Model training method, consciousness state recognition method, device and equipment
By acquiring and processing the characteristics of EEG data, training the consciousness state recognition model is solved, and the problem of large amount of data required for model training in the prior art is improved.
Patent Information
- Application Number
- CN202411049076.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-22
AI Technical Summary
In the prior art, training of consciousness state recognition models requires a large amount of clinical data, resulting in high cost and low efficiency.
By obtaining multiple EEG data of the subject in different test scenarios and their corresponding conscious state change values, frequency domain transformation and feature data calculations are performed, and these feature data are used as training samples to identify the preset consciousness states. Training is carried out, and model training is performed using adaptive neural fuzzy inference system and backpropagation algorithm.
The number of clinical data required to train the model is reduced, the training efficiency of the consciousness state recognition model is improved, and the accuracy of the model is improved.
Smart Images

Figure CN119004103A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a model training method, a consciousness state recognition method, a device and equipment. Background Art
[0002] From the patient's brain waves, we can extract the change from awake to anesthetized during anesthesia, or the change from responsive to unresponsive during injury, or the change of consciousness in other scenarios with stimulation sources. Extracting changes in consciousness state often uses artificial intelligence methods to obtain the corresponding instance model, so as to realize the recognition of changes in consciousness state through the instance model, but this instance model usually requires a large amount of clinical data for model training, which makes this method relatively costly. Summary of the invention
[0003] The main technical problem solved by the present invention is to reduce the amount of clinical data required for training a consciousness state recognition model and to improve the training efficiency of the consciousness state recognition model.
[0004] According to the first aspect, an embodiment provides a model training method, characterized by comprising:
[0005] Acquire multiple pieces of EEG data of the subject in different test scenarios and consciousness state change values corresponding to the multiple pieces of EEG data;
[0006] For each piece of the EEG data, perform frequency domain transformation on it to obtain corresponding EEG frequency domain data, and calculate characteristic data of the EEG data based on the EEG frequency domain data;
[0007] The characteristic data of the plurality of EEG data are used as training sample data, and the corresponding consciousness state change values are used as labels to train the preset consciousness state recognition model to obtain a trained consciousness state recognition model.
[0008] In some embodiments, the step of calculating characteristic data of the EEG data based on the EEG frequency domain data includes:
[0009] Calculating the energy value of the EEG frequency domain data;
[0010] Dividing the EEG frequency domain data into frequency band data in a plurality of different frequency ranges;
[0011] Calculate the energy value of each frequency band data respectively;
[0012] The characteristic data of the EEG data is calculated based on the energy value of the EEG frequency domain data and the energy values of all the frequency band data.
[0013] In some embodiments, the EEG data has N feature data and the frequency band data has N, where N is an integer greater than 1; the feature data of the EEG data is calculated based on the energy value of the EEG frequency domain data and the energy values of all the frequency band data, including:
[0014] The nth characteristic data of the EEG data is obtained by calculating the ratio between the energy value of the nth frequency band data among the N frequency band data and the energy value of the EEG frequency domain data, wherein n=1, 2...N.
[0015] In some embodiments, the nth characteristic data of the EEG data is calculated according to the following formula:
[0016] F n =20×ln(E n / E0)
[0017] Among them, F n represents the nth feature data of the EEG data, E n represents the energy value of the nth frequency band data among the N frequency band data, and E0 represents the energy value of the EEG frequency domain data.
[0018] In some embodiments, the consciousness state change value includes an anesthesia state change value and / or an injury state change value, wherein the anesthesia state change value is used to determine the state change of the subject from awake to anesthetized during the anesthesia process, and the injury state change value is used to determine the state change of the subject from responsive to unresponsive during the face of injury.
[0019] In some embodiments, the consciousness state recognition model is an adaptive neural fuzzy inference system, and the training of the preset consciousness state recognition model includes:
[0020] A back propagation algorithm is used to adjust the premise parameters in the adaptive neural fuzzy reasoning system, and a least squares estimation algorithm is used to adjust the conclusion parameters in the adaptive neural fuzzy reasoning system.
[0021] According to the second aspect, an embodiment provides a method for recognizing a state of consciousness, characterized by comprising:
[0022] Obtaining the patient’s EEG data;
[0023] Performing frequency domain transformation on the patient's EEG data to obtain corresponding EEG frequency domain data, and calculating characteristic data of the patient's EEG data based on the EEG frequency domain data;
[0024] The characteristic data of the patient's EEG data is input into a consciousness state recognition model trained by a model training method to obtain the patient's consciousness state change value.
