An epilepsy prediction system based on synchronous acquisition of electroencephalogram - electrogastrogram signals

Abstract

The present invention relates to the field of disease prediction, and particularly to an epilepsy prediction system based on synchronously acquired electroencephalogram-electrogastrography signals, comprising: a data acquisition module for acquiring data, where the data includes electroencephalogram data and electrogastrogram data; a first analysis module for obtaining the average electrode inconsistency index of a first channel and a second channel based on a first model according to time series data; and a prediction module for comparing the average electrode inconsistency index of a subject with a preset evaluation threshold so as to predict epilepsy patients. The system provided by the present invention can more accurately and efficiently distinguish epilepsy patients from healthy people and can be used as an auxiliary diagnostic tool for primary screening of epilepsy in a large population (such as communities, physical examination centers).

Description

Technical Field

[0001] The present invention relates to the field of disease prediction, and particularly to an epilepsy prediction system based on synchronously collected electroencephalogram-electrogastrograph signals. Background Art

[0002] Epilepsy is a chronic neurological disorder that affects approximately 70 million people globally, and is listed by the World Health Organization as one of the five major neuropsychiatric diseases that require key prevention and treatment globally, and is also a key area of focus in the global brain science program. Epileptic seizures are caused by abnormal electrical discharges of a group of brain cells, and different parts of the brain can become the sites of abnormal electrical discharges. Its characteristics are recurrent attacks, sudden onset and termination, and it is difficult to capture the electroencephalogram during the seizure period. During an epileptic seizure, a certain part of the body or the entire body often experiences brief involuntary twitching, sometimes accompanied by loss of consciousness and urinary and fecal incontinence. The seizures range from extremely brief loss of consciousness or muscle reflexes to severe and persistent convulsions. The frequency of seizures can also vary, from less than one seizure per year to several seizures per day.

[0003] Currently, the basis for the diagnosis of epilepsy includes clinical seizures and epileptiform discharges on electroencephalograms. On the one hand, the assessment of electroencephalogram abnormalities may be subjective and not very sensitive. For example, in routine electroencephalogram recordings, epileptiform discharges (such as spikes, sharp waves, spike-slow complexes, and sharp-slow complexes, etc.) are only present in 29%-55% of patients. On the other hand, a large proportion of epilepsy patients have very short seizure durations, and even if they undergo electroencephalogram examinations, it is very difficult to capture their abnormal electrical discharges. Therefore, the current existing technology based on electroencephalograms can only detect the epileptiform discharges of subjects, and thus is applicable to the diagnosis of epileptic seizures, but is difficult to be used for the screening of epilepsy.

[0004] It is reported that up to 70% of epilepsy patients can achieve seizure - free status through the proper use of anti - epileptic drugs. Thus, "early detection and early treatment" of epilepsy has great practical significance. However, there are at least ten million epilepsy patients in China, but only about one - third of them receive correct and adequate treatment. The Global Burden of Disease Study estimates that the burden of epilepsy in China is 1.6 million disability - adjusted life years (DALYs), accounting for 12% of the global total and 95% of the East Asian region. The reasons for the above - mentioned results may lie in the long - term misunderstandings, fears, and prejudices of epilepsy patients towards epilepsy, which lead to a considerable number of people in China being reluctant to go to medical institutions for further examination even if they have epileptic seizures; and the existing epilepsy diagnosis is based on the detection of epileptiform discharges in subjects, which causes epilepsy patients without epileptiform discharges during electroencephalogram (EEG) detection to be missed by the existing technology. For another part of the patients, they may suffer from non - epileptic paroxysmal diseases (such as neurotic seizures, transient ischemic attacks), but often worry that they have epilepsy, and it is difficult to distinguish from epilepsy only based on clinical manifestations. Correct diagnosis is a prerequisite for effective treatment, and the existing technology is difficult to distinguish which are non - epilepsy patients and which are epilepsy patients only based on the patient's oral description and an EEG showing no abnormal discharges, especially in a large population, such as in communities and physical examination centers.

[0005] Chinese patent application CN117481666A discloses a method for coupling gastric - brain electrical signals at rest, which reflects the degree of bidirectional influence between the original gastric electrical signal and the original brain electrical signal by extracting the phase - amplitude coupling, phase - phase coupling, and amplitude - amplitude coupling indices, as well as transfer entropy between the original gastric electrical signal and the original brain electrical signal in different frequency bands. However, its application scenario is limited to regulating gastric function and it is difficult to achieve disease diagnosis. Summary of the Invention

[0006] The present invention provides an epilepsy prediction system based on synchronously collected electroencephalogram - gastrointestinal electrical signals, which is characterized by including:

[0007] A data acquisition module for acquiring data, where the data includes electroencephalogram data and gastrointestinal electrical data. The electroencephalogram data is collected from a first channel, and the gastrointestinal electrical data is collected from a second channel; the data includes sample data from a sample population and subject data from a subject;

