Noninvasive-minimally invasive electroencephalogram signal recording method for deep brain region activity
By calibrating the model using deep and superficial EEG signals collected in a previous time period, deep EEG signals can be predicted and generated in a non-invasive state, thus solving the problems of data integrity and security in deep EEG recording and achieving a recording method with high accuracy and low risk.
Patent Information
- Application Number
- CN202511979276.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-25
- Publication Date
- 2026-03-24
AI Technical Summary
Existing deep brain electroencephalography (DBE) techniques suffer from issues of data integrity and security, particularly the immune response resulting from invasive procedures and the instability of long-term recording.
By collecting real deep EEG signals and superficial EEG signals in a previous time period for model calibration, and using spatiotemporal graph generation and multi-layer feature extraction, deep EEG signals in a non-invasive state can be predicted and generated, thereby improving the integrity and security of the recording.
This improved the data integrity and security of deep EEG signal recording, reduced the deep electrode implantation time, and enhanced the accuracy and reliability of the recording.
Smart Images
Figure CN121714282A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electroencephalogram (EEG) signal processing technology, and in particular to a non-invasive-minimally invasive EEG signal recording method for deep brain region activity. Background Technology
[0002] Because of its irreplaceable core role in the localization and resection strategies for refractory epileptic foci, the analysis of circuit mechanisms in Parkinson's disease and motor disorders, and the intervention of circuits related to mental illnesses (such as DBS treatment for depression and obsessive-compulsive disorder), deep brain electroencephalography (DBE) recording has gradually gained attention from relevant practitioners.
[0003] Currently, the relevant technologies typically involve using stereotactic navigation to precisely implant slender electrodes with multiple ring-shaped or point-like contact points into structures such as the hippocampus, amygdala, thalamus, globus pallidus, and subthalamic nucleus. This allows for the recording of synchronized firing, network oscillations, and cross-regional coupling processes of local neural groups, thereby achieving deep brainwave recording. However, because deep electrode implantation is an invasive neurosurgical procedure, the integrity of EEG data from short-term deep brainwave recordings is not high, and long-term deep brainwave recordings are prone to triggering immune responses in the organism (such as chronic inflammation and glial scar formation), making the safety of deep brainwave recording questionable.
[0004] Therefore, the problems with the relevant technologies still need to be solved and optimized. Summary of the Invention
[0005] The purpose of this invention is to at least partially solve one of the technical problems existing in the related art.
[0006] Therefore, one objective of this invention is to provide a non-invasive / minimally invasive method for recording deep brain region activity, which can effectively improve the integrity and safety of deep brain signal recording.
[0007] To achieve the above-mentioned technical objectives, the technical solutions adopted in the embodiments of this application include: In a first aspect, embodiments of this application provide a non-invasive / minimally invasive method for recording electroencephalogram (EEG) signals of deep brain region activity, comprising: The deep EEG signal and the first superficial EEG signal were acquired from the subject in a first time period under minimally invasive conditions, and the second superficial EEG signal was acquired from the subject in a second time period under non-invasive conditions; the first time period was earlier than the second time period. Based on the deep EEG real signal and the first superficial EEG signal, the initialized EEG signal prediction model is calibrated to obtain the calibrated EEG signal prediction model. The second superficial EEG signal is input into the calibrated EEG signal prediction model for deep prediction, and the deep EEG prediction signal of the subject in the second time period is obtained. Deep brainwave recording data is obtained based on the actual deep brainwave signals and the predicted deep brainwave signals.
[0008] In addition, the method according to the above embodiments of this application may also have the following additional technical features: Further, in one embodiment of this application, the step of calibrating the initialized EEG signal prediction model based on the deep EEG real signal and the first superficial EEG signal to obtain a calibrated EEG signal prediction model includes: Spatiotemporal graph generation processing is performed on the actual deep EEG signals to obtain a deep EEG spatiotemporal graph; The first superficial EEG signal is processed to generate a spatiotemporal map, resulting in a superficial EEG spatiotemporal map. Based on the deep EEG spatiotemporal map and the shallow EEG spatiotemporal map, the initialized EEG signal prediction model is calibrated to obtain the calibrated EEG signal prediction model.
