Neural electrophysiological positive problem modeling method, device and electronic equipment

Abstract

The present invention discloses a neuroelectrophysiological positive problem modeling method, device and electronic device, wherein the method comprises: obtaining MRI data of a subject of a target object, and constructing a head geometry model and a head surface measurement electrode distribution model according to the subject MRI data; wherein the head geometry model includes the position information of the fontanelle tissue of the target object; constructing a source model of the target object, and determining a conduction matrix in the positive problem model according to the head geometry model, the head surface measurement electrode distribution model and the source model; constructing a first model according to historical subject data, and constructing a second model according to the conduction matrix, and determining an error model between the first model and the second model; and constructing a third model matching the subject MRI data based on the conduction matrix and the error model. The technical solution provided by the present application can improve the modeling accuracy of the positive problem model.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to a method, device and electronic equipment for modeling a neural electrophysiological positive problem. Background Art

[0002] Neurofunctional imaging with high temporal and spatial resolution is of great significance for infant brain science. To achieve high temporal and spatial resolution neurofunctional imaging (i.e., neuroelectrophysiological source imaging), it is necessary to construct an EEG / magnetic positive problem model. However, few existing neuroelectrophysiological positive problem modeling methods are specifically designed for infants and young children, and most methods are designed for adults. Therefore, under the existing technology, how to construct an accurate model of infant neuroelectrophysiological positive problems is a current problem. Summary of the invention

[0003] In view of this, the embodiments of the present invention provide a method, device and electronic device for modeling a neuro-electrophysiological positive problem, which can improve the modeling accuracy of the positive problem model.

[0004] The present invention provides a neural electrophysiological positive problem modeling method, the method comprising:

[0005] Acquire MRI data of a target subject, and construct a head geometry model and a head surface measurement electrode distribution model based on the MRI data; wherein the head geometry model includes location information of the fontanelle tissue of the target subject;

[0006] Constructing a source model of the target object, and determining a conduction matrix in a positive problem model according to the head geometry model, the head surface measurement electrode distribution model and the source model;

[0007] constructing a first model according to historical test subject data, and constructing a second model according to the conduction matrix, and determining an error model between the first model and the second model;

[0008] Based on the conduction matrix and the error model, a third model matching the subject MRI data is constructed.

[0009] In one embodiment, constructing the brain geometry model includes:

[0010] Performing brain tissue segmentation on the target object to obtain a brain tissue segmentation result, wherein the brain tissue segmentation result includes at least one of scalp, skull, cranial suture, gray matter, white matter, cerebrospinal fluid, eye socket, fontanelle, dura mater, arachnoid mater, and pia mater;

[0011] Segmenting a hydrogel region matching the fontanelle tissue from the subject's MRI data;

[0012] Identifying the outer curved surface of the skull tissue in the brain tissue segmentation result, and using the area covered by the projection of the hydrogel area on the outer curved surface of the skull tissue as the position information of the fontanelle tissue;

[0013] In one embodiment, the fontanelle tissue is identified as follows:

[0014] Determining the normal direction of each point in the hydrogel region, and projecting the hydrogel region onto the outer curved surface of the skull tissue along the normal direction;

[0015] The skull tissue contained in the closed area of ​​the skull outer surface covered by the projection is regarded as the identified fontanelle tissue.

[0016] In one embodiment, the head surface measurement electrode distribution model is constructed in the following manner:

[0017] The hydrogel region of the sensor is segmented from the MRI data of the subject, and the center position of the sensor is identified in combination with the shape and size of the sensor, so as to construct a head surface measurement electrode distribution model according to the center position.

[0018] In one embodiment, constructing the source model of the target object includes:

[0019] Segment the cerebral cortex and reconstruct the surface of the segmented cerebral cortex;

[0020] Based on the surface reconstruction results of the cerebral cortex, the vertex / center of gravity of each triangular facet on the surface is set as an electric dipole.

