A method for sorting and optimizing brain cognitive function leads under G load based on double feature adaptive fusion

Abstract

The application discloses a G-load based brain cognitive function lead sorting and optimization method based on double-feature adaptive fusion, and relates to the field of aviation medicine and biomedical technology. Gradient G-load is applied by a three-axis manned centrifuge, and multi-channel electroencephalogram signals are collected and preprocessed. Arrangement entropy and TBR parameters of each lead are extracted, and the Spearman correlation coefficient of the parameters and the G-load is calculated. A single-lead comprehensive correlation coefficient is obtained based on the correlation coefficient strength adaptive weighting, and a group final comprehensive correlation coefficient is obtained by fusing the median and the sign consistency. The leads are sorted and optimized according to the absolute value of the coefficient. The application adopts double-feature complementary evaluation, adaptive weight fusion and robust cross-individual fusion, can realize G-load based electroencephalogram lead optimization, reduces the data processing amount, improves the cognitive monitoring accuracy and robustness, and is suitable for dynamic evaluation of brain cognitive function of aerospace personnel.

Description

Technical Field

[0001] This invention relates to the fields of aviation medicine and biomedical technology, and more specifically to a method for sorting and optimizing brain cognitive function leads based on dual-feature adaptive fusion under G-load. Background Technology

[0002] Currently, flight cognitive ability refers to the ability of pilots to organically combine good attention, judgment, decision-making, and spatial orientation abilities, and apply them during flight. It is the ability to comprehensively process various flight information. It is the foundation of flight technical level, a comprehensive reflection of the pilot's psychological qualities, and a core element for successfully completing flight missions and ensuring flight safety. Pilots should possess good perception, attention, memory, spatial orientation, and maneuvering abilities. In the aerospace field, in order to gain a more comprehensive and in-depth understanding of the cognitive state of subjects during flight and under G-load, some scholars have begun to conduct experiments using manned centrifuges to study changes in physiological signals, including electroencephalography (EEG), under G-load. For neuroscientists, EEG is an effective tool for measuring brain activity in the cerebral cortex. As a non-invasive technique for monitoring the brain, it has the characteristics of high temporal resolution and relatively low cost.

[0003] Since EEG directly reflects the electrophysiological activity of the cerebral cortex, researchers attempt to discover cognitive changes in EEG through appropriate variations, and then use these characteristics to predict and evaluate brain cognitive function under G-load. Currently, the commonly used number of electrodes for EEG acquisition includes 16, 32, and 64 leads. EEG measurements are mostly performed under static conditions. To gain a more comprehensive understanding of information from various brain regions, the number of electrodes is trending upwards, potentially expanding to 128, 256, or even 512 leads. However, under dynamic conditions, especially under centrifuge + Gz load, EEG signals are susceptible to various interferences. The installation and fixation of electrodes and lead wires, as well as the installation and fixation of the measuring instrument and its components, require specialized methods. The workload for electrode installation and fixation is relatively large, and the installation process is complex and cumbersome. Furthermore, not all lead data is necessary or reflects changes in cognitive function. Some leads may carry useless or redundant information, and there are also correlations and mutual information between leads. Therefore, selecting an appropriate number of leads can reduce the amount of EEG data processing and improve efficiency. Based on this, it is necessary to identify leads that better reflect the state of brain cognitive function and contribute significantly to changes in cognitive function under centrifuge G-load conditions. Then, based on this, leads can be sorted and simplified. Simplification techniques will have significant practical application value in improving EEG cognitive function measurement techniques and EEG data analysis techniques in this special environment under G-load conditions.

[0004] Entropy and TBR (Theta / beta power ratio) were chosen as the basic analytical indicators of EEG cognitive function because entropy reflects the complexity of neural information processing and is sensitive to changes in cognitive load, while TBR is a classic alertness indicator widely used in aviation fatigue monitoring. These two indicators reflect cognitive function status from different dimensions—time-domain complexity and frequency-domain power balance—and are highly complementary. The two parameters comprehensively assess cognitive neural changes under G-load; entropy reveals changes in information processing patterns, while TBR reflects the regulation of alertness. The dual-parameter approach avoids the limitations of a single indicator, providing a more comprehensive assessment of cognitive state, while also avoiding the computational complexity and redundancy of too many parameters, thus improving analytical efficiency. The permutation entropy parameter has a simple calculation process and high anti-interference capability and good robustness. It only compares adjacent values, thereby reducing computation time. Permutation entropy is used to analyze changes in nonlinear time series, describing signal complexity, and is independent of signal amplitude and phase, exhibiting good noise resistance. Recently, some researchers have applied permutation entropy to the life sciences, proposing a method for separating and reconstructing life signals based on particle swarm optimization-optimized variational mode decomposition (PSO-VMD). Permutation entropy and its multi-scale application are now widely used in biomedicine, mechanical troubleshooting, sleep monitoring, and other fields. The Theta / beta ratio (TBR) is often referred to as an attention index, used to assess the level of concentration during a task experiment; generally, a lower TBR value indicates greater concentration. Some researchers have conducted preliminary analyses of EEG cognitive function under centrifuge G-load, comparing results from the perspectives of EEG time-domain, frequency-domain, and parameter changes. They found that the changes in permutation entropy and TBR parameters are consistent with the changes in EEG time-domain beta waves. Permutation entropy and the energy changes in various brain bands are basically consistent, reflecting the actual changes in the brain well. The changes in TBR and permutation entropy parameters are also largely consistent. TBR can effectively explain cognitive changes under specific tasks from the perspective of focus. When fusing two parameters, traditional multi-parameter methods often use fixed weights or simple averaging, failing to consider the differences in information content among the parameters and being susceptible to extreme values. It is evident that existing technologies suffer from drawbacks such as redundancy in EEG acquisition leads under dynamic G-load conditions, severe signal interference, limitations of single evaluation indicators, reliance on manual parameter fusion settings, and poor robustness in integrating cross-subject data.

