Method and medium for analyzing the contribution of erp components in the assessment of the state of consciousness

Abstract

The application discloses a method and a medium for sharing contribution of ERP components in consciousness state evaluation. The method comprises the following steps: introducing a preset disturbance in a time window corresponding to a target ERP component to obtain multi-lead scalp electroencephalogram data after shielding, for multi-lead scalp electroencephalogram data to be analyzed; inputting the multi-lead scalp electroencephalogram data to be analyzed and the multi-lead scalp electroencephalogram data after shielding into a trained neural network model respectively, performing a consciousness state evaluation operation, and outputting a first prediction result and a second prediction result composed of prediction probabilities of each lead; calculating an absolute difference value between the first prediction result and the second prediction result as a prediction difference; and determining the contribution of the target ERP component in the consciousness state evaluation based on the prediction difference. By using the scheme of the application, quantitative analysis of the contribution of each ERP component in the consciousness state evaluation can be realized, and the accuracy of the consciousness state evaluation is improved.

Description

Technical Field

[0001] This application generally relates to the interdisciplinary field of biomedical engineering and artificial intelligence. More specifically, this application relates to a method and a computer-readable storage medium for analyzing the contribution of event-related potential (ERP) components in the assessment of states of consciousness. Background Technology

[0002] Differential diagnosis of patients with disorders of consciousness is a major challenge in clinical medicine. Accurately determining a patient's state of consciousness, such as vegetative state (VS) and minimally conscious state (MCS), is crucial for developing treatment plans, assessing prognosis, and making ethical decisions. Currently, clinical practice primarily relies on behavioral assessments based on question-and-answer sessions and cognitive response detection methods based on electroencephalography (EEG) to identify a patient's state of consciousness. Among these methods, event-related potentials (ERPs), particularly P300 and N1, are widely considered important biomarkers for revealing residual consciousness in patients. P300 is a positive wave that appears approximately 250-500 ms after stimulus presentation and is generally regarded as a marker of cognitive processing of significant stimuli in the brain; N1 is a negative wave that appears approximately 80-150 ms after stimulus presentation and is associated with early sensory processing in the sensory cortex.

[0003] Currently, ERP-based diagnostic methods for disorders of consciousness mainly focus on detecting the "presence" or changes in waveform characteristics of specific ERP components. For example, in the oddball paradigm, patients with MCS are more likely to exhibit a significant P300 response compared to patients with VS. However, existing technologies have significant limitations: firstly, current methods only focus on the presence or absence of ERP components without exploring the actual contribution of different ERP components (such as P300 and N1) to the discrimination of consciousness states; secondly, there is a lack of systematic methods to quantify the changes in the discriminative ability of the consciousness assessment system when ERP components are artificially masked, resulting in the inability to accurately locate key diagnostic indicators. This technological gap makes it difficult for clinicians to fully utilize the rich diagnostic information contained in ERP components, limiting further improvements in diagnostic accuracy.

[0004] In view of this, there is an urgent need to provide a scheme for analyzing the contribution of ERP components in the assessment of consciousness, so as to achieve quantitative analysis of the contribution of each ERP component in the assessment of consciousness and improve the accuracy of the assessment of consciousness. Summary of the Invention

[0005] In order to at least address one or more of the technical problems mentioned above, this application proposes a scheme in several aspects for analyzing the contribution of event-related potential (ERP) components in the assessment of state of consciousness.

[0006] In a first aspect, this application provides a method for analyzing the contribution of event-related potential (ERP) components in the assessment of a state of consciousness, comprising: acquiring multi-lead scalp EEG data to be analyzed; introducing a preset perturbation within a time window corresponding to a target ERP component in the multi-lead scalp EEG data to be analyzed, thereby masking the target ERP component and obtaining masked multi-lead scalp EEG data; inputting the multi-lead scalp EEG data to be analyzed and the masked multi-lead scalp EEG data into a trained neural network model, respectively, to perform a state of consciousness assessment operation, thereby outputting a first prediction result and a second prediction result composed of the prediction probabilities of each lead; calculating the absolute difference between the first prediction result and the second prediction result as a prediction difference; and determining the contribution of the target ERP component in the assessment of a state of consciousness based on the prediction difference.