[0008] A data pre - processing module for pre - processing the data to obtain corresponding time - series data;

[0009] A first analysis module for obtaining the average electrode inconsistency index of the first channel and the second channel based on a first model according to the time - series data; the first model includes: where JSD represents the JS divergence scoring function, t represents time, ait represents the time series data of the corresponding channels of the first channel and the second channel recorded for the subject at time point t seconds, m represents the total number of channels, and n represents the total duration of the time series data of the corresponding channels; ICI represents the inconsistency index of the corresponding channel;

[0010] A prediction module, configured to compare the average electrode inconsistency index of the subject with a preset evaluation threshold. When the average electrode inconsistency index of the subject is not lower than the preset evaluation threshold, the subject is predicted to be an epilepsy patient;

[0011] wherein, the first channel includes F3, C4, P4, O2, T7, FC1, CP1, CP6, TP9, TP10, F2, C1, C2, P2, FC3, FC4, CP3, CP4, PO4, P6, PO8, CPz, and POz; the second channel includes the gastric body, lesser curvature of the stomach, gastric antrum, ascending colon, transverse colon, descending colon, and rectal channels.

[0012] In some embodiments, the system further includes a second analysis module, configured to convert the time series data into probability density distribution data.

[0013] In some embodiments, the second analysis module is further configured to obtain the inconsistency index of the corresponding channel based on the second model according to the probability density distribution data. The second model includes: wherein, ICI represents the inconsistency index function, D KL represents the KL divergence function, t represents time, a it represents the time series data of the corresponding channel recorded for the subject at time point t seconds, b it represents the time series data of the corresponding channel recorded for the healthy population at time point t seconds, n is the total duration of the time series data of the corresponding channel, P(a it ) represents the sample distribution, and Q(b it ) represents the reference distribution.

[0014] In some embodiments, the system further includes a database, configured to store the data obtained by the data acquisition module, the data preprocessing module, the first analysis module, the second analysis module, and the prediction module, and to generate a corresponding sample data set.

[0015] In some embodiments, the system further includes an evaluation threshold setting module. The evaluation threshold determination module randomly obtains the average electrode inconsistency index of 50 - 70% of the epilepsy patient samples in the sample data set and calculates the critical value of epilepsy occurrence, and sets the average electrode inconsistency index corresponding to the smallest critical value as the evaluation threshold.

[0016] In some embodiments, the electroencephalogram (EEG) data and the electrogastrogram (EGG) data are synchronously collected.

[0017] In some embodiments, the acquisition time of the EEG data and the EGG data is at least 5 minutes.

[0018] In some embodiments, the preprocessing includes filtering the EEG data to filter out EEG signals below 1 Hz and above 30 Hz, and normalizing the EGG data and the filtered EEG data to obtain the time series data.

[0019] In some embodiments, the data is collected before or after a meal.

[0020] In some embodiments, the data is collected from subjects in a normal state. As used herein, "normal state" means that the EEG activity of the subject is normal, that is, there are no detectable epileptiform discharges in the electroencephalogram of the subject. In some embodiments, when the subject is a healthy person (i.e., a non-epileptic patient), the normal state is any state of the subject. In some embodiments, when the subject is an epileptic patient, the normal state is the period during which the subject is not having an epileptic seizure (or the interictal period).

[0021] In some embodiments, the subject is an epileptic patient.

[0022] In some embodiments, the epileptic patient has a focal or generalized origin.

[0023] Compared with the prior art, the beneficial effects of the present invention at least include the following aspects:

[0024] The present invention provides a seizure prediction system based on synchronously collected EEG-EGG signals. The prior art methods for processing EEG signals and gastric electrical signals are to perform phase-amplitude coupling, phase-phase coupling, and amplitude-amplitude coupling on EEG signals and gastric electrical signals in different frequency bands. There has been no report in the prior art on using EEG signals and gastric electrical signals for seizure prediction and diagnosis.

[0025] Different from the existing conventional ideas for extracting frequency bands from electroencephalogram (EEG) signals, the present invention realizes accurate prediction of epilepsy by collecting EEG data and electrogastrogram (EGG) data corresponding to the streamlined EEG channels and EGG channels (i.e., the first channel and the second channel of the present invention) and combining processing means such as the inconsistency index. It should be emphasized that, on the one hand, the present invention breaks through the general idea of simultaneously selecting channels with relatively high correlation in EEG channels and EGG channels. By streamlining the channels with relatively high correlation in EEG channels and replacing them with specific EGG channels, the channels are streamlined, reducing the actual operation difficulty while improving the accuracy and sensitivity of diagnosing epilepsy. On the other hand, the present invention does not detect epileptic seizures in epileptic patients by detecting whether abnormal discharges occur in the electroencephalogram, but diagnoses whether a subject is an epileptic patient based on the EEG (and EGG) of the subject in the normal state.