[0009] Furthermore, in one embodiment of this application, spatiotemporal map generation processing is performed on the target EEG signal to obtain the target EEG spatiotemporal map, including: The target EEG signal is bandpass filtered to obtain a first intermediate signal; The first intermediate signal is subjected to artifact removal to obtain the second intermediate signal; Based on the multi-channel time series corresponding to the target EEG signal, a spatiotemporal map is constructed from the second intermediate signal to obtain the target EEG spatiotemporal map. Wherein, if the target EEG signal is the deep EEG real signal, then the target EEG spatiotemporal map is the deep EEG spatiotemporal map; or, if the target EEG signal is the first shallow EEG signal, then the target EEG spatiotemporal map is the shallow EEG spatiotemporal map.
[0010] Furthermore, in one embodiment of this application, the step of constructing a spatiotemporal map of the second intermediate signal based on the multi-channel time series corresponding to the target EEG signal to obtain the target EEG spatiotemporal map includes: Obtain the target time window; Based on the target time window, the multi-channel time series is slicing to obtain several time slices; Based on each time slice and the second intermediate signal, a target EEG spatiotemporal map corresponding to each time slice is obtained.
[0011] Further, in one embodiment of this application, the step of calibrating the initialized EEG signal prediction model based on the deep EEG spatiotemporal map and the shallow EEG spatiotemporal map to obtain the calibrated EEG signal prediction model includes: The shallow EEG spatiotemporal map is segmented into spatiotemporal blocks to obtain an embedded vector sequence; Multi-layer spatiotemporal feature extraction is performed on the embedded vector sequence to obtain a spatiotemporal feature sequence; The spatiotemporal feature sequence is recombined and predicted to obtain a deep spatiotemporal prediction map; Based on the deep EEG spatiotemporal map and the deep spatiotemporal prediction map, the parameters of the initialized EEG signal prediction model are updated to obtain the calibrated EEG signal prediction model.
[0012] Furthermore, in one embodiment of this application, the step of performing spatiotemporal map segmentation on the shallow EEG spatiotemporal map to obtain an embedding vector sequence includes: The shallow EEG spatiotemporal map is divided into image blocks to obtain several non-overlapping EEG matrix blocks; Projection mapping is performed on each of the EEG matrix blocks to obtain several EEG embedding vectors; Position encoding is performed on all the EEG embedding vectors to obtain the embedding vector sequence.
[0013] Further, in this embodiment of the application, the step of performing multi-layer spatiotemporal feature extraction on the embedded vector sequence to obtain a spatiotemporal feature sequence includes: Obtain the input feature sequence, which is the embedded vector sequence or the output feature sequence in the previous spatiotemporal feature extraction process; Multi-head self-attention feature extraction is performed on the input feature sequence to obtain the intermediate feature sequence; Based on the input feature sequence, feature mapping processing is performed on the intermediate feature sequence to obtain the spatiotemporal feature sequence.
[0014] Secondly, embodiments of this application provide a non-invasive / minimally invasive electroencephalogram (EEG) signal recording system for deep brain region activity, comprising: The first processing unit is used to acquire the deep EEG real signal and the first superficial EEG signal collected by the subject in a first time period and under minimally invasive condition, and the second superficial EEG signal collected by the subject in a second time period and under non-invasive condition; the first time period is earlier than the second time period. The second processing unit is used to perform model calibration on the initialized EEG signal prediction model based on the deep EEG real signal and the first shallow EEG signal to obtain the calibrated EEG signal prediction model. The third processing unit is used to input the second superficial EEG signal into the calibrated EEG signal prediction model for deep prediction, and obtain the deep EEG prediction signal of the subject in the second time period. The fourth processing unit is used to obtain deep brainwave recording data based on the actual deep brainwave signal and the predicted deep brainwave signal.
[0015] Thirdly, embodiments of this application also provide an electronic device, including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor performs the method described above.
[0016] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a processor-executable program, which, when executed by the processor, is used to implement the above-described method.
[0017] Fifthly, embodiments of this application also provide a computer program product, which includes a computer program stored in a computer-readable storage medium. A processor of an electronic device reads the computer program from the computer-readable storage medium and executes the computer program, causing the electronic device to perform the method described above.