[0021] In one embodiment, constructing a brain geometry model includes:

[0022] Performing brain tissue segmentation on the target object, and performing cranial suture segmentation based on the brain tissue segmentation result;

[0023] Map the suture segmentation result to the external surface of the skull to obtain the mapping curve of the suture on the external surface of the skull;

[0024] Combining standard statistical data of the fontanelle and a mapping curve of the cranial suture on the outer surface of the skull, the position information of the fontanelle tissue is calculated on the skull;

[0025] A brain geometric structure model is constructed according to the brain tissue segmentation result and the position information of the fontanelle tissue.

[0026] In one embodiment, the error model is determined in the following manner:

[0027] Generate a posterior probability density distribution conditioned on the source model, and convert the posterior probability density distribution into a probability distribution model that conforms to a Gaussian distribution; the probability distribution model includes a depth coefficient of the dipole source;

[0028] The probability distribution model is obtained by calculating the depth coefficient of the dipole source, and the error model is characterized by the probability distribution model.

[0029] Another aspect of the present invention provides a neurophysiological positive problem modeling device, the device comprising:

[0030] A model building unit, used to obtain the MRI data of the target subject, and build a head geometry model and a head surface measurement electrode distribution model according to the MRI data of the target subject; wherein the head geometry model includes the position information of the fontanelle tissue of the target subject;

[0031] a conduction matrix determination unit, configured to construct a source model of the target object and determine a conduction matrix in a positive problem model according to the head geometry model, the head surface measurement electrode distribution model and the source model;

[0032] an error determination unit, configured to construct a first model according to historical test subject data, and to construct a second model according to the conduction matrix, and to determine an error model between the first model and the second model;

[0033] A model building unit is used to build a third model matching the subject MRI data based on the conduction matrix and the error model.

[0034] Another aspect of the present invention provides an electronic device, which includes a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the above-mentioned neurophysiological positive problem modeling method is implemented.

[0035] On the other hand, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the above-mentioned neurophysiological positive problem modeling method is implemented.

[0036] The technical solution provided in this application improves the modeling accuracy of the positive problem model by constructing a personalized real brain geometry model and a head surface electrode distribution model based on the target object's subject MRI and a conduction matrix determined by the source model, and combining the error model determined by the conduction matrix. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0038] Figure 1 A schematic diagram showing the steps of the method for modeling a neural electrophysiological positive problem of the present invention is shown;

[0039] Figure 2 A schematic diagram of functional modules of a neurophysiological positive problem modeling device in one embodiment of the present invention is shown;

[0040] Figure 3 A schematic structural diagram of an electronic device in one embodiment of the present invention is shown. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0042] EEG (non-invasive collection outside the scalp) / EMG has the advantages of no radiation, no invasiveness, high temporal resolution, and bedside access. It is especially suitable for infant brain function analysis and is an important means for infant brain science. However, it is undeniable that the spatial resolution of EEG / EMG is low, and the use of EEG / EMG alone for auxiliary diagnosis has certain limitations. Therefore, in order to solve this problem, high spatial resolution neuroimaging methods such as MR and CT are often combined with EEG / EMG high temporal resolution neuroimaging methods to achieve high temporal and spatial resolution neurofunctional imaging. The above process can be referred to as neuroelectrophysiological source imaging, and its principle is: on the basis of constructing a real head model based on structural images, the electrical activity of the electrophysiological source of the cerebral cortex is inverted and calculated from the potential / magnetic flux generated by the biological neural activity observed on the scalp, thereby completing high temporal and spatial resolution neurofunctional imaging. Among them, the premise of imaging is to construct an EEG / EMG positive problem model, that is, based on the source model, the geometric model of the brain, the distribution model of the head surface measurement electrodes, and the electrical / magnetic permeability of each brain tissue, a model for calculating the potential / magnetic flux distribution on the head surface. Therefore, accurately constructing an EEG / MRI positive problem model is of utmost importance to achieve high temporal and spatial resolution neural functional imaging (i.e., neuroelectrophysiological source imaging).

