Derivation method of effective connection information, training method and device of digital twin brain model

Abstract

The application discloses a brain effective connection information derivation method, a digital twin brain model training method and device, and equipment. The method comprises the following steps: determining disturbance data of a first brain node in at least two brain nodes and brain neural data of the at least two brain nodes in a time sequence; obtaining first predicted neural data of the at least two brain nodes at a next time through a digital twin brain model based on the disturbance data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence; and determining brain effective connection information from the first brain node to a second brain node based on the first predicted neural data of the at least two brain nodes at the next time and second predicted neural data of the at least two brain nodes at the next time. The time sequence prediction network is trained based on a training data set to obtain the digital twin brain model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of brain neuroscience, and in particular to a method for deriving brain effective connection information, a method and device for training a digital twin brain model. BACKGROUND

[0002] Effective connectivity (EC) can represent the causal interaction between brain regions and is the basis for understanding brain information processing. The effective connectivity can be obtained by experimental methods and data-driven methods.

[0003] However, the common experimental method is not suitable for human whole brain stimulation and observation; the common data-driven effective connectivity inference method has high computational complexity based on the model method, and the model-free method can only distinguish whether the directed connection exists.

[0004] It can be seen that the common experimental method and data-driven method cannot obtain accurate brain effective connection information. SUMMARY

[0005] The embodiments of the present application provide a method for deriving brain effective connection information, a method and device for training a digital twin brain model, which can obtain accurate brain effective connection information.

[0006] The technical scheme of the embodiments of the present application is as follows:

[0007] In a first aspect, the embodiments of the present application provide a method for deriving brain effective connection information, the method comprising:

[0008] determining perturbation data of a first brain node in at least two brain nodes and brain neural data of the at least two brain nodes in a time sequence;

[0009] obtaining first predicted neural data of the at least two brain nodes at a next time point based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence through a digital twin brain model, wherein the digital twin brain model is obtained based on a time sequence prediction network;

[0010] determining brain effective connection information from the first brain node to a second brain node based on the first predicted neural data of the at least two brain nodes at the next time point and second predicted neural data of the at least two brain nodes at the next time point, wherein the second brain node is other than the first brain node in the at least two brain nodes; and the second predicted neural data represents the predicted data obtained without adding the perturbation data.

[0011] In the embodiments of the present application, based on the digital twin brain model obtained through training, the response of the brain under the perturbation stimulation is predicted by adding perturbation data to the brain nodes, and the effective connection information of the brain is inferred. Wherein, the digital twin brain model is obtained through training based on the time series prediction network, therefore, the digital twin brain model can effectively and accurately predict the dynamic changes of the brain neural data, so as to obtain more accurate brain effective connection information.

[0012] In a second aspect, the embodiments of the present application provide a training method of a digital twin brain model, the method comprises:

[0013] training a time series prediction network based on a training data set to obtain a digital twin brain model; wherein the digital twin brain model is used to predict the neural data of the next time based on the brain neural data of the brain nodes within a time series;

[0014] The training data set comprises neural training data corresponding to a preset time length p+1 of at least two brain nodes; wherein the neural training data corresponding to the preset time length p+1 comprises neural training data from the m-p time to the m time, and the training data set further comprises neural training data of the at least two brain nodes at the m+1 time; p is an integer greater than or equal to 0, and m is an integer greater than p;

[0015] The training of the time series prediction network based on the training data set to obtain the digital twin brain model comprises:

[0016] Through the time series prediction network, based on the neural training data of the at least two brain nodes from the m-p time to the m time, the prediction data of the at least two brain nodes at the m+1 time is obtained;

[0017] Based on the prediction data of the at least two brain nodes at the m+1 time and the neural training data of the at least two brain nodes at the m+1 time, the time series prediction network is corrected to obtain the digital twin brain model.

[0018] In the embodiments of the present application, the time series prediction network can be used to learn and predict the dynamic changes of the brain neural signal, and obtain a digital twin brain model with time series prediction function. Wherein, the digital twin brain model can effectively and accurately predict the dynamic changes of the brain neural data.

[0019] In a third aspect, the embodiments of the present application provide a brain effective connection information derivation device, the brain effective connection information derivation device comprises:

[0020] determining unit, configured to determine disturbance data of a first brain node of at least two brain nodes and brain neural data of the at least two brain nodes in a time sequence;

[0021] an obtaining unit, configured to obtain first predicted neural data of the at least two brain nodes at a next time through a digital twin brain model based on the disturbance data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence; wherein the digital twin brain model is obtained based on a time sequence prediction network;

[0022] The determining unit is further configured to determine brain effective connection information between the first brain node and a second brain node based on the first predicted neural data of the at least two brain nodes at the next time and second predicted neural data of the at least two brain nodes at the next time; wherein the second brain node is a brain node other than the first brain node among the at least two brain nodes; and the second predicted neural data represents predicted data obtained without adding the disturbance data.

[0023] In a fourth aspect, an embodiment of the present application provides a training device of a digital twin brain model, the training device of the digital twin brain model comprising:

[0024] a training unit, configured to train a time sequence prediction network based on a training data set to obtain a digital twin brain model; wherein the digital twin brain model is used to predict neural data at a next time based on brain neural data of brain nodes in a time sequence;

[0025] The training data set comprises neural training data corresponding to a preset time length p+1 of at least two brain nodes; wherein the neural training data corresponding to the preset time length p+1 comprises neural training data from an m-p time to an m time, and the training data set further comprises neural training data of the at least two brain nodes at an m+1 time; p is an integer greater than or equal to 0, and m is an integer greater than p;

[0026] The training of the time sequence prediction network based on the training data set to obtain the digital twin brain model comprises:

[0027] obtaining, through the time sequence prediction network, predicted data of the at least two brain nodes at the m+1 time based on the neural training data of the at least two brain nodes from the m-p time to the m time;

[0028] Based on the predicted data of the at least two brain nodes at the m+1 time point and the neural training data of the at least two brain nodes at the m+1 time point, the time series prediction network is corrected to obtain the digital twin brain model.

[0029] In a fifth aspect, an embodiment of the present application provides a computer device, the computer device comprising: a processor and a memory; wherein,

[0030] The memory is configured to store a computer program capable of running on the processor.

[0031] The processor is configured to execute the method according to the first aspect or the second aspect when the computer program is running.

[0032] In a sixth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a program, and the program is executed by a processor to implement the method according to the first aspect or the second aspect.

[0033] In a seventh aspect, an embodiment of the present application provides a computer program product, which comprises a computer program or instructions, and the computer program or instructions are executed by a processor to implement the method according to the first aspect or the second aspect.

[0034] Therefore, in the embodiment of the present application, the digital twin brain model is obtained based on the time series prediction network, and therefore, the digital twin brain model can effectively and accurately predict the dynamic changes of brain neural data, thereby obtaining more accurate brain effective connection information. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 Implementation process of the brain effective connection information derivation method proposed in the embodiment of the present application Figure One ;

[0036] Figure 2 Schematic diagram of the brain node proposed in the embodiment of the present application Figure One ;

[0037] Figure 3 Implementation process of the brain effective connection information derivation method proposed in the embodiment of the present application Figure Two ;

[0038] Figure 4 Derivation of brain effective connection information proposed in the embodiment of the present application Figure One ;

[0039] Figure 5 Derivation of brain effective connection information proposed in the embodiment of the present application Figure Two ;

[0040] Figure 6 A schematic diagram of derivation of brain effective connectivity information for an embodiment of the present application Figure Three ;

[0041] Figure 7 A schematic diagram of derivation of brain effective connectivity information for an embodiment of the present application Figure Four ;

[0042] Figure 8 A schematic diagram of derivation of brain effective connectivity information for an embodiment of the present application Figure Five ;

[0043] Figure 9 A schematic diagram of implementation of training of a digital twin brain model for an embodiment of the present application

[0044] Figure 10 A schematic diagram of a brain node for an embodiment of the present application Figure Two ;

[0045] Figure 11 A schematic diagram of analysis of neural data under real stimulation

[0046] Figure 12 A schematic diagram of implementation of derivation of brain effective connectivity information based on a digital twin brain model obtained through training for an embodiment of the present application

[0047] Figure 13 A schematic diagram of a composition structure of a device for deriving brain effective connectivity information for an embodiment of the present application

[0048] Figure 14 A schematic diagram of a composition structure of a device for training a digital twin brain model for an embodiment of the present application

[0049] Figure 15 A schematic diagram of a composition structure of a computer device for an embodiment of the present application DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. It can be understood that the specific embodiments described herein are only used to explain the related application, and not to limit the application. In addition, it should be noted that, for the convenience of description, only the parts related to the application are shown in the drawings.

[0051] Brain network models are often used to depict the correlation between different regions of the brain and the conversion of brain states, and are an important method for understanding the working mechanism of the brain under different scales and states. Exploring the causal relationship between different regions in the brain network is crucial for in-depth understanding of the advanced cognitive functions of the brain and locating the lesions of brain functional diseases.

[0052] Causal relationships in brain networks describe how one brain region directly influences other brain regions. Constructing causal brain networks requires determining the size, direction of such influences, and distinguishing excitatory from inhibitory influences, which are also called effective connections.

[0053] However, the dynamic changes of brain networks are highly nonlinear, and constructing linear dynamics models with finite state variables cannot accurately describe the complexity of the dynamics of the whole brain network, and the direction and size of information flow between brain regions and the causality of their interactions are not clear. Therefore, how to accurately establish a brain network dynamics model to describe the complex dynamic processes in the brain is a great challenge.

[0054] Currently, effective connections can be obtained by experimental methods and data-driven methods.

[0055] Experimental methods obtain effective connections by perturbing the brain signals of a region through electrical or magnetic stimulation and observing how this perturbation propagates to other brain regions. However, it is technically infeasible to simultaneously apply stimulation and observation across the whole brain. Therefore, many studies attempt to infer effective connections from neural data, such as functional magnetic resonance imaging data and electroencephalography data.

[0056] Data-driven methods include model-based methods and model-free methods.

[0057] Model-based methods typically parameterize effective connections in a generative model and fit neural signals to estimate effective connections based on model assumptions. For example, dynamic causal modeling uses biophysical models in which effective connections are parameters in neural dynamics, transforming generated neural signals into blood oxygen level dependent signals using hemodynamic functions for fitting functional magnetic resonance imaging data. However, although dynamic causal modeling has been widely used to estimate effective connections between a small number of regions, this approach is not suitable for studying whole-brain effective connections that consider signals from all brain regions due to its high computational complexity in parameter estimation. Another limitation of such model-based methods is their dependence on model assumptions, which can lead to significant inference bias if the model does not match the actual brain dynamics.

[0058] Model-free methods do not rely on explicit assumptions about underlying neural dynamics and use statistical methods to infer effective connections from observed neural signals. For example, Granger causality uses time series analysis to determine how much past neural activity in one region can predict future neural activity in another region. This method can be used to infer the strength and direction of effective connections, but it cannot distinguish between excitatory and inhibitory connections. In addition, due to conceptual and performance limitations, the effectiveness of Granger causality analysis is also controversial.

[0059] In addition to the statistical-based methods, some deep learning-based causal inference models have been proposed in recent years. Such neural network-based causal inference models generally represent the brain network with a graph structure, use data-driven methods to treat the effective connections in the neural signals as hidden tasks, and use explicit tasks such as neural signal prediction or classification to drive the learning of the connection relationships between different brain regions in the brain network. Such data-driven causal inference neural network models have achieved excellent performance in many fields such as physical system trajectory prediction, and have great application potential for inferring effective connections in the brain. However, most of these methods can only distinguish whether a directed connection exists, and cannot depict the weight and sign of the connection, and the individual differences of the obtained connections are small, which limits their application in target selection for precise neural regulation.

