesthesia depth recognition is performed according to the electroencephalogram, blood oxygen saturation, integrated heart rate and skin conductance according to the anesthesia monitoring precision coefficient, anesthesia depth monitoring results are obtained and received, and the anesthesia depth monitoring results are displayed.

[0064] Further, through the head-mounted module and the wristband module, the user wearing the device is monitored for electroencephalogram, blood oxygen saturation, integrated heart rate and skin conductance, and the electroencephalogram, blood oxygen saturation, integrated heart rate and skin conductance are obtained, comprising:

[0065] Through the head-mounted module and the wristband module, the user wearing the device is monitored for electroencephalogram, blood oxygen saturation, integrated heart rate and skin conductance, and the electroencephalogram, blood oxygen saturation, integrated heart rate and skin conductance are obtained, comprising:

[0066] Further, the physiological characteristic information and surgical characteristic information of the user are obtained, comprising:

[0067] According to a plurality of physiological characteristic categories, the physiological characteristic information of the user is collected; the length of the operation time and the difficulty level of the operation are collected as the surgical characteristic information.

[0068] Further, according to the physiological characteristic information and surgical characteristic information, an anesthesia monitoring precision coefficient for monitoring the anesthesia depth of the user is configured, comprising: ​

[0069] constructing an anesthesia monitoring precision configurator; inputting the physiological feature information and the surgical feature information into the anesthesia monitoring precision configurator, and outputting an anesthesia monitoring precision coefficient.

[0070] Further, the anesthesia monitoring precision configurator is constructed, comprising:

[0071] According to the user's surgical anesthesia history data, a sample physiological feature information set and a sample surgical feature information set are collected; according to the corresponding anesthetic dosage of each sample physiological feature information and sample surgical feature information, a sample anesthesia monitoring precision set is labeled; the sample physiological feature information set, the sample surgical feature information set and the sample anesthesia monitoring precision set are used as supervised training data, and machine learning is used to construct and train the anesthesia monitoring precision configurator.

[0072] Further, according to the anesthesia monitoring precision coefficient, the anesthesia depth is identified according to the electroencephalogram signal, the blood oxygen saturation, the integrated heart rate and the skin conductance, the anesthesia depth monitoring result is obtained and received, and the anesthesia depth monitoring result is displayed, comprising:

[0073] An integrated machine learning is used to construct an anesthesia depth identifier, wherein the anesthesia depth identifier comprises a plurality of anesthesia depth identification paths; the anesthesia monitoring precision coefficient is multiplied by the number of the plurality of anesthesia depth identification paths and is rounded to obtain the number of anesthesia monitoring paths, at least one anesthesia depth identification path is randomly selected according to the number of anesthesia monitoring paths; the electroencephalogram signal, the blood oxygen saturation, the integrated heart rate and the skin conductance are input into the at least one anesthesia depth identification path, at least one anesthesia depth is identified and obtained, the mean value is calculated to obtain the anesthesia depth monitoring result; the anesthesia depth monitoring result transmitted by the cloud server is received and displayed.

[0074] Further, an integrated machine learning is used to construct an anesthesia depth identifier, comprising: according to the user's anesthesia record data, a sample electroencephalogram signal set, a sample blood oxygen saturation set, a sample integrated heart rate set, a sample skin conductance set and a sample anesthesia depth set are collected, and a anesthesia depth monitoring training sample set is obtained by combination; the anesthesia depth monitoring training sample set is divided to obtain a plurality of anesthesia depth monitoring training samples; the plurality of anesthesia depth monitoring training samples are respectively used to train a plurality of anesthesia depth identification paths based on integrated machine learning, and an anesthesia depth identifier is obtained.