[0018] The advantages and beneficial effects of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application: This application discloses a non-invasive / minimally invasive EEG signal recording method for deep brain region activity. The method acquires real deep EEG signals and a first superficial EEG signal from a subject during a first time period under minimally invasive conditions, and a second superficial EEG signal from the subject during a second time period under non-invasive conditions; the first time period is earlier than the second time period. Based on the real deep EEG signal and the first superficial EEG signal, an initialized EEG signal prediction model is calibrated to obtain a calibrated EEG signal prediction model. The second superficial EEG signal is input into the calibrated EEG signal prediction model for deep prediction, obtaining the subject's predicted deep EEG signal during the second time period. Deep EEG recording data is obtained based on the real deep EEG signal and the predicted deep EEG signal. This method calibrates the EEG signal prediction model using the actual deep EEG signals and the first superficial EEG signals of the subject in a previous time period under minimally invasive conditions. Based on the calibrated EEG signal prediction model, it performs deep prediction on the second superficial EEG signals of the subject in the current time period under non-invasive conditions. This predicts and generates relatively reliable deep EEG signals in the second time period. It can ensure the integrity of the EEG data recorded in deep EEG while reducing the implantation time of deep electrodes in the subject, thereby effectively improving the safety of deep EEG recording. Attached Figure Description
[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following description is provided with accompanying drawings of the relevant technical solutions in the embodiments of this application or the prior art. It should be understood that the accompanying drawings described below are only for the purpose of clearly illustrating some embodiments of the technical solutions in this application. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.
[0020] Figure 1 A flowchart illustrating a non-invasive / minimally invasive electroencephalogram (EEG) signal recording method for deep brain region activity, provided as an embodiment of this application; Figure 2 A schematic diagram of the framework of a non-invasive-minimally invasive EEG signal recording system for deep brain region activity provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0021] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.
[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.
[0023] Currently, the relevant technologies typically involve using stereotactic navigation to precisely implant slender electrodes with multiple ring-shaped or point-like contact points into structures such as the hippocampus, amygdala, thalamus, globus pallidus, and subthalamic nucleus. This allows for the recording of synchronized firing, network oscillations, and cross-regional coupling processes of local neural groups, thereby achieving deep brainwave recording. However, because deep electrode implantation is an invasive neurosurgical procedure, the integrity of EEG data from short-term deep brainwave recordings is not high, and long-term deep brainwave recordings are prone to triggering immune responses in the organism (such as chronic inflammation and glial scar formation), making the safety of deep brainwave recording questionable.
[0024] Furthermore, some related technologies bridge the gap between the scalp, cortex, and deep brain regions through multi-level simultaneous recording and signal relationship modeling. Specifically, this involves simultaneously collecting scalp EEG (electroencephalography), ECoG (electrocorticography), and deep brain electroencephalography during preoperative epilepsy assessment or SEEG monitoring, and utilizing power spectral analysis, coherence and phase synchronization indices, Granger causality, and phase... Amplitude coupling (phase) Methods such as amplitude coupling (PAC) can depict the correlation patterns of signals at different levels within specific frequency bands and time windows. However, these methods remain at the level of statistical correlation and low-dimensional linear / quasi-linear analysis, often assuming that the relationships between signals are relatively simple, stable, or can be described by a few parameters. They are insufficient to fully characterize the nonlinear dynamics, multi-scale oscillatory coupling, state dependence, and cross-regional time-varying interactions prevalent in real brain networks, resulting in low accuracy in deep brain electroencephalography (EEG) recordings. Specifically, they have the following limitations: 1. Aliasing effect of volumetric conduction: Surface electrodes record mixed signals from multiple deep and cortical sources. Simple linear deconvolution or source localization often has significant errors under non-ideal conditions. Specifically, deep EEG signals pass through multiple layers of tissue, including cerebrospinal fluid, meninges, skull, and scalp, during their propagation to the surface. They are subject to significant volumetric conduction and spectral filtering, resulting in severe attenuation and aliasing of high-frequency and focal information. Surface electrodes record mixed signals from multiple sources, and the contributions of deep structures are often masked by the broader and larger-amplitude cortical activity.