[0026] In large-scale screening, even for experienced clinicians, it is difficult to determine whether a subject is an epileptic patient based solely on the subject's oral description of the symptoms of "epileptic" seizures under time pressure and lack of diagnostic basis. The accuracy and sensitivity of the prediction model obtained based on the streamlined channels and specific algorithms of the present invention are both at a relatively high level, which can capture and magnify the differences between epileptic patients and healthy people with almost the same EEG activities under conventional methods (such as visual observation), and then distinguish them, and is applicable to assisting in diagnosing subjects who may not show abnormal discharges in the electroencephalogram in large-scale screening (such as healthy people with symptoms similar to epileptic seizures but not epileptic patients, and epileptic patients with infrequent epileptic seizures).

[0027] Based on this, the system provided by the present invention not only requires short data collection time, but also can more accurately and efficiently distinguish epileptic patients from healthy people, and can be used as an auxiliary diagnostic tool for primary screening of epileptic patients in a large population (such as communities, physical examination centers). BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to actual scale. Obviously, the following described drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to these drawings without creative efforts.

[0029] Figure 1 Schematic diagram of the positions of the streamlined first channel and second channel of the present invention;

[0030] Figure 2 Result graph of the performance of different prediction models obtained for the present invention;

[0031] Figure 3 Flowchart of the analysis method for simultaneously collected electroencephalogram-electrogastrogram signals provided by the present invention;

[0032] Figure 4 Schematic diagram of an epilepsy prediction system based on simultaneously collected electroencephalogram-electrogastrogram signals provided by the present invention.

[0033] 100 is a prediction system, 102 is a data acquisition module, 104 is a data preprocessing module, 106 is a first analysis module, 108 is a second analysis module, 110 is a prediction module, 112 is a database, and 114 is an evaluation threshold setting module. Detailed implementation manners

[0034] To make the objectives, 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 with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] In this article, suffixes such as "module", "component", or "unit" used to represent elements are only for the convenience of describing the present invention, and they have no specific meaning in themselves. Therefore, "module", "component", or "unit" can be used interchangeably.

[0036] In this article, terms such as "upper", "lower", "inner", "outer", "front", "rear", "one end", "the other end", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as a limitation of the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0037] In this article, unless otherwise clearly specified and limited, terms such as "installed", "provided with", "connected", etc. shall be understood in a broad sense. For example, "connected" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium, and can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0038] As used herein, "and / or" includes any and all combinations of one or more of the listed associated items.

[0039] As used herein, "a plurality of" means two or more, i.e., it includes two, three, four, five, etc.

[0040] It should be noted that, as used herein, the term "comprises", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or apparatus comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or apparatus. Without further limitation, an element defined by the statement "comprising a..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising such element.

[0041] As used in this specification, the term "about" typically means + / - 5% of the stated value, more typically + / - 4% of the stated value, more typically + / - 3% of the stated value, more typically + / - 2% of the stated value, even more typically + / - 1% of the stated value, and even more typically + / - 0.5% of the stated value.

[0042] In this specification, certain embodiments may be disclosed in a format that is within a certain range. It should be understood that this description of "within a certain range" is merely for convenience and brevity and should not be construed as a rigid limitation on the disclosed range. Accordingly, the description of a range should be considered to have specifically disclosed all possible sub-ranges and the individual numerical values within that range. For example, the description of the range 1 - 6 should be considered to have specifically disclosed sub-ranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6, etc., as well as the individual numbers within that range, such as 1, 2, 3, 4, 5, and 6. The above rules apply regardless of the breadth of the range.

[0043] Example 1

[0044] 1.1 Data set

[0045] Subjects: In this example, 102 epilepsy patients with focal or generalized onset (57 males, age 33.82 ± 12.65) and 46 healthy controls (26 males, age 31.00 ± 8.45 years) were included.

[0046] 1.1.1 Electroencephalogram (EEG) recording

[0047] The subjects sat in a quiet room and were prepared for the test. Five-minute resting-state electroencephalogram (EEG) data were obtained from all subjects using 64 Ag / AgCl electrodes (BrainProducts, Munich, Germany), which were placed according to the 10 / 20 system with an impedance < 10 kΩ. The EEG data were amplified and collected at a sampling rate of 1000 Hz using an actiCAmp amplifier (BrainVision system amplifier; Brain Products, Munich, Germany). The subjects were instructed to fix their gaze at the center of the screen and to avoid body movements as much as possible during the EEG recording.

