A system for assessing the level of consciousness of an injured person

Abstract

The application discloses a kind of to the consciousness level evaluation system for wounded person, it is related to personnel consciousness evaluation grading technical field, including the acquisition module for obtaining the electroencephalogram signal and the photoplethysmogram signal of two channels of brain of wounded person, the digitization processing module of corresponding module electric signal digitization conversion, feature extraction module extracts time domain feature, frequency domain feature and entropy from electroencephalogram digital signal, from the ratio of pulse wave digital signal, calculate blood oxygen saturation and calculate blood perfusion index, consciousness evaluation module passes through fusion feature vector input consciousness state assessment model, output continuous consciousness state score, and with multiple threshold mapping obtains consciousness grade.The system of the application carries out time domain, frequency domain and nonlinear analysis respectively in the feature extraction of electroencephalogram signal and pulse wave signal, and blood oxygen and blood perfusion degree calculation, realizes the quantitative description and grading determination to the consciousness level of wounded person.

Description

Technical Field

[0001] This invention relates to the field of personnel consciousness assessment and grading technology, specifically a consciousness level assessment system for injured personnel. Background Technology

[0002] Current methods for assessing consciousness, such as electroencephalography (EEG), event-related potentials (ERPs), or functional brain imaging (FBI), involve applying external stimuli to the subject and identifying brain activation patterns to determine whether cognitive processing is occurring in seemingly unconscious patients. However, for injured individuals with some level of motor function, or those who can occasionally cooperate or exhibit some voluntary responses, activation of specific brain regions under specific stimuli is insufficient to differentiate levels of consciousness.

[0003] For injured individuals who can occasionally cooperate and have spontaneous reactions, the problem is that the brain's activation pattern may be a passive reflex or sensory processing, rather than a specific determination of whether there will be a conscious reaction next. For example, for two injured individuals with different spontaneous reaction abilities, under the same experimental variables, the amplitude of the event-related potentials detected are similar. Analysis of activation patterns and intensity alone cannot distinguish this difference in consciousness.

[0004] Considering that for injured persons with some mobility, consciousness is not only about whether there is a response in brain electrical activity, but also whether the brain's response can cause physiological coordination, the current brain neuro-consciousness assessment lacks the ability to detect whether it is accompanied by a real physiological coupling response, so as to achieve the determination and dynamic assessment of the level of consciousness of injured persons with physiological response capabilities. Summary of the Invention

[0005] 1) Technical problems to be solved

[0006] This invention provides a system for assessing the level of consciousness of wounded soldiers, which can simultaneously collect, process and analyze the multimodal physiological signals of wounded soldiers, and output a consciousness state score and its level.

[0007] (ii) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a system for assessing the level of consciousness of wounded soldiers, comprising:

[0009] The acquisition module is used to acquire the electroencephalogram (EEG) signals and dual-channel photoplethysmography (PPG) signals from the brain of the injured person, and output the corresponding analog electrical signal sequence. The EEG signals are acquired by electrodes to collect the temporal potential of brain electrical activity, reflecting the state of group firing of neurons in the cerebral cortex. The dual-channel PPG signals are recorded as blood volume pulsation curves under different optical paths, which are used to calculate the ratio of oxyhemoglobin to deoxyhemoglobin.

[0010] The digital processing module performs preprocessing on each type of analog electrical signal sequence, including filtering, noise reduction, and time synchronization, and then converts and outputs EEG digital signals and pulse wave digital signals, respectively.

[0011] The feature extraction module extracts time-domain features, frequency-domain features, and entropy values ​​from the EEG digital signal. It calculates blood oxygen saturation based on the ratio of the dual-channel pulse wave digital signals and calculates the blood perfusion index based on the pulse waveform integral and amplitude change. After normalizing the multi-domain features of the EEG digital signal with the blood oxygen saturation and blood perfusion index, a fused feature vector is formed.

[0012] The consciousness assessment module receives the fused feature vector and inputs it into a preset consciousness state assessment model. The consciousness state assessment model identifies the input signal pattern through trained feature weights and outputs a consciousness state score. The consciousness state score is then compared sequentially with a set multi-level threshold to obtain the corresponding consciousness level.

[0013] In a feasible embodiment, the acquisition module receives potential changes caused by the discharge of neurons in the cerebral cortex of the injured person through electrodes placed on the surface of the injured person's scalp. The amplification circuit outputs a continuous voltage sequence as an analog signal sequence of the electroencephalogram (EEG) signal, which is used to record the electrical activity state of the brain in the time dimension.

[0014] In a feasible embodiment, the acquisition module sets up two optical paths with different wavelengths and records the changes in light absorption intensity caused by the change in blood volume with the cardiac cycle under the two optical paths with different wavelengths, thereby obtaining the pulsation waveform sequence of the corresponding channel, which serves as the analog electrical signal sequence of the dual-channel photoplethysmography pulse wave signal.