[0005] Therefore, how to achieve accurate sorting and optimized optimization of EEG cognitive function leads under G-load, reduce data processing volume, and improve the accuracy and efficiency of cognitive state assessment is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a method for ranking and optimizing brain cognitive function leads based on dual-feature adaptive fusion under G-load, in order to solve the problems existing in the background art.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A method for ranking and optimizing brain cognitive function leads based on dual-feature adaptive fusion under G-load, comprising: S1. A triaxial manned centrifuge was used, and a gradient G load loading curve was set to monitor the physiological signals of the subjects in real time. S2. Collect and store multi-channel EEG signals from the subject under G-load loading; S3. Preprocess the multi-channel EEG signals to obtain preprocessed EEG data; S4. Extract EEG data for a fixed time under different G loads in each lead, and calculate the permutation entropy parameter and TBR parameter; S5. Calculate the correlation coefficient between the load and the corresponding entropy parameter of each lead. r e Correlation coefficient between load and corresponding TBR parameters r t ; S6. Based on the correlation coefficient strength r e , r t Adaptive weighting was performed to obtain the comprehensive correlation coefficient R of each lead for a single subject. S7. The median and sign consistency fusion method was used to calculate the final comprehensive correlation coefficient (RR) for each lead of all subjects. S8. Sort each lead according to the absolute value of the final comprehensive correlation coefficient RR; S9. Select the lead with the highest ranking based on the sorting results to complete the optimization selection.

[0008] Optionally, S1 specifically includes: The specific acceleration curve settings for each run of the triaxial manned centrifuge are as follows: when the centrifuge is stationary, the subject experiences an acceleration of 1G; when the centrifuge starts, it first reaches the baseline at an acceleration rate of 1G / s for several seconds, then reaches the maximum G value set for each run at an acceleration rate of 3G / s for 10-15 seconds, and then decreases to 1G at an acceleration rate of 3G / s, returning to the initial state, and the centrifuge stops; the doctor monitors and records the subject's ear pulse and electrocardiogram signals in real time, and after each run, asks the subject about their subjective visual perception of the surrounding lights and the central light in the cabin, and makes a comprehensive judgment on their endurance based on the subject's facial expressions.

[0009] Optionally, S2 specifically includes: An electroencephalogram (EEG) recorder was installed and fixed inside the cockpit to collect and record EEG signals. When placing the electrodes, a corresponding size of securing mesh cap was used according to the subject's head size. Medical tape was applied to each electrode for tight fixation to prevent loosening during operation. A 16-channel EEG electrode setup was used. Electrode F was taken. p1 F3, C3, P3, O1, F p2 F4, C4, P4, O2, F7, T3, T5, F8, T4, T6; the reference electrode for all electrodes is electrode A1 or A2 located on the same side earlobe, and a unipolar lead measurement method is used; the 16 EEG signals are recorded on an SD card in a specific format.

[0010] Optionally, the formula for calculating the permutation entropy parameter is expressed as:

[0011] In the formula, Represents the permutation in sequence S The probability of its occurrence; It represents the logarithm to the base 2.

[0012] Optionally, the TBR parameter is calculated using the following formula:

[0013] in, for Average power spectral density of the frequency band; for Average power spectral density of the frequency band.

[0014] Optionally, the method for calculating the correlation coefficient in S5 specifically includes: There are two random variables , And on , Sort the elements to get the ranking set of these two elements. , Then the set , Subtracting the corresponding elements in the set yields a ranking difference set. Based on this, random variables , The Spearman rank correlation coefficient between them is determined by , or It is obtained through calculation; the calculation formula is:

[0015] In the formula, n is the number of levels. di This represents the rank differences between two paired variables.

[0016] Optionally, the specific process for calculating the comprehensive correlation coefficient R of each lead for a single subject in S6 is as follows: First, calculate the weights of the permutation entropy parameters. :

[0017] When calculating the weights of the TBR parameters :

[0018] Finally, the comprehensive correlation coefficient R of each lead for a single subject was obtained: Optionally, the specific process for calculating the final composite correlation coefficient (RR) for all subjects in each lead is as follows: Data preparation: There are N subjects, and each subject has M leads with correlation coefficients. For each lead j, calculate the median of the N subjects: Calculate the sign consistency ratio for each lead; The final composite correlation coefficient (RR) for all subjects in each lead was calculated based on the median of N subjects and the sign consistency ratio of each lead.

[0019] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method for ranking and optimizing brain cognitive function leads based on dual-feature adaptive fusion under G-load, which has the following beneficial effects: 1. For the first time, cognitive function-related simplification and optimization selection of EEG leads under G-load were performed; 2. The present invention effectively reduces the randomness and instability of EEG in its design; 3. In the optimization selection of cognitive function leads, this invention selects entropy parameter and TBR as evaluation parameters, realizing the synergistic analysis of G load from two orthogonal dimensions of nonlinear dynamics and spectral energy, overcoming the limitation of one-sided information from a single indicator, and significantly improving the sensitivity and robustness of EEG cognitive function analysis results. 4. In calculating the correlation coefficient between the entropy parameter, TBR parameter, and G load, this invention adopts the Spearman correlation coefficient, which achieves the best balance between robustness, applicability, and practicality.

[0020] 5. The correlation coefficients between the entropy parameter and the TBR parameter and the G-load in this invention. r e and r tAn adaptive weighting method is used for fusion, which automatically assigns weights based on the correlation strength between each parameter and the G-load, avoiding the subjectivity of manually setting thresholds, maximizing the use of effective information, and significantly improving the correlation between the fused parameters and the G-load.

