Model training method and device for brain magnetic system and source tracing and positioning method

Abstract

The invention relates to a model training method and device for a brain magnetic system and a traceability positioning method. The model training method comprises the following steps: acquiring a three-dimensional model of the brain of a subject and a relative position relationship between the brain and a sensor array; performing grid division on a preset region of interest in the three-dimensional model to obtain a plurality of simulation source points; generating at least one group of analog source signals at a plurality of analog source points in the three-dimensional model; determining, based on the at least one set of analog source signals, at least one set of analog sensor signals expected to be acquired by the sensor array using a forward model, where the forward model is a mathematical model representing a relationship between electromagnetic activity of a source point in the brain and brain magnetic signals acquired by the sensor array, the forward model is determined based on the relative position relationship; and training the deep learning model by taking at least one group of analog sensor signals as input and at least one group of analog source signals as labels so as to construct a brain magnetic traceability model.

Description

Technical Field

[0001] The present disclosure relates to the field of medical devices, and in particular to a model training method and a model training device for a brain magnetic system, a source tracing and positioning method for a brain magnetic system, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] Magnetoencephalography (MEG) is a brain function detection technology that non-invasively detects the magnetic field generated by the brain's electrical activity. It has millisecond and millimeter time resolutions, and the signal is not affected by tissue conductivity and skull thickness. It can be used to accurately locate electrical activity in the cerebral cortex, that is, brain magnetic tracing and localization, thereby assisting brain tumor resection surgery and epilepsy treatment.

[0003] The physical essence of brain magnetic source tracing and positioning is based on the propagation model of brain magnetic signals generated by brain electrophysiological activities. The brain magnetic signals collected by the brain magnetic sensor array are used for reverse solution to determine the source location and intensity of the electrophysiological activities in the cerebral cortex.

[0004] The brain magnetic signals measured by magnetoencephalography decay exponentially with distance, and the low signal-to-noise ratio of the brain magnetic signals makes source localization in deep brain regions challenging. Since the cerebral cortex is closer to the sensors of the brain magnetic system, a signal with a relatively higher signal-to-noise ratio can be obtained. In related technologies, the source estimation method assumes that most of the activities recorded by MEG come from the cerebral cortex, thus limiting the source space to the surface of the cerebral cortex, and rarely involving deep brain regions. However, deep brain regions including the hippocampus and hypothalamus play an important role in brain function, involving movement, emotion, memory, and physiological regulation. Studying brain magnetic signals in deep brain regions will not only contribute to the progress of basic neuroscience, but also provide important support for clinical applications. Therefore, how to achieve source localization in deep brain regions is a current key issue. Summary of the invention

[0005] According to one aspect of the present disclosure, a model training method for a brain magnetometry system is provided, wherein the brain magnetometry system includes a brain magnetometry helmet and a sensor array fixed on the brain magnetometry helmet, the brain magnetometry helmet is located on the head of a subject, and the method includes: obtaining a three-dimensional model of the subject's brain and a relative positional relationship between the brain and the sensor array; gridding a preset area of ​​interest in the three-dimensional model to obtain a plurality of simulated source points; generating at least one group of simulated source signals at the plurality of simulated source points in the three-dimensional model; based on the at least one group of simulated source signals, using a forward model to determine at least one group of simulated sensor signals expected to be collected by the sensor array, wherein the forward model is a mathematical model representing the relationship between the electromagnetic activity of the source points in the brain and the brain magnetometry signals collected by the sensor array, and the forward model is determined based on the relative positional relationship; and training a deep learning model with the at least one group of simulated sensor signals as input and the at least one group of simulated source signals as labels to construct a brain magnetometry tracing model.

[0006] According to another aspect of the present disclosure, a source tracing and positioning method for a brain magnetometry system is provided, wherein the brain magnetometry system comprises a brain magnetometry helmet and a sensor array fixed on the brain magnetometry helmet, the brain magnetometry helmet is located on the head of a subject to be tested, and the method comprises: obtaining a three-dimensional model of the brain of the subject to be tested; obtaining a sensor signal collected by the sensor array; inputting the sensor signal into a brain magnetometry source tracing model according to the present disclosure to obtain a source signal output by the brain magnetometry source tracing model; and mapping the source signal to a preset area of ​​interest of the three-dimensional model of the brain of the subject to be tested to determine the position information in the source signal.

[0007] According to another aspect of the present disclosure, a model training device for a brain magnetometry system is provided, wherein the brain magnetometry system includes a brain magnetometry helmet and a sensor array fixed on the brain magnetometry helmet, the brain magnetometry helmet is located on the head of a subject, and the device includes: an acquisition module, the acquisition module is configured to acquire a three-dimensional model of the subject's brain and a relative position relationship between the brain and the sensor array; a division module, the division module is configured to grid a preset area of ​​interest in the three-dimensional model to obtain a plurality of simulation source points; a generation module, the generation module is configured to generate a plurality of simulation source points in the three-dimensional model at least one group of simulated source signals is generated at the location; a determination module is configured to determine, based on the at least one group of simulated source signals, at least one group of simulated sensor signals expected to be collected by the sensor array using a forward model, wherein the forward model is a mathematical model representing the relationship between the electromagnetic activity of a source point in the brain and the brain magnetic signals collected by the sensor array, and the forward model is determined based on the relative position relationship; and a training module is configured to train a deep learning model using the at least one group of simulated sensor signals as input and the at least one group of simulated source signals as labels to construct a brain magnetic tracing model.

[0008] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program, and the computer program implements the model training method according to the present disclosure when executed by the at least one processor.

