A method and system for automatically identifying background activity based on resting-state electroencephalogram

Abstract

The application discloses a background activity automatic identification method and system based on resting state electroencephalogram (EEG), which automatically identifies the EEG signals of a patient in a closed-eye state in different regions, extracts EEG features of each region and each frequency, analyzes and compares the EEG features by referring to a standardized range, and finally outputs the identification and analysis results as a visual report, so as to facilitate quick manual verification, online editing and modification when necessary, realize the efficiency of EEG automatic interpretation, guarantee the accuracy of result analysis, provide a possibility of interpreting high-dimensional data information from original two-dimensional EEG efficiently and quickly, and provide a potential technical means for non-EEG professionals who are not familiar with EEG data analysis and processing. The application can realize the automation of the identification process, improve the identification efficiency, thereby reducing heavy and repetitive mechanical labor, facilitating medical staff to analyze and judge the result report, improving the work efficiency, and saving the labor cost.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of resting state electroencephalogram analysis and recognition, and particularly relates to a resting state electroencephalogram based background activity automatic recognition method and system. BACKGROUND

[0002] Electroencephalogram (EEG) is a non-invasive, non-invasive neural electrical signal acquisition technology, which can reflect the electrical activity of the cluster point of brain neurons at a very high time precision. Compared with time-consuming and high-cost detection methods, it has the advantages of high time resolution, short examination time, low cost, convenience and easy operation. In recent years, electroencephalogram has become a powerful tool for evaluating changes in brain function. With the development of brain function imaging technology, electroencephalogram signals are closely related to nervous system diseases such as epilepsy, cerebrovascular disease, insomnia, anxiety and depression, Parkinson's disease, etc. Through energy and brain connection analysis at the electrode level and source level, high-dimensional electroencephalogram features can be extracted for the auxiliary diagnosis, efficacy prediction and prognosis of some diseases.

[0003] Limitations of manual processing of electroencephalogram signals: The interpretation of the background activity of electroencephalogram signals requires several years of professional training and certain clinical experience. The feature extraction of electroencephalogram signals requires familiarity with the software operation process of Matlab and EEGlab, making it difficult to widely popularize electroencephalogram analysis and interpretation. The one-key automatic processing and analysis system of the present application solves the problem of user limitation.

[0004] Time-consuming and labor-intensive manual recognition: Electroencephalogram is most commonly used in clinical applications to assist in the diagnosis of epilepsy. After collecting the electroencephalogram of the patient, the background activity and epileptic waves of the electroencephalogram are identified by manual recognition. Epileptic waves have good characteristics and are easy to identify, but the identification of background activity often requires a lot of manpower.

[0005] Manual recognition of electroencephalogram requires electroencephalogram specialists to carefully identify and check each 10-second data. Generally, 8 hours of electroencephalogram recording takes several hours, and requires experienced electroencephalogram physicians who have been trained for a long time. The large amount of electroencephalogram data requires a long processing time, and the requirements for hardware and operating personnel are relatively high. The current manual EEG processing mode limits the efficiency of data analysis.

[0006] Manual recognition has subjectivity and inconsistency: Since there is a lack of objective quantitative indicators for EEG results, the judgment is greatly influenced by the individual experience level and comprehensive analysis ability of the doctor and technician. The understanding of the diagnostic criteria by the reviewer is not consistent, often with varying degrees of subjectivity and bias, resulting in inconsistency in the judgment of EEG results. A domestic report showed that the consistency rate of EEG result judgment by high-seniority doctors with more than 10 years of experience in EEG was 73%, and the consistency rate between high-seniority doctors and low-seniority doctors with less than 5 years of experience in the profession was 64%.

[0007] The machine learning classification recognition based on the electroencephalogram signal features has high efficiency, but lacks unified specific feature standards and has poor generalization ability: the original electroencephalogram is subjected to time-frequency conversion, feature extraction, and training combined with a machine learning model, so that the recognition efficiency is high, but the accuracy, specificity and generalization ability need large sample verification, and the recognition result is greatly affected by the sample for model training.

[0008] Therefore, the prior art has defects and needs to be improved. SUMMARY

[0009] The purpose of the present application is to overcome the deficiencies of the prior art and provide an automatic background activity recognition method and system based on resting-state electroencephalogram.

