A whole-brain dynamics modeling method for epilepsy based on key propagation zones

Abstract

The present invention belongs to the technical field of brain dynamics modeling, and specifically relates to a whole-brain dynamics modeling method for epilepsy based on key propagation areas, comprising the following steps: S1, data preprocessing and network construction; S2, precise positioning of key propagation areas for epilepsy; S3, construction of a high-precision whole-brain dynamics model for epilepsy. The present invention first constructs an epileptogenic network communication model, calculates the participation of each brain region in the abnormal discharge propagation process of epilepsy, and accurately locates the key position and optimal propagation point of epileptic seizures; secondly, integrates the key propagation areas for epilepsy, and constructs and optimizes a high-precision whole-brain network dynamics model for epilepsy. Finally, the present invention is applied to real epilepsy fMRI data to evaluate its effectiveness. The whole-brain dynamics modeling method for epilepsy based on key propagation areas provided by the present invention not only has important medical value for the study of the neural mechanism of focal epilepsy, but also promotes the development of high-precision whole-brain computational model construction and parameter optimization technology.

Description

Technical Field

[0001] The present invention belongs to the technical field of brain dynamics modeling, and in particular relates to a whole-brain dynamics modeling method for epilepsy based on key propagation zones. Background Art

[0002] Focal epileptic seizures typically originate in a limited region of the cerebral cortex and then propagate to larger areas of the cerebral cortex and subcortical structures. Long-term, frequent seizures can lead to severe cognitive impairment in patients. For patients with focal epilepsy whose seizures are uncontrollable with medication, surgical resection of the epileptogenic zone is a clinically effective cure for epilepsy, but its success relies on precise preoperative assessment to avoid irreversible neurological deficits. Furthermore, the diverse seizure types of epilepsy patients exacerbate the complexity of epilepsy surgery. High-precision whole-brain network dynamics models not only provide a theoretical basis for studying the dynamic mechanisms of epileptic seizures and propagation but also aid in clinical applications such as preoperative planning for epilepsy surgery, precise localization of lesions, and improved surgical efficacy. This model not only provides new insights into the network dynamics mechanisms in healthy individuals or patients with brain diseases, but also promotes the rapid development of research directions such as the integration of multimodal neurophysiological data and non-invasive treatment methods.

[0003] Research based on epilepsy dynamics models has deepened our understanding of epilepsy at various levels. At the brain circuit and system level, Cui, Guo, and others used a local neural circuit model of the hippocampal dentate gyrus to elucidate the mechanism of the characteristic discharges during temporal lobe epilepsy. In addition, this type of model has been used to simulate epileptic seizures under different conditions and test treatment hypotheses, such as identifying potential therapeutic targets. At the whole-brain level, Meijer et al. used the Wilson-Cowan model to study the internal mechanisms of focal epilepsy pathology. Chakraborty et al. established patients with focal lesions with fusion lesions based on the DMF whole-brain model to determine the guiding role of compensatory mechanisms in the dynamic recovery of connectivity patterns. In addition, this type of model has shown direct clinical application value in identifying epileptogenic zones in drug-resistant patients undergoing resection surgery.

[0004] Currently, research on large-scale brain dynamics is still in its infancy, and the models established still face many key technical and theoretical issues that need to be addressed. At the theoretical level, existing network computational models struggle to systematically describe the dynamics of epileptic seizures, propagation, and control. At the critical technical level, existing computational models, due to their "high-dimensional parameter complexity," make it difficult to distinguish whether differences are caused by parameter selection or the nature of the model itself. Furthermore, due to the assumptions of "local or global consistency of lesions" or "abnormal excitability of epileptogenic zones," they lack a description and integration of the internal details of the whole-brain network structure of epilepsy. Consequently, the models established are not strictly whole-brain computational models of epilepsy.

[0005] Numerous studies have demonstrated that epileptic seizures are initiated by abnormal neuronal firing in the epileptogenic zone, which is then recruited through a large-scale network to other brain regions known as propagation zones. Although epilepsy typically has an epileptogenic zone, the most influential downstream "propagation points" within the epileptogenic network hold greater potential for inhibiting the continued spread of seizure activity. Analyzing and leveraging the information implicit in these downstream "propagation points" is crucial for gaining a deeper understanding of the mechanisms of epilepsy propagation, identifying key brain regions for epilepsy intervention, and exploring methods to alleviate epileptic seizures. Summary of the Invention

[0006] In response to the above technical problems, the present invention provides a whole-brain dynamics modeling method for epilepsy based on key propagation zones. Taking focal epilepsy as the research object and based on structure-function data, it aims to study how to accurately locate the key propagation zones of epilepsy, construct and optimize a high-precision whole-brain dynamics model of epilepsy, and realize full-parameter space simulation.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0008] A whole-brain dynamics modeling method for epilepsy based on key propagation zones comprises the following steps:

[0009] S1, data preprocessing and network construction;

[0010] S2. Accurately locate the key epilepsy transmission area;

[0011] S3. Construct a high-precision whole-brain dynamic model of epilepsy.

