Children epilepsy feature recognition method based on multi-modal data

Abstract

According to the children epilepsy feature recognition method based on the multi-modal data, the multi-modal data comprehensive epileptic seizure prediction model is constructed, and higher prediction accuracy is obtained. Meanwhile, parameters of the whole brain nerve field model are optimally fitted by using a deep learning method according to actually-measured EEG data, optimal parameters of the model are obtained in combination with the actually-measured EEG data, and the constructed multi-scale modeling system has a reasonable biological basis and a solid theoretical basis.

Description

Technical Field

[0001] The present invention relates to the technical fields of neurodynamics theory and artificial intelligence deep learning, and particularly relates to a method for identifying children's epilepsy characteristics based on multi-modal data. Background Art

[0002] In the prior art, the explanations of children's epilepsy and related terms are as follows: Children's epilepsy: A neurological syndrome caused by complex etiologies, recurrent, paroxysmal, and temporary brain function disorders during the period from 0 to 18 years old; Multi-modal medical data: A data set that integrates electroencephalogram signals and data such as electrocardiogram and electromyogram, and the correlation information with diseases; Multi-scale epilepsy neural mechanism model: A full-brain neural field model under a multi-scale modeling system that fits the states and parameters of the full-brain neural field model, and can reconstruct the electroencephalogram activities in normal and epileptic states, and reproduce the electroencephalogram characteristics of various epileptic states including the pre-epileptic seizure period; Automatic recognition and diagnosis model for children's epilepsy: A computer model that can automatically recognize the electroencephalogram of children with epilepsy based on comprehensive characteristic indicators that effectively represent the characteristics of epilepsy diseases; Prediction model for children's epilepsy seizures based on multi-modal clinical data: A comprehensive epilepsy seizure prediction model for multi-modal data constructed based on the complementary degree of multi-modal data and their characteristics for the diagnosis and treatment effects of epilepsy, and the comprehensive characteristic indicators of multi-modal epilepsy data; Neural mechanism and auxiliary diagnosis and treatment model for children's epilepsy diseases: A control intervention algorithm model for children's epilepsy for clinical application developed through optimal model parameters, based on objective data characteristics such as electroencephalogram signals and neural mechanism models; Nonlinear dynamics characteristic system: A comprehensive characteristic indicator used clinically to automatically identify children's epilepsy diseases through EEG, and can effectively represent the characteristics of epilepsy diseases.

[0003] In the prior art, the general methods for identifying children's epilepsy are as follows: Pay attention to the characteristic clinical manifestations of children's epilepsy seizures and record the characteristics; Clinically, electroencephalogram (EEG) is generally used as the basic tool for diagnosing children's epilepsy diseases; Conduct research on the neurodynamics mechanism of epilepsy based on the central nervous system of the cerebral cortex; Auxiliary treatment algorithms for automatic recognition, seizure prediction, and epileptogenic focus localization of children's epilepsy.

[0004] In the prior art, the following problems still exist: 1. Manual interpretation of electroencephalograms of children with epilepsy is time-consuming, boring, and involves subjective judgment factors of doctors; 2. Currently, the research on the neurodynamics mechanism of epilepsy is mainly limited to single-scale research and cannot explain problems from the overall level of the brain; 3. At present, most of the research on auxiliary treatment algorithms such as automatic recognition, seizure prediction, and epileptogenic focus localization for childhood epilepsy only uses scalp electroencephalograms of patients. However, epileptic seizures not only involve changes in electroencephalograms but also changes in other physiological signals such as electrocardiograms and electromyograms; 4. Using electroencephalograms alone to assist in the clinical diagnosis and treatment of epilepsy still has certain limitations. Electroencephalograms have high diagnostic value for childhood epilepsy, but there are also clinical cases showing that normal electroencephalograms cannot completely rule out epilepsy.

[0005] Therefore, it is necessary to integrate multi-scale neuroelectrophysiological signals of childhood epilepsy, explore the neurodynamics mechanism of epilepsy at the microscopic, mesoscopic, and macroscopic scales, and use big data to depict the pathogenesis and evolution law of childhood epilepsy at a deeper level. In order to assist in improving the current neurodynamics mechanism of epilepsy through feature recognition methods and overcome the limitations of single-scale research that cannot consider problems from the overall level of the brain. Summary of the Invention

[0006] In view of this, the present invention aims to propose a method for identifying features of childhood epilepsy based on multi-modal data. The present invention integrates electroencephalogram signals and multi-modal data such as electrocardiogram and electromyogram, effectively improving the accuracy of constructing an auxiliary diagnosis and treatment model for childhood epilepsy.

[0007] The present invention applies neurodynamics and deep learning methods to the research of childhood epilepsy. Compared with traditional machine learning algorithms, it is difficult to comprehensively generalize through small sample data. Deep learning, based on the imitation and design of the human brain structure, can complete tasks such as processing massive complex data, automatically extracting abstract features, and adaptive learning.

[0008] The technical solution of the present invention is as follows: A method for identifying features of childhood epilepsy based on multi-modal data, characterized by including the following: S1. Construct multi-scale neuro-mechanism dynamics models at the microscopic, mesoscopic, and macroscopic scales, and obtain optimal model parameters by fitting real electrophysiological data through the model output; S2. Based on objective data features such as electroencephalogram signals and neuro-mechanism models, establish an early recognition model for childhood epilepsy and a comprehensive recognition model for the types of childhood epilepsy; S3. Use deep learning methods for cross-modal data fusion analysis, study cross-modal associations, and improve the accuracy and stability of the epileptic seizure prediction model.