[0021] 6. When fusing the correlation coefficient R of each lead for all subjects, this invention adopts a fusion method based on median and sign consistency to obtain the comprehensive correlation coefficient RR of each lead, effectively eliminating individual differences and outlier interference, ensuring stable and reliable fusion results, and preserving the true physiological relationship between parameters. This significantly improves the accuracy and robustness of analysis of multiple subjects and provides a robust integration scheme for multi-lead group EEG analysis. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the method flow provided by the present invention; Figure 2 The topographic distribution of the correlation coefficients between cognitive function parameters and G-load of the subjects' electroencephalogram leads provided by the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] This invention discloses a method for ranking and optimizing brain cognitive function leads based on dual-feature adaptive fusion under G-load, comprising the following steps: Step 1. Design of Test Operation Mode The main equipment used in the experiment was a novel triaxial high-performance manned centrifuge. The centrifuge's main arm was 8 meters long and featured triaxial acceleration. The acceleration curve for each run was specifically set as follows: when the centrifuge was stationary, the subject experienced an acceleration of 1G. Upon startup, the centrifuge initially accelerated at a rate of 1G / s to reach the baseline for several seconds, then accelerated at a rate of 3G / s to reach the maximum G value set for each run, lasting 10-15 seconds. Afterward, the acceleration rate decreased to 1G at a rate of 3G / s, returning to the initial state, and the centrifuge stopped. Directly exposing the subject to a large G value was dangerous; therefore, the maximum G value for each run started at 2.5G and increased in increments of 0.5G until the subject reached their endurance limit or exhibited signs of needing to stop. Doctors monitored and recorded the subject's ear pulse and electrocardiogram signals in real time. After each run, the subject was asked about their subjective visual perception of the ambient and central lights within the cabin, and their endurance was comprehensively assessed based on their facial expressions.

[0026] The flowchart for the acquisition and processing of electroencephalogram (EEG) signals is as follows: Figure 1 As shown, the EEG signals were acquired and recorded using a portable EEG device. The portable EEG device was installed and fixed inside the centrifuge cabin. The EEG signals first entered an interference box inside the cabin for preliminary interference processing, and then entered an amplification and recording box inside the cabin, where they were recorded in a specific format onto an SD card for further processing. The recorded EEG data underwent data conversion, noise reduction, and filtering preprocessing. Each subject experienced different G-loads. First, the EEG data corresponding to each 10-second G-load period were identified. For the EEG data during the 10-second plateau period of each load, permutation entropy and TBR parameters were calculated to obtain the values ​​of different G-loads and their corresponding entropy and TBR parameters. The correlation coefficients between the load and the entropy parameters and TBR parameters of the corresponding 16 EEG leads were calculated, and correlation statistics tables of G-load and entropy parameters and G-load and TBR parameters were created. An adaptive weighting method was used to weight these two correlation coefficients for each lead to obtain a comprehensive correlation coefficient R. Then, a topographic map distribution of the EEG cognitive function correlation between each subject's 16 leads and the corresponding comprehensive correlation coefficient R of the 16 leads was created. Finally, for all subjects, the final comprehensive correlation coefficient RR for each lead was obtained using the median method proposed in the study. The median method is that RR is equal to the product of the median of the comprehensive correlation coefficients of each subject and the sign consistency. Finally, the obtained RRs of the 16 leads were sorted according to the absolute value of RR, and the top-ranked leads were selected.

[0027] Acquiring EEG signals under dynamic conditions, especially under +Gz loading, is quite challenging, primarily because EEG signals are easily interfered with by various factors such as electromyography (EMG), electrooculography (EOG), body movement, rotation, and power supply, potentially containing numerous artifacts. Therefore, it is crucial to eliminate or minimize these interferences at their source. EMG and body movement significantly impact EEG signals, especially EMG. During G loading, subjects often involuntarily exert force to resist fainting, inevitably generating EMG. Therefore, the experimental design requires subjects to exert minimal force during G loading and to maintain their original posture with their backs pressed firmly against the chair back to minimize the impact of body swaying. Considering that speaking, blinking, and body movement also significantly affect the signal and easily create artifacts, some of which are difficult to remove, the experimenter is required to avoid talking to the subjects during G loading and to minimize blinking and other facial movements. Considering that factors such as thinking and concentration affect the quality of EEG signals, subjects are required to relax their bodies as much as possible, maintain their original state, and avoid intentionally focusing on visible instruments such as dials and dashboards during G loading. Since this invention primarily focuses on EEG changes under the influence of +Gz, the acceleration values ​​in the Gx and Gy directions were controlled to be as small as possible when designing the acceleration curves. In this invention, the actual Gx component was controlled below 1, and the Gy component was controlled below 0.5, achieving the requirements designed at the beginning of the experiment.

[0028] Step 2. Acquisition of multi-channel EEG signals under +Gz action

[0029] Electroencephalogram (EEG) signal measurements are mostly used in static situations, and less frequently in dynamic situations, mainly because the signals are extremely weak and easily interfered with, making dynamic measurements more difficult. If the electrodes are not properly fixed or become loose during the process, the collected data is limited. In this experiment, the centrifuge requires high-speed and triaxial motion, making these requirements even more stringent. Appropriate electrode fixation methods were employed in the experiment. A portable EEG recorder was installed and fixed inside the cabin to collect and record EEG signals. When placing the electrodes, appropriate size fastening caps were fitted according to the different head sizes of the subjects, and medical tape was applied to each electrode for secure fixation to prevent loosening during operation. The EEG electrode placement followed the international 10 / 20 standard, using 16 electrodes, with electrode F... p1 ,F3,C3,P3,O1,F p2The reference electrodes for all electrodes were A1 or A2 located on the same side of the earlobe, and a unipolar lead measurement method was used. In the confined space of the cockpit, the interference box and amplification box were fixed in suitable locations where EEG signals could be easily acquired and recorded under +GHz conditions. The 16 channels of EEG signals were recorded in a specific format on an SD card at a sampling frequency of 128Hz. After secondary conversion by specialized software, they were converted into text files, filtered, and then processed further.

[0030] Step 3. Noise reduction processing of EEG signals

[0031] Because EEG signals are very weak and dynamic EEG signals contain various artifacts and interference, appropriate methods need to be selected for preprocessing to remove these artifacts and improve the performance and effectiveness of data feature extraction. This study uses digital filtering to remove various interferences from EEG signals. Currently, digital filtering technology is widely used and has become one of the most fundamental and important research areas in real-time signal processing. Filtering EEG signals and extracting various rhythms based on a reasonable filter order is a commonly used method in digital filtering. The purpose of a digital filter is to remove noise from the signal to the maximum extent possible by designing an effective filter. When selecting the filter type, the emphasis to be emphasized during filtering and the characteristics of the signal to be filtered should be considered. When selecting the filter order, the minimum order of the filter must first be calculated. As the filter order increases, the resulting waveform will be delayed, and the amplitude-frequency response characteristics within the passband will deteriorate. Therefore, based on the calculated minimum order, the filter order should not be too high.