[0012] The method for data preprocessing and network construction in S1 is:

[0013] DWI was performed according to the preprocessing pipeline provided on the HCP official website. The cortical gray matter was divided into 68 anatomically distinct brain regions based on the Desikan-Killiany atlas to define the structural network. One million streamlines were assigned to each subject for deterministic fiber tract tracking. The distance between two gray matter brain regions was defined as the length of the fiber tract between them, and the structural connectivity was defined as the number of fiber tracts between them. Connections with fewer than five fiber tracts were pruned to reduce the high false positive rate of the probabilistic traction map. The individual-level structural connectivity matrix was thresholded, retaining only the top 20% of the strongest structural connections.

[0014] The data preprocessing is rs-fMRI data preprocessing, which is performed using the DPARSF toolbox. The processing steps include: removing the data of the first 10 time points, time layer correction, head motion correction, smoothing, filtering, and removing the influence of covariates on the signal; segmenting according to the Desikan-Killiany atlas, averaging the time series of voxels located in the same gray matter, obtaining Pearson correlations between time series based on 68 cortical brain regions, and constructing functional connectivity.

[0015] The method for accurately locating the key epilepsy propagation area in S2 comprises the following steps:

[0016] S2.1, quantitative estimation of signal transmission between neuronal pairs;

[0017] S2.2. Define strategies for the spread of epilepsy;

[0018] S2.3, normalize the node weights;

[0019] S2.4. Based on the PDs method, the epileptic brain regions are spatially divided into the epileptogenic zone (EZ), the critical propagation zone (PZ), and the healthy zone (HZ).

[0020] The method for quantitatively estimating signal transmission between neuron pairs in S2.1 is:

[0021] The key epilepsy propagation areas are quantified by calculating the participation of each brain region in the propagation of abnormal epileptic discharges. First, the epileptogenic network communication model W is constructed, coupling the healthy subjects' SC with the empirical epilepsy FC, while taking into account both structural and functional information. This matrix transforms the network of structural connectivity into a quantitative estimate of signal transmission between neuronal pairs:

[0022] W ij =bSC ij ·wFC ij

[0023] Where i represents the epileptogenic area, j represents other brain areas, if i = j, then W ij =0, bSC ij is the structural connectivity between the epileptogenic region and other brain regions, i.e., sparse binary; wFC ij It is the functional connectivity between the empirical epileptogenic area and other brain areas, that is, sparse weighting.

[0024] The method for defining the epilepsy propagation and diffusion strategy in S2.2 is:

[0025] PDs uses the epileptogenic zone as the seed node, i.e., the signal source. Abnormal excitation signals propagate simultaneously along multiple front ends of the network. The specific process of propagation and diffusion is as follows: First, starting from the seed node, all paths of length 1 are divided, i.e., W, and the signal is sent to the node directly coupled to it; second, all paths of length 2 starting from the seed node are divided, i.e., W 2 , all nodes that receive the signal source then propagate the signal to all their adjacent brain areas. All paths of length 2 include the path back to the signal source and the path that re-propagates the signal between the signal source neighbors. There are connections between the paths. Similarly, the set of all paths with a step length of n is divided, that is, W n ; Ultimately, the signal propagated from the seed node to other nodes is composed of all paths of length 1, 2, ..., n between them;

[0026] k-order matrix W k The (i, j)th item represents the number of paths of length k connecting a pair of nodes i and j. The participation degree between nodes i and j is:

[0027]

[0028] The participation of a propagation from seed node i to node j can be calculated as the weighted sum of the total number of times node j participated in the epilepsy propagation, where the contribution of each path is weighted according to 1 / k!, ensuring that short paths have higher participation than long paths; therefore, PDs are mainly carried out along short paths, and the overall contribution of inefficient paths quickly disappears as the paths become longer.

[0029] The method for normalizing the node weights in S2.3 is:

[0030]

[0031] in, is the connection strength of node i, and S=diag(s i ).

[0032] The method for spatial division of epileptic brain regions in S2.4 is:

[0033] Taking the epileptogenic zone as the center, the brain area directly coupled with it is defined as the super-spreading area, because it participates in every path of the epilepsy propagation process. Excluding the epileptogenic zone i and the super-spreading area, the remaining brain areas are regarded as the key propagation area candidate area j. The top 25% of all brain areas involved are defined as the key propagation area:

[0034]

[0035] in, is the participation of the upper quartile of the candidate areas of the key propagation area. If the participation of node j is greater than Then the node is a critical propagation area, otherwise it is considered a healthy area;

[0036] According to the PDs method, the spatial distribution of epileptic brain regions includes the epileptogenic zone (EZ), the key propagation zone (PZ), and the healthy zone (HZ).