[0010] The technical solution of the present application is as follows: an automatic background activity recognition method based on resting-state electroencephalogram is provided, comprising the following steps:

[0011] Step 1: using an electroencephalogram signal acquisition device to obtain the resting-state electroencephalogram rsEEG signal of a patient;

[0012] Step 2: using the ARTIST full-automatic denoising method to preprocess the resting-state electroencephalogram rsEEG signal data of the patient in a closed-eye state for 3 minutes;

[0013] Step 3: calculating and extracting the average time proportion, average amplitude and bilateral amplitude difference of each frequency band and different brain regions;

[0014] Step 4: analyzing and interpreting the extracted electroencephalogram features, and sequentially discriminating in the conditions of severe background abnormality, moderate background abnormality, mild background abnormality, borderline electroencephalogram and normal electroencephalogram features;

[0015] Step 5: automatically generating a visual report according to the discrimination result, presenting the conclusion of electroencephalogram background activity recognition, and the quantitative electroencephalogram features of each frequency band and different brain regions and their reference ranges;

[0016] Step 6: manually quickly verifying the visual report generated in step 5, and modifying and editing the report content with deviation if necessary.

[0017] Further, the specific steps of step 2 are as follows:

[0018] Step 2.1: let the patient keep a closed-eye state for a period of time, and select 3 minutes of resting-state electroencephalogram rsEEG signal data from the collected signals;

[0019] Step 2.2: removing the direct current drift in the resting-state electroencephalogram rsEEG signal data;

[0020] Step 2.3: Remove eye movement interference in the selected signal data of the closed eye state;

[0021] Step 2.4: Reduce the sampling rate to 250 Hz and adjust the band-pass filter to 1-45 Hz, and replace the bad channels;

[0022] Step 2.5: Segment the data, each segment has a duration of 2s;

[0023] Step 2.6: Remove the bad channels after segmentation and interpolate;

[0024] Step 2.7: After analyzing the independent components, remove the false components and use the average reference value.

[0025] Further, the specific steps of step 3 are as follows:

[0026] Step 3.1: Divide the whole brain into four regions, wherein the first region includes the frontal and anterior temporal regions, the second region includes the central and middle temporal regions, the third region includes the parietal region, and the fourth region includes the occipital and posterior temporal regions;

[0027] Step 3.2: According to the calculation method of frequency, identify the signals of different frequencies of each channel of electroencephalogram over time, divide the frequency into five frequency bands, namely α wave, β wave, γ wave, δ wave and θ wave, and mark the identified different frequency waves with different colors, then the time length ratio of each frequency band to the total signal time length is the average time ratio of each frequency band;

[0028] Step 3.3: Calculate the peak-to-peak value of each identified brain wave and save it as the amplitude of each wave. The average peak-to-peak value of all brain waves in each frequency band is the average amplitude of this frequency band;

[0029] Step 3.4: Take the midline of the brain as the boundary, divide the average amplitude difference of a certain frequency band in the left and right hemispheres by the low value of the average amplitude of the same frequency band in the left and right hemispheres, reflect the symmetry of the left and right hemispheres, and thus obtain the bilateral amplitude difference;

[0030] Step 3.5: Extract the average time ratio, average amplitude, and bilateral amplitude difference electroencephalogram features of each region in each frequency band according to steps 3.2-3.4.

[0031] Further, the diagnosis criteria of normal electroencephalogram in step 4 are as follows:

[0032] ① The average time ratio of δ wave in each region is 0-2%, and the average amplitude is 0-50μV;

[0033] ② The average time ratio of θ wave in each region is 0-15%, and the average amplitude is 0-50μV;

[0034] ③ The average time proportion of alpha wave in the first and second regions is 40-100%, the average time proportion in the third region is 45-100%, and the average time proportion in the fourth region is 50-100%, and the average amplitude in each region is 0-100μV;

[0035] ④ The average time proportion of beta wave in each region is 0-40%, and the average amplitude is 0-20μV;

[0036] ⑤ The bilateral amplitude difference of each frequency band in the first and second regions is 0-30%, the bilateral amplitude difference in the third region is 0-40%, and the bilateral amplitude difference in the fourth region is 0-100%.