[0043] However, few existing neuroelectrophysiological positive problem modeling methods are specifically designed for infants and young children, and most methods are designed for adults. According to existing research, infants have special brain structure and organization fontanelles (strictly speaking, the fontanelles include the anterior fontanelle and the posterior fontanelle, and the posterior fontanelle is usually closed shortly after birth. Therefore, the focus is mainly on the anterior fontanelle). Directly applying adult neuroelectrophysiological positive problem modeling and solution methods that do not consider the fontanelle to infants will produce certain errors. Therefore, a neuroelectrophysiological positive problem modeling and solution method for infants is proposed.

[0044] As mentioned above, neuroelectrophysiological source imaging includes two key links: neuroelectrophysiological positive problem modeling and solution. Among them, neuroelectrophysiological positive problem modeling is the premise of neuroelectrophysiological source imaging, and the error of the positive problem model will directly affect the accuracy of source imaging. The existing positive problem modeling methods can be roughly divided into two categories according to whether the real brain geometry is used: the first category is a spherical or ellipsoidal model without a priori real brain geometry; the second category is a personalized positive problem model with a priori real brain geometry. Obviously, the model accuracy of the second category of methods is much higher than that of the first category. Theoretically, the more accurate the priori model of the real brain geometry is, the more accurate the positive problem model is. However, the priori modeling of the real brain geometry is calculated based on the results of brain structure imaging and segmentation modeling, that is, the accuracy of brain structure imaging and segmentation modeling directly affects the accuracy of individual brain structure prior modeling, and can also be said to directly affect the accuracy of neuroelectrophysiological positive problem modeling. For the imaging and segmentation modeling of the brain structure of infants, combining CT and MRI can obtain more accurate results. However, because CT has a certain degree of radiation, it is best not to collect radioactive CT on infants. Therefore, in most cases, MRI can only be used to image and segment the infant's brain structure. However, the existing MRI imaging contrast of the infant's special brain tissue fontanelle is extremely low, and the existing MRI segmentation modeling method cannot achieve the segmentation modeling of the infant's fontanelle, and it is almost impossible to achieve it manually. It can be seen that the accuracy of constructing a real infant brain geometric model based on MRI is restricted by the inability to segment the fontanelle.

[0045] In some application scenarios, such as when examining the brain of an infant, CT is not suitable for infants due to its radiation, and the resolution of magnetic resonance imaging (MRI) is low due to the presence of the fontanelle, so a neurophysiological positive problem model for infants is needed.

[0046] In view of this, the present application proposes a neuroelectrophysiological positive problem modeling method, wherein the expression of the positive problem model can be expressed as Y=AX+E, wherein Y is the modeling result, A is the conduction matrix in the constructed infant neuroelectrophysiological positive problem model, X is the source model in the constructed infant neuroelectrophysiological positive problem model, and E is the error model in the constructed infant neuroelectrophysiological positive problem model.

[0047] See also Figure 1 , a neurophysiological positive problem modeling method provided in one embodiment of the present application may include the following multiple steps.

[0048] S1: Acquire MRI data of a target subject, and construct a head geometry model and a head surface measurement electrode distribution model based on the MRI data; wherein the head geometry model includes location information of the fontanelle tissue of the target subject. The target subject refers to the head of the infant being tested.

[0049] S3: constructing a source model of the target object, and determining a conduction matrix in a positive problem model according to the head geometry model, the head surface measurement electrode distribution model and the source model.

[0050] S5: constructing a first model according to historical test data, and constructing a second model according to the conduction matrix, and determining an error model between the first model and the second model.

[0051] S7: Based on the conduction matrix and the error model, construct a third model that matches the subject's MRI data.