[0009] According to another aspect of the present disclosure, one or more computer-readable storage media are provided, on which instructions are stored, and the instructions, in response to being executed by one or more processors, prompt the one or more processors to execute the model training method according to the present disclosure.

[0010] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program implements the model training method according to the present disclosure when executed by a processor.

[0011] In the present disclosure, a simulated source signal emitted by the brain is simulated by constructing a three-dimensional model, and the simulated sensor signal collected by the sensor array is estimated by a forward model to form a training set to train the deep learning model. The brain magnetic source tracing model formed by training can be used for brain magnetic source tracing, which shows strong robustness in noise suppression, and can reduce the influence of the signal with low signal-to-noise ratio detected by the sensor on the source tracing positioning, thereby improving the accuracy of source tracing positioning, and is suitable for source tracing positioning in deep brain areas. In addition, the brain magnetic source tracing model formed by training is an end-to-end model, which takes the sensor signal of the sensor space as input and the source signal of the brain space as output. It can directly use the signal collected by the sensor of the magnetoencephalogram, and generate a source signal that can be directly mapped to the brain space without any other processing.

[0012] These and other aspects of the disclosure will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Further details, features and advantages of the present disclosure are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0014] Figure 1 A flowchart of a model training method for a magnetoencephalogram system according to some exemplary embodiments of the present disclosure is shown;

[0015] Figure 2 Shows Figure 1 A flowchart of the steps of obtaining a three-dimensional model and a relative position relationship in a model training method;

[0016] Figure 3 A flow chart of a method for tracing and localizing a brain magnetoencephalogram system according to some exemplary embodiments of the present disclosure is shown;

[0017] Figure 4 A schematic block diagram of a model training device for a magnetoencephalogram system according to an exemplary embodiment of the present disclosure is shown;

[0018] Figure 5 A schematic block diagram showing a source tracing and positioning device for a brain magnetoencephalography system according to an exemplary embodiment of the present disclosure is shown; and

[0019] Figure 6 An example configuration of an electronic device is shown that may be used to implement the methods described herein. DETAILED DESCRIPTION

[0020] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present application. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0021] Source localization of magnetoencephalography is an ill-posed inverse problem, that is, estimating brain activity at potentially thousands of locations from sensors at a small number of locations, and an infinite number of source activity patterns may appear as a sensor signal distribution. With the rapid development of deep learning models, it has provided state-of-the-art results for X-ray computed tomography, magnetic resonance image reconstruction, natural image restoration, etc., and has shown strong robustness in noise suppression. Therefore, deep learning models can be used to handle ill-posed inverse problems such as magnetoencephalography source localization.

[0022] In the present disclosure, a simulated source signal emitted by the brain is simulated by constructing a three-dimensional model, and the sensor signal collected by the sensor array is estimated by a forward model to form a training set to train the deep learning model. The brain magnetic source tracing model formed by training can be used for brain magnetic source tracing, which shows strong robustness in noise suppression, and can reduce the influence of the signal with low signal-to-noise ratio detected by the sensor on the source tracing positioning, thereby improving the accuracy of source tracing positioning, and is suitable for source tracing positioning in deep brain regions. In addition, the brain magnetic source tracing model formed by training is an end-to-end model, which takes the sensor signal of the sensor space as input and the source signal of the brain space as output. It can directly use the signal collected by the sensor of the magnetoencephalogram, and generate a source signal that can be directly mapped to the brain space without any other processing.

[0023] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0024] Figure 1A flow chart of a model training method 1000 for a brain magnetic system according to some exemplary embodiments of the present disclosure is shown. Among them, the brain magnetic system may include: a brain magnetic helmet, a sensor array fixed on the brain magnetic helmet, and the brain magnetic helmet is located on the head of the subject. The sensor array is used to collect brain magnetic signals from the brain of the subject to be tested (the brain is also the cerebrum, located in the head). When stimulated, the human brain can generate electromagnetic activity at one or some source field points (i.e., a certain position point in the brain), and the electromagnetic activity can be detected by the sensor array, thereby outputting brain magnetic signals. In the process of brain magnetic source tracing and positioning, the mapping relationship between the source space of the brain of the subject to be tested and the sensor space of the sensor array (for example, a forward model, which describes the relationship between the electromagnetic activity of the source point in the brain and the brain magnetic signal collected by the sensor array) can be reversely solved to obtain the location information of the source field point in the brain that generates the electromagnetic activity (i.e., related to the brain magnetic signal).

[0025] like Figure 1 As shown, the model training method 1000 includes: step S101, obtaining a three-dimensional model of the subject's brain and the relative position relationship between the brain and the sensor array; step S102, gridding the preset interest area in the three-dimensional model to obtain multiple simulated source points; step S103, generating at least one group of simulated source signals at multiple simulated source points in the three-dimensional model; step S104, based on at least one group of simulated source signals, using a forward model to determine at least one group of simulated sensor signals expected to be collected by the sensor array, wherein the forward model is a mathematical model that represents the relationship between the electromagnetic activity of the source point in the brain and the brain magnetic signals collected by the sensor array, and the forward model is determined based on the relative position relationship; and step S105, using at least one group of simulated sensor signals as input and at least one group of simulated source signals as labels, training the deep learning model to construct a brain magnetic tracing model.