[0037] The method for constructing a high-precision whole-brain dynamics model of epilepsy in S3 is:

[0038] The high-precision whole-brain dynamic model of epilepsy based on the key propagation area is:

[0039]

[0040] Where s represents the subject, i represents the epileptogenic zone of the subject, j represents the candidate propagation zone of the subject, and i≠j; the model includes the epileptogenic zone and the key propagation zone, A state signal representing neuronal activity generated by abnormal discharges in the epileptogenic zone. It represents the state signal of neuronal activity caused by abnormal discharge in the key propagation area. The model allows the existence of multiple key propagation areas. j PDs are used to determine whether brain region j is a key transmission area;

[0041] The dynamical model of a single epileptic brain region consists of state variables at three different time scales; among them, state variables x1 and y1 explain the rapid oscillation epileptiform activity during epileptic seizures, state variables x2 and y2 describe the peak-fluctuation epileptiform activity, and state variable z explains slow processes such as changes in extracellular ion concentration, energy consumption, and tissue oxygenation. The switching between the epileptic seizure state and the interictal state is realized by the state variable z. As the dielectric constant z decreases, the system shifts toward an unstable saddle node, causing the interictal state to lose stability and the system to shift to the seizure state. As the dielectric constant z increases, the system passes through an isoclinic bifurcation, leaving only stable fixed points, and the system shifts from the seizure state to the interictal state.

[0042] The dynamic equation of a single brain region is:

[0043]

[0044] Among them, τ0, τ2, I1, I2, and γ are all constant parameters; τ0 and τ2 are the time scale characteristics of the dielectric constant, I1 and I2 are the passive currents that set the operating point of the epileptic patient; the network nodes are coupled by the dielectric constant z, and K is defined i,j =λ·wSC ij ,λ is used to control the scaling factor of the global coupling strength of the network,λ<0, parameter exploration is required;wSC ijIt is the structural connection between brain regions i and j, that is, sparse weighting; the key parameters ωE(i) and δP(i) of the dielectric constant z are used to control the excitability value of the brain region; E(i) represents the excitability value contribution of the epileptogenic region i to the abnormal discharge behavior of the whole brain, and P(i) represents the excitability value contribution of the propagation region i to the abnormal discharge behavior of the whole brain; ω is used to determine whether the brain region i is an epileptogenic region, and δ is used to judge whether the brain region i is a key propagation region;

[0045] According to the PDs method, the spatial division of each brain region in epilepsy is into the epileptogenic zone EZ, the key propagation zone PZ, and the healthy zone HZ; when embedded in the epileptic brain network, each brain region has an excitability value, which quantifies the ability of each brain region to trigger epileptic seizures, that is, the critical value is x0; the brain regions in EZ have a higher excitability value satisfying E>x0 + 0.5, while the excitability of the nodes in PZ satisfies x0<P<x0 + 0.5, and the excitability value of the brain regions in HZ is lower than x0.

[0046] The beneficial effects of the present invention compared with the prior art are as follows:

[0047] Taking focal epilepsy as the research object, based on structure-function data, the present invention first constructs an epileptogenic network communication model, proposes an epilepsy propagation and diffusion strategy, calculates the participation degree of each brain region in the process of epileptic abnormal discharge propagation respectively, and accurately locates the key position and the best propagation point of epileptic seizures; secondly, integrates the key propagation regions of epilepsy, constructs and optimizes a high-precision whole-brain network dynamics model of epilepsy, and clarifies the functional role played by the key propagation regions in the process of epileptic abnormal discharge behavior. Finally, the present invention is applied to real epileptic fMRI data to evaluate its effectiveness. The method for constructing a whole-brain dynamics model of epilepsy based on key propagation regions provided by the present invention not only has important medical value for the research on the neural mechanism of focal epilepsy, but also promotes the development of high-precision whole-brain computational model construction and parameter optimization technology. Brief Description of the Drawings

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, other implementation drawings can be obtained by extension according to the provided drawings without creative efforts.

[0049] The structures, proportions, sizes, etc. illustrated in this specification are intended solely to complement the contents disclosed herein and to facilitate understanding and reading by persons skilled in the art. They are not intended to limit the conditions under which the present invention may be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportions, or adjustments in sizes, without affecting the efficacy and objectives of the present invention, shall remain within the scope of the technical contents disclosed herein.

[0050] Figure 1 Schematic diagram of the spatial distribution of epileptic brain networks;

[0051] Figure 2 It is the structural and functional network connection matrix diagram;

[0052] Figure 3 This is a map of the degree of participation of epilepsy in each brain region;

[0053] Figure 4 A high-precision whole-brain dynamics model parameter exploration diagram;

[0054] Figure 5 is the similarity graph between the simulated FC and empirical FC networks;

[0055] Figure 6 Functional signal diagrams with and without considering the propagation area;

[0056] Figure 7 Comparison diagram of simulated FC and empirical FC network characteristics. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of this application, not all the embodiments. These descriptions are only to further illustrate the features and advantages of the present invention, rather than to limit the claims of the present invention. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0058] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following embodiments are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0059] This is accomplished by the following steps:

[0060] 1. Data Preprocessing and Network Construction

[0061] DWI was performed according to the preprocessing pipeline provided on the HCP official website. The cortical gray matter was divided into 68 anatomically distinct brain regions based on the Desikan-Killiany atlas, defining a structural network. For each subject, one million streamlines were assigned for deterministic fiber tract tracing. The distance between two gray matter regions was defined as the length of the fiber tract between them, and structural connectivity was defined as the number of fiber tracts between them. Connections with fewer than five fiber tracts were pruned to reduce the high false-positive rate of probabilistic traction maps. The individual-level structural connectivity matrix was thresholded, retaining only the top 20% of the strongest structural connections. This process generates a sparse, weighted, undirected brain structural network for each subject, which is then used to reconstruct a whole-brain epilepsy model.

[0062] rs-fMRI data were preprocessed using the DPARSF toolbox. Key processing steps included removing the first 10 time points, performing temporal slice correction, motion correction, smoothing, filtering, and removing the influence of covariates. Segmentation was performed according to the Desikan-Killiany atlas, and the time series of voxels within the same gray matter were averaged. Functional connectivity was constructed based on Pearson correlations between time series across 68 cortical regions.

[0063] 2. Accurately locate the key epilepsy transmission area

[0064] The present invention quantifies the key epilepsy transmission zones by calculating the degree of involvement of each brain region in the propagation of abnormal epileptic discharges. First, an epileptogenic network communication model W is constructed, coupling (dot product) the healthy subjects' SC with the empirical epilepsy FC, while taking into account both structural and functional information. This matrix transforms the network of structural connectivity into a quantitative estimate of signal transmission between pairs of neurons:

[0065] W ij =bSC ij ·wFC ij (1)

[0066] Where i represents the epileptogenic area, j represents other brain areas, if i = j, then W ij =0. bSC ij is the structural connectivity between the epileptogenic region and other brain regions (sparse binary), wFC ij It is the functional connectivity relationship between the empirical epileptogenic zone and other brain regions (sparse weighting).

[0067] Secondly, the propagation and diffusion strategy (PDs) is defined. This strategy aims to solve the problem that the "existence or absence of an edge" or "distance" between other brain regions and the epileptogenic zone does not necessarily reflect the degree of epilepsy propagation and diffusion from the epileptogenic zone to other brain regions.

[0068] PDs uses the epileptogenic zone as the seed node (signal source), and the abnormal excitation signal propagates simultaneously along multiple front ends of the network. The specific process of propagation and diffusion is as follows: first, starting from the seed node, all paths of length 1 (i.e., W) are divided and the signal is sent to the node directly coupled to it; second, all paths of length 2 (i.e., W) starting from the seed node are divided and the signal is sent to the node directly coupled to it. 2 ), all nodes that receive the signal source then propagate the signal to all their adjacent brain regions. Importantly, this process is independent of whether a node has received the signal. All paths of length 2 include paths that trace back to the signal source and paths that retransmit the signal between the signal source's neighbors (if there is a connection between them). Similarly, the set of all paths with a step length of n (i.e., W n ). Ultimately, the signal propagated from the seed node to other nodes consists of all paths of length 1, 2, ..., n between them.

[0069] k-order matrix W k The (i, j)th item of represents the number of paths of length k connecting a pair of nodes i and j. The participation degree between nodes i and j is:

[0070]

[0071] The participation of a propagation from seed node i to node j can be calculated as the weighted sum of the total number of times node j participated in the epilepsy propagation, where the contribution of each path is weighted according to 1 / k!, ensuring that short paths have higher participation than long paths. Therefore, PDs are mainly transmitted along short paths, and the overall contribution of inefficient paths quickly disappears as the paths become longer.

[0072] Since PDs quantify the involvement of each brain region in the propagation of epileptic abnormal discharges by the total weight of their connections rather than by the number of nodes, in order to avoid the undue influence of excessive node weights, they are normalized to 3:

[0073]

[0074] in, is the connection strength of node i, and S=diag(s i ).

[0075] Taking the epileptogenic zone as the center, the brain regions directly coupled to it are defined as super-spreading zones, as they participate in every path of the epilepsy propagation process. Excluding the epileptogenic zone i and the super-spreading zones, the remaining brain regions are considered as candidate key propagation zones j. The top 25% of all brain regions involved are defined as key propagation zones:

[0076]

[0077] in, is the participation of the upper quartile of the candidate areas of the key propagation area. If the participation of node j is greater than Then the node is a critical propagation area, otherwise it is considered a healthy area.

[0078] According to the PDs method, the spatial distribution of epileptic brain regions is as follows Figure 1 As shown, it includes the epileptogenic zone EZ, the critical propagation zone PZ and the healthy zone HZ.