[0052] The conductive matrix construction provided in one embodiment of the present application includes the following steps:

[0053] S21: Acquire the MRI data of the target object. Specifically including:

[0054] Determine the shape, size and number of sensors to be used through the EEG instruction manual;

[0055] Prepare a developing gel that matches the shape and size of the anterior fontanelle and the EEG sensor placed on the infant's head;

[0056] Specifically, the baby's hair is first removed to prepare the skin; the baby's anterior fontanelle and the sensor area are marked with a non-toxic and easily removable colored pen; the marks are recorded with a camera or other method, and the shape and size of the anterior fontanelle and the sensor are extracted based on a computer vision method, thereby preparing the desired hydrogel.

[0057] After the prepared hydrogel is placed at a corresponding position according to shape and size, an MRI scan is performed on the target object.

[0058] S23: Construct a geometric structure model of the brain, including:

[0059] In this embodiment, the brain tissue of the target object is segmented based on existing methods / software (such as fieldtrip, Kangcheng Ruinao, etc.) to obtain brain tissue segmentation results (such as scalp, skull, cranial sutures, gray matter, white matter, cerebrospinal fluid, eye sockets, fontanelles, dura mater, arachnoid mater, pia mater, etc.), and then the inner and outer surfaces of the brain tissue are reconstructed.

[0060] Since the hydrogel of the anterior fontanelle is not on the scalp surface, its imaging result cannot accurately match the actual brain result. Therefore, the area covered by the projection of the hydrogel area on the outer curved surface of the skull tissue is used as the position information of the fontanelle tissue.

[0061] Specifically, segmenting a hydrogel region matching the fontanelle tissue from the subject MRI data;

[0062] Determining the normal direction of each point in the hydrogel region, and projecting the hydrogel region onto the outer curved surface of the skull tissue along the normal direction;

[0063] The outer curved surface of the skull tissue in the brain tissue segmentation result is identified, and the area covered by the projection of the hydrogel area on the outer curved surface of the skull tissue is used as the position information of the fontanelle tissue.

[0064] According to the brain tissue segmentation result and the position information of the fontanelle tissue, the geometric structure model of the brain is constructed.

[0065] S25: Construction of the head surface measurement electrode distribution model, including:

[0066] The hydrogel region of the sensor is segmented from the MRI data of the subject, and the center position of the sensor is identified in combination with the shape and size of the sensor, so as to obtain the electrode distribution at the location of the sensor, and then construct a head surface measurement electrode distribution model.

[0067] S27: Build a source model of the target object.

[0068] The source model is a neuron discharge model. The cerebral cortex is segmented based on software such as Kangcheng Ruinao, and the surface of the segmented cerebral cortex is reconstructed.

[0069] Based on the surface reconstruction results of the cerebral cortex, the vertex / center of gravity of each triangular facet on the surface is set as an electric dipole.

[0070] S29: Based on the above-mentioned brain geometry model, the head surface measurement electrode distribution model, the source model and the electrical / magnetic permeability of each brain tissue, the finite element / boundary element and other calculation methods are used to complete the construction of the conduction matrix in the direct problem according to the electromagnetic field theory.

[0071] In another embodiment of the present application, the conductive matrix construction provided includes the following steps:

[0072] S31: routinely obtaining an MRI of the target object, such as directly performing an MRI scan on the target object.

[0073] S33: Construct a geometric structure model of the brain.

[0074] Based on the acquired MRI, the brain tissue of the target object is segmented, and then the cranial suture is segmented, and the cranial suture segmentation result is mapped to the external surface of the skull to obtain a mapping curve of the cranial suture on the external surface of the skull;

[0075] The approximate size of the fontanelle of the target object and the mapping curve of the cranial suture on the outer surface of the skull are calculated based on standard statistical data corresponding to the statistical values ​​of the fontanelle sizes of a large number of infants of different months of age in history, so as to complete the identification of the position information of the fontanelle tissue on the skull.

[0076] S35: Construction of the head surface measurement electrode distribution model.

[0077] Obtained using standard sensor distribution and MRI registration methods.

[0078] S37: Build a source model of the target object.