[0048] 1.1.2 Gastrointestinal electrical signal (EGG) recording

[0049] Gastrointestinal myoelectric activity was measured using an 8-channel gastrointestinal electromyograph (XDJ-S8, Hefei Kaili Co., Ltd., Hefei, China). The subjects were informed to avoid alcohol, spicy or irritating foods for at least three days and to fast for more than 6 hours before the examination. The measurements were taken in the supine position. Four gastric electrodes (reflecting the body of the stomach, antrum of the stomach, lesser curvature of the stomach, greater curvature of the stomach) and four intestinal electrodes (reflecting the ascending colon, transverse colon, descending colon, rectum) were placed on the abdominal skin (Hanjie Co., Shanghai, China). The specific placement sites of the above electrodes were as follows: Body of the stomach: 3 to 5 cm to the left and 1 cm upward from the midpoint of the line connecting the xiphoid process and the umbilicus; Antrum of the stomach: 2 to 4 cm to the right of the midpoint of the line connecting the xiphoid process and the umbilicus; Lesser curvature of the stomach: at the upper 1 / 2 of the midpoint of the line connecting the xiphoid process and the umbilicus; Greater curvature of the stomach: at the lower 1 / 2 of the midpoint of the line connecting the xiphoid process and the umbilicus. Ascending colon: 2 to 4 cm to the right at the level of the umbilicus; Transverse colon: 1 cm below the umbilicus; Descending colon: 2 to 4 cm to the left at the level of the umbilicus; Rectum: below the coccyx of the back. The ground electrode was placed on the medial ankle of the right calf, and the reference electrode was placed on the medial wrist of the right hand. The subjects were informed to keep quiet during the test and not to move or talk. In addition, after the EGG recording 6 minutes before the meal, a meal-time functional load test was performed with approximately 200 kcal of food.

[0050] 1.2 Prediction algorithm based on Jensen–Shannon divergence (JSD)

[0051] 1.2.1 Data preprocessing

[0052] The dataset was divided into a training set and a test set at a ratio of 7:3. For independent training, only EEG or EGG signals were used. For EEG-EGG multimodal fusion training, EEG and EGG signals were used.

[0053] First, use EEGLAB in MATLAB to filter the original EEG signals to maintain data integrity by eliminating signals below 1 Hz and above 30 Hz. Second, normalize the filtered EEG data and the original EGG signals. The feature matrices of the preprocessed EEG time series or the preprocessed EGG time series are as follows.

[0054]

[0055] Among them, m and n respectively represent the dimensions of the feature matrices of the EEG signals or EGG signals. m is the total number of electrodes, so the i-th row represents the recording per second at electrode e i . n is the total duration of the EEG time series signal or the EGG time series signal, that is, the entire EEG data or EGG data contains n seconds, so the t-th column represents the EEG data or EGG data measured at each electrode at the time point of t seconds. Therefore, a it is the EEG signal or EGG signal of electrode e i recorded at the time point of t seconds.

[0056] 1.2.2 Convert the preprocessed EEG time series or the preprocessed EGG time series into probability density distribution data

[0057] Convert the data of each electrode into probability density distribution data. In the training set, use the time series data a it of each electrode to fit the sample distribution P(a it ), where the mean u1 i and the standard deviation σ1 i are used as sample features. In addition, for the healthy control data in the training set, fit the time series data b it of each electrode to the reference distribution Q(b it ), where the mean u2 i and the standard deviation σ2 i are used as the reference features for discrimination.

[0058]

[0059] Among them, P(a it ) represents the sample distribution, a it represents the time series data of each electrode of the sample, u1 i represents the mean of the sample, and σ1 i represents the standard deviation of the sample.

[0060]

[0061] Among them, Q(b it ) represents the reference distribution, b itTime series data of each electrode representing the reference, u2 i Mean value representing the reference, σ2 i Standard deviation representing the reference.

[0062] When the difference between the sample features and the reference features is small, it means there is no substantial difference between the sample distribution and the reference distribution, thus indicating no epilepsy. On the contrary, when the difference is quite large, it means there is a significant difference between the sample distribution and the reference distribution, indicating the presence of epilepsy.

[0063] 1.2.3 Construction of inconsistency index based on JSD

[0064] JS divergence is used to quantify the deviation of the sample distribution from the reference distribution to measure the difference between different states (i.e., epilepsy and healthy control); the inconsistency index (ICI) for each electrode is calculated to represent the difference between time points.

[0065]

[0066] Among them, D KL Represents the KL divergence function, P(a it ) represents the sample distribution, Q(b it ) represents the reference distribution, a it Represents the time series data of the corresponding channel of the sample recorded at time point t seconds, b it Represents the time series data of the corresponding channel of the reference recorded at time point t seconds, t represents time, and T represents the total time.

[0067]

[0068] ICI represents the inconsistency index function, P(a it ) represents the sample distribution, Q(b it ) represents the reference distribution, a it Represents the time series data of the corresponding channel of the sample recorded at time point t seconds, b it Represents the time series data of the corresponding channel of the reference recorded at time point t seconds, t represents time, and n is the total duration of the time series data of the corresponding channel.

