Automatic joint registration method for brain MRI image and MEG sensor

Abstract

The invention provides an automatic joint registration method for a brain MRI image and an MEG sensor. The automatic joint registration method comprises the following steps: acquiring three-dimensional face point cloud data; calculating the curvature of each point; performing comprehensive calculation based on the distance and the curvature to obtain a curvature weighted distance based on curvature weighting; sequentially selecting points from the three-dimensional face point cloud data based on the curvature weighted distance to form a sampling point set; extracting feature points based on the sampling point set; and performing registration of the three coordinate systems based on the feature points. According to the configuration, the curvature is introduced when the sampling point set is selected, so that the face salient region can be better focused, the selected sampling points are more representative, and the subsequent registration effect is improved. According to the method, automatic registration of the magnetoencephalogram equipment coordinate system, the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system can be achieved, and the magnetoencephalogram technology can better play the effect in diagnosis and treatment of diseases such as epilepsy.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to an automatic joint registration method for brain MRI images and MEG sensors. Background Art

[0002] Epilepsy is a common chronic neurological disease. Abnormal discharges of brain neurons cause patients to experience recurrent epileptic-like symptoms. In addition to anti-epileptic drugs, surgical resection can also be performed for drug-resistant epilepsy. Therefore, it is crucial to accurately locate the epileptic lesion during preoperative evaluation. The gold standard for localizing epileptic brain areas is intracranial electroencephalography. The signals obtained in this way are more accurate, but invasive methods are often accompanied by certain risks. In clinical testing, electroencephalography (EEG) and magnetoencephalography (MEG) are two commonly used epilepsy detection methods. EEG provides information on epileptic discharge activity by recording changes in scalp potentials, but it is easily affected by tissues such as the scalp and skull, resulting in low spatial resolution. Magnetoencephalography uses superconducting quantum interference devices (SQUIDs) or optically pumped magnetometers (OPMs) to record changes in the magnetic field caused by neuronal discharges, which can more intuitively reflect brain activity and has high temporal and spatial resolutions.

[0003] Compared with electroencephalography, magnetoencephalography has the advantages of non-contact and no electrode-induced errors, and its magnetic field signal is not significantly attenuated by the skull and scalp, so it can capture the neural activity of epileptic lesions more accurately. Source localization analysis based on magnetoencephalography data can provide powerful assistance for surgical operations or precision treatment of epilepsy patients. However, in the process of source localization, the precise position of the magnetoencephalography sensor relative to the brain is crucial to the final result. MEG source imaging requires the position and orientation of the sensor relative to the scalp to be determined in advance. Since the sensor position and the brain anatomical image are usually stored in different coordinate systems, accurate registration must be performed to achieve alignment between the two. In the current commercial MEG source imaging system, all sensors are fixed at known positions in a rigid helmet, and registration is performed by unifying the helmet space and the MRI image space. The conventional operation is to manually move the electromagnetic digitizer to locate some anatomical landmarks on the head to define the head coordinate system. The digitizer can establish a one-to-one correspondence between the located feature points and the feature points obtained from the MRI image, thereby realizing the conversion from the anatomical image to the head coordinate system.

[0004] The existing joint registration methods of sensors and brain MRI images have certain limitations. They usually require cumbersome manual operations, which makes it difficult to ensure the registration accuracy and increase time and labor costs. First, the conventional registration process usually relies on manual operations or customized accessories (such as personalized 3D printed helmets, semi-rigid helmets), and requires manual selection of anatomical landmarks and measurement of sensor depth. These operations are not only time-consuming and complicated, but also require extremely high accuracy, which significantly increases the workload. Secondly, the registration process is susceptible to human errors. For example, the positioning method based on the head position indicator (HPI) coil has high accuracy, but the strict requirements for operation make it extremely vulnerable to the limitations of the small size and easy movement of the coil, and a slight deviation can lead to significant errors. In addition, the method using color marker sensors needs to be manually removed after imaging, which may cause the helmet to deform, further affecting the registration accuracy. Finally, the existing methods have a low degree of automation and still rely on manual intervention or semi-automatic tools. Even with the use of optical scanning and 3D structured light facial surface matching technology, landmarks still need to be manually selected at key steps, limiting processing efficiency and consistency of results.