[0032] The EEG artifacts in this invention mainly include electromyography (EMG), electrooculography (EOG), body movement, rotation, baseline drift, and power supply. When subjects experience +Gz exposure in the cabin, they inevitably experience tension and make certain counter-movements, thus being significantly affected by EMG, primarily in the 35.8-51Hz frequency range. EOG is difficult to remove as it may be mixed within multiple EEG frequency bands, easily causing loss of useful information during removal. Therefore, based on experimental experience, signals below 0.5Hz are primarily removed. The AC power frequency is concentrated around 50Hz. Electrode fixation and baseline drift easily generate low-frequency slow waves below 0.5Hz and 0.2Hz. The interference generated by centrifuge rotation is strong and cannot be ignored. Based on the maximum achievable G value, centrifuge arm radius, and the conversion formula for centripetal acceleration, it can be calculated that the interference of centrifuge rotation on the signal is mainly below 0.5Hz. In addition, power supply, magnetic fields, and body movement also have varying degrees of influence. Based on the above considerations, the lower limit of the filter is set at approximately 0.5Hz, and the upper limit at approximately 35Hz. For EEG signals with a sampling frequency of 128Hz, according to... (8-13Hz) (13-25Hz, (0.5-4Hz), Based on the segmentation standard of (4-8Hz) and considering the filtering out low-frequency interference such as baseline drift, electrode noise, and rotation below 0.5Hz, as well as high-frequency interference such as EMG and power supply concentrated in the 35.8-51Hz band, a second-order Butterworth bandpass filter is designed to perform bandpass filtering on EEG signals from 0.5-35Hz. A second-order Butterworth bandstop filter is used to filter out 50Hz power frequency interference, with a stopband of 49.5Hz-50.5Hz. While filtering out power frequency interference, the useful signal components are preserved to the maximum extent. At the same time, a FLTFILT zero-phase digital filter is used for zero-phase filtering. Its key difference from ordinary filters is that it does not introduce phase delay, avoids phase distortion, keeps the time position of the signal waveform unchanged, and does not delay. This ensures that the signal time is accurate when calculating entropy parameters and TBR parameters, and will not be misaligned or delayed due to filtering.

[0033] The results after filtering show that various digital filtering methods can effectively remove EMG, power frequency, and other high-frequency interference from EEG signals under G load. They also have a significant effect on slow wave interference in the low-frequency band, such as baseline drift, electrode interference, and rotation. The signal-to-noise ratio after denoising is greatly improved.

[0034] Step 4. Calculation of arrangement entropy parameters and TBR parameters under different G loads

[0035] This invention selects entropy parameters and TBR (Theta / beta power ratio) as basic analytical indicators of EEG cognitive function because entropy parameters reflect the complexity of neural information processing and are sensitive to changes in cognitive load, while TBR is a classic indicator of alertness and is widely used in aviation fatigue monitoring. These two indicators reflect cognitive function status from different dimensions—time-domain complexity and frequency-domain power balance—and have good complementarity. For each subject and each lead, 10-second EEG data were selected under different G-load conditions, and the permutation entropy parameter and TBR parameter for the corresponding 10-second EEG data were calculated.

[0036] (1) Permutation Entropy Calculation: The permutation entropy parameter analysis method is a nonlinear method for detecting signal abrupt changes. Permutation entropy can be used to analyze changes in nonlinear time series, describe signal complexity, and is independent of the amplitude and phase of the signal, exhibiting good noise resistance. In a more active state, the EEG signal is more complex and less periodic, reflected in a higher permutation entropy value. Permutation entropy refers to the value of a symbol sequence of length n. In this context, the negative logarithm of the number of distinct permutations is given. The number of permutations refers to the total number of distinct sequences in a sequence of n symbols, i.e., the number of permutations of the sequence. The formula for permutation entropy can be expressed as:

[0037] In the formula, Represents the permutation in sequence S The probability of its occurrence; It represents the logarithm to the base 2; multiplying it by the negative of the probability gives the negative logarithm.

[0038] (2) TBR calculation: The TBR parameter is commonly used to determine the level of concentration in a subject. TBR is the ratio of the average power in the theta band to the average power in the beta band. This ratio reflects the relative intensity of brain electrical activity in different frequency bands, helping to analyze the electrical activity characteristics of the brain in different functional states.

[0039]

[0040] in, for Average power spectral density of the frequency band (μV) 2 / Hz). for Average power spectral density of the frequency band (μV) 2 / Hz); By calculating the corresponding permutation entropy and TBR parameters for 10s EEG data under different G loads for each subject, the entropy parameter values ​​and TBR values ​​for each of the 16 leads of each subject under different G loads can be obtained.

[0041] Step 5. Correlation analysis of entropy parameters and TBR parameters with G-load.

[0042] The correlation analysis between the entropy parameter and the TBR parameter and the G-load is represented by their correlation coefficients, r. The correlation coefficient is a very important statistical indicator of the degree of correlation between two phenomena, usually denoted by r. This invention uses the Spearman method to calculate the correlation coefficients between the entropy parameter and the TBR parameter and the G-load.

[0043] The absolute value |r| reflects the degree of correlation; the closer to 1, the higher the correlation, and the closer to 0, the lower the correlation. r > 0 indicates that the two correlated phenomena change in the same direction; if r < 0, it indicates a negative correlation, meaning the two correlated phenomena change in opposite directions. r = 0 indicates no correlation. Calculation formula: Suppose two random variables , ,right , Sort the elements (either in ascending or descending order) to obtain the ranking sets of these two elements. , Then the set , Subtracting the corresponding elements in the set yields a ranking difference set. Based on this, random variables , The Spearman rank correlation coefficient between them can be obtained from , or It is obtained through calculation. The calculation formula is:

[0044] In the formula, n is the number of levels. d i This represents the rank differences between two paired variables.

[0045] Based on the above formula for calculating the correlation coefficient, and the entropy and TBR parameter values ​​corresponding to different loads, the correlation coefficient between the load and the corresponding entropy parameter of each of the 16 EEG leads was calculated, taking different G loads and their corresponding EEG parameter values ​​as two variables. r e Correlation coefficient between load and corresponding TBR parameters r t .