[0007] In some embodiments, the preset perturbation is Gaussian noise perturbation; and for the multi-lead scalp EEG data to be analyzed, Gaussian noise perturbation is introduced within the time window corresponding to the target ERP component to mask the target ERP component, thereby obtaining masked multi-lead scalp EEG data, using the following formula:

[0008]

[0009] in, This represents the multi-lead scalp EEG data to be analyzed. This represents the masked multi-lead scalp EEG data; Represents a binary mask matrix. The corresponding time window requires the introduction of Gaussian noise perturbation. This corresponds to other time windows that do not require the introduction of Gaussian noise perturbation; This represents a Gaussian noise disturbance with a mean of 0 and a variance of . It follows a normal distribution.

[0010] In some embodiments, the target ERP component includes a P300 component and / or an N1 component, wherein the time window corresponding to the P300 component is 250ms to 500ms after stimulation, and the time window corresponding to the N1 component is 80ms to 150ms after stimulation.

[0011] In some embodiments, when the target ERP component includes P300 and N1 components, Gaussian noise perturbation is introduced within the time window corresponding to the target ERP component to obtain masked multi-lead scalp EEG data, using the following formula:

[0012]

[0013] in, This represents the multi-lead scalp EEG data to be analyzed. This represents the masked multi-lead scalp EEG data; The binary mask matrix representing the time window corresponding to the P300 component; The binary mask matrix representing the time window corresponding to the N1 component; This represents the Gaussian noise disturbance introduced within the time window corresponding to the P300 component, which has a mean of 0 and a variance of . The normal distribution; This represents the Gaussian noise disturbance introduced within the time window corresponding to component N1, with a mean of 0 and a variance of . It follows a normal distribution.

[0014] In some embodiments, the absolute difference between the first prediction result and the second prediction result is calculated as the prediction difference using the following formula:

[0015]

[0016] in, Indicates the difference in predictions. This represents the first prediction result output by the neural network model after the multi-lead scalp EEG data to be analyzed is input into the neural network model for training. This represents the prediction result output by the neural network model trained with masked multi-lead scalp EEG data, assuming the target ERP components include P300 and N1 components. This represents the prediction result output by the neural network model trained with masked multi-lead scalp EEG data as input, assuming the target ERP component only includes the P300 component. The prediction results output by the neural network model after masking multi-lead scalp EEG data is input into the target ERP component, which only includes the N1 component.

[0017] In some embodiments, determining the contribution of the target ERP component in the state of consciousness assessment based on the predicted difference includes: determining whether the predicted difference is greater than 0; if the predicted difference is greater than 0, the information provided by the deterministic P300 component and the N1 component in the state of consciousness assessment is complementary, and the contribution of the P300 component and the N1 component to the state of consciousness assessment when acting together is greater than the sum of their individual contributions; if the predicted difference is less than 0, determining that the information provided by the P300 component and the N1 component in the state of consciousness assessment is redundant, and the contribution of the P300 component and the N1 component to the state of consciousness assessment when acting together is less than the sum of their individual contributions.

[0018] In some embodiments, for the multi-lead scalp EEG data to be analyzed, Gaussian noise perturbation is introduced within the time window corresponding to the target ERP component to mask the target ERP component, and the masked multi-lead scalp EEG data is obtained by: dividing each lead into multiple lead groups according to the spatial location of each lead; performing the following operations for each lead group to obtain the masked multi-lead scalp EEG data corresponding to each lead group: introducing Gaussian noise perturbation into the scalp EEG data of each lead in the current lead group within the time window corresponding to the target ERP component, while keeping the scalp EEG data of each lead in other lead groups unchanged, to obtain the masked multi-lead scalp EEG data corresponding to the current lead group.

[0019] In some embodiments, obtaining the prediction difference corresponding to each lead group includes: inputting the masked multi-lead scalp EEG data corresponding to each lead group into the trained neural network model, performing a consciousness state assessment operation, and outputting a second prediction result corresponding to each lead group; calculating the absolute difference between the first prediction result and the second prediction result corresponding to each lead group, and using it as the prediction difference corresponding to each lead group.

[0020] In some embodiments, the method further includes: mapping the predicted differences corresponding to each lead group to an electroencephalogram (EEG) thermogram based on the spatial location of each lead, so as to visualize the contribution of the target ERP components of each lead group to the assessment of consciousness state.