[0005] Current methods for co-registering MEG sensors with MRI images are expensive in terms of time and manpower and cannot achieve acceptable accuracy. The main obstacle is the tedious manual operation and the human errors it introduces. Here, we automate the entire process to improve efficiency while avoiding human errors.

[0006] In summary, in the prior art, the registration of the MEG device coordinate system, the MEG head coordinate system and the MRI coordinate system is difficult, requiring manual marking of feature points, and limits the effectiveness of MEG technology in the diagnosis and treatment of diseases such as epilepsy. Summary of the invention

[0007] The purpose of the present invention is to provide an automatic joint registration method for brain MRI images and MEG sensors, so as to realize the automatic registration of the magnetoencephalography device coordinate system, the magnetoencephalography head coordinate system and the magnetic resonance imaging coordinate system, solve the registration difficulty problem existing in the prior art, and help the magnetoencephalography technology to better play its effect in the diagnosis and treatment of diseases such as epilepsy.

[0008] In order to solve the above technical problems, the present invention provides an automatic joint registration method for brain MRI images and MEG sensors, the automatic joint registration method is used to uniformly register the magnetoencephalography device coordinate system, the magnetoencephalography head coordinate system and the magnetic resonance imaging coordinate system; the automatic joint registration method includes: acquiring three-dimensional face point cloud data; calculating the curvature of each point; performing a comprehensive calculation based on distance and curvature to obtain a curvature-weighted distance weighted by curvature; selecting points from the three-dimensional face point cloud data in sequence based on the curvature-weighted distance to form a sampling point set; extracting feature points based on the sampling point set; and, based on the feature points, registering the magnetoencephalography device coordinate system, the magnetoencephalography head coordinate system and the magnetic resonance imaging coordinate system.

[0009] Optionally, the step of calculating the curvature of each point includes: selecting a preset number of neighboring points of the point to be calculated to form a local neighborhood, fitting the surface through principal component analysis, and solving the principal curvature using the second-order derivative matrix of the fitted surface k 1 and k 2 The curvature includes the mean curvature H and Gaussian curvature K. H and K are calculated according to the following formula: H = (k 1 +k 2 ) / 2, K = k 1 ·k 2 .

[0010] Optionally, the calculation formula of the curvature weighted distance is d·ω(p i ), where d represents the distance, ω(p i ) represents the weighting factor of the i-th point, ω(p i )=1+αH i +βK i , H i and K i They represent the mean curvature H and Gaussian curvature K of the i-th point respectively, and α and β are hyper parameters.

[0011] Optionally, the step of selecting points from the three-dimensional face point cloud data in sequence based on the curvature weighted distance to form a sampling point set includes: randomly selecting a point in the three-dimensional face point cloud data to form an element of the sampling point set; calculating the curvature weighted distance between the remaining points in the three-dimensional face point cloud data and each point in the current sampling point set, wherein the smallest curvature weighted distance is used as the set distance between this point and the sampling point set; selecting the point with the largest set distance from the remaining points in the three-dimensional face point cloud data and adding it to the sampling point set; and repeating the selection steps until the number of elements in the sampling point set reaches the target number.

[0012] Optionally, the feature points include nose tip feature points, cheek feature points and eye socket feature points; the step of aligning the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system based on the sampling point set includes: selecting the point with the largest curvature as the nose tip feature point; based on the geometric morphology analysis of the point cloud, selecting two feature points in the symmetrical area as the eye socket feature points, and the eye socket feature point is located at the center of the eye socket; and, selecting a point whose curvature is less than a preset curvature and whose distance from the nose tip feature point is greater than a preset distance as the cheek feature point.

[0013] Optionally, the step of aligning the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system based on the sampling point set also includes: adding geometric constraints to assist in extracting the nose tip feature points, the cheek feature points and the eye socket feature points based on geometric symmetry and a preset relative position relationship.