[0026] The brain magnetism tracing model formed in the above manner can be used for brain magnetism tracing. For example, the sensor signal collected by the sensor array is input into the brain magnetism tracing model to obtain the source signal corresponding to the sensor signal output by the model, and then the source signal is mapped to the anatomical structure model of the brain to observe the distribution of the source signal, thereby obtaining the location information of the tracing location. Since the brain magnetism tracing model adopts a deep learning model, it shows strong robustness in noise suppression, can reduce the influence of the signal with low signal-to-noise ratio detected by the sensor on the tracing location, thereby improving the accuracy of the tracing location, and is suitable for tracing location in deep brain areas.

[0027] In some embodiments, the three-dimensional model of the brain is an anatomical structure model of the brain, which can be formed by using images of the anatomical structure of the head acquired by medical imaging equipment (e.g., magnetic resonance imaging equipment, computed tomography equipment, etc.).

[0028] In some embodiments, the relative position relationship between the brain and the sensor array may include the relative position relationship between the preset region of interest in the brain and the sensor array. Among them, the preset region of interest may include the whole brain, one of the hippocampus and the thalamus, or other deep brain regions. In the case where the preset region of interest includes the whole brain, the complexity of model training may be increased, but the above-mentioned method of constructing a training set (i.e., generating simulated source signals and simulated sensor signals) and training a deep learning model according to the present disclosure can alleviate the above-mentioned problems.

[0029] In some embodiments, Figure 2 As shown, step S101, obtaining a three-dimensional model of the subject's brain and the relative positional relationship between the brain and the sensor array may include: step S1011, acquiring an anatomical structure image of the subject's head through a medical imaging device; step S1012, performing brain segmentation and reconstruction on the anatomical structure image of the head to generate a three-dimensional model of the brain; and step S1013, aligning the scalp surface of the head with the position of the sensor array to obtain the relative positional relationship between the subject's brain and the sensor array. The medical imaging device may be a nuclear magnetic resonance imaging device, etc. Any method in the prior art may be used to perform brain segmentation and reconstruction on the anatomical structure image acquired by the medical imaging device, for example, a nuclear magnetic resonance imaging image processing method, etc. In some examples, registering the scalp surface of the head with the position of the sensor array may include: performing spatial registration to obtain the positional relationship between the scalp surface of the head and the sensor array (for example, the positional relationship between the scalp surface of the head and the sensor array can be obtained by optical or electromagnetic positioning methods); thereafter, based on the relative positional relationship between the brain magnetic helmet and the sensor array thereon (the brain magnetic helmet and the sensor array thereon are relatively fixed), the positional relationship between the scalp surface of the head and the sensor array can be obtained. In some examples, a three-dimensional model of the brain-sensor array is established based on the anatomical structure image of the head acquired by the medical imaging device and the spatial registration information obtained by registering the scalp surface of the head with the position of the sensor array, so as to determine the relative positional relationship between the brain (specifically, the preset area of ​​interest) and the sensor array according to the three-dimensional model. Alternatively, the relative positional relationship between the brain (specifically, the preset area of ​​interest) and the sensor array can also be determined based on the spatial registration information obtained by registering the scalp surface of the head with the position of the sensor array and the three-dimensional model of the brain.

[0030] In some embodiments, at least one group of simulated source signals may include multiple groups of simulated source signals, and step S103, generating at least one group of simulated source signals at multiple simulated source points in the three-dimensional model may include: performing the step of generating a group of simulated source signals at multiple simulated source points in the three-dimensional model multiple times to obtain multiple groups of simulated source signals. Among them, a group of simulated source signals (or, each group of simulated source signals) may include simulated source signals at each simulated source point in multiple simulated source points. In order to simulate the real electromagnetic activity of each source point in the brain, each time at least part of the simulated source points in the multiple simulated source points are activated, the multiple simulated source points are simulated with simulated source signals. These simulated source signals can be stronger signals, weaker signals, or even signals as weak as zero. Alternatively, a group of simulated source signals may also include simulated source signals at some of the simulated source points in the multiple simulated source points. By performing the step of generating a group of simulated source signals at multiple simulated source points in the three-dimensional model multiple times, multiple groups of simulated source signals can be obtained to facilitate more effective training of the deep learning model.

[0031] In some embodiments, performing the step of generating a set of simulated source signals at multiple simulated source points in the three-dimensional model multiple times to obtain multiple sets of simulated source signals may include: applying a simulated signal to at least one simulated source point among the multiple simulated source points as a seed source each time the step of generating simulated source signals is performed to obtain the simulated source signal at each simulated source point among the multiple simulated source points. The signal generated at the simulated source points around each simulated source point as the seed source decays according to a Gaussian distribution as the distance from the simulated source point increases, and the simulated signal has a time dimension. In order to simulate the real electromagnetic activity of the brain, a region growing method can be used to simulate multiple source points that generate source signals and the generated source signals in a preset area of ​​interest of the three-dimensional model. In other words, at each simulation, a region growing method is used to form at least one source cluster from multiple simulated source points in the three-dimensional model. Each source cluster includes a simulated source point as a seed source and simulated source points around it that meet certain conditions. The simulated source points included in each source cluster may be completely different, or may be partially the same and partially different (at this time, the simulated source signals of the simulated source points simultaneously included in multiple source clusters may be the superposition of signals assigned by multiple source clusters). Specifically, at least one of the multiple simulated source points is used as a seed source to apply a simulated signal to form at least one source cluster, each source cluster includes a seed source (a seed source can be one simulated source point or multiple simulated source points), and the signal generated at the simulated source point in the source cluster decays according to a Gaussian distribution as the distance from the seed source (i.e., the simulated source point as the seed source) included in the source cluster increases (i.e., the signal generated at the simulated source point in the source cluster is distributed with an attenuation pole moment that obeys a Gaussian distribution as the distance from the seed source increases), thereby simulating the source point of electromagnetic activity and the generated signal in a three-dimensional model. A specific way to perform the above steps may be, for example, as follows: at least one source cluster is formed by a regional growth method for each simulation, and the signal generated at the simulated source point in the source cluster is distributed with an attenuation pole moment that obeys a Gaussian distribution as the distance from the seed source increases (each source cluster can be one simulated source point as a seed source, or more than two simulated source points as seed sources), thereby generating multiple sets of data; a time dimension is added to the generated multiple sets of data, and the activation signal of the seed source of the source cluster is a simulated signal containing multiple sampling points. Alternatively, other methods may be used to simulate the simulated source signals generated at multiple simulated source points in the three-dimensional simulation.