[0046] Step 6. Entropy parameters, TBR parameters, and correlation coefficients with G-load. r e and r t Adaptive weighting

[0047] The correlation coefficients between the entropy parameter and TBR parameter of each of the 16 leads and the G-load were calculated. r e and r t Adaptive weighting is performed to obtain the correlation coefficient R for each lead. This invention proposes a two-parameter adaptive weighted fusion method for EEG based on relative importance: the weight allocation and the correlation strength of each parameter with the cognitive task are proportional, achieving fully data-driven dynamic fusion.

[0048] Let the entropy parameter weight be... :

[0049] The weights of the TBR parameters are: :

[0050] The correlation coefficient R of each lead after fusion:

[0051] In the experiment, the correlation coefficient R of some leads after fusion was 0. The one with the larger absolute value of the entropy parameter and TBR parameter corresponding to that lead was selected for subsequent calculation and processing.

[0052] The adaptive weighting method in this invention has the following advantages: ① Completely data-driven, with objective and reliable results. The weight allocation is based entirely on the correlation coefficients of the entropy parameter and the TBR parameter, eliminating the need for manually setting thresholds or empirical values ​​and removing subjective biases. The weights are allocated according to the amount of information, and the formula uses the original values ​​of the correlation coefficients re and rt, rather than their absolute values. The direction information of the correlation is preserved during the calculation, ensuring the objectivity and repeatability of the results. ② Dynamically adapt to individual and task differences This method can automatically adapt to the brain response patterns of different subjects and the needs of different cognitive tasks. Although this invention only loads the G-load and does not involve specific cognitive tasks, if subsequent experiments add specific cognitive tasks along with the G-load, this method will be more advantageous, that is, it can automatically adapt to different subjects and different cognitive tasks and generate customized optimal weights for each lead and individual. ③ The calculation is simple and the explanation is intuitive. The weight calculation method and formula are concise and clear. The weights directly reflect the importance of each parameter, and the results are easy to understand and highly interpretable. ④ The results are continuous and smooth, and the robustness is strong. The weighting results change continuously with the correlation coefficient, avoiding the jump problem of calculation based on a threshold. The calculation results are not sensitive to outliers and have good robustness to data fluctuations. ⑤ The method has good scalability This invention employs a two-parameter feature fusion method, which can be easily extended to multi-parameter fusion scenarios. It is computationally efficient and suitable for feature analysis and research of EEG in various environments.

[0053] For each subject's 16 leads, the correlation coefficients between each lead and the corresponding two leads under different G-loads were calculated. r e and r t After calculating the comprehensive correlation coefficient R, statistical tables were created for G-load, EEG entropy parameters, and TBR parameters, as well as for G-load and the two correlation coefficients. r e and r tA statistical table of correlation coefficients R for each lead was created, and a topographic map of the correlation coefficients between G-load and cognitive function leads for each subject was generated based on the table. This more clearly shows the magnitude of the correlation between changes in G-load and cognitive function parameters in different parts of the brain, and observes the patterns of cognitive parameter changes in the 16 EEG leads of different subjects. A cognitive function correlation topographic map was generated for each of the five subjects. The topographic map distribution of the correlation coefficients between cognitive function parameters and G-load in one subject's EEG leads is shown below. Figure 2 As shown in the figure. Based on the correlation coefficient topographic map distribution, the correlation coefficients R of each lead for multiple subjects were integrated to perform subsequent lead sorting and optimization.

[0054] Step 7. When fusing the correlation coefficients R of each lead for all subjects using a fusion method based on median and sign consistency, obtain the composite correlation coefficient RR for each lead.

[0055] To integrate EEG analysis results from multiple subjects, a fusion method based on median and sign consistency was employed to calculate the composite correlation coefficient (RR) for each lead from five subjects. First, the median correlation coefficient for each lead from the five subjects was calculated as an intensity assessment. Then, the sign consistency ratio was calculated to assess the reliability of the results. Finally, the composite correlation coefficient was the product of the median and the sign consistency ratio. This method is both robust against outlier interference and can represent consistent patterns across multiple subjects, providing a robust integration scheme for multi-lead population EEG analysis. Specific calculation process: 1. Data Preparation: There are N subjects, and each subject has M leads with correlation coefficients. ,

[0056] in, R represents the correlation coefficient of the j-th lead for the i-th subject.

[0057] For each lead j, calculate the median of the N subjects:

[0058] 3. Symbol Consistency Assessment

[0059] Calculate the sign consistency ratio for each lead. :

[0060] in, The number of subjects with a positive correlation in lead j. The number of subjects with negative correlation in lead j; 4. Calculate the final composite parameters of each lead.

[0061] The correspondence between SC values ​​and the number of positive and negative correlation coefficients is shown in Table 1. Table 1 Correspondence Table

[0062] In this invention, N is 5 and M is 16. The advantages of using a fusion method based on median and signed consistency are: ① Strong resistance to outliers, robust results Fusion methods based on median and sign consistency are insensitive to extreme values ​​and can effectively resist the interference of abnormal data from individual subjects, ensuring the stability of group results. ② Quantitative identification of consistent patterns yields reliable results. The sign consistency ratio is used, which is the larger of the absolute values ​​of the proportion of positively correlated subjects and the proportion of negatively correlated subjects. This quantifies the consistency of results among multiple subjects, can truly distinguish stable EEG patterns from random fluctuations, and reduces the impact of random errors on the results. ③ Retains directional information of relevance, making the explanation more intuitive. In the calculation of the comprehensive correlation coefficient, the raw value of the median of all subjects is used, that is, the original sign of the median is preserved, so the result can clearly indicate the positive or negative correlation between EEG activity and cognitive load. ④ Highly efficient in calculation and highly applicable The calculation method is relatively simple. The median only requires sorting and comparison, and the sign consistency only requires counting and comparison. The calculation is efficient, suitable for large-scale data analysis of multi-lead and multi-subject, easy to reproduce, and widely applicable.

[0063] Step 8. Perform the final ranking of the 16 leads based on the absolute value of the composite correlation coefficient (RR) for each lead of all subjects.