[0075] It should be noted that the above-mentioned embodiment sequences of the present application are merely for description only, but not for representing the advantages and disadvantages of the embodiments. And the above-mentioned embodiments of the present specification have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0076] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0077] The specification and drawings are merely exemplary of the present application, and any and all modifications, variations, combinations or equivalents that are within the scope of the present application should be included. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. Wearable anesthesia depth monitoring device, characterized in that, The device includes a head-mounted module, a wrist-mounted module, and an anesthesia depth monitoring system, and the system includes: A physiological signal monitoring unit, configured to monitor EEG signals, blood oxygen saturation, integrated heart rate, and skin conductance of a user wearing the device through the head-mounted module and the wrist-mounted module, and obtain EEG signals, blood oxygen saturation, integrated heart rate, and skin conductance, wherein the head-mounted module and the wrist-mounted module are respectively equipped with an EEG sensor, a blood oxygen saturation sensor, an integrated heart rate sensor, and a skin conductance sensor; a characteristic information acquisition unit, configured to acquire physiological characteristic information and surgical characteristic information of the user; an anesthesia monitoring accuracy configuration unit, configured to transmit the EEG signal, blood oxygen saturation, integrated heart rate, skin conductance, physiological characteristic information, and surgical characteristic information to a cloud server, and configure an anesthesia monitoring accuracy coefficient for monitoring the depth of anesthesia of the user based on the physiological characteristic information and surgical characteristic information; The anesthesia monitoring accuracy configuration unit is further used to: Collecting sample physiological characteristic information sets and sample surgical characteristic information sets based on the user's surgical anesthesia history data; According to the anesthetic dosage corresponding to each sample's physiological characteristic information and sample surgical characteristic information, the sample anesthesia monitoring accuracy set is obtained by annotation; Using the sample physiological feature information set, the sample surgical feature information set, and the sample anesthesia monitoring accuracy set as supervised training data, and using machine learning to construct and train an anesthesia monitoring accuracy configurator; Inputting the physiological characteristic information and the surgical characteristic information into the anesthesia monitoring accuracy configurator, and outputting an anesthesia monitoring accuracy coefficient; An anesthesia depth identification unit is used to identify the anesthesia depth according to the anesthesia monitoring accuracy coefficient, the EEG signal, the blood oxygen saturation, the integrated heart rate, and the skin conductance, obtain and receive the anesthesia depth monitoring result, and display the anesthesia depth monitoring result.

2. The wearable anesthesia depth monitoring device according to claim 1, characterized in that: The physiological signal monitoring unit is further used for: Through the head-mounted module and the wristband module, the user wearing the device is monitored for EEG signals, blood oxygen saturation, integrated heart rate, and skin conductance, to obtain raw EEG signals, raw blood oxygen saturation, raw integrated heart rate, and raw skin conductance; The original EEG signal, original blood oxygen saturation, original integrated heart rate and original skin conductance are subjected to noise filtering preprocessing to obtain an EEG sensor, a blood oxygen saturation sensor, an integrated heart rate sensor and a skin conductance sensor.

3. The wearable anesthesia depth monitoring device according to claim 1, characterized in that: The feature information acquisition unit is further configured to: Collecting physiological characteristic information of the user according to multiple physiological characteristic categories; The length of operation and the level of difficulty of the operation were collected as surgical characteristic information.

4. The wearable anesthesia depth monitoring device according to claim 1, characterized in that: The anesthesia depth identification unit is further used for: An anesthesia depth identifier is constructed using integrated machine learning, wherein the anesthesia depth identifier includes multiple anesthesia depth identification paths; multiplying the anesthesia monitoring accuracy coefficient by the number of the multiple anesthesia depth identification paths and rounding the result to obtain the number of anesthesia monitoring paths, and randomly selecting at least one anesthesia depth identification path according to the number of anesthesia monitoring paths; Inputting the EEG signal, blood oxygen saturation, integrated heart rate, and skin conductance into at least one anesthesia depth identification path, identifying and obtaining at least one anesthesia depth, and calculating an average to obtain an anesthesia depth monitoring result; Receive the anesthesia depth monitoring result transmitted by the cloud server and display it.

5. The wearable anesthesia depth monitoring device according to claim 4, characterized in that: When the anesthesia depth identification unit adopts integrated machine learning to construct an anesthesia depth identifier, it also includes: According to the user's anesthesia record data, a sample EEG signal set, a sample blood oxygen saturation set, a sample integrated heart rate set, a sample skin conductance set, and a sample anesthesia depth set are collected, and the combined results are used to obtain an anesthesia depth monitoring training sample set; Dividing the anesthesia depth monitoring training sample set to obtain multiple anesthesia depth monitoring training samples; The plurality of anesthesia depth monitoring training samples are respectively used to train a plurality of anesthesia depth recognition paths based on integrated machine learning to obtain an anesthesia depth identifier.

Images