[0069] By calculating the ICI, the fluctuations at all time points caused by the reference distribution are quantified. The average ICI score for each electrode can be used as a quantitative measurement to evaluate the degree of distribution difference over time. When the average ICI score is small, it indicates that there is no substantial difference between the sample distribution of the electrode and the reference distribution of the electrode. On the contrary, when the average ICI score is large, it indicates that there is a significant difference between the sample distribution of the electrode and the reference distribution of the electrode.

[0070] 1.2.4 Identification of DNB Electrodes

[0071] Based on the DNB (Dynamic Network Biomarker) theory, assuming that a set of known molecules emits signals of the pre-disease state of complex diseases before significant transitions, the average ICI score of each electrode is calculated. Subsequently, the above scores are ranked according to the relative importance at critical moments. The fluctuations caused by the reference distribution between each electrode obtained according to the above equation at all time points are then used to sort the ICI scores of each electrode.

[0072] In this embodiment, the selection of DNB electrodes is divided into two different strategies. For the individual training of EEG, the electrodes ranked in the top 50% according to the highest average score are classified as DNB electrodes.

[0073] For the EEG-EGG multimodal fusion training, a total of 78 channels are included (62 EEG channels and 16 EGG channels). First, the top 30 channels with the highest average score for each individual are selected. Then, if a certain channel appears in more than half of the individuals in the training set, it is used as a DNB electrode. Since different electrodes may be associated with different regions of the brain or stomach, the higher the score of a specific electrode in the entire array, the more important its relationship with epileptic activity.

[0074] 1.2.5 Calculation of JSD for the Global Time Series

[0075] The average electrode inconsistency index of the global time series is defined as follows:

[0076]

[0077] where JSD represents the JS divergence scoring function, t represents time, a it represents the time series data of the corresponding channel recorded at t seconds, m represents the total number of channels, n represents the total duration of the time series data of the corresponding channel; ICI represents the inconsistency index of the corresponding channel.

[0078] The JSD score is used to quantify the distribution differences between all electrodes, thereby helping to identify any significant deviations between distributions. The higher the score, the more obvious the differences between the distributions. When the difference exceeds this value, it indicates that the difference has escalated to a critical level, thus predicting epilepsy. To find an appropriate threshold level for each patient, the following training scenario can be adopted: 70% of the epileptic data is used as training data to obtain the threshold. For each epileptic case in the training data set, according to the JSD score, the critical value in the DNB theory for each seizure can be obtained, and the minimum value of these critical values is used as the threshold. When the score at a certain time point exceeds this threshold, the subject is diagnosed with epilepsy.

[0079] 1.3 Correlation Analysis of EEG Channels and EGG Channels

[0080] According to the following equation, the probability distribution of each channel was evaluated by analyzing the EEG data of 62 channels and the EGG data of 16 channels for each individual:

[0081]

[0082] The mean and standard deviation of the density functions P(a it ) and P(a it ) obtained from the EEG and EGG channels were used to calculate the P - value. The correlation between the EEG and EGG channels was determined by comparing the magnitudes of their P - values. A P - value less than 0.05 indicates no significant difference between the two, thus supporting the hypothesis of a connection between them. To identify the EEG channels with the strongest correlation with the EGG channels, the top three EEG channels with the smallest statistical deviation from the distribution of the 16 EGG channels were selected for each sample.

[0083] Example 2

[0084] It should be noted that in the present invention, the selection of DNB electrodes is carried out in two different ways. For individual EEG training, the top 50% of the electrodes with the highest average ICI score are selected as DNB electrodes according to the total number of electrodes used. For EEG - EGG multimodal fusion training, the top 30 channels with the highest average ICI score for each individual in the training set are selected. If a channel appears in more than half of the training set, it is designated as a common DNB electrode for epilepsy prediction. As a result, as Figure 1 shown, the 30 channels obtained by the present invention include: 23 EEG channels (the first channel): F3, C4, P4, O2, T7, FC1, CP1, CP6, TP9, TP10, F2, C1, C2, P2, FC3, FC4, CP3, CP4, PO4, P6, PO8, CPz, and POz; 7 EGG channels (the second channel): body of stomach channel (i.e., electrode 1), lesser curvature of stomach channel (i.e., electrode 2), gastric antrum (i.e., electrode 4), ascending colon channel (i.e., electrode 5), transverse colon channel (i.e., electrode 6), descending colon channel (i.e., electrode 7), and rectal channel (i.e., electrode 8).

[0085] Figure 2 In, EEG represents a prediction model trained using the data collected from 62 EEG channels by the method of the present invention; EEG *denotes a prediction model trained using the method of the present invention with data collected from the top 50% of electrodes having the highest average ICI scores among 62 EEG channels; EEG+EGG denotes a prediction model trained using the method of the present invention with data collected from 78 channels (i.e., 62 EEG channels and 16 EGG channels); EEG+EGG * denotes a prediction model trained using the method of the present invention with data collected from the first channel and the second channel.