[0025] According to the third aspect, an embodiment provides a consciousness state recognition device, characterized by comprising:
[0026] A feature extraction module is used to obtain the patient's EEG data, perform frequency domain transformation on the patient's EEG data to obtain corresponding EEG frequency domain data, and calculate feature data of the patient's EEG data based on the EEG frequency domain data;
[0027] The consciousness state recognition module is used to input the characteristic data of the patient's electroencephalogram data into the consciousness state recognition model trained by the model training method to obtain the patient's consciousness state change value.
[0028] According to a fourth aspect, an embodiment provides a model training device, characterized in that it includes:
[0029] Memory, used to store programs;
[0030] A processor is used to implement a model training method by executing a program stored in the memory.
[0031] According to the fifth aspect, an embodiment provides a computer-readable storage medium, characterized in that a program is stored on the medium, and the program can be executed by a processor to implement a model training method.
[0032] According to the model training method, device, equipment and computer-readable storage medium of the above-mentioned embodiment, since the characteristic data of multiple EEG data are calculated, the characteristic data of multiple EEG data are used as training sample data, and the corresponding consciousness state change value is used as a label, the preset consciousness state recognition model is trained to obtain a trained consciousness state recognition model. The characteristic data of multiple EEG data can accurately reflect the characteristics of EEG data, and reduce the correlation between the number of EEG data and the accuracy of model training when the consciousness state recognition model is directly trained with multiple EEG data, thereby improving the efficiency of model training. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 This is a flow chart of the model training method of an embodiment of the present application;
[0034] Figure 2 A reference diagram of EEG data and consciousness state change values of an embodiment;
[0035] Figure 3 A reference diagram of EEG data and consciousness state change values of another embodiment;
[0036] Figure 4A flowchart of calculating characteristic data of EEG data according to EEG frequency domain data in an embodiment;
[0037] Figure 5 A flowchart of a method for recognizing a state of consciousness according to an embodiment;
[0038] Figure 6 The figure is a schematic diagram of the structure of a consciousness state recognition device according to an embodiment. DETAILED DESCRIPTION
[0039] The present invention is further described in detail below by specific embodiments in conjunction with the accompanying drawings. Wherein similar elements in different embodiments adopt associated similar element numbers. In the following embodiments, many detailed descriptions are for making the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different situations, or can be replaced by other elements, materials, methods. In some cases, some operations related to the present application are not shown or described in the specification, this is to avoid the core part of the present application being overwhelmed by too much description, and for those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations according to the description in the specification and the general technical knowledge in the art.
[0040] In addition, the features, operations or characteristics described in the specification can be combined in any appropriate manner to form various implementations. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a required sequence, unless otherwise specified that a certain sequence must be followed.
[0041] The serial numbers of the components in this document, such as "first", "second", etc., are only used to distinguish the objects described and do not have any order or technical meaning. The "connection" and "coupling" mentioned in this application, unless otherwise specified, include direct and indirect connections (couplings).
[0042] Please refer to Figure 1 In an embodiment of the present invention, a model training method is proposed, including steps S10 to S30, which are described in detail below.
[0043] Step S10: Obtain multiple pieces of EEG data of the subject in different test scenarios and the consciousness state change values corresponding to the multiple pieces of EEG data.
[0044] In some embodiments, the consciousness state change value includes an anesthesia state change value and / or an injury state change value, wherein the anesthesia state change value is used to determine the state change of the subject from awake to anesthetized during the anesthesia process, and the injury state change value is used to determine the state change of the subject from responsive to unresponsive during the process of facing injury. Different test scenarios refer to scenarios of different anesthesia levels faced by the subject during the anesthesia process or scenarios of different injury levels faced by the subject during the process of facing injury.