[0037] Further, the diagnostic criteria of the borderline electroencephalogram in step 4 are as follows:

[0038] Any of the following abnormal manifestations is a borderline electroencephalogram:

[0039] ① The average time proportion of delta wave in each region is 2-5%, and the average amplitude is >50μV;

[0040] ② The average time proportion of theta wave in each region is 15-30%, and the average amplitude is >50μV;

[0041] ③ The average time proportion of alpha wave in the first and second regions is 10-40%, the average time proportion in the third region is 10-45%, and the average time proportion in the fourth region is 10-50%, and the average amplitude in each region is >100μV;

[0042] ④ The average time proportion of beta wave in each region is 40-100%, and the average amplitude is 20-50μV;

[0043] ⑤ The bilateral amplitude difference of each frequency band in the first and second regions is 30-50%, the bilateral amplitude difference in the third region is 40-60%, and the bilateral amplitude difference in the fourth region is 100-150%.

[0044] Further, the diagnostic criteria of the mild background abnormality in step 4 are as follows:

[0045] Any of the following abnormal manifestations is a mild background abnormality:

[0046] ① The average time proportion of delta wave in each region is 5-10%;

[0047] ② The average time proportion of theta wave in each region is 30-75%;

[0048] ③ The average time proportion of alpha wave in each region is 5-10%;

[0049] IV. The average amplitude of beta waves in each region is > 50 μV;

[0050] V. The bilateral amplitude difference of each frequency band in the first and second regions is 50-75%, in the third region is 60-85%, and in the fourth region is 150-200%.

[0051] Further, the diagnostic criteria for moderate background abnormalities in step 4 are as follows:

[0052] Any of the following abnormal manifestations is a moderate background abnormality:

[0053] I. The average time proportion of delta waves in each region is 10-95%;

[0054] II. The average time proportion of theta waves in each region is 75-95%;

[0055] III. The average time proportion of alpha waves in each region is 0-5%;

[0056] IV. The bilateral amplitude difference of each frequency band in the first and second regions is > 75%, in the third region is > 85%, and in the fourth region is > 200%.

[0057] Further, the diagnostic criteria for severe background abnormalities in step 4 are as follows:

[0058] Any of the following abnormal manifestations is a severe background abnormality:

[0059] I. The average time proportion of delta waves in each region is 95-100%;

[0060] II. The average time proportion of theta waves in each region is 95-100%.

[0061] The present application also provides an automatic background activity recognition system, comprising: a data management module, a data storage module, and a data processing and analysis module connected to each other, the data management module comprising: a data editing and modification module and an information input module, the data processing and analysis module comprising: a preprocessing module, a frequency recognition and color identification module, a brain region division module, a feature extraction and calculation module, a comparison module, and a visualization module, the data management module, the preprocessing module, the frequency recognition and color identification module, and the brain region division module being connected to the feature extraction and calculation module, the feature extraction and calculation module being connected to the comparison module, the comparison module being connected to the visualization module, the visualization module being connected to the data storage module, and the data storage module being connected to the data editing and modification module.

[0062] Further, the data management module is a computer group, the data storage module is a disk array storage, and the data processing and analysis module is a processor.

[0063] With the above scheme, the present application extracts the brain electrical characteristics of different brain regions and different frequencies by system automation, including average time proportion, average amplitude, bilateral amplitude difference, and analyzes and compares the brain electrical characteristics by referring to the standardized range, and finally outputs the visual report to realize the classification of brain electrical background activity. Subsequently, manual rapid verification can be performed, and online modification and editing can be performed if necessary, which not only meets the efficiency of automatic identification of the system, but also guarantees the accuracy of the result analysis, and also increases the flexibility of the application of the system, thereby providing the possibility of efficiently extracting high-dimensional data information from original two-dimensional brain electrical activity, and providing potential brain electrical analysis technical means for non-brain electrical professional physicians and scientific researchers. Compared with the traditional manual identification processing method, the present application realizes the automation of the brain electrical background activity identification process, separates the electroencephalogram physician from the tedious and time-consuming background activity identification and description work, so as to invest in report result auditing and depth interpretation of paroxysmal abnormal activity, improve work efficiency, and save labor cost. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 The flow chart of the present application.

[0065] Figure 2 The brain electrical partition diagram.

[0066] Figure 3 The color marking identification diagram of different frequencies of brain electrical signals.

[0067] Figure 4 The average time proportion, average amplitude and bilateral amplitude difference of each brain region and each frequency band wave and its reference range.

[0068] Figure 5 The automatic brain electrical report classification reference standard table.

[0069] Figure 6 The system connection diagram of the present application. DETAILED DESCRIPTION

[0070] The present application will be described in detail below in combination with the drawings and specific embodiments.

[0071] Please refer to Figure 1 The present application provides a background activity automatic identification method based on resting state electroencephalogram, comprising the following steps:

[0072] Step 1: Use the electroencephalogram signal acquisition device to obtain the resting state electroencephalogram rsEEG signal of the patient.