[0015] In a feasible embodiment, the digitization processing module is used to preprocess and digitize the analog electrical signal sequence output by the acquisition module; firstly, the EEG potential sequence and the photoplethysmography pulse wave sequence are bandpass filtered respectively to retain the effective components within the set target frequency band and suppress power frequency interference and high frequency noise;

[0016] The analog electrical signal sequence was then denoised by wavelet decomposition to eliminate motion artifacts and environmental noise components.

[0017] After filtering and denoising, the digitization module establishes a unified time reference for the EEG signal and the dual-channel pulse wave signal, performs time synchronization processing on each analog electrical signal sequence, and converts the preprocessed analog signal sequence into discrete digital signals through an analog-to-digital converter circuit, forming EEG digital signals and pulse wave digital signals respectively.

[0018] In a feasible embodiment, the feature extraction module extracts features from the EEG digital signal specifically as follows:

[0019] In the time domain, the feature extraction module processes the EEG digital signal in segments, calculates the mean and variance of each segment, and quantifies the potential fluctuation amplitude; it extracts the instantaneous potential change range by calculating the difference between peak values; and it uses the autocorrelation function to calculate the correlation of the EEG digital signal under different lags, reflecting the periodic change characteristics of the signal.

[0020] In the frequency domain, the feature extraction module performs a Fourier transform on the segmented signal to obtain the amplitude and power spectral density of each frequency component, and then determines the frequency based on the spectral boundaries. , , and Frequency bands are used to calculate the power of each band and its proportion of the total power; approximate entropy is calculated by constructing a delay vector sequence from the EEG signal.

[0021] In a feasible embodiment, for the dual-channel pulse wave digital signal, blood oxygen saturation is calculated by separating the AC component and the DC component and calculating the ratio of different wavelength channels.

[0022] The blood perfusion index is obtained by integrating and calculating the amplitude change of the single-cycle waveform of the pulse wave digital signal.

[0023] In a feasible embodiment, during the fusion stage of the feature extraction module, the time-domain features, frequency-domain features, and approximate entropy of the EEG digital signal, as well as the blood oxygen saturation and blood flow perfusion index of the pulse wave digital signal, are normalized, and the various features are combined into a multi-dimensional fusion feature vector in index order.

[0024] In a feasible embodiment, the consciousness assessment module receives a fused feature vector as input and inputs it into a preset consciousness state assessment model; the consciousness state assessment model adopts a multi-layer structure, and the fused feature vector is sequentially input into each processing layer;

[0025] The input layer of the consciousness state assessment model is responsible for receiving multi-dimensional fused features and performing normalization and linear transformation to ensure uniform feature scale.

[0026] Then, the input features are nonlinearly mapped and pattern learned through several hidden layers. Activation functions are used between the hidden layers to achieve nonlinear transformation, and the contribution of each feature in the output is adjusted through a weight matrix.

[0027] In a feasible embodiment, the output layer of the consciousness state assessment model generates a continuous consciousness state score, which is obtained by combining and weighting the hidden layer features to reflect the comprehensive state of EEG and hemodynamic features. During the determination process, the consciousness assessment module compares the consciousness state score with a set multi-level threshold in sequence, and maps the threshold range where the score falls to different consciousness levels. This mapping maintains a one-to-one correspondence between the score and the level, and the output consciousness level is used to quantify the consciousness level of the injured person.

[0028] (iii) Beneficial effects:

[0029] Compared with the prior art, this invention has the following beneficial effects:

[0030] The system of this invention simultaneously acquires electroencephalogram (EEG) signals and dual-channel photoplethysmography (PPG) signals, and performs feature extraction and fusion of the two types of signals. By making full use of neural electrical activity and hemodynamic information, the system enables the assessment of consciousness level to cover multiple physiological dimensions, thereby enhancing the accuracy and reliability of the judgment.

[0031] The system performs time-domain, frequency-domain, and nonlinear analyses, as well as calculates blood oxygenation and blood perfusion, in the feature extraction of EEG and pulse wave signals to form a structured fusion feature vector. The fusion feature vector is input into the consciousness state assessment model to output a continuous consciousness state score, which is then mapped with multi-level thresholds to obtain the consciousness level, thereby realizing the quantitative description and grading of the patient's consciousness level. Attached Figure Description

[0032] Figure 1 A schematic diagram of a system for assessing the level of consciousness of wounded soldiers, provided in an embodiment of the present invention;

[0033] Figure 2 A schematic diagram illustrating the sequence of each operational process in a system for assessing the level of consciousness of wounded soldiers, provided as an embodiment of the present invention.

[0034] Figure 3 This is a schematic diagram illustrating the process from collecting consciousness-related analysis data to outputting a consciousness rating and grading model in a consciousness assessment system for wounded soldiers, provided in an embodiment of the present invention.