[0014] Optionally, the three-dimensional face point cloud data includes scanned three-dimensional data and three-dimensional reconstructed point cloud data.

[0015] The automatic joint registration method also includes: acquiring the optical scanning three-dimensional data based on a three-dimensional optical scanner; acquiring the three-dimensional reconstructed point cloud data based on magnetic resonance imaging; extracting the feature points of the optical scanning three-dimensional data and setting them as first feature points; extracting the feature points of the three-dimensional reconstructed point cloud data and setting them as second feature points; and, aligning the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system based on the first feature points and the second feature points.

[0016] Optionally, the step of registering the MEG head coordinate system and the MRI coordinate system includes: performing registration based on the first feature point and the second feature point to obtain a rotation matrix R and a translation vector t, and constructing a rigid transformation matrix T:

[0017] The rigid transformation matrix T is applied to the magnetic resonance imaging coordinate system to achieve a rough alignment of the MEG head coordinate system and the magnetic resonance imaging coordinate system; and an iterative closest point algorithm is used to optimize the alignment result to reduce the registration error and improve the accuracy.

[0018] Optionally, the magnetoencephalography device coordinate system and the magnetoencephalography head coordinate system are aligned based on a third feature point preset in the magnetoencephalography device coordinate system and the first feature point.

[0019] Optionally, the step of registering the magnetoencephalography device coordinate system and the magnetoencephalography head coordinate system includes: performing a rough alignment with the first feature point based on the third feature point; and optimizing the alignment result using an iterative closest point algorithm to reduce the registration error and improve the accuracy.

[0020] Compared with the prior art, the present invention provides an automatic joint registration method for brain MRI images and MEG sensors, the automatic joint registration method comprising: obtaining three-dimensional face point cloud data; calculating the curvature of each point; performing comprehensive calculation based on distance and curvature to obtain a curvature-weighted distance weighted by curvature; selecting points from the three-dimensional face point cloud data in sequence based on the curvature-weighted distance to form a sampling point set; extracting feature points based on the sampling point set; and performing registration of three coordinate systems based on the feature points. In such a configuration, curvature is introduced when selecting the sampling point set, so that the facial prominent areas (such as the nose tip and eye sockets) can be better focused, and the selected sampling points are more representative, thereby improving the subsequent registration effect. The present invention can realize the automatic registration of the magnetoencephalogram device coordinate system, the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system, solves the problem of registration difficulty existing in the prior art, and helps the magnetoencephalogram technology to better play its effect in the diagnosis and treatment of diseases such as epilepsy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Those skilled in the art will appreciate that the accompanying drawings are provided for a better understanding of the present invention and do not constitute any limitation on the scope of the present invention. Figure 1 is a flow chart of an automatic joint registration method according to an embodiment of the present invention; Figure 2 It is a schematic diagram of a process of registering a magnetoencephalogram head coordinate system and a magnetic resonance imaging coordinate system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, advantages and features of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the drawings are all in a very simplified form and are not drawn to scale, and are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention. In addition, the structure shown in the drawings is often a part of the actual structure. In particular, the emphasis of each drawing is different, and sometimes different scales are used.

[0023] As used in the present invention, the singular forms "one", "an" and "the" include plural objects, the term "or" is generally used to include the meaning of "and / or", the term "several" is generally used to include the meaning of "at least one", and the term "at least two" is generally used to include the meaning of "two or more". In addition, the terms "first", "second" and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first", "second" and "third" may explicitly or implicitly include one or at least two of the features, "one end" and "the other end" and "the proximal end" and "the distal end" generally refer to two corresponding parts, which include not only the endpoints, and the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral body; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. In addition, as used in the present invention, an element disposed on another element generally only indicates that there is a connection, coupling, cooperation or transmission relationship between the two elements, and the connection, coupling, cooperation or transmission between the two elements may be direct or indirect through an intermediate element, and cannot be understood as indicating or implying the spatial position relationship between the two elements, that is, one element may be in any orientation such as inside, outside, above, below or on one side of another element, unless otherwise clearly indicated in the content. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0024] The core idea of ​​the present invention is to provide an automatic joint registration method for brain MRI images and MEG sensors, so as to realize the automatic registration of the magnetoencephalography device coordinate system, the magnetoencephalography head coordinate system and the magnetic resonance imaging coordinate system, solve the registration difficulty problem existing in the prior art, and help the magnetoencephalography technology to better play its role in the diagnosis and treatment of diseases such as epilepsy.