[0086] As Figure 2 shown, the prediction models obtained by processing EEG or EEG+EGG data using the method of the present invention already show good prediction ability. However, while it requires collecting and calculating more data, the accuracy is difficult to reach 95%. The prediction model constructed based on the electrodes refined from the present invention not only reduces the number of electrodes to be placed, but also improves the performance of predicting epilepsy, observing an accuracy of 98.92% and a sensitivity of 98.82%.

[0087] Correlation analysis between EEG channels and EGG channels.

[0088] In this embodiment, the probability distribution is further used to explore whether there is a certain correlation between different EEG channels and EGG channels in predicting epilepsy. Specifically, this embodiment analyzed the probability distributions of 62 EEG channels and 16 EGG channels for each subject. Among 148 samples, the occurrence frequencies of 62 EEG channels were statistically similar to those of 16 EGG channels. Furthermore, this embodiment extracted the top three EEG channels related to each EGG channel in the subjects, and these channels were defined as the key correlation factors of specific EGG channels.

[0089] The distributions of all gastrointestinal electrogram and electroencephalogram channels were compared for similarity, and the number of occurrences of eight gastrointestinal electrogram leads that did not show significant differences from the electroencephalogram channel distributions was calculated in 148 subjects. The more times, the higher the correlation between the electroencephalogram lead and the gastrointestinal electrogram. This embodiment determined that compared with 16 EGG channels, the top 20% of EEG channels did not show statistically significant distribution differences, and they were designated as highly correlated EEG channels, as shown in Table 1.

[0090] Table 1: Statistical analysis of the occurrence frequencies of electroencephalogram channels ranked in the top 20% related to gastrointestinal electrogram in the results of 148 samples

[0091]

[0092] The results show that most of the EEG channels shown in Table 1 are not included in the first channel. Based on this, the present invention believes that the predictive effect of most of the EEG channels shown in Table 1 on epilepsy can be replaced by the relevant electrogastrogram channels (i.e., the above-mentioned second channel).

[0093] Embodiment III

[0094] See Figure 3 , the present invention provides an analysis method for synchronously collected electroencephalogram-electrogastrogram signals, including the following steps:

[0095] S101 Input the synchronously collected electroencephalogram data and electrogastrogram data of the subject;

[0096] In some embodiments, the electroencephalogram data is collected from the first channel, and the electrogastrogram data is collected from the second channel.

[0097] In the present invention, "channel", "electrode" and "lead" have the same meaning unless otherwise specified.

[0098] In some embodiments, the first channel includes F3, C4, P4, O2, T7, FC1, CP1, CP6, TP9, TP10, F2, C1, C2, P2, FC3, FC4, CP3, CP4, PO4, P6, PO8, CPz and POz; the second channel includes the gastric body, lesser curvature of the stomach, gastric antrum, ascending colon, transverse colon, descending colon and rectal channels.

[0099] It should be noted that the first channel and the second channel are respectively the refined electroencephalogram channel and electrogastrogram channel of the present invention, and the first channel and the second channel have a low correlation in epilepsy.

[0100] Different from the prior art that couples electrogastrogram signals and electroencephalogram signals, after obtaining the electroencephalogram channels with strong electrogastrogram correlation in epilepsy patients, the present invention instead excludes these electroencephalogram channels from the actually collected electroencephalogram channels, thereby obtaining the above-mentioned first channel. The present invention finds that the predictive effect of these electroencephalogram channels can be replaced by the relevant electrogastrogram channels (i.e., the above-mentioned second channel). After adopting the above-mentioned first channel and second channel with low correlation in predicting epilepsy, not only the difficulty of the acquisition operation is reduced, but also the accuracy of predicting epilepsy patients is improved, which is more convenient for practical application. However, using a lower number of electroencephalogram channels (such as conventional 16 leads, 32 leads, etc.) is not sufficient to cover the entire brain, and it is difficult to predict epilepsy.

[0101] In some embodiments, the acquisition time of the electroencephalogram data and the electrogastrogram data is at least 5 minutes.

[0102] In some embodiments, the data is collected before or after a meal.

[0103] In some embodiments, the data is collected from subjects in a normal state.

[0104] In some embodiments, the subject is an epilepsy patient.

[0105] In some embodiments, the epilepsy patient has focal onset or generalized onset.

[0106] S102 Preprocess the synchronously collected EEG data and EGG data to obtain corresponding time series data;

[0107] In some embodiments, the preprocessing includes filtering the EEG data to filter out EEG signals below 1 Hz and above 30 Hz, and normalizing the EGG data and the filtered EEG data to obtain the time series data.