[0079] 3. High-precision whole-brain dynamics model of epilepsy

[0080] The high-precision whole-brain dynamic model of epilepsy based on the key propagation area is:

[0081]

[0082] Where s represents the subject, i represents the epileptogenic zone of the subject, j represents the candidate propagation zone of the subject, and i≠j. The model contains the epileptogenic zone and the key propagation zone. A state signal representing neuronal activity generated by abnormal discharges in the epileptogenic zone. The state signal of neuronal activity caused by abnormal discharge in the key propagation area. The model allows the existence of multiple key propagation areas, δ j PDs are used to determine whether brain area j is a key transmission area.

[0083] The dynamical model of a single epileptic brain region consists of state variables on three different time scales. State variables x1 and y1 account for the rapid oscillations of epileptiform activity during seizures, state variables x2 and y2 describe the peak-to-fluctuation epileptiform activity, and state variable z accounts for slow processes such as changes in extracellular ion concentration, energy consumption, and tissue oxygenation. The transition between the ictal state and the interictal state is controlled by the state variable z. As the dielectric constant z decreases, the system shifts toward an unstable saddle node, causing the interictal state to lose stability and the system to shift to the ictal state. As the dielectric constant z increases, the system undergoes an isoclinic bifurcation, leaving only stable fixed points, and the system shifts from the ictal state to the interictal state.

[0084] The dynamic equation of a single brain region is:

[0085]

[0086]

[0087] Where τ0, τ2, I1, I2, and γ are constant parameters. τ0 and τ2 are the time-scale characteristics of the dielectric constant, and I1 and I2 are the passive currents that set the operating point of the epileptic patient. The network nodes are coupled by the dielectric constant z, and K is defined i,j= λ·wSC ij , where λ is a scaling factor used to control the global coupling strength of the network (λ < 0, parameter exploration is required), and wSC ij is the structural connection (sparse weighted) between brain regions i and j. The key parameters ωE(i) and δP(i) of the dielectric constant z are used to control the excitatory values of brain regions. E(i) represents the excitatory value contribution of epileptogenic region i to the abnormal discharge behavior of the whole brain, and P(i) represents the excitatory value contribution of propagation region i to the abnormal discharge behavior of the whole brain. ω is used to determine whether brain region i is an epileptogenic region, and δ is used to judge whether brain region i is a key propagation region.

[0088] According to the PDs method, the spatial division of each brain region in epilepsy is into epileptogenic zone EZ, key propagation zone PZ, and healthy zone HZ as Figure 1 shown. When embedded in the epileptic brain network, each brain region has an excitatory value, which quantifies the ability of each brain region to trigger epileptic seizures (the critical value is x0). The brain regions in EZ (red) have a higher excitatory value satisfying E > x0 + 0.5, while the excitatory values of the nodes in PZ (orange) satisfy x0 < P < x0 + 0.5, and the excitatory values of the brain regions in HZ (white) are lower than x0. IV. Specific Embodiments

[0090] In this embodiment, a real epilepsy dataset from 26 drug-resistant epilepsy patients in OpenNeuro and a healthy dataset from 45 retest datasets in ConnectomeDB are used for precise localization of key propagation regions and high-precision whole-brain epilepsy dynamics modeling and optimization experiments. The specific experimental process is as follows:

[0091] 1. Data Preprocessing and Network Construction

[0092] DWI is carried out according to the preprocessing pipeline provided by the HCP official website. Based on the Desikan-Killiany atlas, the cortical gray matter is divided into 68 anatomically different brain regions to define the structural network. One million streamlines are assigned to each subject for deterministic fiber tractography. The distance between two gray matter brain regions is defined as the length of the fiber bundle between them, and the structural connection is defined as the number of fiber bundles between them. Connections with less than 5 fiber bundles are pruned to reduce the high false positive rate of the probabilistic tractography map, and the individual-level structural connection matrix is thresholded to only retain the top 20% of the strongest structural connections. The rs-fMRI data preprocessing is carried out using the DPARSF toolbox. The processing steps include removing the data of the first 10 time points, temporal slice correction, head motion correction, smoothing, filtering, and removing the influence of covariates on the signal. Segmentation is carried out according to the Desikan-Killiany atlas, and the time series of voxels located in the same gray matter are averaged to obtain the Pearson correlation between the time series based on 68 cortical brain regions to construct the functional connection. The structural and functional network connection matrices are as Figure 2 As shown, hSC represents healthy structural network connectivity, hFC represents healthy functional network connectivity, and eFC represents functional network connectivity in epilepsy patients.