[0025] The following description is given with reference to the accompanying drawings.

[0026] Please refer to Figure 1 This embodiment provides an automatic joint registration method for brain MRI images and MEG sensors, the automatic joint registration method is used to uniformly register the magnetoencephalography device coordinate system, the magnetoencephalography head coordinate system and the magnetic resonance imaging coordinate system; the automatic joint registration method includes: S10, obtaining three-dimensional face point cloud data.

[0027] S20, calculating the curvature of each point.

[0028] S30, performing comprehensive calculation based on the distance and the curvature to obtain a curvature-weighted distance weighted by the curvature.

[0029] S40, selecting points from the three-dimensional face point cloud data in sequence based on the curvature weighted distance to form a sampling point set.

[0030] S50: extracting feature points based on the sampling point set.

[0031] And, S60, aligning the MEG device coordinate system, the MEG head coordinate system and the MRI coordinate system based on the feature points.

[0032] The present embodiment is described below in the chronological order of data processing, which is mainly divided into the following four stages: 1) data acquisition; 2) alignment of the MEG device coordinate system and the MEG head coordinate system; 3) extraction of feature points such as the nose tip and eye sockets, and rough alignment; 4) fine alignment combined with the ICP algorithm.

[0033] Phase 1: Data collection.

[0034] In step S10, the 3D face point cloud data includes scanned 3D data and 3D reconstructed point cloud data. The optical scanned 3D data is acquired based on a 3D optical scanner; and the 3D reconstructed point cloud data is acquired based on magnetic resonance imaging.

[0035] In MEG studies, different coordinate systems are used to describe the relative spatial positions of the head, device, and anatomical images. The MEG device coordinate system is predefined during the design of the helmet and is used to describe the fixed geometric layout of the sensor array in the device. The MEG head coordinate system is established by aligning with specific anatomical landmarks (such as the tip of the nose and the auricle) based on measurements from an optical scanner or a 3D point cloud of the head surface. The Magnetic Resonance Imaging (MRI) coordinate system is defined by the positions of three reference points in the MRI image (such as the tip of the nose, left and right auricles).

[0036] In order to unify these coordinate systems, it is necessary to collect three-dimensional data in the corresponding coordinate system. First, use an optical scanner to obtain head point cloud data, and save the data in stl format. Then use FreeSufer software to segment the MRI image, extract the scalp to generate a scalp surface model, and convert it into 3D point cloud data. The data is stored in stl format.

[0037] Stage 2: Alignment of the MEG device coordinate system and the MEG head coordinate system.

[0038] In different coordinate systems, there are corresponding feature points. For ease of understanding, these feature points can be called nose tip feature points, cheek feature points and eye socket feature points. The feature points of the optical scanning three-dimensional data are extracted and set as the first feature points; the feature points of the three-dimensional reconstructed point cloud data are extracted and set as the second feature points; the feature points in the magnetoencephalogram device coordinate system are the third feature points, which are preset and defined by the device manufacturer or device user.

[0039] In step S60, the registration between the magnetoencephalography device coordinate system and the magnetoencephalography head coordinate system is relatively easy. First, three-dimensional data is acquired using a three-dimensional scanner. There are feature points on the helmet. Based on these feature points, the three-dimensional data of the designed helmet and the three-dimensional data of the scanned helmet are roughly registered. Then, the ICP algorithm is used for fine registration to achieve the transformation from the magnetoencephalography device coordinate system to the magnetoencephalography head coordinate system.