[0032] In some embodiments, the analog signal includes at least one of a sinusoidal signal and a center-peak analog signal.

[0033] In some embodiments, step S104, based on at least one group of simulated source signals, using a forward model to determine at least one group of simulated sensor signals that the sensor array is expected to collect may include: based on each group of simulated source signals in multiple groups of simulated source signals, using a forward model to determine a group of simulated sensor signals in multiple groups of simulated sensor signals that corresponds to the group of simulated source signals. In other words, at least one group of simulated source signals in the brain space can be projected into the sensor space through a forward model to estimate the simulated sensor signals corresponding to at least one group of simulated source signals. Specifically, for example, for each group of simulated source signals J, a group of simulated sensor signals Sm=KJ corresponding to the group of simulated source signals J in multiple groups of simulated sensor signals determined using a forward model K. In the above manner, the sensor signals corresponding to the simulated source signals that may be collected by the sensor array can be estimated more accurately.

[0034] In some embodiments, S105, taking at least one group of analog sensor signals as input and at least one group of analog source signals as labels, training the deep learning model to construct a brain magnetism tracing model may include: for each group of analog source signals in multiple groups of analog source signals, taking a group of analog sensor signals corresponding to the group of analog source signals as input and taking the group of analog source signals as labels (that is, output), training the deep learning model to construct a brain magnetism tracing model. That is to say, multiple groups of analog source signals and multiple groups of analog sensor signals exist in pairs in a one-to-one correspondence. When training, each time a group of analog sensor signals corresponding to a group of analog source signals is taken as input, and the group of analog source signals is taken as labels, the deep learning model is trained, so as to facilitate the effective training of the deep learning model. A group of analog sensor signals used for training can be M×Y dimensional, where M is the number of sensors in the sensor array and Y is the number of sampling points in the analog signal. In some other examples, a group of analog sensor signals can also take other dimensions as input according to actual conditions. A set of simulated source signals used for training may be of Q×Y dimensions, where Q is the number of multiple simulated source points obtained by gridding the preset region of interest.

[0035] In some embodiments, the model training method 1000 may further include: adjusting each group of analog sensor signals in at least one group of analog sensor signals using a noise signal. At this time, taking at least one group of analog sensor signals as input and taking at least one group of analog source signals as labels, training the deep learning model to construct a brain magnetism tracing model includes: for each group of analog source signals in at least one group of analog source signals, taking an adjusted group of analog sensor signals corresponding to the group of analog source signals as input and taking the group of analog source signals as labels, training the deep learning model to construct a brain magnetism tracing model. That is, noise can be added to the analog sensor signals to simulate relatively real sensor signals. For example, an adjusted group of analog sensor signals Sm'=Sm+n(t) (where Sm is a group of analog sensor signals Sm=KJ corresponding to a group of analog source signals J determined using a forward model K, and n(t) is a noise signal n(t)). At this time, taking an adjusted group of analog sensor signals Sm' corresponding to the group of analog source signals J as input and taking the group of analog source signals J as labels, the deep learning model is trained.

[0036] In some embodiments, the noise signal includes at least one of Gaussian white noise with different signal-to-noise ratios and signal noise collected by a real sensor.

[0037] In some embodiments, the deep learning model includes a convolutional neural network (CNN), a deep residual network (ResNet), a long short-term memory network (LSTM), VGGNet, a ViT model, etc.

[0038] In the above implementation, the training data set is obtained by computer simulation, thereby ensuring that the data in the training data set is relatively accurate, thereby improving the accuracy of the trained magnetoencephalographic tracing model.

[0039] According to one or more of the above embodiments, taking the preset area of ​​interest as the hippocampus as an example, the model training method may be, for example: obtaining a three-dimensional model of the subject's brain and the relative positional relationship between the hippocampus and the sensor array (including 64 sensors); gridding the hippocampus in the three-dimensional model to obtain 1000 simulated source points; repeatedly executing the step of generating a set of simulated source signals at the 1000 simulated source points in the three-dimensional model to obtain multiple sets of simulated source signals. For example, the specific method of executing the above steps each time may be to form at least one source cluster each time simulation (each source cluster may be a simulated source point as a seed source, or may be more than two simulated source points as seed sources, the source cluster is generated by the regional growth method, and the signal generated at the simulated source point in the source cluster is distributed with the increase of the distance from the seed source. The attenuation pole moment obeys the Gaussian distribution), thereby generating 100,000 sets of data, and then adding the generated data. The time dimension (the activation signal of the seed source of the source cluster is a simulated signal containing 100 sampling points) is used to generate a set of simulated source signals J; for each set of simulated source signals J, a forward model K is used to determine a set of simulated sensor signals Sm=KJ that are expected to be collected by 64 sensors; each set of simulated sensor signals in at least one set of simulated sensor signals is adjusted using a noise signal, that is, a noise signal is added to each set of simulated sensor signals Sm to generate each adjusted set of simulated sensor signals Sm'=KJ+n(t); for each set of simulated source signals J in multiple sets of simulated source signals, a set of adjusted simulated sensor signals Sm' corresponding to the set of simulated source signals J is used as input and the set of simulated source signals J is used as a label to train a deep learning model to construct a brain magnetotelluric tracing model, each set of simulated sensor signals is 64×100 dimensional, and each set of simulated source signals is 1000×100 dimensional. The deep learning model uses a deep residual network (ResNet). The trained EEG source tracing model can predict the data collected by the sensor to obtain the source signal associated with it, that is, output a 1000×100-dimensional source signal. Then, the 1000×100-dimensional source signal can be mapped to the anatomical structure of the hippocampus, and the distribution of the source tracing signal can be observed to perform source tracing positioning.

