High frequency oscillation rhythm detection method and system based on adaptive fuzzy double clustering algorithm

Abstract

The application discloses a high-frequency oscillation rhythm detection method and system based on an adaptive fuzzy double clustering algorithm, and relates to the medical treatment technical field. The particle swarm optimization algorithm is a kind of random optimization algorithm based on group, has high requirements for the initial position of the solution, has strong global search capability, and is not easy to fall into local optimal value, and has strong processing capability for complex samples with large quantity, so that the particle swarm optimization algorithm is used to optimize the fuzzy double clustering algorithm, and the optimized clustering center can be obtained, and the high-frequency oscillation rhythm of the epilepsy electroencephalogram signal can be detected based on the optimized clustering center. The application can improve the detection precision of the high-frequency oscillation rhythm of the epilepsy electroencephalogram signal, and help doctors to diagnose epilepsy and remove the epileptogenic focus.

Description

Technical Field

[0001] This invention relates to the field of electroencephalogram (EEG) signal detection technology, and in particular to a method and system for detecting high-frequency oscillation rhythms based on an adaptive fuzzy bi-clustering algorithm. Background Technology

[0002] Epilepsy is a highly prevalent chronic neurological disorder caused by highly synchronized abnormal neuronal discharges, often causing immense physical and psychological suffering for patients. For refractory patients unresponsive to medication, surgical intervention is usually necessary. High-frequency oscillatory rhythms in electroencephalograms (EEGs) serve as important biomarkers for the epileptic seizure onset zone, and their accurate detection is crucial for preoperative evaluation. The key challenge of this invention is to detect high-frequency oscillatory rhythms characterized by low amplitude, short duration, and non-stationarity, identify pathological rhythms, and then utilize these pathological high-frequency oscillatory rhythms to accurately locate the seizure onset zone, thereby assisting physicians in diagnosing and treating epilepsy.

[0003] Traditional high-frequency oscillation detection requires two or more professionals. These professionals first need to edit the patient's raw EEG data, and then, after signal filtering, use their experience to determine whether the signal is a high-frequency oscillation or an artifact. However, this method has significant drawbacks: (1) manual detection takes too long; analyzing five minutes of single-channel EEG data takes approximately one hour; (2) the judgment relies on experience and lacks objectivity. Different professionals analyzing the same EEG signal may produce significantly different results, resulting in poor uniformity. These drawbacks limit the clinical application of high-frequency oscillation detection in epilepsy.

[0004] The traditional fuzzy C-means algorithm relies on gradient descent to find the optimal solution. However, gradient descent has weak global search capabilities, and its search for the optimal solution depends mainly on its local search capabilities. The algorithm itself has high requirements for the quality of the initial cluster centers, and the search space, i.e., the number of sample points, cannot be too large, otherwise it will seriously affect the final clustering results. Often, many experiments are needed on the same set of sample points to obtain satisfactory clustering results, making the algorithm time-consuming and extremely inefficient.

[0005] In conclusion, developing an automatic detection model for high-frequency oscillation signals with fast detection speed and high accuracy is an important way to promote the clinical application of high-frequency oscillations and efficiently locate the epileptic onset zone. Summary of the Invention

[0006] This invention addresses the problems and shortcomings of existing technologies by providing a high-frequency oscillation rhythm detection method and system based on an adaptive fuzzy bi-clustering algorithm.

[0007] The present invention solves the above-mentioned technical problems through the following technical solution:

[0008] This invention provides a high-frequency oscillation rhythm detection method based on an adaptive fuzzy bi-clustering algorithm, characterized by comprising the following steps:

[0009] S1. Obtain the time series of EEG signals from a patient with epilepsy, and calculate four characteristic data of the EEG signals at different times: Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f. c This forms a set of feature vectors X = (x1, x2, ..., x...). n ), x j Representing the j-th eigenvector: energy operator T j Short-time energy P j Power ratio R j and the centroid of the spectrum f cj n is a positive integer and 1≤j≤n;

[0010] S2. Randomly divide the feature vector set into c clusters, with the cluster center set V = (v1, v2, ..., vc). c Repeat the partitioning process S times to obtain S cluster center sets, where the S cluster center sets represent the S particles in the particle swarm optimization algorithm.