[0073] Step 2: The resting state electroencephalogram rsEEG signal data of the patient in the closed eye state for 3 minutes is pretreated by using the ARTIST full-automatic denoising method. The specific steps are as follows:

[0074] Step 2.1: Keep the patient in a closed-eye state for a period of time, and select 3 minutes of resting-state electroencephalogram rsEEG signal data from the collected signal;

[0075] Step 2.2: Remove the direct current drift in the resting-state electroencephalogram rsEEG signal data;

[0076] Step 2.3: Remove the eye movement interference in the selected signal data in the closed-eye state;

[0077] Step 2.4: Reduce the sampling rate to 250 Hz, and adjust the band-pass filter to 1-45 Hz, and replace the bad channels;

[0078] Step 2.5: Segment the data, each segment is 2s long;

[0079] Step 2.6: Remove the bad channels after segmentation and interpolate;

[0080] Step 2.7: After analyzing the independent components, remove the false components, and use the average reference value.

[0081] Select 3 minutes of resting-state electroencephalogram rsEEG signal data of the patient in a closed-eye state, so as to ensure the stability of the signal data, and facilitate subsequent identification. By using the ARTIST full-automatic denoising method, the noise interference in the selected signal data is removed, so as to improve the identification accuracy of the signal data.

[0082] Step 3: Extract the average time proportion, average amplitude and bilateral amplitude difference of each frequency band in different brain regions. The specific steps are as follows:

[0083] Step 3.1: Please refer to Figure 2 Divide the whole brain into four regions, wherein the first region includes the frontal and anterior temporal regions, the second region includes the central and middle temporal regions, the third region includes the parietal region, and the fourth region includes the occipital and posterior temporal regions. The frontal region channel includes left Fchan (F7, F3) and right Fchan (F8, F4). The anterior temporal region includes FT7 and FT8. The central region channel includes left Cchan (C3) and right Cchan (C4). The middle temporal region includes T7 and T8. The parietal region channel includes left Pchan (P3) and right Pchan (P4). The occipital region channel includes left Ochan (O1) and right Ochan (O2). The posterior head includes the occipital region, and the anterior head includes the other regions mentioned above except the occipital region.

[0084] Step 3.2: Please refer to Figure 3According to the calculation method of frequency, the signals of different frequencies of each channel of electroencephalogram over time are identified, the frequencies are divided into five frequency bands, which are alpha wave, beta wave, gamma wave, delta wave and theta wave, and the identified waves of different frequencies are marked with different colors, and then the time length ratio of the waves in each frequency band to the total time length is the average time ratio of each frequency band. Among them, the frequency range of delta wave is 1-4Hz, the frequency range of theta wave is 4-8Hz, the frequency range of alpha wave is 8-13Hz, the frequency range of beta wave is 13-30Hz, and the frequency range of gamma wave is 30-45Hz. The wave with a frequency greater than 45Hz is not included in the calculation. Frequency refers to the number of times the same cycle waveform repeats in one second, or one wave occupies a fraction of one second, with the unit of Hz or cycle / second (c / s). The frequency range of scalp EEG analysis is usually 0.1-100Hz, and the EEG frequency is divided into five frequency bands: delta wave, theta wave, alpha wave, beta wave and gamma wave. The measurement of frequency refers to the time length from any wave trough to the next wave trough, or from any wave peak to the next wave peak.

[0085] Specifically, on the preprocessed time series signal, the frequency band range of each visible wave is calculated: the code automatically retrieves the time occupied by each wave, i.e. the time length from trough-peak-trough. Among them, the trough is defined as the first appearing local minimum, and the peak is the maximum value appearing thereafter, when the next trough appears, and the amplitude of the falling branch is greater than 1 / 2 of the rising branch, it is judged as an electroencephalogram wave. When the time occupied by the electroencephalogram wave is 0.25s-1s, it is marked as delta wave. When the time occupied by the electroencephalogram wave is between 0.125s and 0.25s, it is marked as theta wave. Similarly, 0.076s-0.125s is alpha wave, 33ms-76ms is beta wave, and 22ms-33ms is gamma wave. The wave with a time less than 22ms is not included in the calculation range. The identified waves of different frequencies are marked with different colors. Then, the time length of the waves in each frequency band is compared with the total time length of the signal, i.e. the average time ratio of each frequency band.