[0045] Please refer to Figure 2 and Figure 3 In some embodiments, the EEG data is the EEG wave data collected from the subject's scalp, and the consciousness state change value corresponding to the EEG data is a value describing the subject's anesthesia state at a certain moment, wherein the EEG data and the consciousness state change value are one-to-one corresponding, and the consciousness state change value can be derived by observing the EEG wave data. For example, when the consciousness state change value is the anesthesia state change value, that is, MGRSSI, the MGRSSI corresponding to EEG data 1 is 100, the MGRSSI corresponding to EEG data 2 is 70, and the MGRSSI corresponding to EEG data n is 0. Or when the consciousness state change value is the injury state change value, that is, MGRNOX, the MGRNOX corresponding to EEG data 1 is 100, the MGRNOX corresponding to EEG data 2 is 60, and the MGRNOX corresponding to EEG data n is 0.
[0046] Step S20: For each piece of EEG data, perform frequency domain transformation on it to obtain corresponding EEG frequency domain data, and calculate feature data of the EEG data based on the EEG frequency domain data.
[0047] In some embodiments, for each piece of EEG data, the EEG frequency domain data corresponding to the EEG data is calculated using the following Fourier transform calculation formula, wherein the Fourier transform calculation method is as follows:
[0048]
[0049] in, represents the corresponding EEG frequency domain data, f(x) represents the EEG data, ω represents the angular frequency, x represents time, i represents the imaginary unit, R n Represents the set of n-dimensional real numbers.
[0050] Please refer to Figure 4 In some embodiments, step S20 calculates characteristic data of the EEG data based on the EEG frequency domain data, including steps S21 to S24, which are described in detail below.
[0051] Step S21: Calculate the energy value of the EEG frequency domain data.
[0052] In some embodiments, the frequency range of the EEG frequency domain data obtained after performing frequency domain transformation on the EEG data is 1-44 Hz, and the corresponding energy value of the EEG frequency domain data at 1-44 Hz is E0.
[0053] Step S22: Divide the EEG frequency domain data into frequency band data in a plurality of different frequency ranges.
[0054] In some embodiments, the EEG frequency domain data may be divided into frequency band data in a plurality of different frequency ranges, for example, it may be divided into frequency band data in a 4-8 Hz frequency range, a 8-13 Hz frequency range, a 11-22 Hz frequency range, and a 33-44 Hz frequency range.
[0055] Step S23: Calculate the energy value of each frequency band data respectively.
[0056] In some embodiments, the energy value E1 of the frequency band data in the frequency range of 4-8 Hz, the energy value E2 of the frequency band data in the frequency range of 8-13 Hz, the energy value E3 of the frequency band data in the frequency range of 11-22 Hz, and the energy value E4 of the frequency band data in the frequency range of 33-44 Hz are calculated respectively.
[0057] Step S24: Calculate the characteristic data of the EEG data based on the energy value of the EEG frequency domain data and the energy values of all frequency band data.
[0058] In some embodiments, the EEG data has N feature data and N frequency band data, where N is an integer greater than 1.
[0059] In some embodiments, characteristic data of EEG data is calculated based on the energy value of EEG frequency domain data and the energy values of all frequency band data, including:
[0060] The nth characteristic data of the EEG data is calculated according to the ratio between the energy value of the nth frequency band data in the N frequency band data and the energy value of the EEG frequency domain data, where n=1, 2...N.
[0061] In some embodiments, the nth characteristic data of the EEG data is calculated according to the following formula:
[0062] F n =20×ln(E n / E0)
[0063] Among them, F n Represents the nth feature data of EEG data, E n It represents the energy value of the nth frequency band data among N frequency band data, and E0 represents the energy value of the EEG frequency domain data.
[0064] Step S30: Using the feature data of multiple EEG data as training sample data and the corresponding consciousness state change values as labels, the preset consciousness state recognition model is trained to obtain a trained consciousness state recognition model.
[0065] In some embodiments, the preset consciousness state recognition model can be an adaptive neural fuzzy inference system, which is a hybrid of an artificial neural network and a fuzzy logic system. During the training process, a back propagation algorithm is used to adjust the premise parameters in the adaptive neural fuzzy inference system, and a least squares estimation algorithm is used to adjust the conclusion parameters in the adaptive neural fuzzy inference system to achieve model training. At the same time, the adaptive neural fuzzy inference system can also automatically generate If-Then rules, where the If-Then rules are a kind of judgment logic that can output different results according to the characteristics corresponding to the received data.
[0066] Please refer to Figure 5 In an embodiment of the present invention, a method for identifying a state of consciousness is provided, including steps S40 to S60, which are described in detail below.