[0060] Therefore, the common experimental methods, such as obtaining effective connections by disturbing the brain signals of a region through electrical or magnetic stimulation and observing how the disturbance propagates to other brain regions, are not applicable to the whole brain of a human being. The common data-driven effective connection inference methods have high computational complexity based on models, and the model-free methods can only distinguish whether a directed connection exists.

[0061] That is, the common experimental methods and data-driven methods cannot obtain accurate brain effective connection information.

[0062] To solve the above problems, in the embodiments of the present application, on the one hand, a time series prediction network can be used to learn and predict the dynamic changes of brain neural signals to obtain a digital twin brain model with time series prediction function; on the other hand, based on the digital twin brain model obtained through training, the brain effective connection information can be inferred by predicting the response of the brain under the disturbance of the brain nodes.

[0063] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0064] An embodiment of the present application provides a brain effective connection information derivation method, which can be applied to a brain effective connection information derivation device or a computer device, and the present application does not make specific limitations. In the following, the brain effective connection information derivation method proposed in the embodiments of the present application will be exemplarily described by taking the brain effective connection information derivation device as an example.

[0065] Further, in the embodiments of the present application,Figure 1 An implementation flowchart of the method for deriving brain effective connectivity information according to an embodiment of the present application Figure One As shown in Figure 1 The method for deriving brain effective connectivity information can include the following steps:

[0066] Step 101, determining perturbation data of a first brain node among at least two brain nodes and brain neural data of the at least two brain nodes within a time sequence.

[0067] In an embodiment of the present application, the device for deriving brain effective connectivity information can first determine the perturbation data of the first brain node among the at least two brain nodes, and can also determine the brain neural data of the at least two brain nodes within the time sequence.

[0068] It can be understood that the method for deriving brain effective connectivity information according to an embodiment of the present application can determine the brain effective connectivity information between different brain nodes, wherein a brain node can represent a brain region, that is, it can be considered that one brain node corresponds to one brain region. A brain can be divided into at least two brain regions, and accordingly, one brain can correspond to at least two brain nodes.

[0069] Exemplarily, in some embodiments, Figure 2 A schematic diagram of a brain node according to an embodiment of the present application Figure One As shown in Figure 2 It is assumed that a biological brain is divided into multiple regions, such as region A, region B, region C, and region D, and accordingly, it can be considered that the brain corresponds to at least brain node a, brain node b, brain node c, and brain node d.

[0070] It can be understood that in an embodiment of the present application, during the derivation of brain effective connectivity information, the brain neural data corresponding to a time sequence of each of the at least two brain nodes can be first acquired. That is, the device for deriving brain effective connectivity information needs to first acquire the brain neural data of each brain node corresponding to multiple time points.

[0071] In an embodiment of the present application, the time sequence can include continuous time sampling intervals or discrete time sampling points, which is not limited in the present application.

[0072] Further, in an embodiment of the present application, the brain neural data of the at least two brain nodes within the time sequence can include brain neural data corresponding to a preset time length p+1 of the at least two brain nodes. Wherein, the preset time length can determine the time length of the brain neural data used in the prediction process when predicting the brain neural data of the next time point. Wherein, p is an integer greater than or equal to 0.

[0073] It can be understood that, in the embodiments of the present application, the time length of the brain neural data used in the prediction process, i.e., the preset time length p+1, can be preset. For example, p=2 is preset, and the preset time length p+1 is 3.

[0074] It can be understood that, in the embodiments of the present application, the brain neural data of the at least two brain nodes within the time sequence can include brain neural data of the at least two brain nodes corresponding to a plurality of time points, which includes but is not limited to brain neural data of the at least two brain nodes corresponding to the preset time length p+1, i.e., the time sequence length corresponding to the time sequence is greater than or equal to the time length of the brain neural data used in the prediction process.

[0075] Exemplarily, in some embodiments, the time sequence length corresponding to the time sequence is 4800, and the time length of the brain neural data used in the prediction process is 3.

[0076] It can be understood that, in the embodiments of the present application, for a determined preset time length, the time length of the data used in the brain neural data prediction is fixed, for example, for the preset time length p+1, the brain neural data of the previous p+1 time points can be used to predict the brain neural data of the next time point.

[0077] That is to say, in the embodiments of the present application, the brain neural data of one or more time points can be used to predict the brain neural data of the next time point. For example, p=2 is preset, and the brain neural data of time point 1, time point 2 and time point 3 can be used to predict the brain neural data of time point 4, the brain neural data of time point 2, time point 3 and time point 4 can be used to predict the brain neural data of time point 5, and the brain neural data of time point 3, time point 4 and time point 5 can be used to predict the brain neural data of time point 6.

[0078] Exemplarily, in some embodiments, assuming that the preset time length is p+1, the time length indicated by the preset time length includes p+1 time points from time point t-p to time point t. Correspondingly, the brain neural data corresponding to the preset time length p+1 includes the brain neural data from time point t-p to time point t, and t is an integer greater than p.

[0079] That is to say, in some embodiments, the time sequence can include continuous p+1 time points from time point t-p to time point t. Correspondingly, the brain neural data of each brain node within the time sequence can include p+1 brain neural data of the brain node corresponding to the p+1 time points.

[0080] Further, in embodiments of the present application, the brain neural data of any one of the brain nodes in the time series can be determined in various manners, which are not specifically limited in the present application.

[0081] Illustratively, in some embodiments, the brain neural data can be acquired through existing bioinformatics databases, wherein the bioinformatics databases include, but are not limited to, one or more of the public data sets such as human fMRI data sets, EEG data sets, MEG data sets, etc.

[0082] Illustratively, in some embodiments, the brain neural data can also be generated through a computational model.

[0083] Further, in embodiments of the present application, the first brain node can be any one of the at least two brain nodes. For example, the first brain node can be brain node a, or it can also be brain node b.

[0084] In embodiments of the present application, during the derivation of the brain effective connection information, a certain perturbation data is also applied to the first brain node, so as to utilize the changes in the activities of the other brain nodes except the first brain node after the increase of the perturbation data to depict the effective connection of the first brain node to the other brain nodes.

[0085] It can be understood that, in embodiments of the present application, the perturbation data of the first brain node can be understood as a virtual perturbation corresponding to the first brain node. The present application does not specifically limit the form of the perturbation data of the first brain node. For example, an N-dimensional virtual perturbation vector corresponding to N brain nodes can be constructed in advance, wherein N is an integer greater than 1, the perturbation data of the first brain node can be an element corresponding to the N-dimensional virtual perturbation vector, the value of the element is a non-zero value, representing the intensity of the virtual perturbation added to the first brain node, and the values of the other elements are 0, i.e., the other brain nodes except the first brain node do not add virtual perturbation.

[0086] Further, in embodiments of the present application, when determining the perturbation data of the first brain node of the at least two brain nodes, the perturbation data of the first brain node can be set based on a preset perturbation intensity.

[0087] Illustratively, in some embodiments, for an N-dimensional virtual perturbation vector P corresponding to N brain nodes, it is assumed that each element p in the N-dimensional virtual perturbation vector P corresponds to a brain node, wherein if the i-th brain node is the first brain node to which the virtual perturbation is applied, when determining the perturbation data of the first brain node, the i-th element p of the first brain node in the N-dimensional virtual perturbation vector P can be selected as the perturbation data of the first brain node. iis set to Δ, where Δ can be a preset perturbation intensity; meanwhile, the i-th element p i other than the i-th element p

[0088] p i = Δ, p k，k≠i = 0 (1);

[0089] Step 102, obtaining, by the digital twin brain model, first predicted neural data of the at least two brain nodes at a next time based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence; wherein the digital twin brain model is obtained based on training of a time series prediction network.

[0090] In the embodiments of the present application, after the perturbation data of the first brain node in the at least two brain nodes and the brain neural data of the at least two brain nodes in the time sequence are determined, the first predicted neural data of the at least two brain nodes at the next time can be further determined by the digital twin brain model based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence, that is, the brain neural data of the other brain nodes is predicted by the digital twin brain model in the case that there is a virtual perturbation in the first brain node.

[0091] Further, in the embodiments of the present application, the digital twin brain model can be obtained based on training of a time series prediction network. Accordingly, the digital twin brain model can be used to determine the brain effective connection information based on the perturbation data of the brain nodes and the brain neural data of the brain nodes in the time sequence.

[0092] That is, in the embodiments of the present application, the digital twin brain model is obtained based on training of a time series prediction network, therefore, the digital twin brain model can effectively predict the dynamic changes of the brain neural signals (brain neural data), and further can obtain more accurate brain effective connection information.

[0093] Further, in the embodiments of the present application, the time series prediction network used for training the digital twin brain model can be an artificial neural network of any type and structure for time series prediction, which is not specifically limited in the present application.

[0094] Exemplarily, in some embodiments, the time series prediction network can use a multilayer perceptron, a convolutional neural network, a recurrent neural network, etc. classical model, and other network architectures suitable for time series prediction can also be selected, which is not specifically limited in the present application.

[0095] Exemplarily, in some embodiments, for a time sequence of continuous p+1 time instants from a t-p time instant to a t time instant, when the first predicted neural data of the at least two brain nodes at the next time instant is obtained based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes within the time sequence by the digital twin brain model, the post-perturbation brain neural data of the at least two brain nodes at the t time instant can be determined based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes at the t time instant first; then the first predicted neural data of the at least two brain nodes at the t+1 time instant can be further obtained based on the post-perturbation brain neural data of the at least two brain nodes at the t time instant and the brain neural data of the at least two brain nodes at the t-p time instant to the t-1 time instant by the digital twin brain model.

[0096] It can be understood that, in the embodiments of the present application, in the process of dynamically predicting the brain neural data by the digital twin brain model, the post-perturbation brain neural data added with virtual perturbation corresponding to the current time instant (such as the t time instant) can be generated according to the perturbation data of the first brain node corresponding to the current time instant and the brain neural data of all brain nodes corresponding to the current time instant.

[0097] Further, in the embodiments of the present application, the generation of the post-perturbation brain neural data at the t time instant based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes at the t time instant can be performed in various ways. For example, including but not limited to, performing mathematical operations on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes at the t time instant, thereby obtaining the post-perturbation brain neural data at the t time instant.

[0098] Exemplarily, in some embodiments, when the post-perturbation brain neural data of the at least two brain nodes at the t time instant is determined based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes at the t time instant, the perturbation data of the first brain node and the brain neural data of the first brain node at the t time instant can be superimposed, and then combined with the brain neural data of the other brain nodes at the t time instant to obtain the post-perturbation brain neural data of the at least two brain nodes at the t time instant.

[0099] It can be understood that in the embodiments of the present application, the brain nerve data of the at least two brain nodes after the disturbance at the t time can be used as the brain nerve data after the disturbance of the digital twin brain model to predict the brain nerve data of the next time, and finally the first predicted nerve data of the at least two brain nodes at the t+1 time can be output. Wherein, the brain nerve data of the at least two brain nodes after the disturbance at the t time is generated based on the disturbance data of the first brain node, so it can be considered that the digital twin brain model is used to predict the brain nerve data in the case of adding virtual disturbance.

[0100] That is, in the embodiments of the present application, the first predicted nerve data of the at least two brain nodes at the t+1 time is the brain nerve data predicted after adding virtual disturbance.