[0040] That is, the MEG device coordinate system and the MEG head coordinate system are aligned based on the third feature point preset in the MEG device coordinate system and the first feature point. Specifically, the method includes: performing a rough alignment based on the third feature point and the first feature point; and optimizing the alignment result using an iterative closest point algorithm to reduce the alignment error and improve the accuracy.

[0041] Stage 3: Extract feature points such as nose tip and eye sockets, and perform rough registration.

[0042] The registration effect of the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system depends on the selection results of the first feature point and the second feature. Different selection results will affect the registration effect.

[0043] Based on this understanding, this embodiment combines the weighting strategy of facial curvature features to efficiently select representative sampling points from the face point cloud. By introducing curvature features into the sampling process, this method can more accurately focus on prominent facial areas (such as the nose tip, eye sockets, and cheeks), while avoiding the problem of other existing methods that tend to over-select edge points.

[0044] Stage 3.1 Data preprocessing.

[0045] The input is 3D face point cloud data with a shape of (N,3), where N is the number of points in the point cloud, and each point contains 3D coordinates (x, y, z). Define the original point cloud set as P={p1, p2, …, pN}, and select m points from it to form the sampling point set S. Statistical filtering is used to remove isolated points and noise points to ensure the quality of point cloud data, and the point cloud coordinates are normalized to scale the point cloud range to [0, 1] for subsequent calculations.

[0046] Randomly select a point p from the point cloud set P i As the initial point, add the sampling point set S, then S={p i That is, step S40 includes: randomly selecting a point in the three-dimensional face point cloud data to form the sampling point set of an element.

[0047] Stage 3.2 calculates the point cloud curvature.

[0048] For each point p i ∈P, select its k nearest neighbor points to form a local area, fit the surface through principal component analysis (PCA), and use the second-order derivative matrix of the fitted surface to solve the principal curvature k 1 and k 2 , thus obtaining the values ​​of H and K. Points with larger curvature usually correspond to prominent areas of the face, such as the tip of the nose and eye sockets, while points with smaller curvature correspond to flat cheek areas.

[0049] That is, step S20 includes: selecting a preset number of neighboring points of the point to be calculated to form a local neighborhood, fitting the surface through principal component analysis, and solving the principal curvature using the second-order derivative matrix of the fitted surface k 1 and k 2 The curvature includes the mean curvature H and Gaussian curvature K. H and K are calculated according to the following formula: H = (k 1 +k 2 ) / 2, K = k 1 ·k 2 .

[0050] Stage 3.3: Farthest point sampling method based on curvature weighted distance.

[0051] In order to better focus on the prominent areas of the face (such as the nose tip and eye sockets), curvature is introduced as a weighting factor based on the original FPS. The improved sampling steps are as follows: ① Calculate the mean curvature and Gaussian curvature for each point and define the weighting factor:

[0052] Among them, ω(p i ) represents the weighting factor of the i-th point, H i and K i They represent the mean curvature H and Gaussian curvature K of the i-th point respectively. α and β are hyperparameters used to adjust the influence of curvature on sampling.

[0053] ②For each point, calculate its weighted minimum distance to the sampling point set S. The weighted distance makes it easier for points with larger curvature to be selected into the sampling set:

[0054] ③Find d' (p j ,S) the largest point p j For the sake of convenience, d' (p j ,S) is called the set distance of the jth point, and it is added to the sampling point set S. Steps ② and ③ are repeated for iterative sampling until the number of sampling points reaches the target number m.

[0055] ④ Output sampling point set S, S={s 1 ,s 2 ,……,s m} contains m points, which are evenly distributed and more concentrated in the salient areas of the face. Compared with the existing methods, the density of the improved sampling points in areas such as the nose tip and eye sockets is higher, and the proportion of edge points and noise points is significantly reduced.

[0056] That is, step S40 also includes: calculating the curvature weighted distance between the remaining points in the three-dimensional face point cloud data and each point in the current sampling point set, wherein the minimum curvature weighted distance is used as the set distance between this point and the sampling point set; selecting the point with the largest set distance from the remaining points in the three-dimensional face point cloud data and adding it to the sampling point set; and repeating the selection step until the number of elements in the sampling point set reaches the target number.