[0086] Step 3.3: Calculate the peak-peak value of each identified electroencephalogram wave and save it as the amplitude of each wave. The average peak-peak value of all electroencephalogram waves in each frequency band is the average amplitude of this frequency band.

[0087] Amplitude refers to the voltage used to describe brain waves, measured in microvolts (μV) (1 μV = 10-6V) as the potential difference between any two electrodes. The height of this voltage is determined by the amplifier scaling voltage, and the voltage value can be determined by the height (mm) of the brain wave.

[0088] The peak-to-peak value of each identified brain wave is calculated and saved as the amplitude of each wave. The average peak-to-peak value of all brain waves in each frequency band is the average amplitude of this frequency band. For example, if 3 theta waves are identified in 10s, with peak-to-peak values of 20μV, 25μV, and 30μV, respectively, the average amplitude of the theta wave frequency band is 25μV.

[0089] Step 3.4: Divide the difference in average amplitude of a certain frequency band in the left and right hemispheres by the lower value of the average amplitude of the same frequency band in the left and right hemispheres, reflecting the symmetry of the left and right hemispheres, i.e. bilateral amplitude difference.

[0090] Step 3.5: Please refer to Figure 4 According to steps 3.2-3.4, the average time proportion, average amplitude, and bilateral amplitude difference of each region in each frequency band are extracted.

[0091] Step 4: Please refer to Figure 5 The extracted EEG features are analyzed and interpreted, and are successively distinguished in the conditions of severe background abnormalities, moderate background abnormalities, mild background abnormalities, borderline EEG, and normal EEG.

[0092] This system is based on adult EEG diagnosis reference standard, and takes the average time proportion, average amplitude, and bilateral amplitude difference as the core parameters for automatic EEG interpretation. First, it interprets whether the collected patient's EEG is normal, if not, it successively interprets whether it meets the conditions of severe background abnormalities, moderate background abnormalities, mild background abnormalities, and borderline EEG.

[0093] Further, the diagnostic criteria for normal EEG in step 4 are as follows:

[0094] ① The average time proportion of δ wave in each region is 0-2%, and the average amplitude is 0-50μV;

[0095] ② The average time proportion of θ wave in each region is 0-15%, and the average amplitude is 0-50μV;

[0096] ③ The average time proportion of α wave in the first and second regions is 40-100%, in the third region is 45-100%, and in the fourth region is 50-100%, and the average amplitude in each region is 0-100μV;

[0097] ④ The average time proportion of β wave in each region is 0-40%, and the average amplitude is 0-20μV;

[0098] The bilateral amplitude difference of each frequency band in the first region and the second region is 0-30%, the bilateral amplitude difference in the third region is 0-40%, and the bilateral amplitude difference in the fourth region is 0-100%. Further, the diagnostic criteria of the borderline electroencephalogram in step 4 are as follows:

[0099] The borderline electroencephalogram has any of the following abnormal manifestations:

[0100] ① The average time proportion of δ wave in each region is 2-5%, and the average amplitude is >50μV;

[0101] ② The average time proportion of θ wave in each region is 15-30%, and the average amplitude is >50μV;

[0102] ③ The average time proportion of α wave in the first region and the second region is 10-40%, the average time proportion in the third region is 10-45%, the average time proportion in the fourth region is 10-50%, and the average amplitude in each region is >100μV;

[0103] ④ The average time proportion of β wave in each region is 40-100%, and the average amplitude is 20-50μV;

[0104] ⑤ The bilateral amplitude difference of each frequency band in the first region and the second region is 30-50%, the bilateral amplitude difference in the third region is 40-60%, and the bilateral amplitude difference in the fourth region is 100-150%. Further, the diagnostic criteria of the mild background abnormality in step 4 are as follows:

[0105] The mild background abnormality has any of the following abnormal manifestations:

[0106] ① The average time proportion of δ wave in each region is 5-10%;

[0107] ② The average time proportion of θ wave in each region is 30-75%;

[0108] ③ The average time proportion of α wave in each region is 5-10%;

[0109] ④ The average amplitude of β wave in each region is >50μV;

[0110] ⑤ The bilateral amplitude difference of each frequency band in the first region and the second region is 50-75%, the bilateral amplitude difference in the third region is 60-85%, and the bilateral amplitude difference in the fourth region is 150-200%. Further, the diagnostic criteria of the moderate background abnormality in step 4 are as follows:

[0111] The moderate background abnormality has any of the following abnormal manifestations:

[0112] ① The average time proportion of δ wave in each region is 10-95%;

[0113] ②θ wave in each region of the average time ratio of 75-95%;

[0114] ③α wave in each region of the average time ratio of 0-5%;

[0115] ④ each frequency band in the first region and the second region of the bilateral amplitude difference is > 75%, in the third region of the bilateral amplitude difference is > 85%, in the fourth region of the bilateral amplitude difference is > 200%.