[0011] For each of the S particles:

[0012] S3. Calculate the membership matrix U = (u_i - u_i) of each eigenvector in the eigenvector set with respect to each cluster center in the particle. ji );

[0013]

[0014]

[0015] v i Let u represent the i-th cluster center. ji Z represents the membership degree of the j-th eigenvector to the i-th cluster center, with a membership degree constraint of 1. Z is the sparse self-representation coefficient of X. j is the j-th column of Z, and α is a parameter that controls the effect of discriminative features;

[0016] S4. Calculate the fitness value of the particle and determine whether the current fitness value of the particle is greater than the previous fitness value. If so, select the current position of the particle to update the individual optimal position. Otherwise, select the previous position of the particle to update the individual optimal position. The position of the particle corresponding to the maximum current fitness value among all particles is taken as the global optimal position.

[0017] The fitness value f is calculated using the following formula:

[0018]

[0019]

[0020] S5. Determine whether the difference between the maximum value in the current iteration fitness value and the maximum value in the previous iteration fitness value is less than a set threshold. If yes, proceed to step S6; otherwise, proceed to step S9.

[0021] S6. The particle corresponding to the maximum value in the current iteration fitness value is the optimal cluster center, and the optimal cluster center set is obtained.

[0022] S7. Based on the obtained optimal cluster center set, the membership matrix of the feature vector set X is obtained according to the formula in step S3. Each feature vector is divided into the class of the optimal cluster center with the largest membership degree to obtain the clustering result.

[0023] S8. Based on the clustering results, box plots were used to perform statistical analysis on the four features in each cluster to detect the high-frequency oscillation rhythm of the epileptic EEG signal. The four features in each cluster refer to the Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f in each cluster after clustering. c ;

[0024] S9. Update the velocity and position of the particles according to the velocity formula and position formula of the particle swarm optimization algorithm to generate a new generation of particle swarm, and proceed to step S3.

[0025]

[0026]

[0027] Where t is initially 0, t represents the t-th iteration, and s∈{1,...,S} is the number of particles. This represents the velocity of the s-th particle in the t-th iteration. Let represent the position of the s-th particle in the t-th iteration, w represent a constant value called inertia weight in the range [0,1], and c1 and c2 are learning factors, generally taken as c1 = c2 ∈ [0,4]. Let g represent the optimal position of the s-th particle in the t-th iteration. t Let r1 and r2 represent the global optimal position in the entire particle swarm in the t-th iteration. r1 and r2 are extracted from the uniform distribution U(0,1) to add a random component to the velocity update to diversify the search.

[0028] The present invention also provides a high-frequency oscillation rhythm detection system based on an adaptive fuzzy double clustering algorithm, which includes a feature acquisition module, a partitioning module, a first calculation module, a second calculation module, a judgment module, an assignment module, a clustering module, an analysis module, and an update module;

[0029] The feature acquisition module is used to acquire the time series of EEG signals of a certain epilepsy patient and calculate four feature data of the EEG signals at different times: Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f. c This forms a set of feature vectors X = (x1, x2, ..., x...). n ), x j Representing the j-th eigenvector: energy operator T j Short-time energy P j Power ratio R j and the centroid of the spectrum f cj n is a positive integer and 1≤j≤n;

[0030] The partitioning module is used to randomly and initially divide the feature vector set into c clusters, with the cluster center set being V = (v1, v2, ..., vc). c Repeat the partitioning process S times to obtain S cluster center sets, where the S cluster center sets represent the S particles in the particle swarm optimization algorithm.