[0101] Exemplarily, in some embodiments, the process of determining the first predicted nerve data of the at least two brain nodes at the t+1 time by the digital twin brain model may refer to the following formula (2):

[0102]

[0103] Wherein, the digital twin brain model can be represented as f(·), P is the disturbance data, which can be an N-dimensional virtual disturbance vector, x (t-p) , x (t-p+1) ,..., x t may represent the brain nerve data from time point (t-p) to time point t.

[0104] Exemplarily, in some embodiments, the process of determining the first predicted nerve data of the at least two brain nodes at the t+1 time by the digital twin brain model may also refer to the following formula (3):

[0105]

[0106] Wherein, θ represents all the parameters in the digital twin brain model that can be optimized by training iteration.

[0107] Further, in the embodiments of the present application, Figure 3 the implementation process of the brain effective connection information derivation method proposed in the embodiments of the present application is shown in Figure Two As shown in Figure 3 , the brain effective connection information derivation method can further include the following steps:

[0108] At step 104, the second predicted neural data of the at least two brain nodes at the next time point is obtained based on the brain neural data of the at least two brain nodes in the time sequence by the digital twin brain model.

[0109] In the embodiments of the present application, after the perturbation data of the first brain node of the at least two brain nodes and the brain neural data of the at least two brain nodes in the time sequence are determined, the second predicted neural data of the at least two brain nodes at the next time point can be further determined based on the brain neural data of the at least two brain nodes in the time sequence by the digital twin brain model, i.e., the brain neural data of the other brain nodes is predicted by the digital twin brain model without the virtual perturbation of the first brain node.

[0110] Exemplarily, in some embodiments, for a time sequence of consecutive p+1 time points from the t-p time point to the t time point, when the first predicted neural data of the at least two brain nodes at the next time point is obtained based on the brain neural data of the at least two brain nodes in the time sequence by the digital twin brain model, the brain neural data of the at least two brain nodes at the t-p time point to the t time point can be used as the input data of the digital twin brain model to predict the brain neural data at the next time point, and finally the second predicted neural data of the at least two brain nodes at the t+1 time point can be output.

[0111] It can be understood that, in the embodiments of the present application, compared with the first predicted neural data, the second predicted neural data of the at least two brain nodes at the t+1 time point can be understood as the predicted brain neural data without the virtual perturbation.

[0112] That is, in the embodiments of the present application, the first predicted neural data can represent the predicted data obtained when the perturbation data is added, and the second predicted neural data represents the predicted data obtained when the perturbation data is not added.

[0113] Exemplarily, in some embodiments, the process of determining the second predicted neural data of the at least two brain nodes at the t+1 time point by the digital twin brain model can refer to the following formula (4):

[0114]

[0115] Wherein, the digital twin brain model can be represented as f(·), x (t-p) , x (t-p+1) ,..., x t may represent the brain neural data from the time point (t-p) to the time point t.

[0116] Exemplarily, in some embodiments, the second predicted neural data of the at least two brain nodes at the t+1 time point is determined by the digital twin brain model The process can also be referred to as formula (5) as follows:

[0117]

[0118] Wherein, θ represents all the parameters in the digital twin brain model that can be optimized through training iteration.

[0119] Step 103, determining the brain effective connection information from the first brain node to the second brain node based on the first predicted neural data of the at least two brain nodes at the next time point and the second predicted neural data of the at least two brain nodes at the next time point; wherein, the second brain node is other brain node in the at least two brain nodes except the first brain node; the second predicted neural data represents the predicted data obtained without adding the disturbance data.

[0120] In the embodiments of the present application, after obtaining the first predicted neural data and the second predicted neural data at the next time point by the digital twin brain model under the conditions of adding virtual disturbance and not adding virtual disturbance respectively, the brain effective connection information from the first brain node to the second brain node can be further determined based on the first predicted neural data of the at least two brain nodes at the next time point and the second predicted neural data of the at least two brain nodes at the next time point.

[0121] It can be understood that in the embodiments of the present application, the second brain node can be other brain node in the at least two brain nodes except the first brain node. For example, the first brain node is brain node a, and the second brain node can be one or more of brain node b, brain node c and brain node d.

[0122] In the embodiments of the present application, the brain effective connection information from the first brain node to the second brain node can be used to determine the corresponding changes generated by the second brain node after the virtual disturbance is applied to the first brain node.

[0123] It can be understood that in the embodiments of the present application, the first predicted neural data represents the predicted data obtained by adding the disturbance data, and the second predicted neural data represents the predicted data obtained without adding the disturbance data. Therefore, based on the first predicted neural data and the second predicted neural data, the brain effective connection information between the first brain node and any other brain node can be determined.

[0124] Further, in embodiments of the present application, the brain effective connection information from the first brain node to the second brain node determined based on the first predicted neural data and the second predicted neural data can include, but is not limited to, an effective connection type, an effective connection direction, and an effective connection strength.

[0125] It can be understood that, in embodiments of the present application, the effective connection type can determine and distinguish the influence nature of the brain effective connection between the two brain nodes. Wherein, the effective connection type can be, but is not limited to, excitatory and inhibitory.

[0126] It can be understood that, in embodiments of the present application, the effective connection direction can determine the transmission direction of the brain effective connection between the two brain nodes.

[0127] It can be understood that, in embodiments of the present application, the effective connection strength can determine the signal size of the brain effective connection between the two brain nodes.

[0128] It can be seen that, in embodiments of the present application, the brain effective connection information between the two brain nodes obtained based on the digital twin brain model not only clarifies the causal interaction in the brain, but also provides effective connection data with directionality, strength and excitatory / inhibitory characteristics.

[0129] Further, in embodiments of the present application, when determining the brain effective connection information from the first brain node to the second brain node based on the first predicted neural data of at least two brain nodes at the next moment and the second predicted neural data of at least two brain nodes at the next moment, the effective connection type can be determined as excitatory in the case that the first predicted neural data of the second brain node at the next moment is greater than the second predicted neural data of the second brain node at the next moment; and the effective connection type can be determined as inhibitory in the case that the first predicted neural data of the second brain node at the next moment is less than the second predicted neural data of the second brain node at the next moment. That is, in embodiments of the present application, the first predicted neural data and the second predicted neural data obtained in the case of increasing virtual disturbance and not increasing virtual disturbance are compared, and the corresponding effective connection type can be further determined according to the comparison result. Wherein, if the first predicted neural data is greater than the second predicted neural data, it can be considered that the neural activity of the second brain node corresponding to the case of existing disturbance data of the first brain node increases, at this time, the corresponding effective connection type can be determined as excitatory; if the first predicted neural data is less than the second predicted neural data, it can be considered that the neural activity of the second brain node corresponding to the case of existing disturbance data of the first brain node decreases, at this time, the corresponding effective connection type can be determined as inhibitory.

[0130] Therefore, in the embodiments of the present application, the effective connection type can reflect the increase or decrease of the neural activity of the other brain nodes after the first brain node is added with the virtual disturbance, and the generated effects correspond to excitatory or inhibitory, respectively.

[0131] Exemplarily, in some embodiments, for a time sequence of continuous p+1 time points from the t-p time point to the t time point, when determining the brain effective connection information of the first brain node to the second brain node based on the first predicted neural data of the at least two brain nodes at the next time point and the second predicted neural data of the at least two brain nodes at the next time point, the effective connection type can be determined as excitatory type in the case that the first predicted neural data of the second brain node at the t+1 time point is greater than the second predicted neural data of the second brain node at the t+1 time point; and the effective connection type can be determined as inhibitory type in the case that the first predicted neural data of the second brain node at the t+1 time point is less than the second predicted neural data of the second brain node at the t+1 time point. That is, in the embodiments of the present application, the first predicted neural data of the second brain node at the t+1 time point is obtained by adding the disturbance data to the first brain node at the t time point and combining the brain neural data of the first brain node and the second brain node from the t-p time point to the t time point; meanwhile, the second predicted neural data of the second brain node at the t+1 time point is obtained directly according to the brain neural data of the first brain node and the second brain node from the t-p time point to the t time point; if the first predicted neural data of the second brain node at the t+1 time point is greater than the second predicted neural data, it can be considered that the neural activity of the second brain node increases, that is, the disturbance data has an excitatory effect on the second brain node, and therefore the effective connection type can be determined as excitatory type; otherwise, if the first predicted neural data of the second brain node at the t+1 time point is less than the second predicted neural data, it can be considered that the neural activity of the second brain node decreases, that is, the disturbance data has an inhibitory effect on the second brain node, and therefore the effective connection type can be determined as inhibitory type.

[0132] Further, in the embodiments of the present application, when determining the brain effective connection information of the first brain node to the second brain node, the effective connection direction can be directly determined as from the first brain node to the second brain node.

[0133] It can be understood that, in the embodiments of the present application, the monitoring of the neural data of the other brain nodes in the case of applying the virtual disturbance to the first brain node can consider that the effective connection direction of the brain effective connection information between the first brain node and the second brain node is from the first brain node to the second brain node.

[0134] Further, in the embodiments of the present application, when the brain effective connection information from the first brain node to the second brain node is determined based on the first predicted neural data of the second brain node at the next time and the second predicted neural data of the second brain node at the next time, the intensity information of the first predicted neural data of the second brain node at the t+1 time can also be determined, and the intensity information of the second predicted neural data of the second brain node at the t+1 time can also be determined; then, the effective connection intensity is determined based on the intensity information of the first predicted neural data and the intensity information of the second predicted neural data.

[0135] It can be understood that in the embodiments of the present application, the first predicted neural data of the second brain node at the t+1 time is obtained by adding the perturbation data to the first brain node at the t time and combining the brain neural data of the first brain node and the second brain node from the t-p time to the t time; at the same time, the second predicted neural data of the second brain node at the t+1 time is obtained directly according to the brain neural data of the first brain node and the second brain node from the t-p time to the t time; the intensity information of the first predicted neural data and the intensity information of the second predicted neural data are determined respectively, and then the effective connection intensity is further determined based on the intensity information of the two predicted neural data. For example, the intensity information of the first predicted neural data and the intensity information of the second predicted neural data can be subjected to difference operation, and the effective connection intensity is determined according to the difference result.

[0136] Further, in the embodiments of the present application, after the intensity information of the first predicted neural data and the intensity information of the second predicted neural data are determined respectively and compared, if the comparison result is that the intensity information of the first predicted neural data is equal to the intensity information of the second predicted neural data, it can be considered that whether the perturbation data is added to the first brain node or not, the predicted neural data corresponding to the second brain node will not change, that is, the perturbation data added to the first brain node will not affect the second brain node, at this time, it can be considered that there is no brain effective connection information between the first brain node and the second brain node.

[0137] Exemplarily, in some embodiments, assuming that the brain node i is the first brain node and the brain node j is the second brain node, the process of determining the brain effective connection information EC ij between the two brain nodes through the digital twin brain model based on the above formulas (3) and (5) can be as follows:

[0138] EC ij = f(x (t-p) , x (t-p+1) ,..., x t + P, θ) j -f(x(t-p) , x (t-p+1) ,..., x t , θ) j (6);

[0139] Exemplarily, in some embodiments, assuming that the brain node i is a first brain node and the brain node j is a second brain node, based on the above formulas (3) and (5), the process of determining the brain effective connection information EC ij between the two brain nodes by the digital twin brain model can also be as follows formula (7):

[0140] EC ij = E t [f(x (t-p) , x (t-p+1) ,..., x t + P, θ) j -f(x (t-p) , x (t-p+1) ,..., x t , θ) j ] (7);

[0141] wherein, E t (·) represents the mean value over time t.