[0025] 2. Nonlinearity and state dependence: Projection of deep structures to cortical output, and the cortex The feedback loops between deep layers exhibit significant nonlinearity and plasticity under different brain states (awake, asleep, before and after an attack, etc.), making it difficult for fixed-parameter linear models to adapt.
[0026] 3. High-dimensional and cross-scale coupling: Deep activity often affects surface signals through multiple frequency bands, multiple time scales and multiple pathways. Simple one-to-one mapping cannot capture this high-dimensional coupling structure.
[0027] It should be noted that the aforementioned related technologies are only used to assist in understanding the technical solutions of this application and do not mean that they belong to the publicly disclosed prior art.
[0028] In view of this, embodiments of this application provide a non-invasive-minimally invasive EEG signal recording method for deep brain region activity. This method calibrates an EEG signal prediction model using real deep EEG signals from a subject in a previous time period under minimally invasive conditions and a first superficial EEG signal. Based on the calibrated EEG signal prediction model, it performs deep prediction on a second superficial EEG signal from the subject in the current time period under non-invasive conditions. This predicts and generates a more reliable deep EEG signal for the second time period. This method ensures the integrity of the recorded deep EEG data while reducing the implantation time of deep electrodes in the subject, thereby effectively improving the safety of deep EEG recording.
[0029] Furthermore, this method, through spatiotemporal graph generation processing of the target EEG signal, can formalize the task of predicting deep EEG signals from shallow EEG signals into an image-to-image reconstruction problem. Then, through the EEG signal prediction model, it learns the nonlinear dynamics, multi-scale oscillatory coupling, state dependence, and cross-regional time-varying interaction relationships that are prevalent in the brain network between the spatiotemporal graphs of deep EEG signals and shallow EEG signals. Specifically, by performing multi-layer spatiotemporal feature extraction on the embedded vector sequence containing position encoding, the EEG signal prediction model can simultaneously focus on information from different brain regions (spatial domain) and different time segments (temporal domain) of the brain network, so as to automatically focus on the surface (i.e., shallow) spatiotemporal features that are highly correlated with deep brain activity in the brain network, which is beneficial to improving the accuracy of deep EEG signal recording.
[0030] Reference Figure 1 In this application embodiment, a non-invasive / minimally invasive EEG signal recording method for deep brain region activity includes: Step 110: Obtain the deep EEG real signal and the first superficial EEG signal collected from the subject in the first time period under minimally invasive condition, and the second superficial EEG signal collected from the subject in the second time period under non-invasive condition; the first time period is earlier than the second time period. In this embodiment, the subject can be an organism (such as a laboratory mouse) or a human subject receiving EEG signal acquisition. The EEG signals acquired in this embodiment are obtained with the subject's authorization. The deep EEG signal can be a matrix-form multichannel deep EEG signal acquired by the subject over a historical time period using deep electrode implantation technology (i.e., deep EEG signal acquired under minimally invasive conditions), specifically, it can be the hippocampal deep LFP / Spike signal. The first superficial EEG signal can be a matrix-form multichannel cortical EEG signal acquired by the subject under minimally invasive conditions using scalp EEG and / or ECoG technologies, such as cortical ECoG signals. This first superficial EEG signal is time-scale aligned with the deep EEG signal. The second superficial EEG signal can be a matrix-form multichannel cortical EEG signal acquired by the subject under the current time period using scalp EEG and / or ECoG technologies (i.e., second superficial EEG signal acquired under non-invasive conditions).
[0031] Step 120: Based on the deep EEG real signal and the first superficial EEG signal, perform model calibration on the initialized EEG signal prediction model to obtain the calibrated EEG signal prediction model. In this embodiment of the application, the model parameters of the initialized EEG signal prediction model can be calibrated and updated based on the deep EEG real signal and the first superficial EEG signal to obtain the calibrated EEG signal prediction model.