[0057] Stage 3.4: Feature point extraction and matching.

[0058] That is, step S60 includes: selecting the point with the largest curvature as the nose tip feature point; based on the geometric morphology analysis of the point cloud, selecting two feature points in the symmetrical area as the eye socket feature points, and the eye socket feature point is located at the center of the eye socket; and selecting a point with a curvature less than a preset curvature and a distance from the nose tip feature point greater than a preset distance as the cheek feature point.

[0059] The preset curvature and the preset distance can be a fixed parameter or a parameter calculated in real time based on the current state of each point according to a preset rule. For example, the curvature of the top 5% is taken as the preset curvature. When selecting, if there are multiple candidate points, other auxiliary logics can also intervene and process.

[0060] Furthermore, step S60 also includes: based on geometric symmetry and a preset relative position relationship, adding geometric constraints to assist in extracting the nose tip feature points, the cheek feature points and the eye socket feature points.

[0061] The automatic joint registration method further includes: registering the MEG head coordinate system and the MRI coordinate system based on the first feature point and the second feature point. Specifically, the registration is performed based on the first feature point and the second feature point to obtain a rotation matrix R and a translation vector t, and a rigid transformation matrix T is constructed:

[0062] The rigid transformation matrix T is applied to the magnetic resonance imaging coordinate system to achieve a rough alignment between the MEG head coordinate system and the magnetic resonance imaging coordinate system.

[0063] Stage 4: Combine ICP algorithm for fine registration.

[0064] On the basis of coarse registration, the iterative closest point algorithm (ICP) is used to optimize the alignment results, reduce the registration error and improve the accuracy. In the fine registration process, the ICP parameters are first initialized, including setting the distance threshold of the matching point pair and the maximum number of iterations. Then, the KD-Tree data structure is used to accelerate the matching process of the closest point. Then, in each iteration, the rotation matrix and displacement vector are calculated by singular value decomposition (SVD) to achieve rigid transformation. Finally, the calculated transformation matrix is ​​applied to the mobile point cloud, its position is updated, and the iteration is repeated until the error converges or the maximum number of iterations is reached.

[0065] That is, step S60 also includes: using an iterative closest point algorithm to optimize the alignment result, reduce the registration error and improve the accuracy.

[0066] The process of registering the MEG head coordinate system with the MRI coordinate system can also be based on Figure 2 To understand.

[0067] In summary, this embodiment provides an automatic joint registration method for brain MRI images and MEG sensors, and the automatic joint registration method includes: obtaining three-dimensional face point cloud data; calculating the curvature of each point; performing comprehensive calculation based on distance and curvature to obtain a curvature-weighted distance weighted by curvature; selecting points from the three-dimensional face point cloud data in sequence based on the curvature-weighted distance to form a sampling point set; extracting feature points based on the sampling point set; and, performing registration of three coordinate systems based on the feature points. In this configuration, curvature is introduced when selecting the sampling point set, so that the facial salient areas (such as the nose tip and eye sockets) can be better focused, and the selected sampling points are more representative, thereby improving the subsequent registration effect. This embodiment can realize the automatic registration of the magnetoencephalogram device coordinate system, the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system, solves the problem of registration difficulties in the prior art, and helps magnetoencephalogram technology to better play its effect in the diagnosis and treatment of diseases such as epilepsy.

[0068] The above description is only a description of the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any changes or modifications made by a person skilled in the art in the field of the present invention based on the above disclosure shall fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for automatic co-registration of brain MRI images and MEG sensors, characterized in that: The automatic joint registration method is used to uniformly register the magnetoencephalography device coordinate system, the magnetoencephalography head coordinate system and the magnetic resonance imaging coordinate system; the automatic joint registration method comprises: Obtain 3D face point cloud data; Calculate the curvature at each point; Based on the comprehensive calculation of distance and curvature, the curvature-weighted distance is obtained after weighting based on curvature; Selecting points from the three-dimensional face point cloud data in sequence based on the curvature weighted distance to form a sampling point set; Extracting feature points based on the sampling point set; and, The magnetoencephalography device coordinate system, the magnetoencephalography head coordinate system and the magnetic resonance imaging coordinate system are aligned based on the feature points.

2. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 1, characterized in that: The step of calculating the curvature of each point comprises: Select a preset number of neighboring points of the point to be calculated to form a local neighborhood, fit the surface through principal component analysis, and use the second-order derivative matrix of the fitted surface to solve the principal curvature k 1 and k 2 The curvature includes the mean curvature H and Gaussian curvature K. H and K are calculated according to the following formula: H=(k1+k2) / 2, K= k1·k2.

3. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 2, characterized in that: The calculation formula of the curvature weighted distance is d·ω(p i ), where d represents the distance, ω(p i ) represents the weighting factor of the i-th point, ω(p i )=1+αH i +βK i , H i and K i They represent the mean curvature H and Gaussian curvature K of the i-th point respectively, and α and β are hyper parameters.

4. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 1, characterized in that: The step of selecting points from the three-dimensional face point cloud data in sequence based on the curvature weighted distance to form a sampling point set comprises: Randomly select a point in the three-dimensional face point cloud data to form the sampling point set of an element; The curvature weighted distance is calculated for each point in the current sampling point set and the remaining points in the three-dimensional face point cloud data, wherein the minimum curvature weighted distance is used as the set distance between this point and the sampling point set; Selecting the point with the largest set distance from the remaining points in the three-dimensional face point cloud data and adding it to the sampling point set; and The selection step is repeated until the number of elements in the sampling point set reaches the target number.

5. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 1, characterized in that: The feature points include nose tip feature points, cheek feature points and eye socket feature points; the step of performing registration of the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system based on the sampling point set includes: Selecting the point with the largest curvature as the nose tip feature point; Based on the geometric analysis of the point cloud, two feature points in the symmetrical area are selected as the orbital feature points, and the orbital feature points are located at the center of the orbit; and A point whose curvature is smaller than a preset curvature and whose distance from the nose tip feature point is greater than a preset distance is selected as the cheek feature point.

6. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 5, characterized in that: The step of registering the MEG head coordinate system and the MRI coordinate system based on the sampling point set further includes: Based on geometric symmetry and a preset relative position relationship, geometric constraints are added to assist in extracting the nose tip feature points, the cheek feature points and the eye socket feature points.

7. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 1, characterized in that: The three-dimensional face point cloud data includes three-dimensional scanning data and three-dimensional reconstruction point cloud data; The automatic joint registration method further comprises: Acquiring the optical scanning three-dimensional data based on a three-dimensional optical scanner; Acquiring the three-dimensional reconstructed point cloud data based on magnetic resonance imaging; Extracting the feature points of the optically scanned three-dimensional data and setting them as first feature points; extracting the feature points of the three-dimensional reconstructed point cloud data and setting them as second feature points; and, The magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system are registered based on the first feature point and the second feature point.

8. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 7, characterized in that: The step of registering the magnetoencephalogram head coordinate system and the magnetic resonance imaging coordinate system comprises: Based on the first feature point and the second feature point, the rotation matrix R and the translation vector t are obtained, and the rigid transformation matrix T is constructed: Applying the rigid transformation matrix T to the magnetic resonance imaging coordinate system to achieve a rough alignment between the MEG head coordinate system and the magnetic resonance imaging coordinate system; and The alignment result is optimized using an iterative closest point algorithm, reducing the registration error and improving the accuracy.

9. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 7, characterized in that: The magnetoencephalography device coordinate system and the magnetoencephalography head coordinate system are aligned based on a third feature point preset in the magnetoencephalography device coordinate system and the first feature point.

10. The method for automatic co-registration of brain MRI images and MEG sensors according to claim 9, characterized in that: The step of registering the MEG device coordinate system and the MEG head coordinate system comprises: Performing rough alignment based on the third feature point and the first feature point; and The alignment result is optimized using an iterative closest point algorithm, reducing the registration error and improving the accuracy.

Images