[0079] The source model is a neuron discharge model. The cerebral cortex is segmented based on software such as Kangcheng Ruinao, and the surface of the segmented cerebral cortex is reconstructed.

[0080] Based on the surface reconstruction results of the cerebral cortex, the vertex / center of gravity of each triangular facet on the surface is set as an electric dipole.

[0081] S39: Based on the above-mentioned brain geometry model, the head surface measurement electrode distribution model, the source model and the electrical / magnetic permeability of each brain tissue, the finite element / boundary element and other calculation methods are used according to the electromagnetic field theory to complete the construction of the conduction matrix in the direct problem.

[0082] The construction of the error model in the infant neurophysiological positive problem model includes:

[0083] Specifically, in practical applications, the first model can be constructed based on the existing infant CT and MRI data and historical data, and its expression is: where e is the error caused by measurement

[0084] Y=A 6 X+e

[0085] The second model is constructed based on the data obtained from the conduction matrix, and its expression is:

[0086] Y′=A 5 X+e

[0087] Based on the conduction matrix and the error model between the first model and the second model, a third model matching the subject MRI data is constructed, and its expression is: (wherein the sum of the error caused by the model and the measurement error is E=ε+e)

[0088] Y″=A 5 X+(YA 5 X)+e=AX 5 +ε+e=AX 5 +E (1)

[0089] Determination of the error model E:

[0090] The modeling of E can be considered as the posterior probability density distribution P(E|X) conditional on X.

[0091] The posterior probability density distribution can be converted into a probability distribution model that conforms to a Gaussian distribution;

[0092] Assume that P(e) and P(ε|X) obey Gaussian distribution, that is,

[0093] P(e)~N(e * ,σ e ), P(ε|X)~N(ε *|X ,σ ε|X )

[0094] To solve ε *|X and σ ε|X , build

[0095] Z follows a Gaussian distribution, that is, P(z)~N(z * ,σ z ),and

[0096]

[0097] in,

[0098]

[0099] Can get

[0100]

[0101]

[0102] Based on formula (1) and the probability distribution model P(X), the above values ​​can be solved by using the Monte Carlo simulation method. The above probability distribution model P(X) can be constructed according to the properties of the source. For example, the L12 norm can be used to solve the prior distribution, that is,

[0103]

[0104] Among them, ||X i ||2 represents the intensity of the i-th dipole source, α represents the probability distribution coefficient of the source intensity, and w i is the depth coefficient of the i-th dipole source, which is solved as follows:

[0105] [w1…w M ]=diag(A T (AA T ) -1 A)

[0106] See also Figure 2 The present application also provides a neurophysiological positive problem modeling device, the device comprising:

[0107] A model building unit, used to obtain the MRI data of the target subject, and build a head geometry model and a head surface measurement electrode distribution model according to the MRI data of the target subject; wherein the head geometry model includes the position information of the fontanelle tissue of the target subject;

[0108] a conduction matrix determination unit, configured to construct a source model of the target object and determine a conduction matrix in a positive problem model according to the head geometry model, the head surface measurement electrode distribution model and the source model;

[0109] an error determination unit, configured to construct a first model according to historical test subject data, and to construct a second model according to the conduction matrix, and to determine an error model between the first model and the second model;

[0110] A model building unit is used to build a third model matching the subject MRI data based on the conduction matrix and the error model.

[0111] See also Figure 3 The present application also provides an electronic device, which includes a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, a method for modeling a neural electrophysiological positive problem is implemented.

[0112] The present application also provides a computer-readable storage medium, which is used to store a computer program. When the computer program is executed by a processor, the method for modeling a neural electrophysiological positive problem is implemented.

[0113] The technical solution provided in this application improves the modeling accuracy of the positive problem model by constructing a personalized real brain geometry model and a head surface electrode distribution model based on the target object's subject MRI and a conduction matrix determined by the source model, and combining the error model determined by the conduction matrix.