[0032] In some embodiments, the step of calibrating the initialized EEG signal prediction model based on the deep EEG real signal and the first superficial EEG signal to obtain a calibrated EEG signal prediction model includes: Spatiotemporal graph generation processing is performed on the actual deep EEG signals to obtain a deep EEG spatiotemporal graph; The first superficial EEG signal is processed to generate a spatiotemporal map, resulting in a superficial EEG spatiotemporal map. Further, the target EEG signal is processed to generate a spatiotemporal map, resulting in the target EEG spatiotemporal map, including: The target EEG signal is bandpass filtered to obtain a first intermediate signal; The first intermediate signal is subjected to artifact removal to obtain the second intermediate signal; In this embodiment, for a matrix-form EEG signal, the matrix rows correspond to the time dimension of the EEG signal (i.e., each row element corresponds to a point in time within a first time period), while the matrix columns correspond to the channel dimension of the EEG signal (i.e., each column element corresponds to one channel of the EEG signal). Specifically, taking a deep EEG signal as an example, bandpass filtering can be applied to the deep EEG signal (e.g., 0.5-300Hz or 1-600Hz), and a 50Hz harmonic notch filter is added to remove power frequency noise from the deep EEG signal. Normalization is then performed to reduce the influence of individual differences in the subject on the signal amplitude, thereby obtaining the first intermediate signal. Artifact removal can be achieved by marking and removing time segments containing large-amplitude motion artifacts from the first intermediate signal using an artifact detection algorithm to obtain the second intermediate signal. Various artifact detection algorithms exist, and will not be elaborated upon here.
[0033] The target EEG signal is the content of the first superficial EEG signal, which is similar to the content of the aforementioned deep EEG signal, and can be easily deduced by analogy.
[0034] Based on the multi-channel time series corresponding to the target EEG signal, a spatiotemporal map is constructed from the second intermediate signal to obtain the target EEG spatiotemporal map. Further, the step of constructing a spatiotemporal map of the second intermediate signal based on the multi-channel time series corresponding to the target EEG signal to obtain the target EEG spatiotemporal map includes: Obtain the target time window; Based on the target time window, the multi-channel time series is slicing to obtain several time slices; Based on each time slice and the second intermediate signal, a target EEG spatiotemporal map corresponding to each time slice is obtained.
[0035] In this embodiment, the multi-channel time series can be a time series representation of a first time period, which includes several time points sorted along the time axis within the first time period; the target time window can be a sliding window of fixed length, and its window length can be any one of 5 seconds, 10 seconds, 20 seconds, etc.
[0036] Understandably, a sliding slice can be achieved by using a target time window to slide slice a multi-channel time series, resulting in several time slices, each containing several consecutive time points from the time series table. For any given time slice, the matrix-form second intermediate signal can be filtered based on all time points within that time slice to obtain the multi-channel values of the second intermediate signal at each time point in that time slice, constructing an H*W two-dimensional matrix. Here, H is the height of the two-dimensional matrix, indicating the number of channels of the second intermediate signal; W is the width of the two-dimensional matrix, indicating the time step of the second intermediate signal (i.e., the time step indicated by the corresponding time slice). This H*W two-dimensional matrix is then considered as the target EEG spatiotemporal map. Specifically, when the second intermediate signal corresponds to a deep EEG signal, the target EEG spatiotemporal map is a deep EEG spatiotemporal map; or, when the second intermediate signal corresponds to a first superficial EEG signal, the target EEG spatiotemporal map is a superficial EEG spatiotemporal map.
[0037] Based on the deep EEG spatiotemporal map and the shallow EEG spatiotemporal map, the initialized EEG signal prediction model is calibrated to obtain the calibrated EEG signal prediction model.
[0038] Further, the step of calibrating the initialized EEG signal prediction model based on the deep EEG spatiotemporal map and the shallow EEG spatiotemporal map to obtain the calibrated EEG signal prediction model includes: The shallow EEG spatiotemporal map is segmented into spatiotemporal blocks to obtain an embedded vector sequence; Further, the step of performing spatiotemporal graph segmentation on the shallow EEG spatiotemporal graph to obtain an embedding vector sequence includes: The shallow EEG spatiotemporal map is divided into image blocks to obtain several non-overlapping EEG matrix blocks; Projection mapping is performed on each of the EEG matrix blocks to obtain several EEG embedding vectors; Position encoding is performed on all the EEG embedding vectors to obtain the embedding vector sequence.