[0040] Taking the preset area of ​​interest as the thalamus as an example, the model training method may be, for example: obtaining a three-dimensional model of the subject's brain and the relative positional relationship between the thalamus and the sensor array (including 128 sensors); meshing the thalamus in the three-dimensional model to obtain 2000 simulated source points; repeatedly executing the step of generating a set of simulated source signals at the 2000 simulated source points in the three-dimensional model to obtain multiple sets of simulated source signals. For example, the specific manner of executing the above steps each time may be that at least one source cluster is formed each time by simulation (each source cluster may be a simulated source point as a seed source, or may be more than two simulated source points as seed sources, the source cluster is generated by the regional growth method, and the signals generated at the simulated source points in the source cluster are distributed with increasing distance from the seed source, and the attenuation pole moment obeys the Gaussian distribution), thereby generating 100,000 sets of data, and then adding a time dimension to the generated data (the seed source of the source cluster). The activation signal is an analog signal containing 1000 sampling points) to generate a set of analog source signals J; for each set of analog source signals J, a forward model K is used to determine a set of analog sensor signals Sm=KJ that are expected to be collected by 128 sensors; each set of analog sensor signals in at least one set of analog sensor signals is adjusted using a noise signal, that is, a noise signal is added to each set of analog sensor signals Sm to generate each adjusted set of analog sensor signals Sm'=KJ+n(t); for each set of analog source signals J in multiple sets of analog source signals, a set of analog sensor signals Sm' corresponding to the set of analog source signals J is adjusted as input and the set of analog source signals J is used as a label to train a deep learning model to construct a brain magnetotelluric tracing model, each set of analog sensor signals is 128×1000 dimensional, and each set of analog source signals is 2000×1000 dimensional. The deep learning model uses a long short-term memory network (LSTM). The trained EEG source tracing model can predict the data collected by the sensor to obtain the source signal associated with it, that is, output a 2000×1000-dimensional source signal. Then, the 2000×1000-dimensional source signal can be mapped to the anatomical structure of the thalamus, and the distribution of the source tracing signal can be observed to perform source tracing positioning.

[0041] Taking the preset area of ​​interest as the whole brain as an example, the model training method may be, for example: obtaining a three-dimensional model of the subject's brain and the relative positional relationship between the whole brain and the sensor array (including 256 sensors); meshing the whole brain in the three-dimensional model to obtain 2500 simulated source points; repeatedly executing the step of generating a set of simulated source signals at the 2500 simulated source points in the three-dimensional model to obtain multiple sets of simulated source signals. For example, the specific manner of executing the above steps each time may be that at least one source cluster is formed each time by simulation (each source cluster may be a simulated source point as a seed source, or may be more than two simulated source points as seed sources, the source cluster is generated by the regional growth method, and the signals generated at the simulated source points in the source cluster are distributed with increasing distance from the seed source. The attenuation pole moment obeys the Gaussian distribution), thereby generating 500,000 sets of data, and then adding a time dimension to the generated data (the seed source of the source cluster). The activation signal is an analog signal containing 1000 sampling points) to generate a set of analog source signals J; for each set of analog source signals J, a forward model K is used to determine a set of analog sensor signals Sm=KJ that are expected to be collected by 256 sensors; each set of analog sensor signals in at least one set of analog sensor signals is adjusted using a noise signal, that is, a noise signal is added to each set of analog sensor signals Sm to generate each adjusted set of analog sensor signals Sm'=KJ+n(t); for each set of analog source signals J in multiple sets of analog source signals, a set of analog sensor signals Sm' corresponding to the set of analog source signals J is adjusted as input and the set of analog source signals J is used as a label to train a deep learning model to construct a brain magnetotelluric tracing model, each set of analog sensor signals is 256×1000 dimensional, and each set of analog source signals is 2500×1000 dimensional. The deep learning model uses a deep residual network (ResNet). The trained brain magnetotelluric source tracing model can predict the data collected by the sensor to obtain the source signal associated with it, that is, output a 2500×1000-dimensional source signal. Then, the 2500×1000-dimensional source signal can be mapped to the anatomical structure of the whole brain, and the distribution of the tracing signal can be observed to perform source tracing positioning.