[0021] In a second aspect, this application provides a computer-readable storage medium storing program instructions for analyzing the contribution of event-related potential (ERP) components in an assessment of a state of consciousness, wherein when the program instructions are executed by a processor, the method for analyzing the contribution of event-related potential (ERP) components in an assessment of a state of consciousness as described in the foregoing first aspect and several embodiments is implemented.

[0022] Using the general method provided above for analyzing the contribution of event-related potential (ERP) components in consciousness assessment, this application embodiment introduces a preset perturbation into the time window corresponding to the target ERP component to obtain masked EEG data. Combined with the absolute difference of the prediction results before and after masking output by the trained neural network model, it realizes a quantitative analysis of the contribution of the target ERP component in consciousness assessment. This breaks through the limitation of existing technologies that rely solely on the "presence" or "absence" of ERP signal waveforms. It systematically reveals the change in the system's ability to discriminate consciousness after the ERP component is masked, thereby accurately quantifying its contribution in consciousness assessment. This provides a more reliable quantitative basis for the differential diagnosis of patients with consciousness disorders and improves the accuracy and scientific nature of consciousness assessment. Attached Figure Description

[0023] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts, wherein:

[0024] Figure 1 An exemplary flowchart of a method for analyzing the contribution of ERP components in a state of consciousness assessment, according to an embodiment of this application, is shown.

[0025] Figure 2 An exemplary flowchart of a method for analyzing the dependence of conscious state assessment on target ERP components in different brain regions, according to an embodiment of this application, is shown.

[0026] Figure 3 An exemplary structural block diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] It should be understood that the terms "comprising" and "including" used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0029] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0030] As used in this specification and claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0031] The specific embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0032] Figure 1 An exemplary flowchart of a method 100 for analyzing the contribution of ERP components in a state of consciousness assessment, which can be implemented according to embodiments of this application, is shown.

[0033] like Figure 1 As shown, in step S101, the multi-lead scalp EEG data to be analyzed (also known as the initial multi-lead scalp EEG data) can be obtained.

[0034] Multi-lead scalp EEG data is potential change data of brain neuron electrical activity collected from multiple scalp electrodes. Its core is the EEG signal information corresponding to multiple leads (electrodes), which is the basic data for recognizing states of consciousness. In practice, the GSN256 lead system can be used to collect multi-lead scalp EEG data. The GSN256 lead system contains 256 scalp leads, which have a standardized spatial arrangement and positional coordinates.

[0035] Next, in step S102, a preset perturbation can be introduced into the time window corresponding to the target ERP component of the multi-lead scalp EEG data to be analyzed, so as to mask the target ERP component and obtain the masked multi-lead scalp EEG data.

[0036] In practice, the preset perturbation can be Gaussian noise perturbation, or other perturbation methods that replace Gaussian noise can be used, such as directly setting the signal of the target time window to zero, replacing the signal of the target time window with an average waveform, or introducing event-related potential (ERP) segments from incoherent patients. These alternative perturbation methods have been widely used in interpretability analysis in fields such as image recognition and speech processing.

[0037] However, in EEG signal analysis, using alternative methods such as direct zeroing or static interpolation can easily disrupt the frequency structure and evoked characteristics of the original EEG signal, potentially leading to misjudgments by the trained model. In contrast, using Gaussian noise for perturbation better reflects the natural fluctuation characteristics of EEG signals and can effectively remove specific ERP components without introducing anomalous artifacts. This method is particularly suitable for processing components with high time-locking characteristics in EEG signals, such as P300 and N1.

[0038] Therefore, Gaussian noise perturbation is introduced within the time window corresponding to the target ERP component to mask the target ERP component, and the masked multi-lead scalp EEG data can be obtained using the following formula:

[0039] (1)

[0040] in, This represents the multi-lead scalp EEG data to be analyzed. This represents the masked multi-lead scalp EEG data; Represents a binary mask matrix. The corresponding time window requires the introduction of Gaussian noise perturbation. This corresponds to other time windows that do not require the introduction of Gaussian noise perturbation; This represents a Gaussian noise disturbance with a mean of 0 and a variance of . It follows a normal distribution.