[0142] Further, in the embodiments of the present application, each of the at least two brain nodes can be traversed, so that the brain effective connection information between any two of the at least two brain nodes can be determined, and further the whole brain effective connection of the at least two brain nodes can be obtained.

[0143] That is to say, in the embodiments of the present application, a virtual disturbance can be applied to each brain node, that is, the disturbance data is added as input information, and then the brain neural data of other brain nodes is predicted by the digital twin brain model respectively, and the brain neural data of other brain nodes predicted when no virtual disturbance is applied is combined, and further the brain effective connection information between any two brain nodes is determined.

[0144] It can be understood that in the embodiments of the present application, after the brain effective connection information between each two brain nodes is determined, the determination of the whole brain effective connection can be further completed.

[0145] Exemplarily, in some embodiments, Figure 4 derivation of brain effective connection information proposed in the embodiments of the present application Figure One , for example, Figure 4As shown, assuming that the brain node a is the first brain node, after a virtual disturbance is applied to the brain node a, the brain nerve data of the brain node a at the t time is superimposed with the corresponding disturbance data, and the brain nerve data of the brain node b, the brain node c and the brain node d at the t time is combined to obtain the brain nerve data of the four brain nodes after the disturbance at the t time as the input of the digital twin brain model; meanwhile, the brain nerve data of the four brain nodes at the t-p time to the t-1 time can also be input into the digital twin brain model. Through the digital twin brain model, the brain nerve data of the four brain nodes at the next time can be predicted, that is, the first prediction nerve data obtained by prediction can include the first prediction nerve data of the four brain nodes at the t+1 time. Meanwhile, the brain nerve data of the four brain nodes at the t-p time to the t time can also be input into the digital twin brain model, and the second prediction nerve data can be obtained by prediction through the digital twin brain model, wherein the second prediction nerve data can include the second prediction nerve data of the four brain nodes at the t+1 time. By comparing the first prediction nerve data and the second prediction nerve data of the brain node b, the brain node c and the brain node d at the t+1 time respectively, it can be determined that the brain nerve data of the brain node b and the brain node c has no change, and the brain nerve data of the brain node d is enhanced, and then it can be determined that the brain effective connection information of the brain node a to the brain node d includes: the effective connection direction is from the brain node a to the brain node d, the effective connection type is excitation type, and the effective connection strength is the strength change value of the brain nerve data of the brain node d.

[0146] Exemplarily, in some embodiments, Figure 5 Derivation of brain effective connection information for embodiments of the present application Figure Two As Figure 5As shown, assuming the brain node b is the first brain node, after a virtual disturbance is applied to the brain node b, the brain nerve data of the brain node b at the t time is superimposed with the corresponding disturbance data, and the brain nerve data of the brain node a, the brain node c and the brain node d at the t time is combined to obtain the brain nerve data of the four brain nodes after the disturbance at the t time as the input of the digital twin brain model; meanwhile, the brain nerve data of the four brain nodes at the t-p time to the t-1 time can also be input into the digital twin brain model. Through the digital twin brain model, the brain nerve data of the four brain nodes at the next time can be predicted, that is, the first predicted nerve data obtained by prediction can include the first predicted nerve data of the four brain nodes at the t+1 time. Meanwhile, the brain nerve data of the four brain nodes at the t-p time to the t time can also be input into the digital twin brain model, and the second predicted nerve data is obtained by prediction through the digital twin brain model, wherein the second predicted nerve data can include the second predicted nerve data of the four brain nodes at the t+1 time. By comparing the first predicted nerve data and the second predicted nerve data of the brain node a, the brain node c and the brain node d at the t+1 time respectively, it can be determined that the brain nerve data of the brain node d has no change, the brain nerve data of the brain node a is weakened, and the brain nerve data of the brain node c is enhanced, and it can be further determined that the brain effective connection information from the brain node b to the brain node a includes: the effective connection direction is from the brain node b to the brain node a, the effective connection type is inhibitory type, and the effective connection strength is the intensity change value of the brain nerve data of the brain node a, and it is determined that the brain effective connection information from the brain node b to the brain node c includes: the effective connection direction is from the brain node b to the brain node c, the effective connection type is excitatory type, and the effective connection strength is the intensity change value of the brain nerve data of the brain node c.

[0147] Exemplarily, in some embodiments, Figure 6 Derivation of brain effective connection information for embodiments of the present application Figure Three As Figure 6As shown, assuming that the brain node c is the first brain node, after a virtual disturbance is applied to the brain node c, the brain nerve data of the brain node c at the t time is superimposed with the corresponding disturbance data, and the brain nerve data of the brain node b, the brain node a and the brain node d at the t time is combined to obtain the brain nerve data of the four brain nodes after the disturbance at the t time as the input of the digital twin brain model; meanwhile, the brain nerve data of the four brain nodes at the t-p time to the t-1 time can also be input into the digital twin brain model. Through the digital twin brain model, the brain nerve data of the four brain nodes at the next time can be predicted, that is, the first prediction nerve data obtained by prediction can include the first prediction nerve data of the four brain nodes at the t+1 time. Meanwhile, the brain nerve data of the four brain nodes at the t-p time to the t time can also be input into the digital twin brain model to obtain the second prediction nerve data by prediction, wherein the second prediction nerve data can include the second prediction nerve data of the four brain nodes at the t+1 time. By comparing the first prediction nerve data and the second prediction nerve data of the brain node b, the brain node a and the brain node d at the t+1 time respectively, it can be determined that the brain nerve data of the brain node b and the brain node d has no change, and the brain nerve data of the brain node a is weakened, and then it can be determined that the brain effective connection information from the brain node c to the brain node a includes: the effective connection direction is from the brain node c to the brain node a, the effective connection type is inhibitory type, and the effective connection strength is the intensity change value of the brain nerve data of the brain node a.

[0148] Exemplarily, in some embodiments, Figure 7 Deduction of brain effective connection information for embodiments of the present application Figure Four As Figure 7As shown, assuming that the brain node d is the first brain node, after a virtual disturbance is applied to the brain node d, the brain nerve data of the brain node d at the t time is superimposed with the corresponding disturbance data, and the brain nerve data of the brain node b, the brain node a and the brain node c at the t time is combined to obtain the brain nerve data of the four brain nodes after the disturbance at the t time as the input of the digital twin brain model; meanwhile, the brain nerve data of the four brain nodes at the t-p time to the t-1 time can also be input into the digital twin brain model. Through the digital twin brain model, the brain nerve data of the four brain nodes a, b, c and d at the next time can be predicted, that is, the first predicted nerve data obtained by prediction can include the first predicted nerve data of the four brain nodes a, b, c and d at the t+1 time. Meanwhile, the brain nerve data of the four brain nodes at the t-p time to the t time can also be input into the digital twin brain model, and the second predicted nerve data can be obtained by prediction through the digital twin brain model, wherein the second predicted nerve data can include the second predicted nerve data of the four brain nodes a, b, c and d at the t+1 time. Comparing the first predicted nerve data and the second predicted nerve data of the brain node b, the brain node a and the brain node c at the t+1 time respectively, it can be determined that the brain nerve data of the brain node b and the brain node a has no change, and the brain nerve data of the brain node c is weakened, and then it can be determined that the brain effective connection information from the brain node d to the brain node c includes: the effective connection direction is from the brain node d to the brain node c, the effective connection type is inhibitory type, and the effective connection strength is the intensity change value of the brain nerve data of the brain node c.

[0149] Exemplarily, in some embodiments, Figure 8 Deduction of brain effective connection information for embodiments of the present application Figure Five As Figure 8 As shown, after the brain node a, the brain node b, the brain node c and the brain node d are traversed according to the above method, the brain effective connection information between each two brain nodes under the condition of applying a virtual disturbance to different brain nodes can be determined, and then the determination of the whole brain effective connection can be completed, wherein the whole brain effective connection can include the effective connection type, the effective connection direction and the effective connection strength between different brain nodes.

[0150] Further, in the embodiments of the present application, since the time sequence length corresponding to the time sequence is greater than or equal to the time length of the brain neural data used in the prediction process, the prediction of the brain effective connection information at multiple time points by the digital twin brain model can be selected, and the average calculation is performed based on the multiple brain effective connection information obtained by the prediction to determine the effective connection in the average sense. The effective connection in the average sense can more accurately and comprehensively reflect the activity between the brain nodes.

[0151] Further, in the embodiments of the present application, the first prediction neural data corresponding to the multiple time points of the at least two brain nodes can be obtained by the digital twin brain model based on the multiple disturbance data corresponding to the multiple time points of the first brain node and the brain neural data of the at least two brain nodes within the time sequence; at the same time, the second prediction neural data corresponding to the multiple time points of the at least two brain nodes can be obtained by the digital twin brain model based on the brain neural data of the at least two brain nodes within the time sequence; then, the multiple brain effective connection information from the first brain node to the second brain node corresponding to the multiple time points can be determined based on the multiple first prediction neural data and the multiple second prediction neural data; finally, the target effective connection information from the first brain node to the second brain node can be determined based on the multiple brain effective connection information.

[0152] It can be understood that, in the embodiments of the present application, the first prediction neural data is the brain neural data predicted after adding a virtual disturbance, and the second prediction neural data is the brain neural data predicted without adding a virtual disturbance.

[0153] It can be understood that, in the embodiments of the present application, the average calculation is performed based on the multiple brain effective connection information from the first brain node to the second brain node corresponding to the multiple time points, and the average result finally obtained is the target effective connection information from the first brain node to the second brain node. The target effective connection information can represent the effective connection in the average sense from the first brain node to the second brain node.

[0154] That is, in the embodiments of the present application, for a large enough time series length, the prediction of the brain neural data for a plurality of optional parameters t can be selected based on the brain neural data within the time series, that is, a plurality of predicted neural data are obtained, including a plurality of first predicted neural data obtained when the increasing perturbation data corresponding to a plurality of time points and a plurality of second predicted neural data obtained when the non-increasing perturbation data, and finally the effective connection in the average sense is determined by using the plurality of first predicted neural data and the plurality of second predicted neural data corresponding to the plurality of time points. Among them, for a time t, the effective connection obtained at a given t can be considered to be dependent on the current state, and for a plurality of time t, the effective connection is calculated at all optional t and averaged, and the effective connection in the average sense (target effective connection information) can be obtained.

[0155] Exemplarily, in some embodiments, it is assumed that the digital twin brain is 3-step prediction 1-step (predicting the brain neural data at the next time point using three consecutive time points of brain neural data), that is, p = 2, the preset time length is 3, and the brain neural data within the time series includes brain neural data at time points 1 to 5, that is, the time series length is 5. Then the brain effective connection information at time points 4, 5 and 6 can be obtained according to the brain neural data at time points 1, 2 and 3, the brain neural data at time points 2, 3 and 4, and the brain neural data at time points 3, 4 and 5, respectively, and the results obtained depend on the brain state represented at different time points, and the results obtained at different time points are not completely the same. Finally, the brain effective connection information at time points 4, 5 and 6 can be used for further mean calculation, and the average result is used as the effective connection in the average sense, that is, the target effective connection information.

[0156] In summary, through the derivation method of brain effective connection information proposed by the above steps 101 to 104, the digital twin brain model can be stimulated by different virtual stimuli, and the response of the brain under stimulation can be predicted by stimulating the digital twin brain model with virtual perturbation, and the whole brain effective connection can be inferred. This can simultaneously consider a large number of brain regions, is suitable for inferring the whole brain effective connection, and can simultaneously depict the sign (effective connection type), weight (effective connection strength) and direction (effective connection direction) of the connection.