[0031] For each of the S particles:

[0032] The first calculation module is used to calculate the membership matrix U = (u_i - u_i) of each feature vector in the feature vector set with respect to each cluster center in the particle. ji );

[0033]

[0034]

[0035] v i Let u represent the i-th cluster center. ji Z represents the membership degree of the j-th eigenvector to the i-th cluster center, with a membership degree constraint of 1. Z is the sparse self-representation coefficient of X. j is the j-th column of Z, and α is a parameter that controls the effect of discriminative features;

[0036] The second calculation module is used to calculate the fitness value of the particle, determine whether the current fitness value of the particle is greater than the previous fitness value, if so, select the current position of the particle to update the individual optimal position, otherwise select the previous position of the particle to update the individual optimal position, and take the position of the particle corresponding to the maximum value of the current fitness value of all particles as the global optimal position.

[0037] The fitness value f is calculated using the following formula:

[0038]

[0039]

[0040] The judgment module is used to determine whether the difference between the maximum value in the current iteration fitness value and the maximum value in the previous iteration fitness value is less than a set threshold. If yes, the assignment module is called; if no, the update module is called.

[0041] The assignment module is used to assign the particle corresponding to the maximum value in the current iteration fitness value as the optimal cluster center, thereby obtaining the optimal cluster center set.

[0042] The clustering module is used to derive the membership matrix of the feature vector set X based on the obtained optimal cluster center set and the membership formula, and to divide each feature vector into the class of the optimal cluster center with the largest membership degree, thereby obtaining the clustering result.

[0043] The analysis module is used to perform statistical analysis on four features in each cluster based on the clustering results using a box plot method, detecting the high-frequency oscillation rhythm of epileptic EEG signals. The four features in each cluster refer to the Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f in each cluster after clustering. c ;

[0044] The update module is used to update the velocity and position of particles according to the velocity formula and position formula of the particle swarm optimization algorithm, generate a new generation of particle swarm, and call the first calculation module.

[0045]

[0046]

[0047] Where t is initially 0, t represents the t-th iteration, and s∈{1,...,S} is the number of particles. This represents the velocity of the s-th particle in the t-th iteration. Let represent the position of the s-th particle in the t-th iteration, w represent a constant value called inertia weight in the range [0,1], and c1 and c2 are learning factors, generally taken as c1 = c2 ∈ [0,4]. Let g represent the optimal position of the s-th particle in the t-th iteration. t Let r1 and r2 represent the global optimal position in the entire particle swarm in the t-th iteration. r1 and r2 are extracted from the uniform distribution U(0,1) to add a random component to the velocity update to diversify the search.

[0048] The positive and progressive effects of this invention are as follows:

[0049] This invention can improve the detection accuracy of high-frequency oscillation rhythms of epileptic EEG signals, helping doctors to diagnose epilepsy and remove epileptogenic foci. Attached Figure Description

[0050] Figure 1 This is a flowchart of a high-frequency oscillation rhythm detection method based on an adaptive fuzzy bi-clustering algorithm, which is a preferred embodiment of the present invention.

[0051] Figure 2 This is a block diagram of a high-frequency oscillation rhythm detection system based on an adaptive fuzzy bi-clustering algorithm, which is a preferred embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, this embodiment provides a high-frequency oscillation rhythm detection method based on an adaptive fuzzy bi-clustering algorithm. Since the particle swarm optimization algorithm is a population-based random optimization algorithm, it has low requirements for the initial position of the solution, strong global search capability, and is not easily trapped in local optima. It has a strong ability to process a large number of complex samples. Therefore, the particle swarm optimization algorithm is used to optimize the fuzzy bi-clustering algorithm to obtain optimized cluster centers. Based on the optimized cluster centers, the high-frequency oscillation rhythm of epileptic EEG signals can be detected.

[0054] Specifically, the following steps are included:

[0055] Step 101: Obtain the time series of EEG signals from a patient with epilepsy, and calculate four characteristic data of the EEG signals at different times: Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f. c This forms a set of feature vectors X = (x1, x2, ..., x...). n ), x j Representing the j-th eigenvector: energy operator T j Short-time energy P j Power ratio R j and the centroid of the spectrum f cj n is a positive integer and 1≤j≤n.