[0041] In embodiments of this application, the target ERP component may include the P300 component and / or the N1 component. Further, the time window corresponding to the P300 component is 250ms to 500ms post-stimulation, and the time window corresponding to the N1 component is 80ms to 150ms post-stimulation.

[0042] When the target ERP components include P300 and N1 components, Gaussian noise perturbation is introduced within the time window corresponding to the target ERP components to obtain masked multi-lead scalp EEG data. The following formula can be used:

[0043] (2)

[0044] in, This represents the multi-lead scalp EEG data to be analyzed. This represents the masked multi-lead scalp EEG data; The binary mask matrix representing the time window corresponding to the P300 component; The binary mask matrix representing the time window corresponding to the N1 component; This represents the Gaussian noise disturbance introduced within the time window corresponding to the P300 component, which has a mean of 0 and a variance of . The normal distribution; This represents the Gaussian noise disturbance introduced within the time window corresponding to component N1, with a mean of 0 and a variance of . It follows a normal distribution.

[0045] It is understandable that Gaussian noise perturbations injected into the two time windows ( and Different intensity parameters can be set separately (e.g., by adjusting the variance). and To simulate the neural background perturbation characteristics corresponding to P300 and N1, a method was implemented. Since the physiological origins and background noise environments of different ERP components vary, using differentiated noise intensities can better reflect the actual neurophysiological scenario, ensuring more accurate masking of the target components.

[0046] Next, in step S103, the multi-lead scalp EEG data to be analyzed and the masked multi-lead scalp EEG data can be input into the trained neural network model to perform a consciousness state assessment operation, so as to output a first prediction result and a second prediction result composed of the prediction probabilities of each lead.

[0047] The aforementioned neural network model can be a one-dimensional convolutional neural network (1D-CNN). This model can learn the distribution characteristics of different states of consciousness in the spatial structure of EEG leads, outputting the predicted probability of each lead. This predicted probability represents the likelihood of belonging to a specific state of consciousness, such as MCS or VS, and ultimately outputs results with clear clinical interpretability. Furthermore, this one-dimensional convolutional neural network can run on computing devices equipped with a Python execution environment, the PyTorch deep learning framework, and the MNE EEG signal processing library.

[0048] In some implementation scenarios, the neural network model can also be a graph neural network (GNN). Specifically, a GNN can treat each EEG lead as a node in a graph, and the spatial adjacency between leads as edges. By automatically learning the information transmission patterns and spatial correlations between nodes, it completes spatial modeling of EEG signals and assessment of consciousness states. However, it should be noted that the internal reasoning process of a GNN has difficulty accurately identifying leads that play a key role in the model's judgment, thus lacking intuitive interpretability in clinical applications.

[0049] It is understandable that neural network models require training data to achieve high performance. In the forward propagation phase of training, this application utilizes multiple multi-lead scalp EEG data containing consciousness state labels to train the neural network model, comparing the model output with the expected true values ​​(i.e., consciousness state labels) to obtain the corresponding loss function. During backpropagation, based on the obtained loss function, this application employs optimization algorithms such as stochastic gradient descent to update the model parameters, thereby reducing the error between the output and the true values ​​and improving the model's accuracy and generalization ability.

[0050] It is important to note that when acquiring multiple multi-lead scalp EEG data, it is possible to obtain multiple multi-lead scalp EEG data from patients with different levels of consciousness, as well as from the same patient at different stages. It is also important to note that multi-lead scalp EEG data collected from the same patient at different stages should be treated as independent samples.

[0051] Furthermore, abnormal lead removal and bandpass filtering can be performed on multiple multi-lead scalp EEG data for preprocessing, resulting in preprocessed multi-lead scalp EEG data for training neural network models. Specifically, when performing abnormal lead removal, low-quality leads (such as those located at the temporal margin or significantly affected by electromyographic contamination) can be filtered out based on lead quality and noise level. When performing bandpass filtering, 0.5–40Hz bandpass spatial filtering can be applied to all multi-lead scalp EEG data, combined with noise processing and multiple linear regression to suppress interference and improve the signal-to-noise ratio.

[0052] In addition, when annotating multiple multi-lead scalp EEG data, 0 or 1 can be used to annotate two different states of consciousness. For example, VS is annotated with 1 and MCS with 0, or MCS is annotated with 1 and VS with 0, and the state of consciousness annotated with 1 is the target state of consciousness.