[0157] The embodiment of the application provides a derivation method of brain effective connection information, a derivation device of brain effective connection information determines disturbance data of a first brain node in at least two brain nodes and brain neural data of the at least two brain nodes in a time sequence; through a digital twin brain model, first predicted neural data of the at least two brain nodes at a next moment is obtained based on the disturbance data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence; wherein the digital twin brain model is obtained by training based on a time sequence prediction network; brain effective connection information of the first brain node to a second brain node is determined based on the first predicted neural data of the at least two brain nodes at the next moment and second predicted neural data of the at least two brain nodes at the next moment; wherein the second brain node is a brain node other than the first brain node in the at least two brain nodes; the second predicted neural data represents predicted data obtained when no disturbance data is added. That is, in the embodiment of the application, based on the digital twin brain model obtained by training, the response of the brain under the disturbance stimulation is predicted by adding the disturbance data to the brain node, and the brain effective connection information is inferred, wherein the digital twin brain model can effectively and accurately predict the dynamic change of the brain neural data, so that more accurate brain effective connection information can be obtained.

[0158] Another embodiment of the application provides a training method of a digital twin brain model, wherein the training method of the digital twin brain model can be applied to a training device of the digital twin brain model or a computer device, and the application does not make specific limitations. In the following, the training method of the digital twin brain model proposed in the embodiment of the application is exemplarily described by taking the training device of the digital twin brain model as an example.

[0159] Further, in the embodiment of the application, Figure 9 The implementation flowchart of the training of the digital twin brain model proposed in the embodiment of the application is shown as Figure 9 The training method of the digital twin brain model can include the following steps:

[0160] Step 901, training a time sequence prediction network based on a training data set to obtain a digital twin brain model; wherein the digital twin brain model is used for predicting neural data at a next moment based on brain neural data of brain nodes in a time sequence.

[0161] In the embodiments of the present application, the training device of the digital twin brain model can use the neural training data of the at least two brain nodes in the time sequence to train the time sequence prediction network, so that the digital twin brain model can be obtained. Among them, based on the digital twin brain model obtained by training the time sequence prediction network, the dynamic change of the neural data can be predicted, that is, the neural data of the next moment can be predicted based on the brain neural data of the brain nodes in the time sequence, so that the determination of the effective brain connection information between the brain nodes can be further performed based on the obtained predicted neural data, and more accurate effective brain connection information can be obtained.

[0162] It can be understood that in the embodiments of the present application, the brain node can represent a brain region, that is, it can be considered that one brain node corresponds to one brain region. The brain can be divided into at least two brain regions, and accordingly, one brain can correspond to at least two brain nodes.

[0163] Exemplarily, in some embodiments, Figure 10 Schematic diagram of brain nodes proposed for embodiments of the present application Figure Two As Figure 10 As shown, assuming that the biological brain is divided into multiple regions, such as region A, region B and region C, and accordingly, it can be considered that the brain at least corresponds to brain node a, brain node b and brain node c.

[0164] Further, in the embodiments of the present application, the neural training data of the at least two brain nodes in the time sequence can include the neural training data of each brain node corresponding to the time sequence, that is, the training device of the digital twin brain model needs to obtain the neural training data of multiple moments corresponding to each brain node.

[0165] In the embodiments of the present application, the time sequence can include continuous time sampling intervals, or can be discrete time sampling points, which are not limited in the present application.

[0166] Further, in the embodiments of the present application, the training data set can include brain neural data corresponding to a preset time length p+1 of the at least two brain nodes; wherein, when predicting the brain neural data of the next moment, the preset time length can determine the time length of the brain neural data used in the prediction process. Wherein, p is an integer greater than or equal to 0.

[0167] It can be understood that in the embodiments of the present application, the time length of the brain neural data used in the prediction process, that is, the preset time length p+1, can be pre-set. For example, if p=2 is pre-set, then the value of the preset time length p+1 is 3.

[0168] It can be understood that, in the embodiments of the present application, the training data set can include brain neural data corresponding to a plurality of time points of the at least two brain nodes, and the brain neural data corresponding to the plurality of time points includes but is not limited to brain neural data corresponding to a preset time length p+1 of the at least two brain nodes, that is, the time sequence length corresponding to the training data set is greater than or equal to the time length of the brain neural data used in the prediction process.

[0169] Exemplarily, in some embodiments, the time sequence length corresponding to the training data set is 5000, and the time length of the brain neural data used in the prediction process is 3.

[0170] It can be understood that, in the embodiments of the present application, for a determined preset time length, the time length of the data used in the prediction of brain neural data is fixed, for example, for a preset time length p+1, the brain neural data of the previous p+1 time points can be used to predict the brain neural data of the next time point.

[0171] That is, in the embodiments of the present application, the brain neural data of one or more time points can be used to predict the brain neural data of the next time point. For example, if p=2 is set in advance, then the brain neural data of time point 1, time point 2 and time point 3 can be used to predict the brain neural data of time point 4, the brain neural data of time point 2, time point 3 and time point 4 can be used to predict the brain neural data of time point 5, and the brain neural data of time point 3, time point 4 and time point 5 can be used to predict the brain neural data of time point 6.

[0172] Exemplarily, in some embodiments, assuming that the preset time length is p+1, the time length indicated by the preset time length includes p+1 time points from time point m-p to time point m. Accordingly, the neural training data corresponding to the preset time length p+1 includes neural training data from time point m-p to time point m, and m is an integer greater than p.

[0173] Further, in the embodiments of the present application, the training data set further includes neural training data of the at least two brain nodes at time point m+1.

[0174] Further, in the embodiments of the present application, various ways can be used to determine the neural training data of any brain node in the time sequence, which is not specifically limited in the present application.

[0175] Exemplarily, in some embodiments, the neural training data can be obtained through an existing bioinformatics database, wherein the bioinformatics database includes but is not limited to one or more of public data sets such as human fMRI data set, EEG data set, MEG data set, etc.

[0176] For example, in some embodiments, neural training data can also be generated using computational models.

[0177] Furthermore, in the embodiments of this application, the time series prediction network used to train the digital twin brain model can be any type and structure of artificial neural network for time series prediction, and this application does not impose any specific limitations.

[0178] For example, in some embodiments, the time series prediction network can use classic models such as multilayer perceptron, convolutional neural network, and recurrent neural network, or other network architectures suitable for time series prediction can be selected. This application does not make specific limitations.

[0179] Furthermore, in the embodiments of this application, when training the time series prediction network based on the training dataset to obtain a digital twin brain model, the time series prediction network can be used to obtain prediction data for at least two brain nodes at time m+1 based on the neural training data of at least two brain nodes from time mp to time m; based on the prediction data of at least two brain nodes at time m+1 and the neural training data of at least two brain nodes at time m+1, the time series prediction network is corrected to obtain the digital twin brain model.

[0180] For example, in some embodiments, when training a time series prediction network based on a training dataset, the predicted data for at least two brain nodes at time m+1 is determined using a digital twin brain model. The process can be referred to the following formula (8):

[0181]

[0182] The time series prediction network can be represented as f(·), x (m-p) x (m-p+1) , ..., x m It can represent neural training data from time point (mp) to time point m.

[0183] For example, in some embodiments, when training a time series prediction network based on a training dataset, the predicted data for at least two brain nodes at time m+1 is determined using a digital twin brain model. The process can be referred to the following formula (9):

[0184]

[0185] Where θ represents all parameters in the time series prediction network that can be optimized through training iterations.

[0186] Further, in the embodiments of the present application, when the time series prediction network is corrected based on the prediction data of the at least two brain nodes at the m+1 time and the neural training data of the at least two brain nodes at the m+1 time to obtain the digital twin brain model, the model parameters are determined based on the prediction data of the at least two brain nodes at the m+1 time, the neural training data of the at least two brain nodes at the m+1 time, and the preset time length corresponding to the training data set, and the digital twin brain model is determined based on the model parameters.

[0187] It can be understood that, in the embodiments of the present application, after the prediction of the neural data at the next time based on the input neural training data by the time series prediction network to obtain the prediction data at the next time, the training of the time series prediction network can be further combined with the neural training data at the next time to complete the training by loss function minimization to obtain the corresponding model parameters, and finally the corresponding digital twin brain model can be obtained by using the model parameters.

[0188] In the embodiments of the present application, the loss function can be defined as any type of objective function, for example, the loss function can be defined by minimizing the mean square error of time series prediction, and the present application does not make specific limitation.

[0189] Exemplarily, in some embodiments, it is assumed that the time series prediction network is trained by minimizing the mean square error of time series prediction, that is, the loss function is defined as the following formula (10):

[0190]

[0191] Wherein, T represents the number of prediction data of the at least two brain nodes corresponding to a plurality of times obtained based on the training data set. The loss function L(θ) is minimized by using a random gradient descent method, an adaptive learning rate algorithm and the like, and the network parameters of the time series prediction network suitable for describing the dynamics of large-scale brain networks are obtained, that is, the model parameters are obtained, and then the artificial neural network which can be used as the digital twin brain model can be obtained based on the model parameters, that is, the digital twin brain model is obtained.

[0192] It can be understood that, in the embodiments of the present application, the digital twin brain model is obtained based on the training of the time series prediction network, and therefore the digital twin brain model can effectively predict the dynamic changes of the brain neural signals (brain neural data), and further obtain more accurate brain effective connection information.

[0193] Further, in the embodiments of the present application, since the length of the time series corresponding to the training data set is greater than or equal to the time length of the brain neural data used in the prediction process, the prediction of the data at multiple time points by the time series prediction network can be selected, and the difference between the multiple prediction data obtained by the prediction and the multiple neural training data corresponding to the time points is calculated, and the multiple errors obtained by the calculation are comprehensively considered to determine the model parameters, so as to obtain a digital twin brain model with better prediction performance.

[0194] Further, in the embodiments of the present application, when the time series prediction network is trained based on the training data set to obtain the digital twin brain model, the time series prediction network can be used to obtain multiple prediction data corresponding to multiple time points of at least two brain nodes based on the training data set; based on the multiple prediction data and the neural training data corresponding to the multiple time points of the at least two brain nodes, multiple data errors corresponding to the multiple time points are determined; the model parameters are determined based on the multiple data errors; and the digital twin brain model is determined based on the model parameters.

[0195] That is, in the embodiments of the present application, for a large enough time series length, the neural data of the multiple selectable parameters t can be predicted based on the neural training data in the training data set, i.e., multiple prediction data are obtained respectively, and finally the difference between the multiple prediction data corresponding to the multiple time points and the multiple neural training data is calculated to determine the multiple data errors (multiple errors) corresponding to the multiple time points. The model parameters are determined by comprehensively considering these data errors.

[0196] Exemplarily, in some embodiments, it is assumed that the digital twin brain is 3-step prediction 1-step (using three consecutive time neural training data to predict the brain neural data at the next time), i.e., p=2, the preset time length is 3, and the neural training data in the training data set includes neural training data at time 1 to time 5, i.e., the time series length is 5. Then, the prediction data at time 4 and time 5 can be obtained according to the neural training data at time 1, time 2, and time 3, and the neural training data at time 2, time 3, and time 4, respectively. Then, the error between the prediction data at time 4 and the neural training data, and the error between the prediction data at time 5 and the neural training data are calculated respectively to obtain the data errors at time 4 and time 5. Finally, the model parameters are determined based on the two data errors.

[0197] Further, in the embodiments of the present application, in order to evaluate the effect of the digital twin brain model on the description of large-scale brain network dynamics, the performance of the digital twin brain model obtained by training can be further measured and evaluated.