[0042] Figure 3 A flow chart of a method 3000 for tracing and locating a brain magnetometry system according to some exemplary embodiments of the present disclosure is shown. The brain magnetometry system includes a brain magnetometry helmet and a sensor array fixed on the brain magnetometry helmet, and the brain magnetometry helmet is located on the head of the subject to be tested. Figure 3As shown, the source tracing and positioning method 3000 includes: step S301, obtaining a three-dimensional model of the brain of the object to be tested; step S302, obtaining the sensor signal collected by the sensor array; step S303, inputting the sensor signal into the brain magnetism tracing model constructed according to the model training method 1000 to obtain the source signal output by the brain magnetism tracing model; and step S304, mapping the source signal to the preset interest area of ​​the three-dimensional model of the brain of the object to be tested to determine the position information in the source signal. Among them, step S301, obtaining the three-dimensional model of the brain of the object to be tested is the same as the method of obtaining the three-dimensional model of the subject's brain in the model training method 1000, which will not be described in detail here. In addition, the preset interest area in the model training method 1000 and the preset interest area in the source tracing and positioning method 3000 are the same area. That is to say, when the preset interest area targeted by the model training method 1000 is, for example, the hippocampus, the constructed brain magnetism tracing model is also targeted at the hippocampus, and the source tracing and positioning method implemented by the brain magnetism tracing model is also targeted at the source tracing and positioning of the hippocampus.

[0043] The above-mentioned source tracing and positioning method adopts the source tracing and positioning model disclosed in the present invention, and thus exhibits strong robustness in noise suppression, can reduce the impact of the low signal-to-noise ratio of the signal detected by the sensor on the source tracing and positioning, thereby improving the accuracy of source tracing and positioning, and is suitable for source tracing and positioning in deep brain areas.

[0044] Figure 4 FIG. 4 shows a schematic block diagram of a model training device 4000 for a brain magnetoencephalogram system according to an exemplary embodiment of the present disclosure. Figure 4 As shown, the source tracing and positioning device 4000 may include an acquisition module 401, a division module 402, a generation module 403, a determination module 404, and a training module 405. The acquisition module 401 is configured to obtain a three-dimensional model of the subject's brain and the relative position relationship between the brain and the sensor array. The division module 402 is configured to grid the preset interest area in the three-dimensional model to obtain multiple simulated source points. The generation module 403 is configured to generate at least one set of simulated source signals at multiple simulated source points in the three-dimensional model. The determination module 404 is configured to determine at least one simulated sensor signal expected to be collected by the sensor array using a forward model based on at least one set of simulated source signals, wherein the forward model is a mathematical model representing the relationship between the electromagnetic activity of the source point in the brain and the brain magnetic signal collected by the sensor array, and the forward model is determined based on the relative position relationship. The training module 405 is configured to take at least one set of simulated sensor signals as input and at least one set of simulated source signals as labels, and train the deep learning model to construct a brain magnetic tracing model.

[0045] It should be understood that Figure 4 The various modules of the device 4000 shown in FIG. Figure 1The steps in the method 1000 described above correspond to each other. Therefore, the operations, features and advantages described above for the method 1000 are also applicable to the apparatus 4000 and the modules included therein. For the sake of brevity, some operations, features and advantages are not described in detail here.

[0046] Figure 5 FIG. 5 shows a schematic block diagram of a source tracing and localization device 5000 for a brain magnetoencephalography system according to an exemplary embodiment of the present disclosure. Figure 5 As shown, the source tracing and positioning device 5000 may include a first acquisition module 501, a second acquisition module 502, an input module 503, and a mapping module 504. The first acquisition module 501 is configured to acquire a three-dimensional model of the brain of the subject to be tested. The second acquisition module 502 is configured to acquire the sensor signal collected by the sensor array. The input module 503 is configured to input the sensor signal into the brain magnetism tracing model to obtain the source signal output by the brain magnetism tracing model. The mapping module 504 is configured to map the source signal to a preset area of ​​interest of the three-dimensional model of the brain of the subject to be tested to determine the position information in the source signal.

[0047] It should be understood that Figure 5 The various modules of the device 5000 shown in FIG. Figure 3 The steps in the method 3000 described above correspond to each other. Therefore, the operations, features and advantages described above for the method 3000 are also applicable to the apparatus 5000 and the modules included therein. For the sake of brevity, some operations, features and advantages are not described in detail here.

[0048] According to yet another aspect of the present disclosure, an electronic device is provided, including: a memory, a processor, and a computer program stored in the memory, wherein the processor is configured to execute the computer program to implement the method 1000 and / or 3000 described above.

[0049] According to yet another aspect of the present disclosure, a non-transitory computer-readable storage medium storing a computer program is provided, wherein the computer program implements the method 1000 and / or 3000 described above when executed by a processor.

[0050] According to yet another aspect of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the method 1000 and / or 3000 described above is implemented.

[0051] In the following, combined Figure 6 Illustrative examples of such a computer device, non-transitory computer-readable storage medium, and computer program product are described.

[0052] Figure 6An example configuration of an electronic device 6000 is shown that can be used to implement the modules and functions described herein.

[0053] The electronic device 6000 can be a variety of different types of devices, such as a server of a service provider, a device associated with a client (e.g., a client device), a system on a chip, and / or any other suitable electronic device or computing system. Examples of the electronic device 6000 include, but are not limited to, a desktop computer, a server computer, a laptop or netbook computer, a mobile device (e.g., a tablet or phablet device, a cellular or other wireless phone (e.g., a smart phone), a notepad computer, a mobile station), a wearable device (e.g., glasses, a watch), an entertainment device (e.g., an entertainment appliance, a set-top box communicatively coupled to a display device, a game console), a television or other display device, a car computer, and the like. Therefore, the electronic device 6000 can range from a full-resource device with a large amount of memory and processor resources (e.g., a personal computer, a game console) to a low-resource device with limited memory and / or processing resources (e.g., a traditional set-top box, a handheld game console).