[0053] In step S104, the absolute difference between the first prediction result and the second prediction result can be calculated as the prediction difference.

[0054] Specifically, the absolute difference can be calculated using the following formula:

[0055] (3)

[0056] in To predict the difference, This is the first prediction result. This is the second prediction result.

[0057] At step S105, the contribution of the target ERP component to the assessment of state of consciousness can be determined based on the predicted differences.

[0058] In the embodiments of this application, the greater the predicted difference, the higher the contribution of the target ERP component in the assessment of state of consciousness; conversely, the smaller the predicted difference, the lower the contribution.

[0059] Specifically, if A significantly larger value indicates that the model is strongly dependent on the target ERP component, meaning that the target ERP component contributes significantly to the assessment of consciousness. When the target ERP component is obscured or blurred, the model's judgment of the patient's consciousness may shift to a lower level (e.g., from MCS to VS). If the value approaches zero, it indicates that the model's prediction results remain almost unchanged after the target ERP component is masked, suggesting that the model has a weak dependence on the target ERP component and its contribution to the assessment of state of consciousness is low.

[0060] The above-mentioned analysis method based on predictive differences enables a quantitative assessment of the contribution of target ERP components in the assessment of consciousness. This breaks through the limitations of existing technologies that rely solely on the "presence" or "absence" of ERP signal waveforms. It systematically reveals the changes in the system's ability to discriminate consciousness after ERP components are masked, thereby accurately quantifying their contribution in the assessment of consciousness. This provides a more reliable quantitative basis for the differential diagnosis of patients with consciousness disorders and improves the accuracy and scientific nature of the assessment of consciousness.

[0061] In some implementation scenarios, to analyze the independence and synergistic effects of the P300 and N1 ERP components in consciousness state assessment, multiple sets of signal processing conditions can be designed for comparative analysis. Specifically, four processing scenarios can be set for the initial multi-lead scalp EEG data: the first is the initial multi-lead scalp EEG data (without any perturbation), denoted as input signal X, with the corresponding model output prediction result Y; the second is the multi-lead scalp EEG data with only the time window corresponding to the P300 component masked, denoted as input signal X. The corresponding prediction result output by the model is Thirdly, multi-lead scalp EEG data with masking only the time window corresponding to the N1 component are denoted as input signals. The corresponding prediction result output by the model is Fourthly, multi-lead scalp EEG data with simultaneous masking of the time windows corresponding to the P300 and N1 components are denoted as input signals. The corresponding prediction result output by the model is .

[0062] The four types of processed multi-lead scalp EEG data were all input into the same trained neural network model. By recording and comparing the prediction results corresponding to different inputs, the prediction difference value (which can be referred to as the cooperative perturbation difference value) was calculated. The calculation formula is as follows:

[0063] (4)

[0064] This predicted variance value is used to quantify the degree of deviation between the combined effect of the two components and the sum of their individual effects, thereby revealing their relationship in conscious discrimination. Specifically, it can be based on... To analyze the independence and synergistic effect of the two ERP components, P300 and N1, in the assessment of state of consciousness, we need to determine whether they are greater than 0.

[0065] like This indicates that the combined effect of masking the P300 and N1 components on the model's prediction results is greater than the sum of the effects of masking either component individually. This means that the information provided by the two components in assessing consciousness is complementary, exhibiting a cumulative effect; that is, their combined contribution to the determination of consciousness is greater than the sum of their individual contributions. From a clinical perspective, in this case, the patient must simultaneously possess complete early perceptual processing ability (reflected by the N1 component) and late cognitive processing ability (reflected by the P300 component) to be accurately identified by the model as having the corresponding state of consciousness (e.g., MCS).

[0066] like This indicates that the combined effect of masking P300 and N1 components is less than the sum of the effects of masking either component individually. This suggests that the two components have some functional overlap in consciousness assessment, providing redundant information. In other words, the signal characteristics of a single component are sufficient to support the model's judgment of consciousness, without relying on the combined effect of both components. In this case, detecting only one component (such as P300 or N1) is sufficient to meet the basic clinical needs for consciousness assessment.