[0009] Furthermore, the feature recognition method specifically includes the following: Collection of children's epilepsy data and construction of database; Construction of multi-scale neural mechanisms and neural network models for children's epilepsy; Construct a non-linear dynamics feature system to achieve automatic recognition of children's epilepsy.

[0010] Furthermore, the feature recognition method further includes: real-time seizure warning based on clinical electroencephalogram and multi-modal data fusion.

[0011] Furthermore, in step S1, the parameters for constructing the microscopic, mesoscopic, and macroscopic scales include electroencephalogram signals and clinical characteristic data of electrocardiogram and electromyogram.

[0012] Furthermore, in step S1, it also includes non-linear dynamics feature indicators, which include correlation dimension, Lyapunov exponent, Hurst exponent, approximate entropy, sample entropy, wavelet entropy, and fractal characteristic parameters.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. The constructed comprehensive epilepsy seizure prediction model based on multi-modal data has obtained higher prediction accuracy.

[0014] 2. The parameters of the whole-brain neural field model are optimally fitted using deep learning methods based on the measured EEG data, and the optimal parameters of the model are obtained by combining the measured EEG data. The constructed multi-scale modeling system has a reasonable biological basis and a solid theoretical foundation.

[0015] 3. The fusion of multi-modal information can obtain more comprehensive features, improve the robustness of the model, and ensure that the model can still work effectively when some modal data is missing. Specific embodiments

[0016] The following will clearly and completely describe the technical solutions of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention. Embodiment 1

[0017] A method for identifying children's epilepsy features based on multi-modal data, characterized by including the following: S1. Construct a multi-scale neural mechanism dynamics model at the microscopic, mesoscopic, and macroscopic scales, and obtain the optimal model parameters by fitting the real electrophysiological data through the model output; S2. Based on objective data features such as electroencephalogram signals and neural mechanism models, establish an early recognition model for children's epilepsy and a comprehensive recognition model for children's epilepsy types; S3. Use deep learning methods for cross-modal data fusion analysis, study cross-modal associations, and improve the accuracy and stability of the epilepsy seizure prediction model.

[0018] Furthermore, the feature recognition method specifically includes the following: Collection of children's epilepsy data and construction of a database; Construction of multi-scale neural mechanisms and neural network models for children's epilepsy; Construct a non-linear dynamics feature system to achieve automatic recognition of children's epilepsy.

[0019] Furthermore, the feature recognition method also includes: real-time seizure warning based on clinical electroencephalogram and multi-modal data fusion.

[0020] Furthermore, in step S1, the parameters for constructing the microscopic, mesoscopic, and macroscopic aspects include electroencephalogram signals and clinical characteristic data of electrocardiogram and electromyogram.

[0021] Furthermore, in step S1, it also includes non-linear dynamics feature indicators, which include correlation dimension, Lyapunov exponent, Hurst exponent, approximate entropy, sample entropy, wavelet entropy, and fractal characteristic parameters.

[0022] The multi-modal data comprehensive epilepsy seizure prediction model constructed by the present invention has obtained higher prediction accuracy; the parameters of the whole-brain neural field model are optimally fitted using deep learning methods based on the measured EEG data, and the optimal parameters of the model are obtained by combining the measured EEG data. The constructed multi-scale modeling system has a reasonable biological basis and a solid theoretical foundation; the fusion of multi-modal information can obtain more comprehensive features, improve the robustness of the model, and ensure that the model can still work effectively when some modal data is missing.

[0023] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to encompass all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention.

[0024] In addition, it should be understood that although this specification is described in terms of embodiments, not every embodiment contains only an independent technical solution. This narrative manner of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art. It should be noted that the technical features not detailedly described in the present invention can all be implemented by any existing technology.

Claims

1. A method for identifying children's epilepsy characteristics based on multi-modal data, characterized in that including the following: S1. Construct multi-scale neural mechanism dynamics models at the microscopic, mesoscopic, and macroscopic levels, and obtain optimal model parameters by fitting real electrophysiological data with the model output; S2. Based on the objective data features of electroencephalogram signals and neural mechanism models, establish an early recognition model for childhood epilepsy and a comprehensive recognition model for the types of childhood epilepsy; S3. Use deep learning methods for cross-modal data fusion analysis, study cross-modal associations, and improve the accuracy and stability of the seizure prediction model.

2. The feature recognition method according to claim 1, wherein Specifically, it includes the following: Collection of childhood epilepsy data and construction of a database; Construction of multi-scale neural mechanisms and neural network models for childhood epilepsy; Construct a non-linear dynamics feature system to achieve automatic recognition of childhood epilepsy.

3. The feature recognition method according to claim 2, wherein It also includes: Real-time seizure warning based on clinical electroencephalogram and multi-modal data fusion.

4. The feature recognition method according to claim 1, wherein In step S1, the parameters for constructing the microscopic, mesoscopic, and macroscopic levels include electroencephalogram signals and clinical characteristic data of electrocardiogram and electromyogram.

5. The feature recognition method according to claim 4, wherein, It also includes non-linear dynamics feature indicators, which include correlation dimension, Lyapunov exponent, Hurst exponent, approximate entropy, sample entropy, wavelet entropy, and fractal characteristic parameters.