[0035] Figure 4 This is a schematic diagram illustrating the collection of corresponding EEG and PPG data from a wounded person in a system for assessing the level of consciousness of a wounded person provided in an embodiment of the present invention.

[0036] 100. Data Acquisition Module; 200. Digital Processing Module; 300. Feature Extraction Module; 400. Consciousness Assessment Module. Detailed Implementation

[0037] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0038] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0039] Furthermore, the terms "first" and "second" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0040] It should be noted that, where there is no conflict, the features in the embodiments of the present invention can be combined with each other.

[0041] Combination Figures 1 to 4 This system demonstrates a consciousness level assessment system for injured personnel. It performs time-domain, frequency-domain, and nonlinear analyses on features extracted from electroencephalogram (EEG) and pulse wave signals, and calculates blood oxygenation and perfusion, forming a structured fusion feature vector. The assessment model employs a multi-layered structure for nonlinear mapping and weighted integration of features, improving the ability to recognize different signal patterns and ensuring the stability and consistency of the output results. This system can simultaneously acquire, rapidly process, and analyze the multimodal physiological signals of injured personnel, and instantly output consciousness level scores and grades, providing medical personnel with directly usable consciousness assessment results to support rapid on-site judgment and treatment.

[0042] Specifically, the main function of the acquisition module 100 is to acquire the electroencephalogram (EEG) signal and the dual-channel photoplethysmography (PPG) signal from the brain of the injured person, and output the corresponding analog electrical signal sequence to provide the raw data basis for subsequent digital processing and consciousness assessment.

[0043] Regarding the acquisition of electroencephalogram (EEG) signals, in some embodiments of the present invention, the acquisition module 100 acquires signals by placing electrodes on the surface of the injured person's scalp, such as... Figure 4 As shown, it receives the weak potential changes generated by the collective firing of neurons in the cerebral cortex.

[0044] It should be noted that the amplitude of the EEG signals acquired by the EEG electrodes is usually in the microvolt range. Therefore, it needs to be amplified by a preamplifier circuit to form a continuous voltage sequence, and the output is an analog electrical signal sequence of the EEG signal.

[0045] During signal acquisition, in order to ensure the integrity and accuracy of the signal, the electrode layout needs to cover key brain regions such as the frontal lobe, parietal lobe, and occipital lobe, and appropriate reference electrodes and grounding electrodes should be used to reduce interference and noise.

[0046] It is understandable that the EEG signals collected in this way can reflect the brain's electrical activity state over time, capture the dynamic changes in neuronal population firing patterns, and provide basic information for analyzing neural function and state of consciousness.

[0047] For example, when an injured person is conscious, their brain signals are... The frequency band will exhibit a clear rhythm, while in a state of decreased consciousness or coma, the low frequency... The wave may be enhanced, and these changes can be captured in real time through continuous acquisition.

[0048] In embodiments of the present invention, a high degree of coupling is considered between brain neural activity (reflected by EEG) and blood oxygenation (reflected by PPG). This is because when neuronal populations fire more intensely, their energy demand also increases, and the body automatically increases blood flow to supply oxygen through local vasodilation. Conversely, when conscious activity begins to decline, metabolic demand also decreases, at which point a decrease in blood perfusion can be detected.

[0049] Therefore, it is understandable that this synergistic relationship between nerves and blood flow is the most direct physiological response chain of brain activity. Thus, for the type of injured person targeted by the embodiments of the present invention—who is not in a complete coma but whose level of consciousness is unstable or has partially recovered—it is difficult to comprehensively assess the degree of consciousness recovery based solely on neurological indicators such as EEG when such persons have a certain degree of mobility. However, by further introducing signals reflecting physiological metabolic state, such as blood oxygen and perfusion index, and analyzing the coupling relationship between them and EEG activity, it is possible to determine whether neural activity is accompanied by a real systemic physiological response. By revealing the dynamic correlation between nerves, blood flow, and metabolism, the level of consciousness of the injured person can be more accurately classified, and the degree of recovery can be judged.

[0050] In the embodiments of the present invention, for the acquisition of pulse wave signals, the acquisition module 100 forms a dual-channel optical path by setting two light sources of different wavelengths and a photodetector, and simultaneously records the changes in light absorption intensity caused by changes in blood volume with the cardiac cycle.

[0051] This setup is because different wavelengths of light correspond to the absorption characteristics of oxyhemoglobin (HbO2) and deoxyhemoglobin (Hb) in the blood, respectively. By measuring the changes in light intensity in the two channels, blood oxygen saturation (SPO2) and blood perfusion index (PI) can be calculated.