[0067] In the above embodiments, when masking the acquired multi-lead scalp EEG data to be analyzed, the operation is performed on all leads included in the data. Specifically, a preset perturbation (such as Gaussian noise perturbation) is uniformly injected into the time windows corresponding to the target ERP components (such as P300 and N1) in all leads to analyze the overall contribution of the target ERP components in the assessment of consciousness. This method can reveal the impact of the target ERP components on model decisions at a global level, but it fails to refine the specific roles of different brain regions in the spatial dimension. To further improve the spatial resolution of EEG data in the assessment of consciousness and accurately locate the functional differences of different brain regions in the assessment of consciousness, the masking operation of the target ERP components can be limited to the lead groups corresponding to specific brain regions.

[0068] Based on this, such as Figure 2The diagram shows an exemplary flowchart of a method 200 for analyzing the dependence of consciousness state assessment on target ERP components in different brain regions, as provided in this application embodiment. Figure 2 As shown, in step S201, each lead is divided into multiple lead groups according to its spatial position.

[0069] Specifically, five lead groups can be set up, corresponding to the prefrontal, central, parietal, occipital, and temporal lobes, respectively, labeled R1, R2, R3, R4, and R5. This division helps avoid interference between spatially adjacent but functionally heterogeneous regions during analysis. Next, based on the spatial location of each lead on the scalp, all leads in the GSN256 lead system can be assigned to these lead groups. For example, lead group R1 is the prefrontal lead group, containing leads such as Fp1 and Fp2 corresponding to the prefrontal lobe; lead group R2 is the central lead group, containing leads such as C3 and C4 corresponding to the central lobe, and so on, ensuring that each lead group corresponds to a specific set of leads for a particular brain region.

[0070] In step S202, the following operations are performed for each lead group to obtain the masked multi-lead scalp EEG data corresponding to each lead group: Gaussian noise perturbation is introduced into the scalp EEG data of each lead in the current lead group within the time window corresponding to the target ERP component, while the scalp EEG data of each lead in other lead groups remain unchanged, so as to obtain the masked multi-lead scalp EEG data corresponding to the current lead group.

[0071] Next, in step S203, the masked multi-lead scalp EEG data corresponding to each lead group are input into the trained neural network model to perform a consciousness state assessment operation, so as to output the second prediction result corresponding to each lead group.

[0072] Next, in step S204, the absolute difference between the first prediction result and the second prediction result corresponding to each lead group is calculated, and used as the prediction difference corresponding to each lead group.

[0073] The prediction difference for each lead group is used to quantify the degree of change in the model's assessment of the state of consciousness after the target ERP component of a specific brain region is masked. The larger the prediction difference value, the more significant the impact of the target ERP component of the brain region corresponding to that lead group on the model's decision, that is, the higher the model's dependence on the target ERP component of that brain region.

[0074] As an example, if the prediction discrepancies in the prefrontal cortex lead groups The predicted difference corresponding to R3 in the parietal lobe brain region lead group This indicates that the target ERP components in the prefrontal cortex contribute more significantly to the model's assessment of consciousness, and the model relies more heavily on the response information of the target ERP components in the prefrontal cortex leads. Statistical comparison of the prediction differences across lead groups reveals the brain regions that play a crucial role in consciousness assessment. Therefore, if the prediction difference value in the prefrontal cortex is significantly higher than that in other brain regions after statistical comparison, it indicates that the model is highly dependent on the response information of the target ERP components in the prefrontal cortex leads.

[0075] After obtaining the predicted differences for each lead group, the perturbation intensities of all lead groups (i.e., the predicted differences corresponding to each lead group) can be further constructed into a lead perturbation atlas. This atlas is then mapped to a standard GSN256 electrode coordinate template and visualized as an EEG thermogram. The color depth of different lead groups in the atlas corresponds to the magnitude of the perturbation intensity, intuitively showing the degree of change in the model output in different brain regions after the target ERP component time window is occluded. Additionally or optionally, for key lead groups with significant perturbation intensity, the response intensity differences of each lead within them can be further analyzed to refine the contribution of different leads in specific brain regions to consciousness assessment and improve the precision of spatial resolution.