[0108] S103 Based on the first model, obtain the average electrode inconsistency index of the first channel and the second channel;

[0109] In some embodiments, the first model includes: where JSD represents the JS divergence scoring function, t represents time, a it represents the time series data of the corresponding channels of the first channel and the second channel recorded by the subject at time point t seconds, m represents the total number of channels, n represents the total duration of the time series data of the corresponding channels; ICI represents the inconsistency index of the corresponding channel.

[0110] In some embodiments, the corresponding channels refer to F3, C4, P4, O2, T7, FC1, CP1, CP6, TP9, TP10, F2, C1, C2, P2, FC3, FC4, CP3, CP4, PO4, P6, PO8, CPz, POz, gastric body channel, gastric lesser curvature channel, gastric antrum channel, ascending colon channel, transverse colon channel, descending colon and rectal channel. In some embodiments, the total duration refers to the total duration of the synchronously collected EEG data and EGG data. For example, when the acquisition time is 5 min, the total duration is 5 min.

[0111] Based on the above first channel and second channel, the present invention not only reduces the difficulty of the acquisition operation, but also reduces the amount of data that needs to be analyzed and calculated, improves the calculation efficiency, and maintains the accuracy and sensitivity at a relatively high level.

[0112] In some embodiments, the method further includes S201 converting the time series data into probability density distribution data.

[0113] In some embodiments, the method further includes S202 obtaining, according to the probability density distribution data, an inconsistency index of a corresponding channel based on a second model, where the second model includes: where ICI represents an inconsistency index function, D KL represents a KL divergence function, t represents time, a it represents the time series data of the corresponding channel of the subject recorded at the t-th second, b it represents the time series data of the corresponding channel of a healthy population recorded at the t-th second, n is the total duration of the time series data of the corresponding channel, P(a it ) represents a sample distribution, and Q(b it ) represents a reference distribution.

[0114] S104 compares the average electrode inconsistency index of the subject with a preset evaluation threshold. When the average electrode inconsistency index of the subject is not lower than the preset evaluation threshold, the subject is predicted to be an epilepsy patient.

[0115] In some embodiments, the method for obtaining the preset evaluation threshold includes S203 randomly obtaining the average electrode inconsistency index of 50%-70% of the cognitive impairment patient samples in the sample dataset and calculating their epilepsy occurrence critical values (based on the DNB theory), and setting the average electrode inconsistency index corresponding to the smallest critical value as the evaluation threshold.

[0116] Embodiment 4

[0117] Referring to Figure 4 , the present invention provides an epilepsy prediction system 100 based on synchronously acquired electroencephalogram-electrogastrogram signals, which is characterized by including:

[0118] A data acquisition module 102, configured to acquire data, where the data includes electroencephalogram data and electrogastrogram data;

[0119] In some embodiments, the electroencephalogram data and the electrogastrogram data are synchronously acquired.

[0120] In some embodiments, the acquisition time of the electroencephalogram data and the electrogastrogram data is at least 5 minutes.

[0121] In some embodiments, the data is acquired before or after a meal.

[0122] In some embodiments, the electroencephalogram data is acquired from a first channel, and the electrogastrogram data is acquired from a second channel.

[0123] In some embodiments, the data includes sample data from a sample population and subject data from a subject.

[0124] In some embodiments, the sample population includes a healthy population and epilepsy patients.

[0125] In some embodiments, the data is collected from a subject in a normal state.

[0126] In some embodiments, the subject is an epilepsy patient.

[0127] In some embodiments, the epilepsy patients are of focal origin or generalized origin.

[0128] In some embodiments, the first channels include F3, C4, P4, O2, T7, FC1, CP1, CP6, TP9, TP10, F2, C1, C2, P2, FC3, FC4, CP3, CP4, PO4, P6, PO8, CPz, and POz; the second channels include the body of the stomach, lesser curvature of the stomach, antrum of the stomach, ascending colon, transverse colon, descending colon, and rectal channels.

[0129] The data preprocessing module 104 is configured to preprocess the data to obtain corresponding time series data;

[0130] In some embodiments, the preprocessing includes filtering the electroencephalogram data to filter out electroencephalogram signals below 1 Hz and above 30 Hz, and normalizing the electrogastrogram data and the filtered electroencephalogram data to obtain the time series data.

[0131] The first analysis module 106 is configured to obtain the average electrode inconsistency index of the first channels and the second channels based on a first model according to the time series data;

[0132] In some embodiments, the first model includes: where JSD represents the JS divergence scoring function, t represents time, a it represents the time series data of the corresponding channels of the first channels and the second channels recorded at the time point t seconds of the subject, m represents the total number of channels, n represents the total duration of the time series data of the corresponding channels; ICI represents the inconsistency index of the corresponding channels.

[0133] The second analysis module 108 is configured to convert the time series data into probability density distribution data.

[0134] In some embodiments, the second analysis module 108 is further configured to obtain the inconsistency index of each sub-channel based on a second model according to the probability density distribution data.