[0093] 2. Accurate positioning of key communication areas

[0094] This example uses a patient with focal epilepsy in the left temporal pole as an example to accurately locate the key propagation area. Among them, brain area 32 is the epileptogenic area. Taking the epileptogenic area as the seed node, the PDs algorithm is used to calculate the participation of each brain area in the abnormal discharge propagation process of epilepsy, as shown in the following figure: Figure 3 As shown in A. Red represents the right hemisphere and blue represents the left hemisphere. The results showed that the participation of each brain region in the propagation of abnormal epileptic discharges was distributed symmetrically in the hemisphere, indicating that the left and right parts of the same brain region have similar propagation and diffusion capabilities. Statistical tests were performed on the whole brain, left hemisphere, and right hemisphere, as shown in the figure below. Figure 3 As shown in Figure B. The results showed that at the whole-brain level, the average whole-brain participation of the epilepsy group was significantly lower than that of the healthy control group (N=45, t=-13.216, p=0.000), indicating that the overall diffusion capacity of epilepsy patients was significantly reduced. At the left hemisphere level, the average whole-brain participation of the epilepsy group was significantly lower than that of the healthy control group (N=45, t=-13.804, p=0.000); at the right hemisphere level, the average whole-brain participation of the epilepsy group was significantly higher than that of the healthy control group (N=45, t=8.461, p=0.000). This indicates that when abnormal discharges occur in the left temporal lobe, the diffusion capacity of the left hemisphere of the epilepsy group is significantly reduced, while the diffusion capacity of the right hemisphere provides partial compensation. Excluding the two super-spreading areas directly connected to the epileptogenic zone (the participation of brain areas 29 and 34 is much greater than that of other brain areas), the participation of each brain area is sorted in descending order. The brain areas with participation higher than the upper quartile (dark color) are the key propagation areas involved in the abnormal discharge process of epilepsy, such as Figure 3 Figure C shows the division of the epileptogenic zone, propagation zone, and healthy zone, with the size of the ball indicating the degree of involvement. The results showed that the critical propagation zone was primarily located near the epileptogenic zone and exhibited significant left-hemisphere symmetry.

[0095] 3. Parameter optimization of high-precision whole-brain dynamics model

[0096] This example uses a patient with focal epilepsy in the left temporal pole as an example to optimize the parameters of a high-precision whole-brain dynamics model. Among them, brain area 32 is the epileptogenic zone. The constant parameters are set according to existing literature, where τ0 = 2857, τ2 = 10, I1 = 3.1, I2 = 0.45, γ = 0.01, and x0 = -2.05. The parameter x0 controls the excitability of brain tissue and can autonomously induce epileptic seizures. Without considering the influence of the propagation zone PZ, this example explores the parameters of the scaling factor λ that controls the global coupling strength of the network and the excitability value E of the epileptogenic zone, and optimizes the model parameters by calculating the network similarity between the simulated FC and the empirical FC. The global coupling strength λ represents the coupling strength between network nodes, and the excitability value E of the epileptogenic zone represents the ability of the epileptogenic zone to autonomously induce epilepsy.

[0097] Parameter exploration such as Figure 4 As shown, Figure 4 A represents the relationship between the scaling factor controlling the global coupling strength of the network, the excitability value, and the network fit. The horizontal axis represents the excitability value of the healthy zone (HZ), and the vertical axis represents the excitability value of the epileptogenic zone (EZ). Under a uniform parameter scale, warm colors indicate increasing network similarity, while cool colors indicate N / A similarity, indicating an inappropriate parameter setting. The results showed no significant relationship between the scaling factor λ controlling the global coupling strength of the network and the network fit, which may be related to the range of the empirical data. Figure 4 B shows the relationship between the excitability of the epileptogenic zone (EZ) and the healthy zone (HZ) and the network fit. The horizontal axis represents the excitability value, with blue representing the relationship between the excitability value and the network fit of the HZ, and red representing the relationship between the excitability value and the network fit of the EZ. The results showed that the HC had the highest fit when the excitability value was -3.05, and the EZ had the highest fit when the excitability value was -0.5.

[0098] 4. Reliability verification of high-precision whole-brain network model

[0099] This example verifies the reliability of the high-precision whole-brain network model from three aspects: the similarity between the simulated FC and the empirical FC network, the network characteristics, and the functional signals. Figure 5 As shown in Figure 2, the results show that when the propagation area is considered, the network similarity between the simulated FC and the empirical FC is significantly higher than that without considering the propagation area. Figure 5 A shows the network similarity graphs with and without considering the propagation area. As the number of propagation areas increases, the network similarity between the simulated FC and the empirical FC increases first and then decreases. The optimal number of propagation areas is 7. Figure 5 B shows the relationship between the number of propagation areas and network similarity. The epileptic whole brain functional signal diagram without considering the propagation area and after considering the propagation area is shown in Figure 2. Figure 6As shown in the figure, red represents the signal of the EZ area, yellow represents the PZ area, and blue represents the HZ area. The results show that the functional signal after adding the propagation area is closer to the empirical epileptic seizure signal. The network properties are compared without considering the propagation area and with considering the propagation area. Figure 7 As shown in the figure, they are clustering coefficient (Cp), local efficiency (Eloc), global efficiency (Eg) and shortest path length (Lp), E, ​​withPZ and nonPZ are the network properties under empirical data, considering the propagation area and not considering the propagation area, respectively. The results show that the network properties considering the propagation area are closer to the empirical data.

[0100] The above only describes in detail the preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in this field without departing from the purpose of the present invention, and various changes should be included in the scope of protection of the present invention.