[0076] The above combination Figure 2 The method described herein, which analyzes the dependence of consciousness state assessment on target ERP components in different brain regions, divides multi-lead EEG data into specific lead groups according to brain regions, selectively masks target ERP components (such as P300 and N1) in different brain regions, and calculates the difference in prediction before and after masking. This method can accurately quantify the contribution of each brain region's ERP components to consciousness state assessment, significantly improving the spatial resolution of the assessment. It not only overcomes the limitation of traditional global analysis in distinguishing brain region specificity, but also intuitively reveals the core role of key brain regions (such as the prefrontal and parietal lobes) in consciousness state recognition through the visualization method of EEG thermography. This not only enhances the interpretability of model decisions, but also provides a quantitative basis for the diagnosis of consciousness disorders that combines temporal characteristics and spatial localization.

[0077] Next, combine Figure 3 An electronic device 300 provided in an embodiment of this application will be described by way of example. Figure 3 As shown, the electronic device 300 in this application embodiment may include a processor 301, a memory 302, and a communication bus 303.

[0078] In specific embodiments, the processor 301 described above can be at least one of the following: Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), CPU, controller, microcontroller, and microprocessor. It is understood that for different devices, the electronic device used to implement the above processor function can also be other types, and this embodiment does not specifically limit it.

[0079] In this embodiment, the communication bus 303 is used to establish communication between the processor 301 and the memory 302; the memory 302 stores program instructions for analyzing the contribution of ERP components in the assessment of state of consciousness; when the processor 801 executes the program instructions stored in the memory 802, it implements the combination of this application. Figures 1 to 2 The method described is for analyzing the contribution of ERP components in the assessment of state of consciousness.

[0080] The above combination Figure 3 This document describes an electronic device that can be used to execute program instructions for analyzing the contribution of ERP components in the assessment of state of consciousness, as described in this application. It should be understood that the device structure or architecture described herein is merely exemplary, and the implementation methods and entities of this application are not limited thereto, but can be modified without departing from the spirit of this application. It is understood that the description of the various embodiments in this disclosure emphasizes the differences between the various embodiments, and their similarities or corresponding parts can be referred to mutually. For the sake of brevity, this disclosure will not elaborate on each one.

[0081] Based on the foregoing description in conjunction with the accompanying drawings, those skilled in the art will understand that the embodiments of this application can also be implemented by software programs. Therefore, this application also provides a computer-readable storage medium. This computer-readable storage medium stores program instructions for analyzing the contribution of ERP components in the assessment of state of consciousness, and these program instructions can be used to implement the embodiments of this application. Figures 1 to 2 The method described is for analyzing the contribution of ERP components in the assessment of state of consciousness.

[0082] It should be noted that although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart can be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0083] While numerous embodiments of this application have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will arise for those skilled in the art without departing from the spirit and intent of this application. It should be understood that various alternatives to the embodiments of this application described herein may be employed in the practice of this application. The appended claims are intended to define the scope of protection of this application and therefore cover equivalents or alternatives within the scope of these claims.

[0084] Furthermore, the collection and acquisition of various data in this application comply with relevant laws and regulations and are authorized by the data providers. Any organization or individual that needs to obtain external data shall obtain authorization in accordance with the law and ensure data security, and shall not illegally collect, use, process, or transmit unauthorized or unprotected data, nor shall it illegally buy, sell, provide, or disclose unauthorized or unprotected data.

Claims

1. A method for analyzing the contribution of event-related potential (ERP) components in assessment of state of consciousness, characterized in that, include: Acquire multi-lead scalp EEG data to be analyzed; For the multi-lead scalp EEG data to be analyzed, a preset perturbation is introduced within the time window corresponding to the target ERP component to mask the target ERP component and obtain the masked multi-lead scalp EEG data. The multi-lead scalp EEG data to be analyzed and the masked multi-lead scalp EEG data are respectively input into the trained neural network model to perform a consciousness state assessment operation, so as to output a first prediction result and a second prediction result composed of the prediction probabilities of each lead. The absolute difference between the first prediction result and the second prediction result is calculated as the prediction difference; Based on the predicted discrepancies, the contribution of the target ERP component to the assessment of state of consciousness is determined.

2. The method according to claim 1, characterized in that, The preset perturbation is Gaussian noise perturbation; and for the multi-lead scalp EEG data to be analyzed, Gaussian noise perturbation is introduced within the time window corresponding to the target ERP component to mask the target ERP component, and the masked multi-lead scalp EEG data is obtained using the following formula: in, This represents the multi-lead scalp EEG data to be analyzed. This represents the masked multi-lead scalp EEG data; Represents a binary mask matrix. The corresponding time window requires the introduction of Gaussian noise perturbation. This corresponds to other time windows that do not require the introduction of Gaussian noise perturbation; This represents a Gaussian noise disturbance with a mean of 0 and a variance of . It follows a normal distribution.