[0064] Based on the comprehensive correlation coefficient (RR) of all subjects across 16 leads, the absolute value of the RR was calculated. All leads were then ranked according to their absolute values, and the leads with the highest rankings were selected for simplification or optimization. The ranking was based on the absolute value of RR rather than RR itself. The main reason is that after obtaining the comprehensive correlation coefficient, both positive and negative values ​​are observed, indicating that changes in EEG cognitive function state and load changes under +Gz are not entirely consistent. The ranking focuses on the closeness of the cognitive function parameter in a lead to load changes, i.e., the sensitivity of the cognitive function parameter in a lead to load changes, and the magnitude of the cognitive function parameter in a lead relative to load changes, rather than focusing more on the sign of the correlation coefficient. The cognitive function parameters and loadings of the two leads are both very strong, but their signs are opposite. In this case, it cannot be arbitrarily said that the correlation coefficient of the lead with a negative correlation coefficient is weaker than that of the lead with a positive correlation coefficient. That is, when ranking, more consideration should be given to the strength of the correlation between the cognitive function parameters and the load changes in that lead. Under G load, EEG is subject to various interferences such as body movement, electromyography, rotation, and magnetic fields, and has a certain degree of instability and specificity. Therefore, under G load, the instability and specificity of EEG changes need to be effectively eliminated. Therefore, ranking based on the absolute value of the correlation coefficient, rather than the value itself, is more scientific and reasonable.

[0065] Step 9. Ordering and Optimization Selection of Leads under Centrifuge G-Load

[0066] Based on the final result after lead sorting, and according to simplification requirements, 8 leads, 6 leads, or 4 leads are selected as the simplified selection result.

[0067] Based on this method, following its steps and procedures, the 16 leads of EEG data acquired under centrifugation + Gz conditions were sorted and simplified. The final sorting result, in order of ranking, is as follows: O 1, T 6, C 4, F P2 , O 2, C 3, F 4, P 4, T 5, F 8, P 3, T 3, F 7, F 3, T 4, F P1 If simplified to 6 leads, the selected 6 leads are: O 1, T6, C 4, F P2 , O 2, C 3; The selected 8 leads are O 1, T 6, C 4, F P2 , O 2, C 3, F 4, P 4.

[0068] Parameter and result verification of the sorting and optimization selection process: The intrinsic changes of EEG entropy parameters and TBR parameters: Based on the intrinsic values ​​of entropy parameters and TBR parameters under different G-loads, for entropy parameters, under G-load, the intrinsic values ​​of entropy parameters in each EEG lead are relatively large, almost all above 0.8. It is generally considered that an entropy value of 0.7-0.8 is the high-efficiency working zone, and 0.8-0.9 begins to enter the cognitive load limit zone, indicating that brain neural activity under G-load has entered a relatively complex and high-load working mode. This suggests that the changes in entropy parameters under G-load have a certain degree of stability, or relatively strong stability. TBR... The parameters varied considerably, with some differences among the subjects. Most values ​​were below 10, with a few above 10. Generally, a TBR of 1-5 corresponds to a relaxed state, 5-10 to a moderate cognitive load, and above 10 to a high cognitive load and high concentration state. The TBR values ​​in the study were consistent with empirical values, indicating that most subjects were making an effort under G-load and were in a high state of concentration and effort. However, some leads showed insufficient stability, which may be related to the instability and specificity of EEG under G-load due to various interferences such as body movement, electromyography, rotation, and magnetic fields. Correlation coefficients of entropy parameters, TBR parameters, and G loads: 1. According to the judgment criteria of Spearman correlation coefficients: r > 0.5 indicates a strong correlation, 0.3 < r < 0.5 indicates a moderate correlation, r < 0.3 indicates a weak correlation, and r < 0.1 means the correlation is very small. Judging from the values of the correlation coefficients, most are moderately correlated. Even for the same subject, different leads show different intensities, indicating that the electroencephalogram (EEG) self-cognitive function parameters are affected by the loads and have a certain correlation with the loads, but the correlation is not strong or stable. However, individual leads show a strong correlation, with the absolute value of the maximum greater than 0.9, approaching 1, but there are also individuals with a value possibly < 0.1, showing a weak correlation. Overall, the entropy parameters show a positive correlation, and the TBR parameters show a negative correlation. For all subjects, most leads show a consistent negative correlation, but there are also individual leads with inconsistent positive and negative polarities, indicating that their change regularity is not very strong. Analyzing the reasons, because EEG data is easily interfered by various signals under dynamic conditions. Under the action of the centrifuge + Gz, the equipment needs to rotate at high speed and move in three axial directions, and the EEG signal is more vulnerable to various interferences such as electromyogram, electrocardiogram, electrooculogram, body movement, baseline drift, electrode interference, power supply, rotation, etc. The amplitude of individual artifact signals may be several times or even dozens of times that of the EEG signal, which enhances the randomness of the EEG signal and also affects the correlation analysis of EEG cognitive functions, reducing the correlation degree between the changes of EEG cognitive function parameters and the loads or making it unstable.

[0069] Results Verification and Analysis: The selected leads predominantly showed right-side bias. The reasons are twofold: First, under +Gz loading, the centrifugal force places the person in a semi-reclined position with their left arm down. Due to the influence of systemic circulation, the right side of the body is more affected by ischemia than the left, meaning the right side is more affected by G-endurance. This results in a more significant increase in the complexity and activity of the right-side EEG with increasing G-load, while a more pronounced decrease in focus is also observed. This manifests as increased permutation entropy and decreased TBR parameters. Right-side cognitive function leads show relatively stronger changes, and the correlation coefficient between EEG parameters and load is relatively larger. In other words, the overall correlation between changes in cognitive function parameters on the right side of the brain and load is stronger than on the left. Second, it is related to the inherent characteristics of EEG changes. This invention involves sorting and optimizing cognitive function EEG under G-load, with the selected leads predominantly on the right side, consistent with previous research results. Lateralization of brain function response under G-load is an important research topic in aerospace medicine. Previous studies have shown that the right cerebral hemisphere plays a dominant role in spatial orientation, gravity perception, and stress response. Our findings indicate that under G-load, the changes in EEG parameters in the right leads (including the central, occipital, temporal, and frontal lobes) are more significant than in the left, manifested as higher overall sensitivity scores and more pronounced parameter fluctuations. This finding is highly consistent with the reported phenomenon of the right hemisphere being more sensitive to G-load, supporting the right-lateralized response pattern of spatial cognition and stress response under +Gz conditions. The similarity in the selection of cognitive function leads under different conditions suggests that the research results are largely related to the brain's own inherent characteristics, and related literature validates the results of this study.