[0198] Further, in the embodiments of the present application, when the performance of the digital twin brain model obtained by training is measured and evaluated, the determination coefficient corresponding to the digital twin brain model can be determined based on the training data set and the prediction data set corresponding to the training data set; the first functional connection matrix predicted by the digital twin brain model is determined based on the brain neural data and the random noise, and the matrix correlation coefficient between the first functional connection matrix and the second functional connection matrix corresponding to the training data set is determined; the model performance parameter of the digital twin brain model is determined based on the determination coefficient and the matrix correlation coefficient.

[0199] It can be understood that, in the embodiments of the present application, the training data set can include neural training data of at least two brain nodes corresponding to T time points, and the prediction data set can include prediction data of at least two brain nodes corresponding to T time points, that is, the prediction data set can include prediction data of T time points corresponding to neural training data of T time points.

[0200] Further, when the determination coefficient corresponding to the digital twin brain model is determined based on the training data set and the prediction data set corresponding to the training data set, the data mean of at least two brain nodes corresponding to T time points is determined based on the neural training data of at least two brain nodes corresponding to T time points; the determination coefficient corresponding to the digital twin brain model is determined based on the data mean, the neural training data corresponding to T time points and the prediction data corresponding to T time points.

[0201] For example, in some embodiments, for the determination coefficient R 2 , introduced for evaluating the digital twin brain model obtained by training, the following formula (11) can be used for calculation:

[0202]

[0203] Wherein, N represents the number of nodes (the number of brain nodes) of the large-scale brain network, x i，t represents the value of node i in the brain neural data at time point t, represents the value of node i predicted by the digital twin brain model at time point t, represents the mean value of node i in the brain neural data at all time points, and T represents the length of the brain neural data used for training the time series prediction network.

[0204] It can be understood that, in the embodiments of the present application, the determination coefficient R 2 obtained by calculation can be used to evaluate the performance of the digital twin brain model, wherein the better the prediction effect of the digital twin brain model on the brain neural data, the better the digital twin brain model can depict the state transition relationship of the brain network.

[0205] Further, in the embodiments of the present application, the first functional connectivity matrix predicted by the digital twin brain model can be determined based on the brain neural data and the random noise, and meanwhile, the second functional connectivity matrix can be obtained based on the training data set, i.e., the second functional connectivity matrix corresponding to the training data set is determined, and then the matrix correlation coefficient can be further determined according to the first functional connectivity matrix and the second functional connectivity matrix. The matrix correlation coefficient can be used to determine the correlation between the first functional connectivity matrix and the second functional connectivity matrix.

[0206] It can be understood that, in the embodiments of the present application, the functional connectivity matrix is the correlation matrix between different brain node activities. The first functional connectivity matrix can be understood as the correlation matrix between different brain node activities predicted by the digital twin brain model obtained by training, and the second functional connectivity matrix can be understood as the correlation matrix between different brain node activities calculated based on the training data set.

[0207] Further, in the embodiments of the present application, when the first functional connectivity matrix is determined based on the brain neural data and the random noise, the digital twin brain model can be used to generate data driven by the random noise, so as to generate the corresponding first functional connectivity matrix.

[0208] For example, in some embodiments, the all-zero data can be used as the initial value (i.e., x 0..p = 0), and the next simulation data can be generated through the formula x t+1 = f(x (t-p)..t + δ (t-p)..t , θ) until the length of the simulation data is consistent with the length of the brain neural data used for training the model. The δ t is the random noise obeying the normal distribution with 0 as the mean value.

[0209] It can be understood that, in the embodiments of the present application, the functional connectivity matrix is reconstructed by the digital twin brain model, and then the matrix correlation coefficient between the first functional connectivity matrix reconstructed by the digital twin brain model and the second functional connectivity matrix corresponding to the training data set is determined, which can determine the reconstruction effect of the functional connectivity matrix. The better the reconstruction effect of the functional connectivity matrix is, the better the digital twin brain model can depict the functional dependence relationship between the brain nodes.

[0210] That is, in the embodiments of the present application, after the digital twin brain model is obtained through training, in order to evaluate the effect of the digital twin brain model in describing the dynamics of large-scale brain networks, the coefficient of determination and the matrix correlation coefficient can be introduced to measure and evaluate the performance of the digital twin brain model. For example, the coefficient of determination using brain neural data prediction, the correlation coefficient (matrix correlation coefficient) between the functional connectivity matrix of the data generated by the digital twin brain model under the driving of random noise and the functional connectivity matrix of the brain neural data (training data set) used to train the model to measure the effect of the digital twin brain model.

[0211] Further, in the embodiments of the present application, after the digital twin brain model is completed, the digital twin brain model can be used to further determine the effective connection information between different brain nodes according to the derivation method of brain effective connection information proposed in the above embodiments in the case of increasing virtual disturbance, and then the whole brain effective connection of at least two brain nodes can be obtained.

[0212] In summary, through the training method of the digital twin brain model proposed in the above step 901, the dynamic changes of brain neural signals are learned and predicted using an artificial neural network (such as a time series prediction network), the artificial neural network is used as a digital twin brain model, the digital twin brain model is trained by reconstructing or predicting brain neural data, and the prediction error and the functional connection of the generated data can be used as an index to evaluate the digital twin brain model. In this way, it can be applied flexibly to different modal neural signals without relying on specific model assumptions and without the need for a large number of adjustments.

[0213] The embodiments of the present application provide a training method of a digital twin brain model. A training device of the digital twin brain model trains a time series prediction network based on a training data set to obtain a digital twin brain model. The digital twin brain model is used to predict neural data at the next time based on brain neural data of brain nodes in a time series. The training data set includes neural training data of at least two brain nodes in a time series. That is, in the embodiments of the present application, a time series prediction network can be used to learn and predict the dynamic changes of brain neural signals to obtain a digital twin brain model with time series prediction function. The digital twin brain model can effectively and accurately predict the dynamic changes of brain neural data.

[0214] Based on the above embodiments, another embodiment of the present application proposes a method for deriving brain effective connectivity information, and a method for training a digital twin brain model. On the one hand, the method for training a digital twin brain model can use an artificial neural network to fit large-scale neural dynamics, construct a digital twin brain model, and depict the functional dependence relationship between different brain regions. On the other hand, the method for deriving brain effective connectivity information can use the digital twin brain model obtained by training to apply a virtual stimulus to the digital twin brain model, observe the response of other regions to the stimulus to depict the effective connectivity between brain nodes, and predict the response mode of the brain in different states, so as to depict the symbol (effective connection type), weight (effective connection strength) and direction (effective connection direction) of the effective connection in the whole brain.

[0215] Figure 11 For the analysis of neural data under real stimulation, as shown in FIG. 1, in general, the effective connectivity can be evaluated by a neural stimulation experiment, which disturbs a specific brain region by a real stimulus and monitors the neural response of other regions, and records the synchronous neural data, thereby directly proving the causal relationship. However, such a method limits the scalability of whole-brain analysis due to its invasiveness. Figure 11

[0216] Further, in the embodiments of the present application, the digital twin brain model can be constructed using an artificial neural network, wherein the artificial neural network can be a time series prediction network.

[0217] Exemplarily, in some embodiments, the large-scale brain network dynamics is represented as formula (12) using an artificial neural network f(·):

[0218]

[0219] Wherein, the artificial neural network f(·) can use a multi-layer perceptron, a convolutional neural network, a recurrent neural network, etc. classical model, and other network architectures suitable for time series prediction can also be selected. θ represents all parameters in the artificial neural network that can be optimized by training iteration, x (t-p)..t represents the brain neural data from time point (t-p) to time point t, represents the brain neural data at time point (t+1) predicted by the artificial neural network, and the value of the hyperparameter p is determined by the nature of the brain neural data.

[0220] ​Exemplarily, in some embodiments, the artificial neural network is trained by minimizing the mean square error of the time series prediction, i.e., defining the loss function as formula (10). Wherein, T represents the length of the brain neural data used to train the artificial neural network. The artificial neural network parameters suitable for describing the dynamics of large-scale brain networks can be obtained by minimizing the loss function L(θ) using optimization methods such as stochastic gradient descent, adaptive learning rate algorithm, etc., and the artificial neural network can be obtained as a digital twin brain model.

[0221] Further, in the embodiments of the present application, after obtaining the digital twin brain model, the digital twin brain model can be evaluated by the prediction error and the functional connectivity of the generated data.

[0222] It can be understood that, in the embodiments of the present application, to evaluate the effect of the digital twin brain model in describing the dynamics of large-scale brain networks, the determination coefficient of the brain neural data prediction, the correlation coefficient between the functional connectivity matrix of the data generated by the digital twin brain model under the driving of random noise and the functional connectivity matrix of the brain neural data used to train the model (training data set) can be used to measure the effect of the surrogate model.

[0223] Exemplarily, in some embodiments, the determination coefficient R 2 can be calculated by formula (12), wherein N represents the number of nodes of the large-scale brain network, x i,t represents the value of node i at time point t in the brain neural data, represents the value of node i at time point t predicted by the digital twin brain model, represents the mean value of node i at all time points in the brain neural data. The better the prediction effect of the digital twin brain model on the brain neural data, the better the surrogate model can describe the state transition relationship of the brain network.

[0224] Exemplarily, in some embodiments, the functional connectivity matrix is a correlation matrix between different node activities. When generating data using the digital twin brain model under the driving of random noise, the all-zero data can be used as the initial value (i.e., x 0..p = 0), and the next step of simulation data is generated by formula x t+1 = f(x (t-p)..t + δ (t-p)..t , θ) until the length of the simulation data is consistent with the length of the brain neural data used to train the model. Wherein, δ t is a random noise obeying normal distribution with mean value of 0. The better the reconstruction effect of the digital twin brain model on the functional connectivity matrix, the better the surrogate model can describe the functional dependence relationship between nodes.

[0225] Further, in embodiments of the present application, after obtaining the digital twin brain model with time series prediction function through training, the effective connections of the whole brain can be inferred by virtually perturbing the digital twin brain model.

[0226] It can be understood that after obtaining and verifying the effectiveness of the digital twin brain model, the effective connections of the brain network can be obtained by applying virtual perturbation to the model. When characterizing the effective connections of a given node to other nodes, the virtual perturbation can be represented as a slight lifting of the signal of the given node in the last time point of the brain neural data after perturbation, and the effective connections of the node to other nodes can be characterized by observing the influence of the virtual perturbation on the output of the surrogate model.

[0227] Exemplarily, in some embodiments, the effective connection EC ij can be obtained by formula (7), wherein E t (·) represents the mean value over time t, f(·) j represents the component j of the output of the surrogate model, and the virtual perturbation P is a p i = Δ, p k，k≠i = 0 or an N-dimensional virtual perturbation vector required by other conventional experimental paradigms, wherein Δ is the virtual perturbation strength applied to node i.

[0228] Further, in embodiments of the present application, the effective connections of the whole brain can be obtained by traversing all nodes i and j. The effective connections obtained by the derivation method of brain effective connection information proposed in the present application can be brain effective connection information with sign (effective connection type), weight (effective connection strength) and direction (effective connection direction), and can be used to guide precise neural regulation.