[0052] Specifically, in some embodiments of the present invention, a light source illuminates the forehead or other acquisition area by emitting light of a specific wavelength, and a photodetector receives the light signal after absorption by tissue and blood. With the heartbeat, the periodic changes in blood volume cause changes in light absorption intensity, forming a pulsating waveform sequence. The dual-channel design allows for the measurement of oxyhemoglobin in one channel and deoxyhemoglobin in the other, thereby enabling ratio calculation and enhancing the accuracy of blood oxygen saturation measurement.

[0053] Understandably, the dual-channel PPG signals acquired in this way can reflect changes in the hemodynamics of the injured person. For example, the blood perfusion index is relatively stable at rest, while the amplitude and waveform of the pulsatile waveform change significantly when consciousness is impaired or blood flow is restricted. These signals provide important hemodynamic support for assessing consciousness status.

[0054] In summary, the analog electrical signal sequence ultimately output by the acquisition module 100 includes a continuous electroencephalogram (EEG) signal sequence and a dual-channel pulse wave (PPG) signal sequence. The EEG signal provides temporal information on neural electrical activity, while the PPG signal provides information on blood oxygenation and blood perfusion. Both types of signals provide a synchronous, raw data foundation for the subsequent digital processing, feature extraction, and consciousness assessment modules 400 of the system.

[0055] The main function of the digitization processing module 200 in the system is to preprocess and digitize the analog electrical signal sequences of the EEG signals and dual-channel photoplethysmography (PPG) signals output by the acquisition module 100, in order to generate digital signals that can be used for feature extraction and consciousness assessment. The technical implementation of this module includes three core components: filtering, denoising and time synchronization, and analog-to-digital conversion.

[0056] Specifically, in the filtering stage, the digital processing module 200 applies bandpass filters to both the EEG signal and the pulse wave signal. For the EEG signal, the bandpass filter is set within the typical effective frequency range of EEG (e.g., 0.5Hz-50Hz) to preserve the main frequency components of the EEG signal and suppress high-frequency noise and power frequency interference.

[0057] For pulse wave signals, bandpass filters can select frequency bands based on cardiac cycle characteristics to filter out low-frequency drift and high-frequency noise. The filtered signal retains the effective characteristics of the original physiological waveform, which is convenient for subsequent feature calculations.

[0058] After filtering, the digitization processing module 200 further denoises the signal. Common methods include, but are not limited to, wavelet decomposition, which divides the signal into sub-bands of different scales, removes noise components through thresholding, and then reconstructs the signal.

[0059] For electroencephalogram (EEG) signals, this processing can eliminate the influence of poor electrode contact, electromyography (EMG) interference, or external environmental noise on the signal; for pulse wave signals, denoising can reduce motion artifacts and optical path interference, making the waveform smoother and more stable. After denoising, the key fluctuation characteristics and pulsation patterns of the signal are preserved, while reducing analysis errors.

[0060] To ensure the correspondence of multimodal signals, the module establishes a unified time reference for EEG signals and dual-channel pulse wave signals. In some embodiments of the present invention, the signal sequences are aligned by calibrating the sampling clock and interpolating to ensure that EEG and pulse wave signals at the same time point correspond.

[0061] It is important to understand that time synchronization is crucial for fusion feature extraction. For example, when analyzing the relationship between specific EEG rhythms and hemodynamic changes, ensuring that features are aligned in the time dimension can improve the accuracy of consciousness state judgment.

[0062] After filtering, denoising, and time synchronization, the analog signal sequence is converted into a discrete digital signal by an analog-to-digital converter (ADC). The EEG signal and pulse wave signal are converted into EEG digital signal and pulse wave digital signal, respectively. In this way, the digital signal can be directly input into a computer or embedded processing module for subsequent feature extraction and model evaluation.

[0063] In summary, the digitization module 200 can output stable and reliable multimodal digital signals. Specifically, the output EEG digital signal retains the original potential fluctuations and frequency domain characteristics, reducing environmental and electromyographic interference; the output pulse wave digital signal retains cardiac cycle characteristics and pulse amplitude changes, enhancing the accuracy of blood oxygenation and perfusion analysis. The time synchronization processing ensures the consistency of characteristics of different signals at the same time point, supporting multimodal fusion analysis.

[0064] For example, when a wounded person is in a state of diminished consciousness, low-frequency EEG signals... While the pulse wave amplitude may increase, the pulse wave signal amplitude may decrease. These changes can be fully captured by the signal output synchronously by the digital processing module 200, providing high-quality data support for feature extraction and consciousness assessment.

[0065] Regarding the feature extraction module 300 in the system, its core function is to extract multi-dimensional features from EEG digital signals and dual-channel pulse wave digital signals, and then normalize and fuse them to form a fused feature vector that can be used for consciousness state assessment.