[0070] Result verification and analysis: The selected 6 leads are O1, T6, C4, F P2 Analysis results: The basic basis of this method is the correlation coefficient between G-load and EEG entropy parameters and the correlation coefficient between G-load and TBR parameters. Based on their fused correlation coefficients and the topographic distribution of the correlation coefficients between the entropy parameters and TBR parameters of the 16 leads, the leads under the load are sorted and optimized to select the required leads. According to the results, the 6 leads with more significant changes are O1, T6, C4, and F. P2The values ​​for O2 and C3 are the highest, with O1 showing a value of -0.572. Leads exhibiting significant changes are mostly distributed in the occipital, temporal, central, and frontal lobes of the brain, primarily located on the right posterior and right anterior sides. The reasons for this are twofold: First, it may be related to the unique semi-reclining position under +Gz influence, where the body is positioned with the left arm down and the right arm up. Under +Gz influence, due to centrifugal force, the body is in a semi-reclining position with the left arm down. Affected by systemic circulation, the right side of the body experiences greater ischemia than the left, resulting in more pronounced changes in the complexity and activity of the EEG on the right side. The right posterior and right anterior leads show stronger changes, and the correlation coefficients between EEG cognitive function and load are relatively higher in the right posterior and right anterior leads. In other words, the overall correlation between changes in EEG cognitive function parameters and load is stronger on the right side of the brain than on the left. Second, it may be related to the inherent characteristics of EEG changes and functional regulation. This invention reveals that the top 6 leads are O1, T6, C4, F... P2 The results for O2 and C3 not only identify key brain regions, but their ranking also aligns closely with recent theories regarding the allocation of neural resources and processing levels under cognitive load. Related literature suggests that leads in the prefrontal cortex, such as the FPI lead, may be related to the modulation of tension under load or higher-level functional activities. The conclusions of this invention indicate that leads showing significant changes are mostly distributed in the left anterior and right posterior sides of the brain, close to areas such as the frontal and temporal regions, which is largely consistent with previous related research findings.

[0071] 1. Top of the List: Occipital Lobe (O1) and Right Temporal Lobe (T6) – Early Processing and Load of Sensory Input: The prominent position of O1 (left occipital lobe), ranking first, strongly supports the recent view that "cognitive load is primarily reflected in the regulatory cost of sensory input." Existing research using magnetoencephalography (MEG) has found that cognitive load in visual tasks modulates the neural population activity of the primary visual cortex (corresponding to O1) earliest and most strongly, manifested as enhanced high-frequency gamma activity and disruption of alpha rhythms; this modulation even precedes behavioral responses. This suggests that high-load tasks require more neural resources for "gain control" from the initial stages of information processing. The inclusion of T6 (right posterior temporal lobe) is particularly crucial. The posterior temporal lobe is a high-level processing hub for visual information flow, responsible for object recognition and spatial processing. Existing research using rapid periodic visual stimulation combined with EEG studies shows that the neural representation intensity of complex visual features in the right temporo-occipital junction (encompassing T6) is significantly reduced due to concurrent memory tasks (cognitive load), reflecting competitive occupation of resources in this area. O1 and T6 ranked first and second, respectively, clearly outlining the core role of the pathway from primary visual processing to advanced visual cognition when bearing cognitive load.

[0072] 2. Involvement of core control nodes: right-hemisphere dominant central and prefrontal cortex (C4, F) P2): C4 (center right) ranked third and fourth, along with F P2 The right prefrontal cortex represents the integration of sensorimotor functions and the incorporation of higher cognitive control nodes. This sequence (sensory processing preceding control) aligns with the processing model of "bottom-up resource demand triggering top-down control." Studies have found that mu / alpha rhythmic inhibition (reflecting sensorimotor readiness) in the right central cortex is correlated with task difficulty, and its temporal dynamics lag behind visual responses in the occipital lobe. Related meta-analyses indicate that high-load tasks involving alertness, attention maintenance, and conflict resolution more consistently activate the right prefrontal and parietal networks. P2 The inclusion of (right forehead) is direct evidence of the involvement of this right hemisphere executive control network. C4 and F P2 The emergence of these lobes may mark a transition in the brain from processing input information (occipital-temporal lobe) to organizing and coordinating responses (central area) and exercising overall control (prefrontal lobe).

[0073] 3. The appearance of symmetrical leads and whole-brain network integration (O2, C3): The appearance of O2 (right occipital) and C3 (left central), ranking fifth and sixth respectively, completes the symmetrical map of brain region involvement. This strongly suggests that when cognitive load reaches a certain level, it triggers extensive mobilization of bilateral networks throughout the brain. The appearance of O2 echoes O1, indicating that visual load is bilateral. The appearance of C3 is symmetrical with C4, possibly reflecting the somatic spatial component of the task involving bilateral sensorimotor coordination or working memory. Existing research confirms that under high cognitive load, tasks that are originally lateralized can induce compensatory activation in the contralateral hemisphere to recruit more neural resources, manifested as an increase in the symmetry of EEG activity. In this invention, C3, as the last central lead to enter the list, may precisely reflect this load-induced transhemispheric resource recruitment phenomenon.

[0074] In summary, the lead ranking order revealed in this invention is not random, but rather depicts the dynamic hierarchical process of cognitive load emerging in the brain: starting at the sensory portal (O1), the load first increases the cost of primary sensory information processing; then it rapidly spreads to higher sensory association areas (T6), indicating that resources for the extraction and representation of complex features are crowded out; then it triggers sensorimotor integration and control (C4, F... P2 The right-hand dominant network is activated to coordinate responses and implement global control, ultimately triggering bilateral mobilization of the entire brain (O2, C3) to recruit a wide range of resources to achieve the task objective, increasing symmetry across the hemispheres. This sequence perfectly echoes the computational theory of "cognitive bottlenecks under parallel processing," where bottlenecks first appear in the information input and processing stages, forcing the control system to intervene and ultimately leading to a reallocation of network resources. The results of this invention, through correlation ranking using a specific method, provide clear and quantifiable electrophysiological evidence for this cutting-edge theory.