[0067] Step S40: Acquire the patient's EEG data.
[0068] Step S50: Perform frequency domain transformation on the patient's EEG data to obtain corresponding EEG frequency domain data, and calculate characteristic data of the patient's EEG data based on the EEG frequency domain data.
[0069] In some embodiments, for each piece of EEG data, the EEG frequency domain data corresponding to the EEG data is calculated using the following Fourier transform calculation formula, wherein the Fourier transform calculation method is as follows:
[0070]
[0071] in, represents the corresponding EEG frequency domain data, f(x) represents the EEG data, ω represents the angular frequency, x represents time, i represents the imaginary unit, R n Represents the set of n-dimensional real numbers.
[0072] In some embodiments, the energy value of the EEG frequency domain data is calculated, the EEG frequency domain data is divided into frequency band data in multiple different frequency ranges, and the energy value of each frequency band data is calculated separately, and the characteristic data of the EEG data is obtained based on the energy value of the EEG frequency domain data and the energy values of all frequency band data.
[0073] In some embodiments, the frequency range of the EEG frequency domain data obtained after the EEG data is frequency-domain transformed is 1-44 Hz, and the corresponding energy value for the EEG frequency domain data at 1-44 Hz is E0. The EEG frequency domain data is divided into frequency band data at multiple different frequency ranges, for example, it can be divided into frequency band data at 4-8 Hz frequency range, 8-13 Hz frequency range, 11-22 Hz frequency range, and 33-44 Hz frequency range, and the energy value E1 of the frequency band data at 4-8 Hz frequency range, the energy value E2 of the frequency band data at 8-13 Hz frequency range, the energy value E3 of the frequency band data at 11-22 Hz frequency range, and the energy value E4 of the frequency band data at 33-44 Hz frequency range are calculated respectively. There are N feature data of EEG data, and there are N frequency band data, where N is an integer greater than 1.
[0074] In some embodiments, characteristic data of EEG data is calculated based on the energy value of EEG frequency domain data and the energy values of all frequency band data, including:
[0075] The nth characteristic data of the EEG data is calculated according to the ratio between the energy value of the nth frequency band data in the N frequency band data and the energy value of the EEG frequency domain data, where n=1, 2...N.
[0076] In some embodiments, the nth characteristic data of the EEG data is calculated according to the following formula:
[0077] F n =20×ln(E n / E0)
[0078] Among them, F n Represents the nth feature data of EEG data, E n It represents the energy value of the nth frequency band data among N frequency band data, and E0 represents the energy value of the EEG frequency domain data.
[0079] Step S60: Input the characteristic data of the patient's EEG data into the consciousness state recognition model trained by the model training method to obtain the patient's consciousness state change value.
[0080] In some embodiments, the obtained value of the patient's consciousness state change can be used as a reliable basis for doctors to evaluate the patient's anesthesia level during surgery. The consciousness state change value can be a statistical value of the patient's state change from full wakefulness to deep anesthesia and sedation represented by a dimensionless value from 100 to 0.
[0081] In some embodiments, the steps of training a consciousness state recognition model using a model training method are as follows:
[0082] Obtain multiple pieces of EEG data of the subject in different test scenarios and the consciousness state change values corresponding to the multiple pieces of EEG data;
[0083] For each piece of EEG data, perform frequency domain transformation on it to obtain corresponding EEG frequency domain data, and calculate characteristic data of the EEG data based on the EEG frequency domain data;
[0084] The characteristic data of multiple EEG data are used as training sample data, and the corresponding consciousness state change values are used as labels to train the preset consciousness state recognition model to obtain a trained consciousness state recognition model.
[0085] Please refer to Figure 6 In an embodiment of the present invention, a consciousness state recognition device is provided, including a feature extraction module 10 and a consciousness state recognition module 20, which are described in detail below.
[0086] The feature extraction module 10 is used to obtain the patient's EEG data, perform frequency domain transformation on the patient's EEG data to obtain corresponding EEG frequency domain data, and calculate feature data of the patient's EEG data based on the EEG frequency domain data.
[0087] In some embodiments, for each piece of EEG data, the feature extraction module 10 calculates the EEG frequency domain data corresponding to the EEG data by using the following Fourier transform calculation formula, wherein the Fourier transform calculation method is as follows:
[0088]
[0089] in, represents the corresponding EEG frequency domain data, f(x) represents the EEG data, ω represents the angular frequency, x represents time, i represents the imaginary unit, R n Represents the set of n-dimensional real numbers.