Claims

1. A whole-brain dynamics modeling method for epilepsy based on key propagation zones, characterized by: The following steps are involved: S1, data preprocessing and network construction; S2. Accurately locate the key epilepsy transmission area, including the following steps: S2.1, quantitative estimation of signal transmission between neuronal pairs; The key epilepsy propagation areas are quantified by calculating the participation of each brain region in the propagation of abnormal epileptic discharges. First, an epileptogenic network communication model W is constructed, coupling the SC of healthy subjects with the FC of experienced epilepsy. Taking both structural and functional information into account, the epileptogenic network communication model W transforms the network of structural connectivity into a quantitative estimate of signal transmission between neuronal pairs: W ij =bSC ij ·wFC ij Where i represents the epileptogenic area, j represents other brain areas; W ij is an element in the epileptogenic network communication model W, which represents the structural and functional coupling connection relationship between brain regions i and j. If i = j, then W ij =0; bSC ij is the structural connectivity between brain regions i and j, i.e., sparse binary; wFC ij is the functional connectivity between brain regions i and j, i.e., sparse weighting; S2.

2. Define strategies for the spread of epilepsy; PDs uses the epileptogenic area as the seed brain region, i.e., the signal source. Abnormal excitation signals propagate simultaneously along multiple front ends of the network. The specific process of propagation and diffusion is as follows: First, starting from the seed brain region, all paths with a length of 1 are divided, i.e., W 1 , sending the signal to the brain area directly coupled with it; secondly, dividing all paths of length 2 starting from the seed brain area, namely W 2 , all brain regions that receive the signal source then propagate the signal to all adjacent brain regions. All paths of length 2 include paths that trace back to the signal source and paths that re-propagate the signal between neighbors of the signal source. There are connections between the paths. Similarly, the set of all paths with a step length of n is divided, that is, W n ; Ultimately, the signal propagating from the seed brain region to other brain regions is composed of all paths of length 1, 2, ..., n between them; k-order matrix W k The (i, j)th item represents the number of paths of length k connecting a pair of brain regions i and j, and the participation degree pd between brain regions i and j ij for: The participation of epileptogenic zone i in the propagation of epilepsy to other brain regions j is calculated as the weighted sum of the total number of times brain region j participates in epilepsy propagation; where k represents the path length between brain regions i and j, and the maximum length is n; Used to quantify the attenuation effect of path length on signal propagation, the contribution of each path is weighted according to 1 / k!, ensuring that short paths participate more strongly than long paths; therefore, PDs are preferentially propagated along short paths, and the overall contribution of inefficient paths quickly disappears as the paths become longer. S2.

3. Normalize brain region weights: in, is the normalized participation; W is the epileptogenic network communication model, It is the normalized form of the epileptogenic network communication model W. The purpose of normalization is to avoid excessive weights in brain regions and thus produce adverse effects. is the connection strength of brain region i, N represents the total number of brain regions, and S = diag(s i ); S2.

4. Based on the PDs method, the epileptic brain regions are spatially divided into the epileptogenic zone (EZ), the critical propagation zone (PZ), and the healthy zone (HZ). Taking the epileptogenic zone as the center, the brain area directly coupled with it is defined as the super-spreading area, because it participates in every path of the epilepsy propagation process. Excluding the empirical epileptogenic zone i and the super-spreading area, the remaining brain area p is regarded as the candidate area of ​​the key propagation area, and the top 25% of the participation of all brain areas is defined as the key propagation area: in, Quantify whether brain area p is a key propagation area; is the participation of the upper quartile of the candidate key propagation area. If the participation of brain area p is greater than or equal to Then the brain area is a critical transmission area, otherwise it is considered a healthy area; S3. Construct a high-precision whole-brain dynamic model of epilepsy.

2. The whole-brain dynamics modeling method for epilepsy based on key propagation zones according to claim 1, characterized in that: The method for data preprocessing and network construction in S1 is: DWI was performed according to the preprocessing pipeline provided on the HCP official website. The cortical gray matter was divided into 68 anatomically distinct brain regions based on the Desikan-Killiany atlas to define the structural network. One million streamlines were assigned to each subject for deterministic fiber tract tracking. The distance between two gray matter brain regions was defined as the length of the fiber tract between them, and the structural connectivity was defined as the number of fiber tracts between them. Connections with fewer than five fiber tracts were pruned to reduce the high false positive rate of the probabilistic traction map. The individual-level structural connectivity matrix was thresholded, retaining only the top 20% of the strongest structural connections.

3. The whole-brain dynamics modeling method for epilepsy based on key propagation zones according to claim 2, characterized in that: The data preprocessing is rs-fMRI data preprocessing, which is performed using the DPARSF toolbox. The processing steps include: removing the data of the first 10 time points, time layer correction, head motion correction, smoothing, filtering, and removing the influence of covariates on the signal; segmenting according to the Desikan-Killiany atlas, averaging the time series of voxels located in the same gray matter, obtaining Pearson correlations between time series based on 68 cortical brain regions, and constructing functional connectivity.