[0075] Simplification and Selection of Multi-Lead EEG: The application results of multi-lead EEG are slightly better than those of single-lead EEG. In actual classification, due to the influence of factors such as electrode cap wearing and offset, single-lead detection has certain biases. Multi-lead EEG can provide a global attribute feature of the brain from the perspective of the overall difference in data, and has a certain degree of stability. Therefore, 8 or 6 leads out of 16 can be selected to detect changes in EEG cognitive function under G-load. Based on this, this invention designs a method for ranking and optimizing the selection of EEG cognitive function leads under G-load. First, the correlation coefficient topography of EEG entropy parameters and TBR parameters with load changes is obtained based on the changes in EEG entropy parameters and TBR parameters, and the degree of change in cognitive function of each lead under G-load is comprehensively evaluated. Then, the total correlation coefficient of each lead with load change is obtained through an adaptive weight fusion method. The correlation coefficient of each lead of all subjects is fused using a fusion method based on median and sign consistency to obtain the comprehensive correlation coefficient of each lead. Finally, the leads are ranked according to the absolute value of the comprehensive correlation coefficient, and the leads are simplified based on the ranking results to select the optimized leads.

[0076] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0077] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for ranking and optimizing brain cognitive function leads based on dual-feature adaptive fusion under G-load, characterized in that, include: S1. A triaxial manned centrifuge was used, and a gradient G load loading curve was set to monitor the physiological signals of the subjects in real time. S2. Collect and store multi-channel EEG signals from the subject under G-load loading; S3. Preprocess the multi-channel EEG signals to obtain preprocessed EEG data; S4. Extract EEG data for a fixed time under different G loads in each lead, and calculate the permutation entropy parameter and TBR parameter; S5, calculating a correlation coefficient of the load of each lead and the corresponding entropy parameter r e and the correlation coefficient of the load and the corresponding TBR parameter r t ; S6、based on the correlation coefficient strength r e 、 r t adaptive weighting, to obtain a single-subject lead comprehensive correlation coefficient R; S7. The median and sign consistency fusion method was used to calculate the final comprehensive correlation coefficient (RR) for all subjects in each lead. S8. Sort each lead according to the absolute value of the final comprehensive correlation coefficient RR; S9. Select the lead with the highest ranking based on the sorting results to complete the optimization selection.

2. The method of G-load based dual-feature adaptive fusion for sorting and selection of brain cognitive function channels according to claim 1, characterized in that, S1 specifically includes: The specific acceleration curve settings for each run of the triaxial manned centrifuge are as follows: when the centrifuge is stationary, the subject experiences an acceleration of 1G; when the centrifuge starts, it first reaches the baseline at an acceleration rate of 1G / s for several seconds, then reaches the maximum G value set for each run at an acceleration rate of 3G / s for 10-15 seconds, and then decreases to 1G at an acceleration rate of 3G / s, returning to the initial state, and the centrifuge stops; the doctor monitors and records the subject's ear pulse and electrocardiogram signals in real time, and after each run, asks the subject about their subjective visual perception of the surrounding lights and the central light in the cabin, and makes a comprehensive judgment on their endurance based on the subject's facial expressions.

3. The method of claim 1, wherein the method is characterized by, S2 specifically includes: An electroencephalogram (EEG) recorder was installed and fixed inside the cockpit to collect and record EEG signals. When placing the electrodes, a corresponding size of securing mesh cap was used according to the subject's head size. Medical tape was applied to each electrode for tight fixation to prevent loosening during operation. A 16-channel EEG electrode setup was used. Electrode F was taken. p1 F3, C3, P3, O1, F p2 F4, C4, P4, O2, F7, T3, T5, F8, T4, T6; the reference electrode for all electrodes is electrode A1 or A2 located on the same side earlobe, and a unipolar lead measurement method is used; the 16 EEG signals are recorded on an SD card in a specific format.

4. The method of ranking and preferring the channels for cognitive function under G load based on dual features adaptive fusion according to claim 1, characterized in that, The formula for calculating the permutation entropy parameter is expressed as follows: In the formula, Represents the permutation in sequence S The probability of its occurrence; It represents the logarithm to the base 2.

5. The method of ranking and preferring the channels for cognitive function under G- load based on dual features adaptive fusion according to claim 1, characterized in that, The formula for calculating the TBR parameter is as follows: wherein is the average power spectral density of the frequency band; is the average power spectral density of the frequency band.

6. The method for ranking and optimizing brain cognitive function leads based on dual-feature adaptive fusion under G-load as described in claim 1, characterized in that, The method for calculating the correlation coefficient in S5 specifically includes: There are two random variables , And on , Sort the elements to get the ranking sets of these two elements. , Then the set , Subtracting the corresponding elements in the set yields a ranking difference set. Based on this, random variables , The Spearman rank correlation coefficient between them is determined by , or It is obtained through calculation; the calculation formula is: where n is the number of ranks, d i is the difference in ranks for the pair of variables.

7. The method of ranking and preferring the channels for cognitive function under G- load based on dual features adaptive fusion according to claim 1, characterized in that, The specific process for calculating the comprehensive correlation coefficient R of each lead for a single subject in S6 is as follows: The permutation entropy parameter weight is calculated first as : In calculating the TBR parameter weight is : Finally, the comprehensive correlation coefficient R of each lead for a single subject was obtained: 。 8. The method of ranking and preferring the channels for cognitive function under G- load based on dual features adaptive fusion according to claim 1, characterized in that, The specific process for calculating the final composite correlation coefficient (RR) for all subjects in each lead is as follows: Data preparation: There are N subjects, and each subject has M leads with correlation coefficients. For each lead j, calculate the median of the N subjects: Calculate the sign consistency ratio for each lead; The final composite correlation coefficient (RR) for all subjects in each lead was calculated based on the median of N subjects and the sign consistency ratio of each lead.

Images