[0066] As mentioned above, for injured persons with a certain level of consciousness, their electroencephalogram (EEG) signals reflect neuronal discharge, while blood oxygenation and perfusion reflect the local metabolic state of the brain. In the embodiments of the present invention, by constructing the coupling relationship between these two, it can also be understood as determining whether EEG activity is accompanied by a physiological response and quantifying the temporal and characteristic correspondence between the two.

[0067] Regarding the choice of dimensions for coupling these two signals, in the embodiments of this invention, coupling is performed on the dimensions of time and features. Specifically, firstly, in the time domain, the presence of synchronous changes or delayed correspondences between EEG signals and blood oxygenation perfusion signals is analyzed. The choice of the time domain is also due to its ability to intuitively demonstrate whether the two types of signals are synchronized in time and whether the delay in blood oxygenation response caused by changes in EEG can be detected.

[0068] Regarding the coupling in the feature domain, this is to consider not only the temporal correspondence between the two types of signals, but also to analyze the correlation between signal features (the frequency domain below). By extracting a deeper relationship between EEG and blood oxygen perfusion, rather than just the original amplitude correspondence of the signals, we can reflect the blood flow (oxygen metabolism) patterns corresponding to different EEG patterns.

[0069] In summary, the coupling in the time domain focuses on the synchronization and lag of signals, while the coupling in the feature domain focuses on the correspondence of patterns. Only by combining the two can we fully characterize the dynamic coordination of the nervous, blood flow, and metabolic systems, thereby achieving a deep quantitative analysis of the level of consciousness of the injured.

[0070] Specifically, for time-domain extraction in EEG, time-domain features are used to quantify the potential changes of EEG signals over time. Extraction methods include, but are not limited to, extracting the mean, numerical variance, peak difference, and autocorrelation function. The mean is the average of the potentials in each time segment of the EEG digital signal, reflecting the overall potential level of the signal; the variance is calculated by the potential variance of each signal segment, representing the magnitude of fluctuation; the peak value and peak-to-peak difference are extracted by the maximum, minimum, and their difference in a short period of time to quantify the instantaneous change amplitude; in some embodiments, the autocorrelation function reflects the periodicity and repetitiveness of the signal by calculating the correlation of signals under different lags.

[0071] For example, when conscious, the mean of EEG signals is close to zero and the variance is moderate; when consciousness declines, low-frequency waves increase, the variance may increase, and peak-to-peak fluctuations intensify.

[0072] Frequency domain features are used to analyze the frequency components and energy distribution of EEG signals. Extraction methods include, but are not limited to, performing Fast Fourier Transform (FFT) on segmented EEG signals to obtain the power spectral density (PSD), followed by segmentation based on spectral boundaries. Band (0.5-4Hz), band (4-8Hz), band (8-13Hz), Frequency bands (13-30Hz) are also included. By calculating the power of each frequency band and its proportion of the total power, indices such as the spectral centroid, bandwidth, and energy entropy can also be calculated.

[0073] Continuing with the example above, in a conscious state, High proportion of wavebands; when consciousness declines The increased proportion of wavebands and the decrease in energy entropy indicate a reduction in the complexity of brain electrical activity.

[0074] In terms of nonlinear analysis, the approximate entropy of EEG signals is calculated by constructing delay vector sequences, quantifying signal complexity and regularity, and capturing nonlinear dynamic characteristics in neural activity. These characteristics can show significant differences as the level of consciousness of the injured person changes; for example, low-frequency components increase and signal complexity decreases when consciousness declines.

[0075] For dual-channel pulse wave digital signals, the system first separates the direct current (DC) component and the alternating current (AC) component. The oxygen saturation (SpO2) can then be calculated using the AC / DC ratio of the two different wavelength channels, with the general formula:

[0076] ;

[0077] This blood oxygen saturation index reflects the proportion of oxyhemoglobin in the total hemoglobin in the blood.

[0078] Then, the perfusion index (PI) is calculated using the integral and amplitude change of the single-cycle waveform to quantify the local hemodynamic status. It should be noted that the calculations of blood oxygen saturation and perfusion index can be directly calculated using their general formulas, and no specific limitations are made here.

[0079] These indicators can capture changes in blood circulation in injured individuals under different states of consciousness, such as decreased pulse wave amplitude or SpO2 in cases of restricted blood flow or low blood oxygenation.

[0080] During the fusion phase, the feature analysis module normalizes the time-domain, frequency-domain, and nonlinear features of the EEG signal with the SpO2 and PI of the pulse wave signal, so that features of different dimensions can be directly combined.

[0081] After normalization, the various features are arranged in index order to form a multi-dimensional fusion feature vector, providing a complete input data in a unified format for the consciousness assessment model.