[0116] Further, the diagnosis criteria of severe background abnormalities in step 4 are:

[0117] Any one of the following abnormal manifestations is a severe background abnormality:

[0118] ①δ wave in each region of the average time ratio of 95-100%;

[0119] ②θ wave in each region of the average time ratio of 95-100%.

[0120] Step 5: automatically generate a visual report according to the judgment result, present the conclusion of the recognition of the background activity of electroencephalogram, and the quantitative electroencephalogram features of different brain regions and their reference ranges;

[0121] This system only classifies and identifies adult EEG background abnormalities, but does not include paroxysmal abnormalities (epileptiform discharges). Epileptiform discharges need to be described by professional personnel in artificial verification of local paroxysmal abnormal activity. Compare the extracted values with the quantitative reference ranges of the classification and identification of electroencephalogram background activity and classify them: (1) normal electroencephalogram, which needs to meet all the values listed in the table; (2) others, which only need to meet any one of the values listed in the table. Then automatically generate a visual report according to the calculation and comparison, classify and identify the electroencephalogram background activity, and describe the report content in detail. The individualized electroencephalogram features calculated automatically are displayed in digital form, and the normal reference values are listed. Red represents an increase in normal value, and blue represents a decrease in normal value.

[0122] Step 6: according to the visual report generated in step 5, artificial rapid verification is carried out, and necessary modifications and edits are made to the report content with deviations. The electroencephalogram technician carries out rapid artificial review according to the visual report result, and can make online modifications and adjustments to the report content if necessary, or can describe the paroxysmal abnormal activity in detail, which increases the flexibility and convenience of the application of this system.

[0123] According to the experiment, compared with the artificial recognition result, in 124 abnormal electroencephalogram reports, 108 cases were recognized as abnormal by artificial recognition, and 124 cases were recognized as abnormal by system recognition; in 170 patients with normal electroencephalogram reported by artificial, 170 cases were recognized as normal by artificial recognition, and 169 cases were recognized as normal by system recognition, and then the reason for not recognizing 1 case as normal was mainly due to imperfect eye movement artifact preprocessing, and after reprocessing, 100% was recognized as normal.

[0124] Sensitivity (SEN): True positive rate, the percentage of actual abnormal patients diagnosed, and the calculation formula is as follows:

[0125] Artificial recognition: (The number of patients diagnosed as positive by artificial recognition ÷ the total number of actual positive patients) * 100% = (108 ÷ 124) * 100% = 87.10%.

[0126] Automatic recognition: (The number of patients diagnosed as positive by automatic recognition ÷ the total number of actual positive patients) * 100% = (124 ÷ 124) * 100% = 100%.

[0127] Specificity (SPE): True negative rate, the percentage of actual disease-free patients correctly identified as disease-free according to the diagnostic criteria, and the calculation formula is as follows:

[0128] Artificial recognition: (The number of patients diagnosed as negative by artificial recognition ÷ the total number of actual negative patients) * 100% = (170 ÷ 170) * 100% = 100%.

[0129] Automatic recognition: (The number of patients diagnosed as negative by automatic recognition ÷ the total number of actual negative patients) * 100% = (169 ÷ 170) * 100% = 99.4%.

[0130] Balanced accuracy (BAC): Considering the influence of the skewness of the number of positive and negative samples in the signal on the accuracy, and the calculation formula is as follows:

[0131] Artificial recognition: 1 / 2 * (87.10% + 100%) * 100% = 93.6%.

[0132] Automatic recognition: 1 / 2 * (100% + 99.4%) * 100% = 99.7%.

[0133] From the above data, the automatic analysis result of the system is generally good in the classification of normal and abnormal electroencephalogram, and has good sensitivity, specificity and balanced accuracy. At the same time, the above data proves that the automatic recognition method provided by the present application has obvious improvement in the accuracy of automatic recognition compared with artificial recognition, and confirms the effectiveness of the automatic recognition method provided by the present application.