3. The method according to claim 2, characterized in that, The target ERP components include the P300 component and / or the N1 component. The time window corresponding to the P300 component is 250ms to 500ms after stimulation, and the time window corresponding to the N1 component is 80ms to 150ms after stimulation.

4. The method according to claim 3, characterized in that, When the target ERP components include P300 and N1 components, Gaussian noise perturbation is introduced within the time window corresponding to the target ERP components to obtain masked multi-lead scalp EEG data, using the following formula: in, This represents the multi-lead scalp EEG data to be analyzed. This represents the masked multi-lead scalp EEG data; The binary mask matrix representing the time window corresponding to the P300 component; The binary mask matrix representing the time window corresponding to the N1 component; This represents the Gaussian noise disturbance introduced within the time window corresponding to the P300 component, which has a mean of 0 and a variance of . The normal distribution; This represents the Gaussian noise disturbance introduced within the time window corresponding to component N1, with a mean of 0 and a variance of . It follows a normal distribution.

5. The method according to claim 3, characterized in that, The absolute difference between the first prediction result and the second prediction result is calculated as the prediction difference using the following formula: in, Indicates the difference in predictions. This represents the first prediction result output by the neural network model after the multi-lead scalp EEG data to be analyzed is input into the neural network model for training. This represents the prediction result output by the neural network model trained with masked multi-lead scalp EEG data, assuming the target ERP components include P300 and N1 components. This represents the prediction result output by the neural network model trained with masked multi-lead scalp EEG data as input, assuming the target ERP component only includes the P300 component. The prediction results output by the neural network model after masking multi-lead scalp EEG data is input into the target ERP component, which only includes the N1 component.

6. The method according to claim 5, characterized in that, Based on the predicted discrepancies, determining the contribution of the target ERP component to the assessment of state of consciousness includes: Determine whether the predicted difference is greater than 0; When the predicted difference is greater than 0, the information provided by the deterministic P300 component and the N1 component in the assessment of state of consciousness is complementary, and the contribution of the P300 component and the N1 component to the assessment of state of consciousness when they work together is greater than the sum of their individual contributions. If the predicted difference is less than 0, it is determined that the information provided by the P300 component and the N1 component in the assessment of consciousness is redundant, and the contribution of the P300 component and the N1 component to the assessment of consciousness when they act together is less than the sum of their individual contributions.

7. The method according to any one of claims 3-6, characterized in that, For the multi-lead scalp EEG data to be analyzed, Gaussian noise perturbation is introduced within the time window corresponding to the target ERP component to mask the target ERP component, resulting in masked multi-lead scalp EEG data including: Based on the spatial location of each lead, each lead is divided into multiple lead groups; Perform the following operations for each lead group to obtain the masked multi-lead scalp EEG data for each lead group: Gaussian noise perturbation is introduced into the scalp EEG data of each lead in the current lead group within the time window corresponding to the target ERP component, while the scalp EEG data of each lead in other lead groups remain unchanged, in order to obtain the masked multi-lead scalp EEG data corresponding to the current lead group.

8. The method according to claim 7, characterized in that, The predicted differences for each lead group include: The masked multi-lead scalp EEG data corresponding to each lead group are input into the trained neural network model to perform consciousness state assessment and output the second prediction result corresponding to each lead group. The absolute difference between the first prediction result and the second prediction result corresponding to each lead group is calculated, and this difference is used as the prediction difference for each lead group.

9. The method according to claim 8, characterized in that, Also includes: Based on the spatial location of each lead, the predicted differences corresponding to each lead group are mapped to EEG thermograms to visualize the contribution of the target ERP components of each lead group to the assessment of consciousness state.

10. A computer-readable storage medium storing program instructions for analyzing the contribution of event-related potential (ERP) components in an assessment of a state of consciousness, wherein when the program instructions are executed by a processor, the method for analyzing the contribution of event-related potential (ERP) components in an assessment of a state of consciousness according to any one of claims 1-9 is implemented.

Images