[0082] One point that needs further explanation is that EEG signals are acquired by collecting potential changes on the scalp surface through electrodes. These potentials reflect the firing state of neuronal populations in the underlying cerebral cortex. As mentioned above, each electrode reflects the neural activity of a specific brain region; therefore, EEG signals are signals with clear spatial localization attributes. In contrast, the use of dual-channel photoplethysmography (PPG) signals, acquired at dual wavelengths (red and infrared light), to calculate the oxygen / deoxygenated hemoglobin ratio and thus obtain blood oxygen saturation, along with waveform integrals or amplitude changes, reflects the local perfusion index, i.e., the degree of blood perfusion. These signals are detected in peripheral areas of the injured person's body, such as the fingers, wrists, earlobes, or forehead. The acquisition of blood oxygen and perfusion signals reflects changes in local microcirculatory blood volume over time, essentially a hemodynamic response. Unlike EEG signals, these are systemic circulatory system parameters.

[0083] In some embodiments of this invention, the focus is on the level of consciousness of injured individuals. This is a systematic indicator, not limited to a specific brain region. The acquisition of blood oxygenation and perfusion signals can be at different locations than that of electroencephalogram (EEG) signals. For example, electrodes can be placed in the frontal and parietal lobes to acquire EEG signals, while blood oxygenation and perfusion signals can be acquired at the fingertips or earlobes of the injured individual. This still allows for the establishment of coupling relationships in the time and feature domains. In other embodiments of this invention, the acquisition sites for blood oxygenation and perfusion signals can be set at the same locations as those for EEG acquisition. However, it is important to understand that regardless of whether the acquisition sites for various signals are consistent, the practical application of coupling EEG signals and dual-channel PPG signals is to capture and determine whether brain activity effectively drives physiological responses, thereby further quantifying the level of consciousness, i.e., the effectiveness of consciousness.

[0084] This fusion method can simultaneously reflect neural electrical activity and hemodynamic state, improving the accuracy and reliability of consciousness level assessment. For easier understanding, Table 1 below shows the corresponding extracted features for each signal.

[0085] Table 1. Corresponding Extracted Features of Various Signals

[0086]

[0087] In summary, it is understandable that EEG characteristics can capture changes in the consciousness state of injured individuals, for example... Increased wave intensity corresponds to decreased consciousness; frequency domain features and entropy values ​​can reflect the activity and complexity of EEG signals. Blood oxygen perfusion features provide hemodynamic information; decreased blood perfusion or reduced SpO2 can indicate abnormal circulatory function or impaired consciousness. Multimodal fusion feature vectors combine neural activity with hemodynamic information, providing a more comprehensive and reliable input for consciousness assessment models, thereby improving assessment accuracy.

[0088] Finally, there is the consciousness assessment module 400 in the system, which is used to receive multimodal fusion feature vectors, map the feature information into quantifiable consciousness scores through a preset consciousness state assessment model, and finally convert them into level outputs to determine the level of consciousness of the wounded.

[0089] The module receives a fused feature vector, which includes the time domain, frequency domain and nonlinear features of the EEG signal, as well as the blood oxygen saturation (SPO2) and blood flow perfusion index (PI) of the pulse wave. In order to ensure that the different feature dimensions are consistent, the input layer normalizes the fused features and maps each feature to the same numerical range (such as 0-1).

[0090] Simultaneously, linear transformations can be performed to adjust the feature scale, adapting the input features to subsequent hidden layer processing. Standardized inputs can prevent certain large numerical features from dominating the model, ensuring that each type of feature can effectively participate in consciousness state assessment.

[0091] Regarding the consciousness state assessment model constructed in this module, the entire assessment model can be viewed as a dual-modal deep neural network, namely a fusion network of EEG and PPG, and its specific architecture is as follows: Figure 3 As shown, the first layer is the input layer, which receives data including EEG feature input and pulse wave blood oxygenation feature input. The EEG feature input includes time domain (mean, variance, peak-to-peak value, autocorrelation), frequency domain (power spectral density), and nonlinearity (approximate entropy). The blood oxygenation feature input includes blood oxygen saturation (SpO2) and blood flow perfusion index (PI). Both sets of features are input after being standardized.

[0092] The input layer data then enters a dual-branch feature encoding layer, which can be understood as the EEG branch and the PPG branch. In some embodiments of the present invention, the encoding layer of the EEG branch uses a one-dimensional convolutional neural network (1D-CNN) to extract local patterns in the time series, and then models the temporal evolution using LSTM. The encoding layer of the PPG branch extracts dynamic trends in blood oxygenation and perfusion fluctuation features through processes including but not limited to convolution or attention mechanisms.

[0093] The data after two-branch encoding is input into the coupling layer. In some embodiments of the present invention, this layer includes, but is not limited to, aligning the temporal features of EEG and the time series of PPG within the same set time window through a cross-modal attention mechanism, and then establishing a bidirectional mapping relationship from EEG to PPG and then from PPG to EEG. The processing of the coupling layer can be understood as learning the delayed coupling strength of the blood flow response caused by changes in EEG activity.