[0056] Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f c The specific calculations are based on existing technology.

[0057] Step 102: Randomly divide the feature vector set into c clusters, with the cluster center set being V = (v1, v2, ..., v...). c Repeat the division S times to obtain S cluster center sets, and the S cluster center sets represent S particles in the particle swarm optimization algorithm.

[0058] For example, the feature vector set is randomly initially divided into 10 clusters, with the cluster center set being V = (v1, v2, ..., v...). 10 ), v 10 The 10th cluster is represented by the cluster center. By repeating the partitioning operation 20 times, 20 cluster center sets can be obtained. Each cluster center set represents a particle in the particle swarm optimization algorithm. The 20 cluster center sets represent 20 particles in the particle swarm optimization algorithm. The particle swarm optimization algorithm can be used to optimize these 20 particles, which is actually to optimize the 20 cluster center sets, thereby obtaining optimized cluster centers.

[0059] For each of the S particles:

[0060] Step 103: Calculate the membership matrix U = (u_i - u_i) of each eigenvector in the eigenvector set with respect to each cluster center in the particle. ji ).

[0061] The membership formula is:

[0062]

[0063]

[0064] v i Let u represent the i-th cluster center. ji Z represents the membership degree of the j-th eigenvector to the i-th cluster center, with a membership degree constraint of 1. Z is the sparse self-representation coefficient of X. j is the j-th column of Z, and α is a parameter that controls the effect of discriminative features.

[0065] Step 104: Calculate the fitness value of the particle, and determine whether the current fitness value of the particle is greater than the previous fitness value. If so, select the current position of the particle to update the individual optimal position. Otherwise, select the previous position of the particle to update the individual optimal position. The position of the particle corresponding to the maximum current fitness value among all particles is taken as the global optimal position.

[0066] The fitness value f is calculated using the following formula:

[0067]

[0068]

[0069] Step 104 can be used to calculate the individual optimal position of each particle and the global optimal position of all particles.

[0070] Step 105: Determine whether the difference between the maximum value of the current iteration fitness value and the maximum value of the previous iteration fitness value is less than the set threshold. If so, it indicates that the optimal cluster center has been reached and proceed to step 106. Otherwise, it indicates that the optimal cluster center has not been reached and proceed to step 109.

[0071] Step 106: The particle corresponding to the maximum value in the current iteration fitness value is the optimal cluster center, and the optimal cluster center set is obtained.

[0072] Step 107: Based on the obtained optimal cluster center set, derive the membership matrix of the feature vector set X according to the formula in step 103, and divide each feature vector into the class of the optimal cluster center with the largest membership degree to obtain the clustering result.

[0073] Step 108: Based on the clustering results, box plots are used to perform statistical analysis on the four features in each cluster to detect the high-frequency oscillation rhythm of the epileptic EEG signal. The four features in each cluster refer to the Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f in each cluster. c This step is existing technology.

[0074] Step 109: Update the velocity and position of the particles according to the velocity and position formulas of the particle swarm optimization algorithm to generate a new generation of particle swarm, and then proceed to step 103.

[0075]

[0076]

[0077] Where t is initially 0, t represents the t-th iteration, and s∈{1,...,S} is the number of particles. This represents the velocity of the s-th particle in the t-th iteration. Let represent the position of the s-th particle in the t-th iteration, w represent a constant value called inertia weight in the range [0,1], and c1 and c2 are learning factors, generally taken as c1 = c2 ∈ [0,4]. Let g represent the optimal position of the s-th particle in the t-th iteration. tLet r1 and r2 represent the global optimal position in the entire particle swarm in the t-th iteration. r1 and r2 are extracted from the uniform distribution U(0,1) to add a random component to the velocity update to diversify the search.

[0078] Cluster centers within a population cannot move randomly without boundaries; they must be confined to a predefined range.