[0135] In some embodiments, the second model includes: where ICI represents an inconsistency index function, D KL represents the KL divergence function, t represents time, a it represents the time series data of the corresponding channel recorded by the subject at the t-th second, b it represents the time series data of the corresponding channel recorded by the healthy population at the t-th second, n is the total duration of the time series data of the corresponding channel, P(a it ) represents the sample distribution, and Q(b it ) represents the reference distribution.

[0136] In some embodiments, the first analysis module 106 and the second analysis module 108 can be combined into an analysis module.

[0137] A prediction module 110, configured to compare the average electrode inconsistency index of a subject with a preset evaluation threshold. When the average electrode inconsistency index of the subject is not lower than the preset evaluation threshold, the subject is predicted to be an epilepsy patient;

[0138] A database 112, configured to store the data obtained by the data acquisition module 102, the data preprocessing module 104, the first analysis module 106, the second analysis module 108, and the prediction module 110, and to generate a corresponding sample data set.

[0139] In some embodiments, the sample data set includes a healthy population data set and an epilepsy patient data set.

[0140] An evaluation threshold setting module 114, which sets the average electrode inconsistency index corresponding to the smallest critical value as the evaluation threshold by randomly obtaining the average electrode inconsistency index of 50-70% of the epilepsy patient samples in the epilepsy sample data set and calculating the critical value of epilepsy occurrence.

[0141] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a computer terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0142] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative rather than restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims. All of these are within the protection scope of the present invention.

Claims

1. An epilepsy prediction system based on synchronous acquisition of electroencephalogram - electrogastrogram signals, characterized in that, Including: A data acquisition module for acquiring data, where the data includes electroencephalogram (EEG) data and electrogastrogram (EGG) data. The data is collected before or after a meal and from subjects in a normal state. The EEG data is collected from a first channel, and the EGG data is collected from a second channel. The data includes sample data from a sample population and subject data from subjects. The EEG data and the EGG data are collected synchronously. A data preprocessing module for preprocessing the data to obtain corresponding time series data. The feature matrix of the time series data is as follows: Where m and n respectively represent the dimensions of the feature matrix of the EEG signal or EGG signal. m is the total number of electrodes, and n is the total duration of the EEG time series signal or EGG time series signal. A first analysis module, configured to obtain an average electrode inconsistency index of the first channel and the second channel based on a first model according to the time series data; the first model includes: where JSD represents a JS divergence scoring function, t represents time, represents the time series data of the corresponding channels of the first channel and the second channel recorded at the time point t seconds of the subject, m represents the total number of channels, and n represents the total duration of the time series data of the corresponding channels; ICI represents the inconsistency index of the corresponding channel; A prediction module for comparing the average electrode inconsistency index of a subject with a preset evaluation threshold. When the average electrode inconsistency index of the subject is not lower than the preset evaluation threshold, the subject is predicted to be an epilepsy patient. Where the first channel is F3, C4, P4, O2, T7, FC1, CP1, CP6, TP9, TP10, F2, C1, C2, P2, FC3, FC4, CP3, CP4, PO4, P6, PO8, CPz, and POz; the second channel is the body of the stomach, lesser curvature of the stomach, antrum of the stomach, ascending colon, transverse colon, descending colon, and rectal channels.

2. The system according to claim 1, characterized in that, The system further includes a second analysis module for converting the time series data into probability density distribution data.

3. The system according to claim 2, wherein The second analysis module is further configured to obtain an inconsistency index of a corresponding channel based on the second model according to the probability density distribution data, where the second model includes: ; where ICI represents an inconsistency index function, represents a KL divergence function, t represents time, represents the time series data of the corresponding channel recorded by the subject at the time point of t seconds, represents the time series data of the corresponding channel recorded by the healthy population at the time point of t seconds, n is the total duration of the time series data of the corresponding channel, represents the sample distribution, represents the reference distribution.

4. The system according to claim 2, characterized in that, The system further includes a database for storing the data obtained by the data acquisition module, the data preprocessing module, the first analysis module, the second analysis module, and the prediction module, and for generating corresponding sample data sets.

5. The system according to claim 4, wherein The system further includes an evaluation threshold setting module. The evaluation threshold setting module sets the average electrode inconsistency index corresponding to the smallest critical value as the evaluation threshold by randomly obtaining the average electrode inconsistency index of 50 - 70% of the epilepsy patient samples in the sample data set and calculating the critical value of epilepsy occurrence.

6. The system according to claim 1, wherein The acquisition time of the EEG data and the EGG data is at least 5 minutes.

7. The system according to claim 1, wherein The preprocessing includes filtering the EEG data to filter out EEG signals below 1 Hz and above 30 Hz, and normalizing the EGG data and the filtered EEG data to obtain the time series data.

Images