[0094] For the fusion layer, temporally coupled features and statistical features of the feature domain are concatenated into a unified fusion vector. This fused feature vector is then sequentially input into several hidden layers (convolutional layers). Each layer undergoes linear combination using weight matrices, and non-linear mapping is achieved through activation functions including, but not limited to, ReLU, Sigmoid, or Tanh. It's important to understand that the weights of each hidden layer are obtained through training and can automatically adjust the contribution of different features to the output of the conscious state. For example, enhancement of low-frequency EEG waves may be assigned higher weights in some hidden layer nodes, while changes in blood perfusion in the pulse wave may have a key weight in another set of nodes.

[0095] For easier understanding, Table 2 below shows the features of each hidden layer, as well as the settings of the corresponding activation function and weight contribution, in one embodiment of the present invention.

[0096] Table 2 Feature settings of each hidden layer in the fusion layer

[0097]

[0098] Finally, regarding the output layer in the model, the output layer performs a weighted summation of the hidden layer features to generate a continuous consciousness state score. The final output is a continuous consciousness state score (e.g., 0-1.00), which represents the current level of neural and hemodynamic coupling. This score reflects the combined state of EEG and hemodynamic features. The higher the value, the higher or more stable the level of consciousness.

[0099] Regarding the model's training method, the training data sources include, but are not limited to, synchronized EEG and PPG data from multiple injured individuals recorded clinically, with each sample having a corresponding consciousness score. Through supervised learning, the training objective is set to minimize the error between the model's output score and the clinically accurate score (the sample corresponds to the set score). For example, MSE loss is used. In some embodiments, a mutual information constraint term is added along with the defined supervised loss, meaning the model maximizes the statistical dependency between EEG and PPG, thereby strengthening coupled learning. Therefore, it can be understood that by gradually teaching the model to identify which EEG activities are accompanied by genuine blood oxygenation responses, its accuracy in distinguishing between mild consciousness impairment (with some autonomy) and pseudo-awakening is improved on the validation set.

[0100] Furthermore, in some embodiments of the present invention, it is also considered to further classify consciousness levels based on set thresholds, such as lucidity, mild impairment, and deep impairment. In conjunction with the embodiments corresponding to Table 2 above, multiple thresholds are set to map score ranges to corresponding consciousness levels. The following Table 3 explains the relationship between score and level mapping.

[0101] Table 3. Mapping Relationship between Consciousness Score Intervals and Levels of Consciousness

[0102]

[0103] For example, during the recovery phase after surgery or trauma emergency treatment, the injured person now has a certain degree of conscious response, and at some point, the patient's brainwaves are detected. The wave intensity increased, but SpO2 decreased. The model calculated a consciousness score of 0.42, corresponding to "mild coma / impaired consciousness," prompting medical staff to intervene immediately. At the same time, the score changes can be continuously tracked to monitor the recovery of consciousness.

[0104] In summary, it is understandable that the assessment model constructed in the consciousness assessment module 400 can automatically output continuous scores and corresponding levels after the fusion of features, without the need for manual observation using behavioral scales, thus reducing subjective human error. Simultaneously considering EEG and hemodynamic features improves the accuracy and stability of consciousness level assessment.

[0105] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. The scope of patent protection of the present invention shall be determined by the claims. Similarly, any equivalent structural changes made based on the description and drawings of the present invention shall also be included within the scope of protection of the present invention.

Claims

1. A system for assessing the level of consciousness of wounded soldiers, characterized in that, include: The acquisition module is used to acquire the electroencephalogram (EEG) signals and dual-channel photoplethysmography (PPG) signals from the brain of the injured person, and output the corresponding analog electrical signal sequence. The EEG signals are acquired by electrodes to collect the temporal potential of brain electrical activity, reflecting the state of group firing of neurons in the cerebral cortex. The dual-channel PPG signals are recorded as blood volume pulsation curves under different optical paths, which are used to calculate the ratio of oxyhemoglobin to deoxyhemoglobin. The digital processing module performs preprocessing on each type of analog electrical signal sequence, including filtering, noise reduction, and time synchronization, and then converts and outputs EEG digital signals and pulse wave digital signals, respectively. The feature extraction module extracts time-domain features, frequency-domain features, and entropy values ​​from the EEG digital signal. It calculates blood oxygen saturation based on the ratio of the dual-channel pulse wave digital signals and calculates the blood perfusion index based on the pulse waveform integral and amplitude change. After normalizing the multi-domain features of the EEG digital signal with the blood oxygen saturation and blood perfusion index, a fused feature vector is formed. The consciousness assessment module receives the fused feature vector and inputs it into a preset consciousness state assessment model. The model internally identifies input signal patterns using trained feature weights and outputs a consciousness state score. This score is then compared sequentially with pre-defined multi-level thresholds to map the corresponding consciousness level. The input layer of the consciousness state assessment model receives EEG feature input and pulse wave blood oxygenation feature input. The EEG feature input includes time-domain features, frequency-domain features, and entropy values. The blood oxygenation feature input includes blood oxygen saturation and blood flow perfusion index. The input layer data then enters a dual-branch feature encoding layer, including an EEG branch and a PPG branch. The EEG branch encoding layer uses a one-dimensional convolutional neural network to extract local patterns in the time series, followed by LSTM modeling of temporal evolution. The PPG branch encoding layer extracts dynamic trends in blood oxygenation and perfusion fluctuation features. The data processed by both branches is then input into a coupling layer, which learns the delayed coupling strength of the blood flow response caused by changes in EEG activity.