[0114] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A neural electrophysiological positive problem modeling method, characterized in that: The method comprises: Acquire MRI data of a target subject, and construct a head geometry model and a head surface measurement electrode distribution model based on the MRI data; wherein the head geometry model includes location information of the fontanelle tissue of the target subject; Constructing a source model of the target object, and determining a conduction matrix in a positive problem model according to the head geometry model, the head surface measurement electrode distribution model and the source model; constructing a first model according to historical test subject data, and constructing a second model according to the conduction matrix, and determining an error model between the first model and the second model; Based on the conduction matrix and the error model, a third model matching the subject MRI data is constructed.

2. The method according to claim 1, characterized in that Building a geometric model of the mind includes: Performing brain tissue segmentation on the target object to obtain a brain tissue segmentation result; the brain tissue segmentation result includes at least one of scalp, skull, cranial suture, gray matter, white matter, cerebrospinal fluid, eye socket, fontanelle, dura mater, arachnoid mater, and pia mater; Segmenting a hydrogel region matching the fontanelle tissue from the subject's MRI data; Identifying the outer curved surface of the skull tissue in the brain tissue segmentation result, and using the area covered by the projection of the hydrogel area on the outer curved surface of the skull tissue as the position information of the fontanelle tissue; A brain geometric structure model is constructed according to the brain tissue segmentation result and the position information of the fontanelle tissue.

3. The method according to claim 2, characterized in that The fontanelle tissue is identified as follows: Determining the normal direction of each point in the hydrogel region, and projecting the hydrogel region onto the outer curved surface of the skull tissue along the normal direction; The skull tissue contained in the closed area of ​​the skull outer surface covered by the projection is regarded as the identified fontanelle tissue.

4. The method according to claim 1, characterized in that The head surface measurement electrode distribution model is constructed in the following manner: The hydrogel region of the sensor is segmented from the MRI data of the subject, and the center position of the sensor is identified in combination with the shape and size of the sensor, so as to construct a head surface measurement electrode distribution model according to the center position.

5. The method according to claim 1, characterized in that: Constructing the source model of the target object includes: Segment the cerebral cortex and reconstruct the surface of the segmented cerebral cortex; Based on the surface reconstruction results of the cerebral cortex, the vertex / center of gravity of each triangular facet on the surface is set as an electric dipole.

6. The method according to claim 1, characterized in that Building a geometric model of the mind includes: Performing brain tissue segmentation on the target object, and performing cranial suture segmentation based on the brain tissue segmentation result; Map the suture segmentation result to the external surface of the skull to obtain the mapping curve of the suture on the external surface of the skull; Combining standard statistical data of the fontanelle and a mapping curve of the cranial suture on the outer surface of the skull, the position information of the fontanelle tissue is calculated on the skull; A brain geometric structure model is constructed according to the brain tissue segmentation result and the position information of the fontanelle tissue.

7. The method according to claim 1, characterized in that The error model is determined as follows: Generate a posterior probability density distribution conditioned on the source model, and convert the posterior probability density distribution into a probability distribution model that conforms to a Gaussian distribution; the probability distribution model includes a depth coefficient of the dipole source; The probability distribution model is obtained by calculating the depth coefficient of the dipole source, and the error model is characterized by the probability distribution model.

8. A neurophysiological positive problem modeling device, characterized in that: The device comprises: A model building unit, used to obtain the MRI data of the target subject, and build a head geometry model and a head surface measurement electrode distribution model according to the MRI data of the target subject; wherein the head geometry model includes the position information of the fontanelle tissue of the target subject; a conduction matrix determination unit, configured to construct a source model of the target object and determine a conduction matrix in a positive problem model according to the head geometry model, the head surface measurement electrode distribution model and the source model; an error determination unit, configured to construct a first model according to historical test subject data, and to construct a second model according to the conduction matrix, and to determine an error model between the first model and the second model; A model building unit is used to build a third model matching the subject MRI data based on the conduction matrix and the error model.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images