[0229] Exemplarily, in some embodiments, Figure 12 The implementation schematic diagram of deriving brain effective connection information based on the digital twin brain model obtained through training in embodiments of the present application is as shown in FIG. 6. Figure 12As shown, in the training process of the digital twin brain model, the number of nodes of the model can correspond to the number of brain nodes, for example, corresponding to four brain nodes of brain node a, brain node b, brain node c and brain node d, there can be four model nodes; the training data set used for model training can include neural training data of brain nodes in a time sequence, for example, resting state functional magnetic resonance imaging data from t time to t+Δt time; taking the neural training data x(t) at t time as input, the prediction data x(t+Δt) at t+Δt time is obtained, combined with the neural training data at t+Δt time, the model can be iterated and corrected, and finally the digital twin brain (digital twin brain model) is obtained. After the digital twin brain model is trained, through the digital twin brain model, combined with the virtual disturbance (disturbance data) applied to different brain nodes, the predicted neural data obtained when the disturbance data is increased and not increased is further compared and analyzed, so as to obtain the brain effective connection information between any two brain nodes, and finally obtain the accurate whole brain effective connection, which includes the effective connection type, the effective connection direction and the effective connection strength.

[0230] Further, the derivation method of brain effective connection information and the training method of digital twin brain model proposed in the embodiments of the present application are simulated and verified by three simulation models, which proves that the method is feasible, and the inference result of the present application on the effective connection is better than that of the traditional data-driven method (Granger causality, dynamic causal model). The derivation method of brain effective connection information proposed in the embodiments of the present application has been successfully applied to four brain atlases, and the whole brain effective connection under different brain atlases is obtained based on resting state functional magnetic resonance imaging data, and the obtained connection has significant difference between autism patients and normal people.

[0231] In summary, in the embodiments of the present application, in the first aspect, an artificial neural network (such as a time series prediction network) is used to learn and predict the dynamic changes of brain neural signals, the artificial neural network is used as a digital twin brain model, and the digital twin brain model is trained by reconstructing or predicting brain neural data, and the prediction error and the functional connection of the generated data can be used as an index to evaluate the digital twin brain model, which can avoid relying on specific model assumptions and can be flexibly applied to different modal neural signals without a large amount of adjustment. In the second aspect, the digital twin brain model can be stimulated by different virtual stimuli, and the whole brain effective connection can be inferred by stimulating the digital twin brain model by virtual disturbance and predicting the response of the brain under stimulation, which can simultaneously consider a large number of brain regions, is suitable for inferring the whole brain effective connection, and can simultaneously depict the symbol, weight and direction of the connection.

[0232] The embodiment of the present application provides a derivation method of brain effective connection information and a training method of a digital twin brain model. The digital twin brain model is obtained by training a time series prediction network. Therefore, the digital twin brain model can accurately predict the dynamic changes of brain neural data, thereby obtaining more accurate brain effective connection information.

[0233] Based on the above embodiment, in another embodiment of the present application, Figure 13 The composition structure schematic diagram of the derivation device of brain effective connection information provided in the embodiment of the present application is shown in FIG. 1. Figure 13 As shown in FIG. 1, the derivation device 130 of brain effective connection information provided in the embodiment of the present application can include a determination unit 1301, an acquisition unit 1302,

[0234] The determination unit 1301 is configured to determine the perturbation data of a first brain node in at least two brain nodes and brain neural data of the at least two brain nodes in a time sequence;

[0235] The acquisition unit 1302 is configured to obtain first predicted neural data of the at least two brain nodes at a next time by a digital twin brain model based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence; wherein the digital twin brain model is obtained by training a time series prediction network;

[0236] The determination unit 1301 is further configured to determine brain effective connection information between the first brain node and a second brain node based on the first predicted neural data of the at least two brain nodes at the next time and second predicted neural data of the at least two brain nodes at the next time; wherein the second brain node is a brain node other than the first brain node in the at least two brain nodes; and the second predicted neural data represents predicted data obtained without adding the perturbation data.

[0237] Further, in the embodiment of the present application, Figure 14 The composition structure schematic diagram of the training device of the digital twin brain model provided in the embodiment of the present application is shown in FIG. 2. Figure 14 As shown in FIG. 2, the training device 140 of the digital twin brain model provided in the embodiment of the present application can include a training unit 1401 and an evaluation unit 1402,

[0238] The training unit 1401 is configured to train a time series prediction network based on a training data set to obtain a digital twin brain model; wherein the digital twin brain model is used for predicting neural data at a next time based on brain neural data of brain nodes in a time sequence;

[0239] The training data set includes neural training data corresponding to a preset time length p+1 of the at least two brain nodes; the neural training data corresponding to the preset time length p+1 includes neural training data from the m-p moment to the m moment, and the training data set further includes neural training data of the at least two brain nodes at the m+1 moment; p is an integer greater than or equal to 0, and m is an integer greater than p;

[0240] The training data set is used to train the time series prediction network to obtain the digital twin brain model, including:

[0241] The time series prediction network is used to obtain prediction data of the at least two brain nodes at the m+1 moment based on the neural training data of the at least two brain nodes from the m-p moment to the m moment;

[0242] The time series prediction network is corrected based on the prediction data of the at least two brain nodes at the m+1 moment and the neural training data of the at least two brain nodes at the m+1 moment, to obtain the digital twin brain model.

[0243] Further, in the embodiment of the present application, the evaluation unit 1402 is configured to determine a determination coefficient corresponding to the digital twin brain model based on the training data set and a prediction data set corresponding to the training data set; determine a first functional connection matrix predicted by the digital twin brain model based on the brain neural data and random noise, and determine a matrix correlation coefficient between the first functional connection matrix and a second functional connection matrix corresponding to the training data set; and determine a model performance parameter of the digital twin brain model based on the determination coefficient and the matrix correlation coefficient.

[0244] In the embodiment of the present application, further, Figure 15 The structural diagram of the computer device according to the embodiment of the present application is shown in FIG. 15. Figure 15 As shown in FIG. 15, the computer device 1500 according to the embodiment of the present application includes a processor 1501 and a memory 1502 storing instructions executable by the processor 1501. Further, the computer device 1500 can further include a communication interface 1503, and a bus 1504 for connecting the processor 1501, the memory 1502 and the communication interface 1503.

[0245] In the embodiments of the present application, the processor 1501 can be at least one of an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a Central Processing Unit (CPU), a controller, a microcontroller, and a microprocessor. It can be understood that the electronic device for implementing the functions of the processor described above can also be other devices for different devices, and the embodiments of the present application are not limited specifically. The computer device 1500 can further include a memory 1502 connected to the processor 1501, wherein the memory 1502 is used to store executable program codes, the program codes including computer operation instructions, and the memory 1502 can include a high-speed RAM memory and can also include a non-volatile memory, for example, at least two disk memories.

[0246] In the embodiments of the present application, the bus 1504 is used to connect the communication interface 1503, the processor 1501, and the memory 1502, and the mutual communication between these devices.

[0247] In the embodiments of the present application, the memory 1502 is used to store instructions and data.

[0248] Further, in the embodiments of the present application, the processor 1501 is configured to: determine perturbation data of a first brain node of at least two brain nodes and brain neural data of the at least two brain nodes in a time sequence; obtain first predicted neural data of the at least two brain nodes at a next time through a digital twin brain model based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence, wherein the digital twin brain model is obtained by training based on a time sequence prediction network; determine brain effective connection information from the first brain node to a second brain node based on the first predicted neural data of the at least two brain nodes at the next time and second predicted neural data of the at least two brain nodes at the next time, wherein the second brain node is a brain node other than the first brain node among the at least two brain nodes; and the second predicted neural data represents predicted data obtained without adding the perturbation data.

[0249] Further, in the embodiments of the present application, the processor 1501 is further configured to train the time series prediction network based on a training data set to obtain a digital twin brain model; wherein the digital twin brain model is used to predict the neural data of the next time based on the neural data of the brain nodes within the time series.

[0250] The training data set includes neural training data corresponding to a preset time length p+1 of at least two brain nodes; wherein the neural training data corresponding to the preset time length p+1 includes neural training data from the m-p time to the m time, and the training data set further includes neural training data of the at least two brain nodes at the m+1 time; p is an integer greater than or equal to 0, and m is an integer greater than p;

[0251] The training of the time series prediction network based on the training data set to obtain the digital twin brain model includes:

[0252] Through the time series prediction network, the prediction data of the at least two brain nodes at the m+1 time is obtained based on the neural training data of the at least two brain nodes from the m-p time to the m time;

[0253] Based on the prediction data of the at least two brain nodes at the m+1 time and the neural training data of the at least two brain nodes at the m+1 time, the time series prediction network is corrected to obtain the digital twin brain model.

[0254] In actual application, the memory 1502 can be a volatile memory such as a random access memory (RAM), or a non-volatile memory such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), or a combination of the above types of memories, and provides instructions and data to the processor 1501.

[0255] In addition, each functional module in the embodiments can be integrated in one processing unit, or each unit can exist physically independently, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional module.

[0256] If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer readable storage medium based on such understanding. The technical solutions of the embodiments essentially or the parts that contribute to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the embodiment method. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0257] The embodiment of the application provides a brain effective connection information derivation device, a digital twin brain model training device and a computer device. The digital twin brain model is obtained based on a time series prediction network, and therefore, the digital twin brain model can accurately predict the dynamic change of brain neural data, so that more accurate brain effective connection information can be obtained.

[0258] Specifically, the program instructions corresponding to the brain effective connection information derivation method in the embodiment can be stored on a storage medium such as an optical disc, a hard disk, a U disk, etc. When the program instructions corresponding to the brain effective connection information derivation method in the storage medium are read by an electronic device or executed, the following steps are included:

[0259] determining perturbation data of a first brain node in at least two brain nodes and brain neural data of the at least two brain nodes in a time sequence;

[0260] obtaining first predicted neural data of the at least two brain nodes at a next time based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes in the time sequence through a digital twin brain model, wherein the digital twin brain model is obtained based on a time series prediction network;

[0261] determining brain effective connection information from the first brain node to a second brain node based on the first predicted neural data of the at least two brain nodes at the next time and second predicted neural data of the at least two brain nodes at the next time, wherein the second brain node is a brain node other than the first brain node in the at least two brain nodes, and the second predicted neural data represents predicted data obtained without adding the perturbation data.

[0262] Specifically, the program instructions corresponding to the training method of the digital twin brain model in the embodiment can be stored on a storage medium such as an optical disc, a hard disk, a U disk, etc. When the program instructions corresponding to the training method of the digital twin brain model in the embodiment are read by an electronic device or executed, the following steps are included:

[0263] training the time series prediction network based on the training data set to obtain a digital twin brain model; wherein the digital twin brain model is used to predict the neural data of the next time based on the neural data of the brain nodes within the time series;

[0264] The training data set includes neural training data corresponding to a preset time length p+1 of at least two brain nodes; wherein the neural training data corresponding to the preset time length p+1 includes neural training data from the m-p time to the m time, and the training data set further includes neural training data of the at least two brain nodes at the m+1 time; p is an integer greater than or equal to 0, and m is an integer greater than p;

[0265] The training of the time series prediction network based on the training data set to obtain a digital twin brain model includes:

[0266] Through the time series prediction network, based on the neural training data of the at least two brain nodes from the m-p time to the m time, the prediction data of the at least two brain nodes at the m+1 time is obtained;

[0267] Based on the prediction data of the at least two brain nodes at the m+1 time and the neural training data of the at least two brain nodes at the m+1 time, the time series prediction network is corrected to obtain the digital twin brain model.

[0268] The embodiments of the present application also provide a computer program product.

[0269] In some embodiments, the computer program product can include a computer program or instructions.

[0270] In some embodiments, the computer program product can be applied to the computer device in the embodiments of the present application, and the computer program instructions enable the computer to execute the corresponding processes realized by the computer device in the various methods of the embodiments of the present application. For brevity, details are not repeated here.