[0079]

[0080]

[0081] Among them, h max and h min Determine the boundary of h, h max h represents the maximum velocity of a particle in a particle swarm. min y represents the minimum velocity value of a particle in a particle swarm. max and y min Determine the boundary of y, y max y represents the position of the largest cluster of particles in the particle swarm. min This represents the minimum cluster position of particles in a particle swarm.

[0082] like Figure 2 As shown, this embodiment also provides a high-frequency oscillation rhythm detection system based on an adaptive fuzzy double clustering algorithm, which includes a feature acquisition module 1, a partitioning module 2, a first calculation module 3, a second calculation module 4, a judgment module 5, an assignment module 6, a clustering module 7, an analysis module 8, and an update module 9.

[0083] Feature acquisition module 1 is used to acquire the time series of EEG signals of a certain epilepsy patient and calculate four feature data of the EEG signals at different times: Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f. c This forms a set of feature vectors X = (x1, x2, ..., x...). n ), x j Representing the j-th eigenvector: energy operator T j Short-time energy P j Power ratio R j and the centroid of the spectrum f cj n is a positive integer and 1≤j≤n.

[0084] Partitioning module 2 is used to randomly and initially divide the feature vector set into c clusters, with the cluster center set being V = (v1, v2, ..., v...). c Repeat the division S times to obtain S cluster center sets, and the S cluster center sets represent S particles in the particle swarm optimization algorithm.

[0085] For each of the S particles:

[0086] The first calculation module 3 is used to calculate the membership matrix U = (u_i) of each feature vector in the feature vector set with respect to each cluster center in the particle. ji ).

[0087]

[0088]

[0089] Among them, v i Let u represent the i-th cluster center. ji Z represents the membership degree of the j-th eigenvector to the i-th cluster center, with a membership degree constraint of 1. Z is the sparse self-representation coefficient of X. j is the j-th column of Z, and α is a parameter that controls the effect of discriminative features.

[0090] The second calculation module 4 is used to calculate the fitness value of the particle, determine whether the current fitness value of the particle is greater than the previous fitness value, if so, select the current position of the particle to update the individual optimal position, otherwise select the previous position of the particle to update the individual optimal position, and take the position of the particle corresponding to the maximum value of the current fitness value of all particles as the global optimal position.

[0091] The fitness value f is calculated using the following formula:

[0092]

[0093]

[0094] The judgment module 5 is used to determine whether the difference between the maximum value in the current iteration fitness value and the maximum value in the previous iteration fitness value is less than a set threshold. If yes, the assignment module 6 is called; if no, the update module 9 is called.

[0095] Module 6 is used to assign the particle corresponding to the maximum value in the current iteration fitness value as the optimal cluster center, thereby obtaining the optimal cluster center set.

[0096] Clustering module 7 is used to derive the membership matrix of the feature vector set X based on the obtained optimal cluster center set and the membership formula, and to divide each feature vector into the class of the optimal cluster center with the largest membership degree, thus obtaining the clustering result.

[0097] Analysis module 8 is used to perform statistical analysis on the four features in each cluster after clustering using the box plot method, based on the clustering results, to detect the high-frequency oscillation rhythm of the epileptic EEG signal. The four features in each cluster refer to the Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f in each cluster after clustering.c .

[0098] The update module 9 is used to update the velocity and position of particles according to the velocity formula and position formula of the particle swarm optimization algorithm, generate a new generation of particle swarm, and call the first calculation module 3.

[0099]

[0100]

[0101] Where t is initially 0, t represents the t-th iteration, and s∈{1,...,S} is the number of particles. This represents the velocity of the s-th particle in the t-th iteration. Let represent the position of the s-th particle in the t-th iteration, w represent a constant value called inertia weight in the range [0,1], and c1 and c2 are learning factors, generally taken as c1 = c2 ∈ [0,4]. Let g represent the optimal position of the s-th particle in the t-th iteration. t Let r1 and r2 represent the global optimal position in the entire particle swarm in the t-th iteration. r1 and r2 are extracted from the uniform distribution U(0,1) to add a random component to the velocity update to diversify the search.

[0102] Cluster centers within a population cannot move randomly without boundaries; they must be confined to a predefined range.