[0090] In some embodiments, the feature extraction module 10 calculates the energy value of the EEG frequency domain data, divides the EEG frequency domain data into frequency band data in multiple different frequency ranges, and respectively calculates the energy value of each frequency band data, and obtains the feature data of the EEG data based on the energy value of the EEG frequency domain data and the energy values of all frequency band data.
[0091] In some embodiments, the frequency range of the EEG frequency domain data obtained after the EEG data is frequency-domain transformed is 1-44 Hz, and the corresponding energy value for the EEG frequency domain data at 1-44 Hz is E0. The EEG frequency domain data is divided into frequency band data at multiple different frequency ranges, for example, it can be divided into frequency band data at 4-8 Hz frequency range, 8-13 Hz frequency range, 11-22 Hz frequency range, and 33-44 Hz frequency range, and the energy value E1 of the frequency band data at 4-8 Hz frequency range, the energy value E2 of the frequency band data at 8-13 Hz frequency range, the energy value E3 of the frequency band data at 11-22 Hz frequency range, and the energy value E4 of the frequency band data at 33-44 Hz frequency range are calculated respectively. There are N feature data of EEG data, and there are N frequency band data, where N is an integer greater than 1.
[0092] In some embodiments, the feature extraction module 10 calculates feature data of the EEG data based on the energy value of the EEG frequency domain data and the energy values of all frequency band data, including:
[0093] The nth characteristic data of the EEG data is calculated according to the ratio between the energy value of the nth frequency band data in the N frequency band data and the energy value of the EEG frequency domain data, where n=1, 2...N.
[0094] In some embodiments, the nth characteristic data of the EEG data is calculated according to the following formula:
[0095] F n =20×ln(E n / E0)
[0096] Among them, F n Represents the nth feature data of EEG data, E n It represents the energy value of the nth frequency band data among N frequency band data, and E0 represents the energy value of the EEG frequency domain data.
[0097] The consciousness state recognition module 20 is used to input the characteristic data of the patient's EEG data into the consciousness state recognition model trained by the model training method to obtain the patient's consciousness state change value.
[0098] In some embodiments, the obtained value of the patient's consciousness state change can be used as a reliable basis for doctors to evaluate the patient's anesthesia level during surgery. The consciousness state change value can be a statistical value of the patient's state change from full wakefulness to deep anesthesia and sedation represented by a dimensionless value from 100 to 0.
[0099] In some embodiments, the steps of the consciousness state recognition module 20 training the consciousness state recognition model by the model training method are as follows:
[0100] Obtain multiple pieces of EEG data of the subject in different test scenarios and the consciousness state change values corresponding to the multiple pieces of EEG data;
[0101] For each piece of EEG data, perform frequency domain transformation on it to obtain corresponding EEG frequency domain data, and calculate characteristic data of the EEG data based on the EEG frequency domain data;
[0102] The characteristic data of multiple EEG data are used as training sample data, and the corresponding consciousness state change values are used as labels to train the preset consciousness state recognition model to obtain a trained consciousness state recognition model.
[0103] In some embodiments, a model training device is provided in an embodiment of the present invention, including a memory for storing a program; a processor for implementing a model training method by executing the program stored in the memory, wherein the model training method is as follows:
[0104] Obtain multiple pieces of EEG data of the subject in different test scenarios and the consciousness state change values corresponding to the multiple pieces of EEG data;
[0105] For each piece of EEG data, perform frequency domain transformation on it to obtain corresponding EEG frequency domain data, and calculate characteristic data of the EEG data based on the EEG frequency domain data;
[0106] The characteristic data of multiple EEG data are used as training sample data, and the corresponding consciousness state change values are used as labels to train the preset consciousness state recognition model to obtain a trained consciousness state recognition model.