[0271] Those skilled in the art will appreciate that embodiments of the application can be further implemented in a computer program product tangibly embodied in a machine-readable storage medium (e.g., memory storage) including instructions that, when executed by a machine (e.g., a processor), cause the machine to perform the steps of embodiments of the application. The terms "machine-readable storage medium" or "computer-readable storage medium" include, but are not limited to, portable or fixed storage devices, optical storage devices, magnetic storage devices, wireline, optical, or other communication links, commonly known as computer communication networks, including the Internet, intranets, local area networks (LANs), wide area networks (WANs), etc. The terms "machine-readable storage medium" or "computer-readable storage medium" also include any medium that is capable of storing or encoding computer readable instructions for execution by a machine (e.g., a processor) and that cause the machine to perform any one or more of the steps that define the procedures described in the detailed description section of the instant disclosure. The terms "machine-readable storage medium" or "computer-readable storage medium" therefore include, but are not limited to, memories (e.g., optical, read-only memories (ROM); flash memories; etc.), floppy disks; compact disks (CD); optical disks; and machine- readable storage media that are external to the computer or other machine or devices.

[0272] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks. Figure One one or more functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks. Figure One one or more functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks.

[0273] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks. Figure One one or more functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks. Figure One one or more functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks.

[0274] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions that execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks. Figure One one or more functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks. Figure One one or more functions specified in the flowchart block or blocks and / or flowchart diagrams block or blocks.

[0275] The foregoing is merely illustrative of the principles of this application and various modifications can be made by those skilled in the art without departing from the scope of the application.

Claims

1. A method for deriving information about the brain's effective connections, characterized in that, The method includes: Determine the perturbation data of a first brain node out of at least two brain nodes and the brain neural data of the at least two brain nodes in a time series; wherein the perturbation data of the first brain node includes an element with a non-zero value in a virtual perturbation vector; Using a digital twin brain model, based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes in the time series, the first predicted neural data of the at least two brain nodes at the next time step is obtained; wherein, the digital twin brain model is obtained by training based on a time series prediction network; Based on the first predicted neural data of the at least two brain nodes at the next time step and the second predicted neural data of the at least two brain nodes at the next time step, the effective brain connection information from the first brain node to the second brain node is determined; wherein, the second brain node is the other brain node among the at least two brain nodes besides the first brain node; the second predicted neural data represents the predicted data obtained without adding the perturbation data.

2. The method according to claim 1, characterized in that, The method further includes: Using the digital twin brain model, based on the brain neural data of the at least two brain nodes in the time series, the second predicted neural data of the at least two brain nodes at the next time step is obtained.

3. The method according to claim 2, characterized in that, The effective brain connectivity information includes effective connectivity types. Determining the effective brain connectivity information from the first brain node to the second brain node based on the first predicted neural data and the second predicted neural data of the at least two brain nodes at the next time step includes: If the first predicted neural data of the second brain node at the next time step is greater than the second predicted neural data of the second brain node at the next time step, the effective connection type is determined to be excitatory. If the first predicted neural data of the second brain node at the next time step is less than the second predicted neural data of the second brain node at the next time step, the effective connection type is determined to be inhibitory.

4. The method according to claim 2, characterized in that, The brain neural data of the at least two brain nodes in the time series includes brain neural data of the at least two brain nodes corresponding to a preset time length p+1; wherein, the brain neural data corresponding to the preset time length p+1 includes brain neural data from time tp to time t, where p is an integer greater than or equal to 0, and t is an integer greater than p. The method involves obtaining first predicted neural data for the at least two brain nodes at the next time step using a digital twin brain model, based on perturbation data of the first brain node and neural data of the at least two brain nodes within a time series. This includes: Based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes at time t, determine the perturbation-resolved brain neural data of the at least two brain nodes at time t. Using the digital twin brain model, based on the perturbed brain neural data of the at least two brain nodes at time t, and the brain neural data of the at least two brain nodes from time tp to time t-1, the first predicted neural data of the at least two brain nodes at time t+1 is obtained.

5. The method according to claim 4, characterized in that, The determination of the perturbation-adjusted brain neural data of the at least two brain nodes at time t, based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes at time t, includes: The perturbation data of the first brain node and the brain neural data of the at least two brain nodes at time t are superimposed to obtain the perturbation-resolved brain neural data of the at least two brain nodes at time t.

6. The method according to claim 4, characterized in that, The method of obtaining second predicted neural data for the at least two brain nodes at the next time step using a digital twin brain model, based on the brain neural data of the at least two brain nodes in the time series, includes: Using the digital twin brain model, based on the brain neural data of the at least two brain nodes from time tp to time t, the second predicted neural data of the at least two brain nodes at time t+1 is obtained.

7. The method according to claim 6, characterized in that, The effective brain connectivity information includes the effective connectivity direction, and determining the effective brain connectivity information from the first brain node to the second brain node includes: The effective connection direction is defined as from the first brain node to the second brain node.

8. The method according to claim 7, characterized in that, The effective brain connectivity information includes effective connectivity types. Determining the effective brain connectivity information from the first brain node to the second brain node based on the first predicted neural data and the second predicted neural data of the at least two brain nodes at the next time step includes: If the first predicted neural data of the second brain node at time t+1 is greater than the second predicted neural data of the second brain node at time t+1, the effective connection type is determined to be excitatory. If the first predicted neural data of the second brain node at time t+1 is less than the second predicted neural data of the second brain node at time t+1, the effective connection type is determined to be inhibitory.

9. The method according to claim 7 or 8, characterized in that, The effective brain connectivity information includes effective connectivity strength. Determining the effective brain connectivity information from the first brain node to the second brain node based on the first predicted neural data and the second predicted neural data of the at least two brain nodes at the next time step includes: Determine the intensity information of the first predicted neural data of the second brain node at time t+1, and determine the intensity information of the second predicted neural data of the second brain node at time t+1; The effective connection strength is determined based on the strength information of the first predicted neural data and the strength information of the second predicted neural data.

10. The method according to any one of claims 1-8, characterized in that, The determination of the perturbation data of the first brain node out of at least two brain nodes includes: The perturbation data of the first brain node is set based on a preset perturbation intensity; The method further includes: Traverse the at least two brain nodes and determine the effective brain connection information between any two brain nodes to obtain the whole-brain effective connection of the at least two brain nodes.

11. The method according to any one of claims 1-8, characterized in that, The method further includes: Using the digital twin brain model, based on multiple perturbation data of the first brain node corresponding to multiple times and brain neural data of the at least two brain nodes in the time series, multiple first predicted neural data of the at least two brain nodes corresponding to the multiple times are obtained respectively. Using the digital twin brain model, based on the brain neural data of the at least two brain nodes in the time series, multiple second predicted neural data corresponding to the at least two brain nodes at the multiple times are obtained respectively; Based on the plurality of first predictive neural data and the plurality of second predictive neural data, determine the plurality of valid brain connections from the first brain node to the second brain node, corresponding to the plurality of time points; Based on the multiple valid brain connectivity information, the target valid connectivity information from the first brain node to the second brain node is determined.

12. A training method for a digital twin brain model, characterized in that, The method includes: A time-series prediction network is trained based on a training dataset to obtain a digital twin brain model; wherein, the digital twin brain model is used to predict the neural data of the next time step based on the brain neural data of the brain nodes in the time series; the digital twin brain model is used to perform the method for deriving effective brain connectivity information as described in any one of claims 1-11; The training dataset includes neural training data for at least two brain nodes corresponding to a preset time length p+1; wherein, the neural training data corresponding to the preset time length p+1 includes neural training data from time mp to time m, and the training dataset also includes neural training data for the at least two brain nodes at time m+1; p is an integer greater than or equal to 0, and m is an integer greater than p. The process of training a time-series prediction network based on a training dataset to obtain a digital twin brain model includes: Using the time series prediction network, based on the neural training data of the at least two brain nodes from the mp-th time to the m-th time, the prediction data of the at least two brain nodes at the (m+1)-th time is obtained; Based on the prediction data of the at least two brain nodes at time m+1 and the neural training data of the at least two brain nodes at time m+1, the time series prediction network is modified to obtain the digital twin brain model.

13. The method according to claim 12, characterized in that, The process of refining the time-series prediction network based on the prediction data of the at least two brain nodes at time m+1 and the neural training data of the at least two brain nodes at time m+1 to obtain the digital twin brain model includes: The model parameters are determined based on the prediction data of the at least two brain nodes at the (m+1)th time, the neural training data of the at least two brain nodes at the (m+1)th time, and the preset time length corresponding to the training dataset. The digital twin brain model is determined based on the model parameters.

14. The method according to claim 12, characterized in that, The process of training a time-series prediction network based on a training dataset to obtain a digital twin brain model includes: Using the time series prediction network, based on the training dataset, multiple predicted data corresponding to multiple time points for the at least two brain nodes are obtained; Based on the multiple prediction data and the neural training data of the at least two brain nodes corresponding to the multiple time points, multiple data errors corresponding to the multiple time points are determined; The model parameters are determined based on the aforementioned multiple data errors; The digital twin brain model is determined based on the model parameters.

15. The method according to claim 13, characterized in that, The method further includes: Based on the training dataset and the prediction dataset corresponding to the training dataset, the determination coefficient corresponding to the digital twin brain model is determined; Using the digital twin brain model, based on the brain neural data and random noise, the first functional connectivity matrix predicted by the digital twin brain model is determined, and the matrix correlation coefficient between the first functional connectivity matrix and the second functional connectivity matrix corresponding to the training dataset is determined. Based on the determination coefficient and the matrix correlation coefficient, the model performance parameters of the digital twin brain model are determined.

16. The method according to claim 15, characterized in that, The training dataset includes neural training data for at least two brain nodes corresponding to T time points, and the prediction dataset includes prediction data for the at least two brain nodes corresponding to T time points, where T is an integer greater than 0. Determining the determination coefficients corresponding to the digital twin brain model based on the training dataset and the prediction dataset corresponding to the training dataset includes: Based on the neural training data of the at least two brain nodes corresponding to T time points, determine the average data of the at least two brain nodes corresponding to T time points; Based on the mean of the data, the neural training data corresponding to T time points, and the prediction data corresponding to T time points, the determination coefficient corresponding to the digital twin brain model is determined.

17. A device for deducing information that effectively connects the brain, characterized in that, The device for deriving effective brain connectivity information includes: A determining unit is configured to determine perturbation data of a first brain node among at least two brain nodes and brain neural data of the at least two brain nodes in a time series; wherein the perturbation data of the first brain node includes an element with a non-zero value in a virtual perturbation vector; The acquisition unit is used to obtain the first predicted neural data of the at least two brain nodes at the next moment based on the perturbation data of the first brain node and the brain neural data of the at least two brain nodes in the time series using a digital twin brain model; wherein, the digital twin brain model is obtained by training based on a time series prediction network; The determining unit is configured to determine effective brain connectivity information between the first brain node and the second brain node based on the first predicted neural data of the at least two brain nodes at the next time step and the second predicted neural data of the at least two brain nodes at the next time step; wherein, the second brain node is the other brain node among the at least two brain nodes besides the first brain node; the second predicted neural data represents the predicted data obtained without adding the perturbation data.

18. A computer device, characterized in that, The computer device includes: a processor and a memory; wherein, The memory is used to store computer programs that can run on the processor; The processor is configured to, when running the computer program, perform the method as described in any one of claims 1-11 or 12-16.

Images