[0103]

[0104]

[0105] Among them, h max and h min Determine the boundary of h, h max h represents the maximum velocity of a particle in a particle swarm. min y represents the minimum velocity value of a particle in a particle swarm. max and y min Determine the boundary of y, y max y represents the position of the largest cluster of particles in the particle swarm. min This represents the minimum cluster position of particles in a particle swarm.

[0106] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but all such changes and modifications fall within the scope of protection of the present invention.

Claims

1. A high-frequency oscillation rhythm detection method based on an adaptive fuzzy bi-clustering algorithm, characterized in that, It includes the following steps: S1. Obtain the time series of EEG signals from a patient with epilepsy, and calculate four characteristic data of the EEG signals at different times: Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f. c This forms a set of feature vectors X=(x1,x2,…,x…). n ), x j Representing the j-th eigenvector: energy operator T j Short-time energy P j Power ratio R j and the centroid of the spectrum f cj n is a positive integer and 1≤j≤n; S2. Randomly divide the feature vector set into c clusters, with the cluster center set V=(v1,v2,…,v…). c Repeat the partitioning process S times to obtain S cluster center sets, where the S cluster center sets represent the S particles in the particle swarm optimization algorithm. For each of the S particles: S3. Calculate the membership matrix U=( ) for each eigenvector in the eigenvector set with respect to each cluster center in the particle. ); ； ； v i Indicates the i-th cluster center. Let X be the membership degree of the j-th eigenvector to the i-th cluster center, with a membership degree constraint of 1, and Z be the sparse self-representation coefficients of X. is the j-th column of Z, and α is a parameter that controls the effect of discriminative features; S4. Calculate the fitness value of the particle and determine whether the current fitness value of the particle is greater than the previous fitness value. If so, select the current position of the particle to update the individual optimal position. Otherwise, select the previous position of the particle to update the individual optimal position. The position of the particle corresponding to the maximum current fitness value among all particles is taken as the global optimal position. The fitness value f is calculated using the following formula: ； ； S5. Determine whether the difference between the maximum value in the current iteration fitness value and the maximum value in the previous iteration fitness value is less than a set threshold. If yes, proceed to step S6; otherwise, proceed to step S9. S6. The particle corresponding to the maximum value in the current iteration fitness value is the optimal cluster center, and the optimal cluster center set is obtained. S7. Based on the obtained optimal cluster center set, the membership matrix of the feature vector set X is obtained according to the formula in step S3. Each feature vector is divided into the class of the optimal cluster center with the largest membership degree to obtain the clustering result. S8. Based on the clustering results, box plots were used to perform statistical analysis on the four features in each cluster to detect the high-frequency oscillation rhythm of the epileptic EEG signal. The four features in each cluster refer to the Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f in each cluster after clustering. c ; S9. Update the velocity and position of the particles according to the velocity formula and position formula of the particle swarm optimization algorithm to generate a new generation of particle swarm, and proceed to step S3. ； ； Where t is initially 0, t represents the t-th iteration, and s∈{1,...,S} is the number of particles. This represents the velocity of the s-th particle in the t-th iteration. Let represent the position of the s-th particle in the t-th iteration, w represent a constant value called inertia weight in the range [0,1], and c1 and c2 are learning factors, where c1 = c2 ∈ [0,4]. This represents the optimal position of the s-th particle at the t-th iteration. Let r1 and r2 represent the global optimal position in the entire particle swarm in the t-th iteration. r1 and r2 are extracted from the uniform distribution U(0,1) to add a random component to the velocity update to diversify the search.

2. The high-frequency oscillation rhythm detection method based on adaptive fuzzy bi-clustering algorithm as described in claim 1, characterized in that, Cluster centers within a population cannot move randomly without boundaries; they must be confined to a predefined range. ； in, and Determine the boundary of h. This represents the maximum velocity value of a particle in a particle swarm. This represents the minimum velocity value of a particle in a particle swarm. and Determine the boundary of y. This represents the position of the largest cluster of particles in the particle swarm. This represents the minimum cluster position of particles in a particle swarm.