[0134] Please refer to Figure 6The application further provides a background activity automatic identification system, comprising a data management module, a data storage module and a data processing and analysis module connected with each other, wherein the data management module comprises a data editing and modifying module and an information input module, and the data processing and analysis module comprises a preprocessing module, a frequency identification and color identification module, a brain region division module, a feature extraction and calculation module, a comparison module and a visualization module.

[0135] The data management module is a computer group, the data storage module is a disk array memory, and the data processing and analysis module is a processor.

[0136] The information input module can facilitate medical staff to check patient information and edit basic information of the patient, such as an outpatient number, a name, a disease type and the like. If a new patient is added, the information input module can also be used to input relevant information of the patient.

[0137] The data processing and analysis module can realize processing of electroencephalogram data. The preprocessing module can realize preprocessing of resting-state electroencephalogram rsEEG signal data, remove interference fragments in the signal data and improve the accuracy of identification. The frequency identification and color identification module can identify and mark different frequency signals of each channel of electroencephalogram over time. The brain region division module can divide the whole electroencephalogram into four brain regions. The output results of the preprocessing module, the frequency identification and color identification module and the brain region division module are taken as input data of the feature extraction and calculation module, so as to extract and calculate three electroencephalogram features of average time proportion, average amplitude and bilateral amplitude difference of each frequency band in different brain regions. Then, the output results of the feature extraction and calculation module are input into the comparison module and then sent to the visualization module, so as to output the comparison and identification results as a visualization report and store the visualization report in the data storage module. Medical staff can check the visualization report stored in the data storage module through the computer group, and can also quickly manually check the report through the data editing and modifying module and edit and modify the report content when necessary to improve the accuracy and authority of the report.

[0138] To sum up, the present application realizes the classification of the EEG background activity by calculating and extracting the EEG characteristics of different brain regions and different frequencies automatically, including the average time proportion, the average amplitude, the bilateral amplitude difference, and the analysis and comparison of the EEG characteristics by referring to the standardized range, and finally outputting the visual report. Then, the rapid manual verification can be carried out, and the online modification and editing can be carried out if necessary. While meeting the efficiency of the automatic recognition of the system, the accuracy of the result analysis is ensured, and the flexibility of the application of the system is increased, so that the possibility of efficiently extracting high-dimensional data information from the original two-dimensional EEG is provided, and the potential EEG analysis technical means are provided for non-EEG professional doctors and scientific researchers. Compared with the traditional manual recognition processing mode, the present application realizes the automation of the EEG background activity recognition process, separates the electroencephalogram doctors from the tedious and time-consuming background activity recognition and description work, so as to invest in the report result auditing and the depth interpretation of the paroxysmal abnormal activity, improve the work efficiency, and save the labor cost.

[0139] The above is only a preferred embodiment of the present application, and is not used to limit the present application. Any modification, equivalent replacement and improvement within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for automatically identifying background activity based on resting-state electroencephalogram, characterized in that, The method comprises the following steps: Step 1: acquiring the resting-state electroencephalogram (rsEEG) signal of a patient by using an electroencephalogram signal acquisition device; Step 2: preprocessing the rsEEG signal data of the patient in a closed-eye state for 3 minutes by using an ARTIST full-automatic denoising method; the specific steps are as follows: Step 2.1: let the patient keep the closed-eye state for a period of time, and select 3 minutes of rsEEG signal data from the collected signals; Step 2.2: remove the direct current drift in the rsEEG signal data; Step 2.3: remove the eye movement interference in the selected signal data in the closed-eye state; Step 2.4: reduce the sampling rate to 250 Hz, adjust the band-pass filter to 1-45 Hz, and replace the bad channels; Step 2.5: segment the data, and each segment has a length of 2 seconds; Step 2.6: remove the bad channels after segmentation and perform interpolation; Step 2.7: after analyzing the independent components, remove the false components, and use the average reference value; Step 3: calculating and extracting the average time ratio, average amplitude and bilateral amplitude difference of each frequency band and different brain regions; The specific steps are as follows: Step 3.1: dividing the whole brain into four regions, wherein the first region includes the frontal and anterior temporal regions, the second region includes the central and middle temporal regions, the third region includes the parietal region, and the fourth region includes the occipital and posterior temporal regions; Step 3.2: according to the calculation method of frequency, identifying the different frequency signals of each channel of the electroencephalogram appearing over time, dividing the frequency into five frequency bands, namely alpha wave, beta wave, gamma wave, delta wave and theta wave, marking the identified different frequency waves with different colors, and then calculating the time length ratio of each frequency band to the total signal length, that is, the average time ratio of each frequency band; Step 3.3: calculating the peak-peak value of each identified electroencephalogram wave and saving it as the amplitude of each wave, and the average peak-peak value of all electroencephalogram waves in each frequency band is the average amplitude of this frequency band; Step 3.4: taking the midline of the brain as the boundary, dividing the left and right hemispheres, and dividing the average amplitude difference of a certain frequency band in the left and right hemispheres by the lower value of the average amplitude of the same frequency band in the left and right hemispheres to reflect the symmetry of the left and right hemispheres, thereby obtaining the bilateral amplitude difference; Step 3.5: extracting the average time ratio, average amplitude and bilateral amplitude difference of each region in each frequency band according to steps 3.2-3.4; Step 4: analyzing and interpreting the extracted electroencephalogram features, and sequentially discriminating in the conditions of severe background abnormalities, moderate background abnormalities, mild background abnormalities, borderline electroencephalogram and normal electroencephalogram features; Step 5: automatically generating a visual report according to the discrimination result, presenting the conclusion of electroencephalogram background activity recognition, and the quantitative electroencephalogram features of different brain regions in each frequency band and the reference range thereof; Step 6: manually quickly verifying the visual report generated in step 5, and modifying and editing the report content with deviation if necessary.