[0054] The electronic device 6000 may include at least one processor 602, memory 604, communication interface(s) 606, a display device 608, other input / output (I / O) devices 610, and one or more mass storage devices 612 that can communicate with each other, such as via a system bus 614 or other appropriate connection.

[0055] The processor 602 may be a single processing unit or multiple processing units, all of which may include a single or multiple computing units or multiple cores. The processor 602 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuits, and / or any device that manipulates signals based on operating instructions. Among other capabilities, the processor 602 may be configured to obtain and execute computer-readable instructions stored in the memory 604, mass storage device 612, or other computer-readable media, such as program code of an operating system 616, program code of an application program 618, program code of other programs 620, and the like.

[0056] The memory 604 and the mass storage device 612 are examples of computer storage media for storing instructions that are executed by the processor 602 to implement the various functions described above. For example, the memory 604 may generally include both volatile memory and non-volatile memory (e.g., RAM, ROM, etc.). In addition, the mass storage device 612 may generally include a hard drive, a solid-state drive, removable media, including external and removable drives, memory cards, flash memory, floppy disks, optical disks (e.g., CDs, DVDs), storage arrays, network attached storage, storage area networks, etc. The memory 604 and the mass storage device 612 may all be collectively referred to herein as memory or computer storage media, and may be non-transitory media capable of storing computer-readable, processor-executable program instructions as computer program code, which may be executed by the processor 602 as a specific machine configured to implement the operations and functions described in the examples herein.

[0057] A plurality of program modules may be stored on a mass storage device 612. These programs include an operating system 616, one or more application programs 618, other programs 620, and program data 622, and they may be loaded into the memory 604 for execution. Examples of such application programs or program modules may include, for example, computer program logic (e.g., computer program code or instructions) for implementing the following components / functions: a model training device 4000 (acquisition module 401, partitioning module 402, generation module 403, determination module 404, training module 405), a source tracing and positioning device 5000 (a first acquisition module 501, a second acquisition module 502, an input module 503, a mapping module 504), a method 1000 (including any suitable steps of the method 1000), a method 3000 (including any suitable steps of the method 3000), and / or other embodiments described herein.

[0058] Although in Figure 6 6000, but modules 616, 618, 620, and 622, or portions thereof, may be implemented using any form of computer-readable media accessible by the electronic device 6000. As used herein, "computer-readable media" includes at least two types of computer-readable media, namely, computer storage media and communication media.

[0059] Computer storage media includes volatile and nonvolatile, removable and non-removable media implemented by any method or technology for storage of information, such as computer readable instructions, data structures, program modules or other data. Computer storage media includes but is not limited to RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD), or other optical storage devices, magnetic cassettes, magnetic tapes, magnetic disk storage devices or other magnetic storage devices, or any other non-transmission media that can be used to store information for access by an electronic device.

[0060] In contrast, communication media may embody computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism. Computer storage media as defined herein does not include communication media.

[0061] The electronic device 6000 may also include one or more communication interfaces 606 for exchanging data with other devices, such as through a network, direct connection, etc., as discussed above. Such communication interfaces may be one or more of the following: any type of network interface (e.g., a network interface card (NIC)), a wired or wireless (such as an IEEE 802.11 wireless LAN (WLAN)) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth™ interface, a Near Field Communication (NFC) interface, etc. The communication interface 606 may facilitate communication within a variety of network and protocol types, including wired networks (e.g., LAN, cable, etc.) and wireless networks (e.g., WLAN, cellular, satellite, etc.), the Internet, etc. The communication interface 606 may also provide communication with an external storage device (not shown) such as a storage array, a network attached storage, a storage area network, etc.

[0062] In some examples, a display device 608 such as a monitor may be included for displaying information and images to the user. Other I / O devices 610 may be devices that receive various inputs from the user and provide various outputs to the user, and may include a touch input device, a gesture input device, a camera, a keyboard, a remote control, a mouse, a printer, an audio input / output device, and the like.

[0063] Although the present disclosure has been illustrated and described in detail in the drawings and in the foregoing description, such illustration and description should be considered illustrative and exemplary rather than restrictive; the present disclosure is not limited to the disclosed embodiments. By studying the drawings, the disclosure and the appended claims, those skilled in the art will be able to understand and implement variations to the disclosed embodiments when practicing the claimed subject matter. In the claims, the word "comprising" does not exclude other elements or steps that are not listed, the indefinite article "a" or "an" does not exclude a plurality, and the term "plurality" means two or more. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Claims

1. A model training method for a brain magnetoencephalogram system, wherein: The brain magnetic system includes a brain magnetic helmet and a sensor array fixed on the brain magnetic helmet, the brain magnetic helmet is located on the head of the subject, and the method includes: Acquiring a three-dimensional model of the subject's brain and a relative positional relationship between the brain and the sensor array; Meshing a preset region of interest in the three-dimensional model to obtain a plurality of simulation source points; generating at least one set of simulated source signals at a plurality of simulated source points in the three-dimensional model; Based on the at least one set of simulated source signals, using a forward model to determine at least one set of simulated sensor signals expected to be collected by the sensor array, wherein the forward model is a mathematical model representing the relationship between the electromagnetic activities of the source points in the brain and the brain magnetic signals collected by the sensor array, and the forward model is determined based on the relative position relationship; and Taking the at least one set of simulated sensor signals as input and the at least one set of simulated source signals as labels, a deep learning model is trained to construct a brain magnetotelluric tracing model.