3. A high-frequency oscillation rhythm detection system based on an adaptive fuzzy bi-clustering algorithm, characterized in that, It includes a feature acquisition module, a partitioning module, a first calculation module, a second calculation module, a judgment module, an assignment module, a clustering module, an analysis module, and an update module; The feature acquisition module is used to acquire the time series of EEG signals of a certain epilepsy patient and calculate four feature data of the EEG signals at different times: Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f. c This forms a set of feature vectors X=(x1,x2,…,x…). n ), x j Representing the j-th eigenvector: energy operator T j Short-time energy P j Power ratio R j and the centroid of the spectrum f cj n is a positive integer and 1≤j≤n; The partitioning module is used to randomly and initially divide the feature vector set into c clusters, with the cluster center set being V=(v1,v2,…,v…). c Repeat the partitioning process S times to obtain S cluster center sets, where the S cluster center sets represent the S particles in the particle swarm optimization algorithm. For each of the S particles: The first calculation module is used to calculate the membership matrix U=( for each feature vector in the feature vector set with respect to each cluster center in the particle). ); ； ； v i Indicates the i-th cluster center. Let X be the membership degree of the j-th eigenvector to the i-th cluster center, with a membership degree constraint of 1, and Z be the sparse self-representation coefficients of X. is the j-th column of Z, and α is a parameter that controls the effect of discriminative features; The second calculation module is used to calculate the fitness value of the particle, determine whether the current fitness value of the particle is greater than the previous fitness value, if so, select the current position of the particle to update the individual optimal position, otherwise select the previous position of the particle to update the individual optimal position, and take the position of the particle corresponding to the maximum value of the current fitness value of all particles as the global optimal position. The fitness value f is calculated using the following formula: ； ； The judgment module is used to determine whether the difference between the maximum value in the current iteration fitness value and the maximum value in the previous iteration fitness value is less than a set threshold. If yes, the assignment module is called; if no, the update module is called. The assignment module is used to assign the particle corresponding to the maximum value in the current iteration fitness value as the optimal cluster center, thereby obtaining the optimal cluster center set. The clustering module is used to derive the membership matrix of the feature vector set X based on the obtained optimal cluster center set and the membership formula, and to divide each feature vector into the class of the optimal cluster center with the largest membership degree, thereby obtaining the clustering result. The analysis module is used to perform statistical analysis on four features in each cluster based on the clustering results using a box plot method, detecting the high-frequency oscillation rhythm of epileptic EEG signals. The four features in each cluster refer to the Teager energy operator T, short-time energy P, power ratio R, and spectral centroid f in each cluster after clustering. c ; The update module is used to update the velocity and position of particles according to the velocity formula and position formula of the particle swarm optimization algorithm, generate a new generation of particle swarm, and call the first calculation module. ； ； Where t is initially 0, t represents the t-th iteration, and s∈{1,...,S} is the number of particles. This represents the velocity of the s-th particle in the t-th iteration. Let represent the position of the s-th particle in the t-th iteration, w represent a constant value called inertia weight in the range [0,1], and c1 and c2 are learning factors, where c1 = c2 ∈ [0,4]. This represents the optimal position of the s-th particle at the t-th iteration. Let r1 and r2 represent the global optimal position in the entire particle swarm in the t-th iteration. r1 and r2 are extracted from the uniform distribution U(0,1) to add a random component to the velocity update to diversify the search.

4. The high-frequency oscillation rhythm detection system based on adaptive fuzzy bi-clustering algorithm as described in claim 3, characterized in that, Cluster centers within a population cannot move randomly without boundaries; they must be confined to a predefined range. ； in, and Determine the boundary of h. This represents the maximum velocity value of a particle in a particle swarm. This represents the minimum velocity value of a particle in a particle swarm. and Determine the boundary of y. This represents the position of the largest cluster of particles in the particle swarm. This represents the minimum cluster position of particles in a particle swarm.

Images