2. The method for automatic recognition of background activity based on resting state electroencephalogram according to claim 1, characterized in that, The diagnostic criteria for normal electroencephalogram in step 4 are as follows: ① The average time ratio of delta wave in each region is 0-2%, and the average amplitude is 0-50 μV; ②θ wave in each area of the average time ratio for 0-15%, the average amplitude of 0-50μV; ③α wave in the first area and the second area of the average time ratio for 40-100%, in the third area of the average time ratio for 45-100%, in the fourth area of the average time ratio for 50-100%, each area average amplitude of 0-100μV; ④β wave in each area of the average time ratio for 0-40%, the average amplitude of 0-20μV; ⑤ each frequency band in the first area and the second area of the bilateral amplitude difference for 0-30%, in the third area of the bilateral amplitude difference for 0-40%, in the fourth area of the bilateral amplitude difference for 0-100%. 3.The method of claim 1, wherein, The diagnostic criteria of the step 4 limit electroencephalogram are as follows: Any one of the following abnormal performance is limit electroencephalogram: ①δ wave in each area of the average time ratio for 2-5%, the average amplitude of >50μV; ②θ wave in each area of the average time ratio for 15-30%, the average amplitude of >50μV; ③α wave in the first area and the second area of the average time ratio for 10-40%, in the third area of the average time ratio for 10-45%, in the fourth area of the average time ratio for 10-50%, each area average amplitude of >100μV; ④β wave in each area of the average time ratio for 40-100%, the average amplitude of 20-50μV; ⑤ each frequency band in the first area and the second area of the bilateral amplitude difference for 30-50%, in the third area of the bilateral amplitude difference for 40-60%, in the fourth area of the bilateral amplitude difference for 100-150%. 4.The method of claim 1, wherein, The diagnostic criteria of the step 4 mild background abnormality are as follows: Any one of the following abnormal performance is mild background abnormality: ①δ wave in each area of the average time ratio for 5-10%; ②θ wave in each area of the average time ratio for 30-75%; ③α wave in each area of the average time ratio for 5-10%; ④β wave in each area of the average amplitude of >50μV; ⑤ each frequency band in the first area and the second area of the bilateral amplitude difference for 50-75%, in the third area of the bilateral amplitude difference for 60-85%, in the fourth area of the bilateral amplitude difference for 150-200%.

5. The method of claim 1, wherein the method further comprises: The diagnostic criteria of the step 4 moderate background abnormality are as follows: Any one of the following abnormal performance is moderate background abnormality: ①δ wave in each area of the average time ratio for 10-95%; ②θ wave in each area of the average time ratio for 75-95%; ③α wave in each area of the average time ratio for 0-5%; ④ each frequency band in the first area and the second area of the bilateral amplitude difference for >75%, in the third area of the bilateral amplitude difference for >85%, in the fourth area of the bilateral amplitude difference for >200%. 6.The method of claim 1, wherein, The diagnostic criteria of the step 4 severe background abnormality are as follows: Any one of the following abnormal performance is severe background abnormality: ①δ wave in each area of the average time ratio for 95-100%; ②θ wave in each area of the average time ratio for 95-100%.

Images