4. The whole-brain dynamics modeling method for epilepsy based on key propagation zones according to claim 1, characterized in that: The method for constructing a high-precision whole-brain dynamics model of epilepsy in S3 is: The high-precision whole-brain dynamic model of epilepsy based on the key propagation area is: in, represents the state signal of whole-brain neurons, a represents the subject, i represents the subject's empirical epileptogenic zone, p represents the subject's candidate propagation zone, and i≠p; the model includes the empirical epileptogenic zone and the key propagation zone, The state signal represents the neuronal activity generated by abnormal discharges in the epileptogenic zone. The state signal representing the neuronal activity generated by the candidate propagation region, It represents the state signal of neuronal activity caused by abnormal discharge of key propagation area. The model allows the existence of multiple key propagation areas. PDs are used to determine whether brain area p is a key transmission area; The dynamic model of a single brain region in epilepsy consists of state variables at three different time scales. Among them, the state variables x1 and y1 explain the rapid oscillatory epileptiform activity during seizures, the state variables x2 and y2 describe the spike-wave epileptiform activity, and the state variable z explains the slow transformation process of extracellular ion levels, tissue oxygen levels, and energy metabolism states. The switching between the seizure state and the interictal period is achieved by the state variable z. As the state variable z decreases, the system shifts towards an unstable saddle-node, leading to the loss of stability in the interictal period and the system turning to the seizure state. As the state variable z increases, the system undergoes a homoclinic bifurcation, leaving only stable fixed points, and the system realizes the shift from the seizure state to the interictal period. The dynamic equation of a single brain region is: Among them τ0, τ2, I 1,u , I 2,u , γ are constant parameters, τ0 and τ2 are the time scale characteristics of the dielectric constant, I 1,u and I 2,u is the passive current that sets the operating point of the epileptic patient; γ is the core parameter of the low-pass filter coupling function, and its value determines g(x 1,u ) slow time scale dynamics; u, v represent any brain region, 1≤u≠v≤N, N represents the total number of brain regions; F u is the dynamics equation responsible for generating the state activity of brain region u, capturing the local neural activity; x 1,u is the fast oscillating state variable of brain area u at time t; x 1,v is the fast oscillating state variable of brain area v at time t; is the fast oscillation state variable x in brain area u 1,u The time derivative of y 1,u is the fast recovery state variable of brain area u at time t; is the brain region u quickly recovering state variable y 1,u The time derivative of x 2,u is the peak-fluctuation state variable of brain region u at time t; is the peak-fluctuation state variable x in brain region u 2,u The time derivative of y 2,u is the peak-fluctuation recovery state variable of brain area u at time t; is the peak-fluctuation recovery state variable y of brain region u 2,u The time derivative of is the slow time scale state variable z of brain region u u The time derivative of z u is the slow time scale state variable of brain region u at time t, which comprehensively characterizes the steady-state characteristics of the system near the epileptic seizure in this brain region; through the state variable z u To couple the network brain areas, define K u,v =λ·wSC uv , represents the influence of brain region v on brain region u; λ is used to control the scaling factor of the global coupling strength of the network, λ<0, parameter exploration is required; wSC uv is the structural connection between brain regions u and v, i.e., sparse weighting; the key parameters ωE(u) and δP(u) are used to control the excitability of brain regions; E(u) represents the excitability contribution of the experienced epileptogenic region to the abnormal discharge behavior of the whole brain; P(u) represents the excitability contribution of the key propagation region to the abnormal discharge behavior of the whole brain; ω is used to determine whether brain region u is an experienced epileptogenic region, and δ is used to determine whether brain region u is a key propagation region; f1(x 1,u ,x 2,u ) is the set of medium-time scale state variables of the model (x 2,u ,y 2,u ) for the fast time scale state variable set (x 1,u ,y 1,u ) is a nonlinear inhibitory function that constrains fast oscillatory activity through negative feedback, ensuring that high-frequency discharges are activated only in the peak-fluctuation waveform phase; f2(x 1,u ,x 2,u ) is the set of medium-time scale state variables of the model (x 2,u ,y 2,u ) dynamic nonlinear driving function to coordinate the generation and switching of peaks and fluctuations; g(x 1,u ) is the set of state variables from the fast time scale in the model (x 1,u ,y 1,u ) to the medium time scale state variable set (x 2,u ,y 2,u ) is a low-pass filtered coupling function to coordinate the peak-fluctuation rhythm generation with the phase synchronization of the fast oscillation, and τ is the variable that traverses the historical moments during the integration process; When embedded in the epileptic brain network, each brain region has an excitability value, which quantifies the ability of each brain region to trigger seizures, that is, the critical value is x0. The brain regions in EZ have a higher excitability value satisfying E > x0 + 0.5, while the excitability of the brain regions in PZ satisfies x0 < P < x0 + 0.5, and the excitability value of the brain regions in HZ is lower than x0.

Images