[0039] In this embodiment, image segmentation can be achieved by dividing the shallow EEG spatiotemporal map into blocks according to a preset segmentation rule, resulting in several non-overlapping EEG matrix blocks, each of the same size. The preset segmentation rule can be that each block covers all channels in the channel dimension and a portion of time points in the time dimension. Specifically, taking a shallow EEG spatiotemporal map as a two-dimensional matrix of H*W as an example, the matrix size of each EEG matrix block can be H*Pw, where Pw is a divisor of W.
[0040] Projection mapping can be achieved by flattening each two-dimensional EEG matrix block into a one-dimensional vector, and then mapping it to the dimension of the hidden layer of the EEG signal prediction model through a learnable linear projection layer to obtain the EEG embedding vector corresponding to each EEG matrix block. Position encoding can be achieved by adding a learnable position vector to each EEG embedding vector, and defining the EEG embedding vector after each position vector embedding as an embedding vector sequence. This allows the model to understand the topological and temporal relationships between different channels of the EEG signal, thereby ensuring that the model can subsequently reconstruct the deep EEG prediction signal based on the input spatiotemporal map.
[0041] Multi-layer spatiotemporal feature extraction is performed on the embedded vector sequence to obtain a spatiotemporal feature sequence; Further, the step of performing multi-layer spatiotemporal feature extraction on the embedded vector sequence to obtain a spatiotemporal feature sequence includes: Obtain the input feature sequence, which is the embedded vector sequence or the output feature sequence in the previous spatiotemporal feature extraction process; Multi-head self-attention feature extraction is performed on the input feature sequence to obtain the intermediate feature sequence; Based on the input feature sequence, feature mapping processing is performed on the intermediate feature sequence to obtain the spatiotemporal feature sequence.
[0042] In this embodiment, the embedded vector sequence can be input into several cascaded feature extraction modules for multi-layer spatiotemporal feature extraction to obtain a spatiotemporal feature sequence. Specifically, for the first feature extraction module, its input can be the embedded vector sequence, and its output can be the output feature sequence from the first spatiotemporal feature extraction process; or, for the second or subsequent feature extraction modules, its input can be the output feature sequence from the previous spatiotemporal feature extraction process. For example, for the third feature extraction module, its input can be the output feature sequence from the second feature extraction module during the second spatiotemporal feature extraction process.
[0043] Understandably, for any spatiotemporal feature extraction process, multi-head self-attention feature extraction can first involve layer normalization of the input feature sequence, then feeding the normalized input feature sequence into a multi-head self-attention (MSA) layer for processing, and finally adding the processed feature sequence element-wise to the input feature sequence to obtain an intermediate feature sequence. Feature mapping can involve layer normalization of the intermediate feature sequence, processing the normalized intermediate feature sequence using a multilayer perceptron (MLP), and then performing element-wise operations on the processed feature sequence and the intermediate feature sequence to obtain the output feature sequence for the current spatiotemporal feature extraction process. Specifically, if the current spatiotemporal feature extraction process is not the last one (i.e., the feature extraction module in the current spatiotemporal feature extraction process is not the last feature extraction module among all feature extraction modules), the process can return to the step of obtaining the input feature sequence; or, if the current spatiotemporal feature extraction process is the last one, the output feature sequence in the current spatiotemporal feature extraction process can be determined as the final time-sequence feature sequence.
[0044] The spatiotemporal feature sequence is recombined and predicted to obtain a deep spatiotemporal prediction map; Based on the deep EEG spatiotemporal map and the deep spatiotemporal prediction map, the parameters of the initialized EEG signal prediction model are updated to obtain the calibrated EEG signal prediction model.
[0045] In this embodiment, feature recombination prediction can involve reshaping each feature vector in the spatiotemporal feature sequence and upsampling the reshaped spatiotemporal feature sequence to obtain a deep spatiotemporal prediction map in matrix form. Parameter updating can be achieved by calculating the mean squared error (MSE) of the deep spatiotemporal prediction map and the deep EEG spatiotemporal map at all time points and channels based on the mean squared error loss (MSE) function, and recording the calculated MSE as the target loss value. Then, based on this target loss value, the model parameters of the initialized EEG signal prediction model are updated using the backpropagation algorithm to obtain a calibrated EEG signal prediction model.