2. The system for assessing the level of consciousness of wounded soldiers according to claim 1, characterized in that, The acquisition module receives potential changes caused by the collective firing of neurons in the cerebral cortex of the injured person through electrodes placed on the surface of the scalp. The continuous voltage sequence output by the amplification circuit is used as the analog signal sequence of the electroencephalogram (EEG) signal to record the electrical activity state of the brain in the time dimension.

3. The consciousness level assessment system for wounded soldiers according to claim 1, characterized in that, The acquisition module sets up two optical paths with different wavelengths, and records the changes in light absorption intensity caused by the change in blood volume with the cardiac cycle under the two optical paths with different wavelengths, to obtain the pulsation waveform sequence of the corresponding channel, which serves as the analog electrical signal sequence of the dual-channel photoplethysmography pulse wave signal.

4. The consciousness level assessment system for wounded soldiers according to claim 1, characterized in that, The digital processing module is used to preprocess and digitize the analog electrical signal sequence output by the acquisition module; firstly, the EEG potential sequence and photoplethysmography pulse wave sequence are bandpass filtered to retain the effective components within the set target frequency band and suppress power frequency interference and high frequency noise; The analog electrical signal sequence was then denoised by wavelet decomposition to eliminate motion artifacts and environmental noise components. After filtering and denoising, the digitization module establishes a unified time reference for the EEG signal and the dual-channel pulse wave signal, performs time synchronization processing on each analog electrical signal sequence, and converts the preprocessed analog signal sequence into discrete digital signals through an analog-to-digital converter circuit, forming EEG digital signals and pulse wave digital signals respectively.

5. The system for assessing the level of consciousness of wounded soldiers according to claim 1, characterized in that, The feature extraction module extracts features from EEG digital signals specifically as follows: In the time domain, the feature extraction module processes the EEG digital signal in segments, calculates the mean and variance of each segment, and quantifies the potential fluctuation amplitude; it extracts the instantaneous potential change range by calculating the difference between peak values; and it uses the autocorrelation function to calculate the correlation of the EEG digital signal under different lags, reflecting the periodic change characteristics of the signal. In the frequency domain, the feature extraction module performs a Fourier transform on the segmented signal to obtain the amplitude and power spectral density of each frequency component, and then determines the frequency based on the spectral boundaries. , , and Frequency bands are used to calculate the power of each band and its proportion of the total power; approximate entropy is calculated by constructing a delay vector sequence from the EEG signal.

6. A system for assessing the level of consciousness of wounded soldiers according to claim 5, characterized in that, For the dual-channel pulse wave digital signal, blood oxygen saturation is calculated by separating the AC component and the DC component and calculating the ratio of different wavelength channels. The blood perfusion index is obtained by integrating and calculating the amplitude change of the single-cycle waveform of the pulse wave digital signal.

7. The consciousness level assessment system for wounded soldiers according to claim 1, characterized in that, The consciousness assessment module receives the fused feature vector as input and inputs it into the preset consciousness state assessment model; the consciousness state assessment model adopts a multi-layer structure, and the fused feature vector is sequentially input into each processing layer; The input layer of the consciousness state assessment model is responsible for receiving multi-dimensional fused features and performing normalization and linear transformation to ensure uniform feature scale. Then, the input features are nonlinearly mapped and pattern learned through several hidden layers. Activation functions are used between the hidden layers to achieve nonlinear transformation, and the contribution of each feature in the output is adjusted through a weight matrix.

8. A system for assessing the level of consciousness of wounded soldiers according to claim 7, characterized in that, The output layer of the consciousness state assessment model generates a continuous consciousness state score, which is obtained by combining and weighting the hidden layer features to reflect the comprehensive state of EEG and hemodynamic features. During the assessment process, the consciousness assessment module compares the consciousness state score with the set multi-level thresholds in sequence. The score is mapped to different consciousness levels according to the threshold range it falls into. This mapping maintains a one-to-one correspondence between the score and the level. The output consciousness level is used to quantify the consciousness level of the injured person.

Images