2. The method according to claim 1, wherein: The at least one set of simulated source signals includes a plurality of sets of simulated source signals, and generating at least one set of simulated source signals at a plurality of simulated source points in the three-dimensional model includes: The step of generating a set of simulated source signals at a plurality of simulated source points in the three-dimensional model is performed multiple times to obtain the plurality of sets of simulated source signals.

3. The method according to claim 2, wherein: The at least one set of analog sensor signals comprises a plurality of sets of analog sensor signals, and wherein, Determining, based on the at least one set of analog source signals, using a forward model, at least one set of analog sensor signals expected to be collected by the sensor array comprises: Based on each group of analog source signals in the plurality of groups of analog source signals, a group of analog sensor signals corresponding to the group of analog source signals in the plurality of groups of analog sensor signals is determined using a forward model, and Taking the at least one set of simulated sensor signals as input and the at least one set of simulated source signals as labels, training a deep learning model to construct a brain magnetotelluric tracing model includes: For each group of simulated source signals in the multiple groups of simulated source signals, the deep learning model is trained with a group of simulated sensor signals corresponding to the group of simulated source signals as input and the group of simulated source signals as labels to construct a brain magnetotelluric tracing model.

4. The method according to claim 2, wherein: Each group of simulated source signals includes a simulated source signal at each of the multiple simulated source points, and the step of generating a group of simulated source signals at the multiple simulated source points in the three-dimensional model is performed multiple times to obtain the multiple groups of simulated source signals, including: Each time the step of generating an analog source signal is performed, an analog signal is applied with at least one of the multiple analog source points as a seed source to obtain the analog source signal at each of the multiple analog source points, wherein the signal generated at the analog source points around each analog source point serving as the seed source decays according to a Gaussian distribution as the distance from the analog source point increases, and the analog signal has a time dimension.

5. The method according to claim 4, wherein: The analog signal includes at least one of a sinusoidal signal and a center-peak analog signal.

6. The method according to any one of claims 1 to 5, wherein: The method further includes: using a noise signal to adjust each group of analog sensor signals in the at least one group of analog sensor signals, and taking the at least one group of analog sensor signals as input and the at least one group of analog source signals as labels, training a deep learning model to construct a brain magnetotelluric tracing model includes: For each group of the at least one group of analog source signals, a deep learning model is trained with a group of adjusted analog sensor signals corresponding to the group of analog source signals as input and the group of analog source signals as labels to construct a brain magnetoencephalography model.

7. The method according to claim 6, wherein: The noise signal includes at least one of Gaussian white noise with different signal-to-noise ratios and signal noise collected by a real sensor.

8. The method according to any one of claims 1 to 5, wherein: Acquiring a three-dimensional model of the subject's brain and a relative positional relationship between the brain and the sensor array includes: Acquiring an anatomical structure image of the subject's head by a medical imaging device; Performing brain segmentation and reconstruction on the anatomical structure image of the head to generate a three-dimensional model of the brain; and The scalp surface of the head is registered with the position of the sensor array to obtain the relative position relationship between the brain of the subject and the sensor array.

9. The method according to claim 8, wherein: The relative positional relationship between the brain and the sensor array includes: the relative positional relationship between a preset region of interest in the brain and the sensor array, and the preset region of interest includes: one of the whole brain, the hippocampus and the thalamus.

10. A method for tracing and locating a brain magnetoencephalogram system, wherein: The brain magnetism system comprises a brain magnetism helmet and a sensor array fixed on the brain magnetism helmet, wherein the brain magnetism helmet is located on the head of the subject to be tested, and the method comprises: Acquiring a three-dimensional model of the brain of the subject to be tested; Acquiring sensor signals collected by the sensor array; Inputting the sensor signal into the brain magnetometry source tracing model according to any one of claims 1 to 9 to obtain a source signal output by the brain magnetometry source tracing model; and The source signal is mapped to a preset region of interest of the three-dimensional model of the brain of the subject to be tested to determine position information in the source signal.

11. A model training device for a brain magnetoencephalogram system, wherein: The brain magnetic system includes a brain magnetic helmet and a sensor array fixed on the brain magnetic helmet, the brain magnetic helmet is located on the subject's head, and the device includes: an acquisition module, the acquisition module being configured to acquire a three-dimensional model of the subject's brain and a relative positional relationship between the brain and the sensor array; A division module, wherein the division module is configured to perform grid division on a preset area of ​​interest in the three-dimensional model to obtain a plurality of simulated source points; A generating module, the generating module being configured to generate at least one set of simulated source signals at a plurality of simulated source points in the three-dimensional model; a determination module, the determination module being configured to determine, based on the at least one set of simulated source signals, at least one set of simulated sensor signals expected to be collected by the sensor array using a forward model, wherein the forward model is a mathematical model representing the relationship between the electromagnetic activities of source points in the brain and the brain magnetic signals collected by the sensor array, and the forward model is determined based on the relative position relationship; and A training module is configured to train a deep learning model using the at least one set of simulated sensor signals as input and the at least one set of simulated source signals as labels to construct a brain magnetotelluric tracing model.

12. An electronic device, comprising: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores a computer program, which, when executed by the at least one processor, implements the method according to any one of claims 1 to 9.

13. One or more computer-readable storage media having instructions stored thereon, wherein the instructions, in response to being executed by one or more processors, cause the one or more processors to perform the method of any one of claims 1-9.

14. A computer program product comprising a computer program, wherein: The computer program implements the method according to any one of claims 1 to 9 when executed by a processor.

Images