[0046] Step 130: Input the second superficial EEG signal into the calibrated EEG signal prediction model for deep prediction to obtain the deep EEG prediction signal of the subject in the second time period. In the embodiments of this application, in a first implementation, the second superficial EEG signal in the second time period can be input into a calibrated EEG signal prediction model, and the deep EEG signal in the second time period can be predicted by the calibrated EEG signal prediction model, which is denoted as the deep EEG prediction signal. In a second implementation, the spatiotemporal map corresponding to the second superficial EEG signal can be input into the calibrated EEG signal prediction model to obtain the deep EEG prediction signal.
[0047] Step 140: Obtain deep brainwave recording data based on the actual deep brainwave signal and the predicted deep brainwave signal.
[0048] In the embodiments of this application, the actual deep brain electrical signals of the first time period can be integrated with the predicted deep brain electrical signals of the second time period to obtain complete deep brain electrical recording data of the subject.
[0049] Reference Figure 2 The present application proposes a non-invasive / minimally invasive EEG signal recording system for deep brain region activity, comprising: The first processing unit 101 is used to acquire the deep EEG real signal and the first superficial EEG signal collected by the subject in the first time period and under minimally invasive condition, and the second superficial EEG signal collected by the subject in the second time period and under non-invasive condition; the first time period is earlier than the second time period. The second processing unit 102 is used to perform model calibration on the initialized EEG signal prediction model based on the deep EEG real signal and the first shallow EEG signal to obtain the calibrated EEG signal prediction model. The third processing unit 103 is used to input the second superficial EEG signal into the calibrated EEG signal prediction model for deep prediction, and obtain the deep EEG prediction signal of the subject in the second time period. The fourth processing unit 104 is used to obtain deep brainwave recording data based on the actual deep brainwave signal and the predicted deep brainwave signal.
[0050] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.
[0051] Reference Figure 3 This application also provides an electronic device, including: At least one processor 201; At least one memory 202 is used to store at least one program; When the at least one program is executed by the at least one processor 201, the at least one processor 201 implements the method embodiment described above.
[0052] Similarly, it can be understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0053] This application also provides a computer-readable storage medium storing a program executable by a processor 201, which, when executed by the processor 201, is used to implement the above-described method embodiments.
[0054] Similarly, the content of the above method embodiments is applicable to the present computer-readable storage medium embodiments. The specific functions implemented by the present computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.
[0055] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0056] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.
[0057] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.
[0058] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional technology for an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.
[0059] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0060] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0061] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0062] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0063] In the foregoing description of this specification, the references to terms such as "one embodiment," "another embodiment," or "some embodiments," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0064] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.
[0065] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.
Claims
1. A non-invasive / minimally invasive method for recording electroencephalogram (EEG) signals of deep brain region activity, characterized in that, include: The deep EEG signal and the first superficial EEG signal were acquired from the subject in the first time period under minimally invasive conditions, and the second superficial EEG signal was acquired from the subject in the second time period under non-invasive conditions. The first time period is earlier than the second time period; Based on the deep EEG real signal and the first superficial EEG signal, the initialized EEG signal prediction model is calibrated to obtain the calibrated EEG signal prediction model. The second superficial EEG signal is input into the calibrated EEG signal prediction model for deep prediction, and the deep EEG prediction signal of the subject in the second time period is obtained. Deep brainwave recording data is obtained based on the actual deep brainwave signals and the predicted deep brainwave signals.
2. The method according to claim 1, characterized in that, The step of calibrating the initialized EEG signal prediction model based on the deep EEG real signal and the first superficial EEG signal to obtain a calibrated EEG signal prediction model includes: Spatiotemporal graph generation processing is performed on the actual deep EEG signals to obtain a deep EEG spatiotemporal graph; The first superficial EEG signal is processed to generate a spatiotemporal map, resulting in a superficial EEG spatiotemporal map. Based on the deep EEG spatiotemporal map and the shallow EEG spatiotemporal map, the initialized EEG signal prediction model is calibrated to obtain the calibrated EEG signal prediction model.