[0107] Those skilled in the art will appreciate that all or part of the functions of the various methods in the above-mentioned embodiments can be implemented by hardware or by computer programs. When all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can be stored in a computer-readable storage medium, and the storage medium can include: read-only memory, random access memory, disk, optical disk, hard disk, etc., and the program is executed by a computer to implement the above-mentioned functions. For example, the program is stored in the memory of the device, and when the program in the memory is executed by the processor, all or part of the above-mentioned functions can be implemented. In addition, when all or part of the functions in the above-mentioned embodiments are implemented by computer programs, the program can also be stored in a storage medium such as a server, another computer, disk, optical disk, flash disk or mobile hard disk, and can be downloaded or copied and saved in the memory of the local device, or the system of the local device is updated, and when the program in the memory is executed by the processor, all or part of the functions in the above-mentioned embodiments can be implemented.
[0108] The above specific examples are used to illustrate the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the art, according to the concept of the present invention, some simple deductions, modifications or substitutions can be made.
Claims
1. A model training method, characterized in that: include: Acquire multiple pieces of EEG data of the subject in different test scenarios and consciousness state change values corresponding to the multiple pieces of EEG data; For each piece of the EEG data, perform frequency domain transformation on it to obtain corresponding EEG frequency domain data, and calculate characteristic data of the EEG data based on the EEG frequency domain data; The characteristic data of the plurality of EEG data are used as training sample data, and the corresponding consciousness state change values are used as labels to train the preset consciousness state recognition model to obtain a trained consciousness state recognition model.
2. The model training method according to claim 1, characterized in that: The step of calculating characteristic data of the EEG data according to the EEG frequency domain data includes: Calculating the energy value of the EEG frequency domain data; Dividing the EEG frequency domain data into frequency band data in a plurality of different frequency ranges; Calculate the energy value of each frequency band data respectively; The characteristic data of the EEG data is calculated based on the energy value of the EEG frequency domain data and the energy values of all the frequency band data.
3. The model training method according to claim 2, characterized in that: The feature data of the EEG data has N pieces, and the frequency band data has N pieces, where N is an integer greater than 1; the feature data of the EEG data is calculated based on the energy value of the EEG frequency domain data and the energy values of all the frequency band data, including: The nth characteristic data of the EEG data is obtained by calculating the ratio between the energy value of the nth frequency band data among the N frequency band data and the energy value of the EEG frequency domain data, wherein n=1, 2...N.
4. The model training method according to claim 3, characterized in that: The nth characteristic data of the EEG data is calculated according to the following formula: F n =20×ln(E n / E0) Among them, F n represents the nth feature data of the EEG data, E n represents the energy value of the nth frequency band data among the N frequency band data, and E0 represents the energy value of the EEG frequency domain data.
5. The model training method according to claim 1, characterized in that: The consciousness state change value includes an anesthesia state change value and / or an injury state change value, wherein the anesthesia state change value is used to determine the state change of the subject from awake to anesthetized during the anesthesia process, and the injury state change value is used to determine the state change of the subject from responsive to unresponsive during the process of facing injury.
6. The model training method according to claim 1, characterized in that: The consciousness state recognition model is an adaptive neural fuzzy inference system, and the training of the preset consciousness state recognition model includes: A back propagation algorithm is used to adjust the premise parameters in the adaptive neural fuzzy reasoning system, and a least squares estimation algorithm is used to adjust the conclusion parameters in the adaptive neural fuzzy reasoning system.
7. A method for identifying a state of consciousness, characterized in that: include: Obtaining the patient’s EEG data; Performing frequency domain transformation on the patient's EEG data to obtain corresponding EEG frequency domain data, and calculating characteristic data of the patient's EEG data based on the EEG frequency domain data; The characteristic data of the patient's EEG data is input into a consciousness state recognition model trained by the model training method according to any one of claims 1 to 6 to obtain the patient's consciousness state change value.
8. A consciousness state recognition device, characterized in that: include: A feature extraction module is used to obtain the patient's EEG data, perform frequency domain transformation on the patient's EEG data to obtain corresponding EEG frequency domain data, and calculate feature data of the patient's EEG data based on the EEG frequency domain data; A consciousness state recognition module is used to input the characteristic data of the patient's EEG data into a consciousness state recognition model trained by the model training method described in any one of claims 1 to 6 to obtain the patient's consciousness state change value.
9. A model training device, characterized in that: include: Memory, used to store programs; A processor, configured to implement the model training method as described in any one of claims 1 to 6 by executing the program stored in the memory.
10. A computer-readable storage medium, characterized in that: The medium stores a program, which can be executed by a processor to implement the model training method as described in any one of claims 1-6.