3. The method according to claim 2, characterized in that, Spatiotemporal graph generation processing is performed on the target EEG signal to obtain the target EEG spatiotemporal graph, including: The target EEG signal is bandpass filtered to obtain a first intermediate signal; The first intermediate signal is subjected to artifact removal to obtain the second intermediate signal; Based on the multi-channel time series corresponding to the target EEG signal, a spatiotemporal map is constructed from the second intermediate signal to obtain the target EEG spatiotemporal map. Wherein, if the target EEG signal is the deep EEG real signal, then the target EEG spatiotemporal map is the deep EEG spatiotemporal map; or, if the target EEG signal is the first shallow EEG signal, then the target EEG spatiotemporal map is the shallow EEG spatiotemporal map.
4. The method according to claim 3, characterized in that, The step of constructing a spatiotemporal map of the second intermediate signal based on the multi-channel time series corresponding to the target EEG signal to obtain the target EEG spatiotemporal map includes: Obtain the target time window; Based on the target time window, the multi-channel time series is slicing to obtain several time slices; Based on each time slice and the second intermediate signal, a target EEG spatiotemporal map corresponding to each time slice is obtained.
5. The method according to claim 2, characterized in that, The step of calibrating the initialized EEG signal prediction model based on the deep EEG spatiotemporal map and the shallow EEG spatiotemporal map to obtain the calibrated EEG signal prediction model includes: The shallow EEG spatiotemporal map is segmented into spatiotemporal blocks to obtain an embedded vector sequence; Multi-layer spatiotemporal feature extraction is performed on the embedded vector sequence to obtain a spatiotemporal feature sequence; The spatiotemporal feature sequence is recombined and predicted to obtain a deep spatiotemporal prediction map; Based on the deep EEG spatiotemporal map and the deep spatiotemporal prediction map, the parameters of the initialized EEG signal prediction model are updated to obtain the calibrated EEG signal prediction model.
6. The method according to claim 5, characterized in that, The step of performing spatiotemporal map segmentation on the shallow EEG spatiotemporal map to obtain an embedding vector sequence includes: The shallow EEG spatiotemporal map is divided into image blocks to obtain several non-overlapping EEG matrix blocks; Projection mapping is performed on each of the EEG matrix blocks to obtain several EEG embedding vectors; Position encoding is performed on all the EEG embedding vectors to obtain the embedding vector sequence.
7. The method according to claim 5, characterized in that, The step of performing multi-layer spatiotemporal feature extraction on the embedded vector sequence to obtain a spatiotemporal feature sequence includes: Obtain the input feature sequence, which is the embedded vector sequence or the output feature sequence in the previous spatiotemporal feature extraction process; Multi-head self-attention feature extraction is performed on the input feature sequence to obtain the intermediate feature sequence; Based on the input feature sequence, feature mapping processing is performed on the intermediate feature sequence to obtain the spatiotemporal feature sequence.
8. A non-invasive / minimally invasive electroencephalogram (EEG) signal recording system for deep brain region activity, characterized in that, include: The first processing unit is used to acquire the deep EEG real signal and the first superficial EEG signal collected by the subject in the first time period and under minimally invasive condition, as well as the second superficial EEG signal collected by the subject in the second time period and under non-invasive condition. The first time period is earlier than the second time period; The second processing unit is used to perform model calibration on the initialized EEG signal prediction model based on the deep EEG real signal and the first shallow EEG signal to obtain the calibrated EEG signal prediction model. The third processing unit is used to input the second superficial EEG signal into the calibrated EEG signal prediction model for deep prediction, and obtain the deep EEG prediction signal of the subject in the second time period. The fourth processing unit is used to obtain deep brainwave recording data based on the actual deep brainwave signal and the predicted deep brainwave signal.
9. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor performs the method as described in any one of claims 1-7.
10. A computer-readable storage medium storing a processor-executable program, characterized in that, The processor-executable program, when executed by the processor, is used to implement the method as described in any one of claims 1-7.