A method for analyzing and processing magnetic resonance images for mental illness examination

Abstract

The present invention relates to the technical field of image analysis, and particularly to a method for analyzing and processing nuclear magnetic resonance images for mental illness examination. An initial group of brain nuclear magnetic resonance images of mental illness patients is obtained, and initial feature points in each image are extracted. By analyzing the similarity of the position distribution of feature points in adjacent images, a group of matching feature points is determined. However, the head movement of the patient during the examination may cause the movement path of the feature points to be uneven, resulting in abnormal feature points. For this reason, the present invention calculates the abnormality index of the initial feature points, distinguishes abnormal feature points from normal feature points, and corrects the positions of the abnormal feature points based on their positional relationship and abnormality index to obtain corrected feature points. Finally, using the normal feature points and the corrected feature points, a corrected group of brain nuclear magnetic resonance images is determined. Thus, it can effectively correct the errors of feature points caused by patient movement, etc., improve the quality of brain nuclear magnetic resonance images, and ensure the reliability of the constructed three-dimensional brain model.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and particularly relates to a method for analyzing and processing nuclear magnetic resonance images for mental illness examination. Background Art

[0002] With the rapid development of modern society, the incidence of mental illness has been increasing year by year, which has seriously affected people's physical and mental health. The diagnosis and treatment of mental illness usually rely on a variety of examination methods. Among them, magnetic resonance imaging (MRI), as a non-invasive and high-resolution medical imaging technology, plays an important role in the research and diagnosis of mental illness. Magnetic resonance imaging can provide detailed high-resolution brain images, allowing doctors and researchers to deeply detect and analyze the changes in brain structure and function related to mental illness, which has far-reaching significance for the scientific research and clinical practice of mental illness.

[0003] When analyzing the brain structure and functional changes of patients using magnetic resonance imaging technology, usually several feature points in magnetic resonance images can be obtained and matched to construct a three-dimensional brain model; however, due to the possible involuntary movement of patients during magnetic resonance scanning, it will cause blurring or ghosting phenomena in magnetic resonance images, resulting in abnormal movement of feature points in magnetic resonance images. Therefore, if the feature points with abnormal movement in magnetic resonance are not corrected and the feature points in the obtained magnetic resonance images are directly matched and a three-dimensional brain model is constructed, the accuracy and reliability of the model will be reduced. Summary of the Invention

[0004] In order to solve the technical problem that if the feature points with abnormal movement in magnetic resonance are not corrected and the feature points in the obtained magnetic resonance images are directly matched and a three-dimensional brain model is constructed, the accuracy and reliability of the model will be reduced, the purpose of the present invention is to provide a method for analyzing and processing nuclear magnetic resonance images for mental illness examination, and the specific technical solution adopted is as follows:

[0005] Obtain an initial group of brain nuclear magnetic resonance images of mental illness patients;

[0006] In each initial brain nuclear magnetic resonance image, obtain initial feature points; in the initial group of brain nuclear magnetic resonance images, analyze the similarity of the position distribution of the initial feature points in every two adjacent initial brain nuclear magnetic resonance images to determine a group of matching feature points in the initial group of brain nuclear magnetic resonance images;

[0007] In the initial group of brain magnetic resonance images, obtain the movement paths of the initial feature points in the matching feature point group; on each movement path, analyze the position changes and path curvature states of the initial feature points, and calculate the anomaly indicators of the initial feature points to distinguish abnormal feature points from normal feature points; in the matching feature point group with abnormal feature points, based on the positional relationship between the abnormal feature points and the normal feature points and the anomaly indicators of the abnormal feature points, determine the corrected feature points corresponding to the abnormal feature points;

[0008] In the initial group of brain magnetic resonance images, based on the normal feature points and the corrected feature points, determine the corrected group of brain magnetic resonance images.

[0009] Furthermore, the method for obtaining the matching feature point group includes:

[0010] Based on the initial group of brain magnetic resonance images, construct a three-dimensional model of the patient's brain, and construct a three-dimensional coordinate system with the centroid of the three-dimensional model of the brain as the origin, so as to obtain the three-dimensional coordinates of each initial feature point in each initial brain magnetic resonance image;

[0011] In each initial brain magnetic resonance image, select an initial feature point as the target feature point, and match the target feature point with each initial feature point in the next adjacent initial brain magnetic resonance image to obtain all feature point pairs of the target feature point;

[0012] In each group of feature point pairs of the target feature point, take the Euclidean distance between the three-dimensional coordinates of the two initial feature points as the distance factor, determine the deviation angle between the two initial feature points according to the relative positional relationship between the two initial feature points, and take the value obtained by performing negative correlation mapping and normalization on the product of the distance factor and the deviation angle between the two initial feature points as the matching coefficient between the two initial feature points;

[0013] In all feature point pairs of the target feature point, take the feature point pair corresponding to the maximum matching coefficient as a group of matching feature point pairs;

[0014] In the initial group of magnetic resonance images, determine all matching feature point groups by synthesizing the corresponding relationships of all matching feature point pairs in every two adjacent initial brain magnetic resonance images.

[0015] Furthermore, the method for obtaining the deviation angle includes:

[0016] In each group of feature point pairs of the target feature point, take the position of the target feature point as the starting point and the position of the other initial feature point as the ending point to obtain the first direction vector;

[0017] In the scanning direction of the initial group of brain magnetic resonance images, obtain a unit vector as the second direction vector;

[0018] Calculate the angular value between the first direction vector and the second direction vector as the deviation angle.

[0019] Further, the method for obtaining the anomaly index includes:

[0020] On each movement path, calculate the curvature value at each initial feature point, and take the product of the curvature value corresponding to each initial feature point and the Euclidean distance between each initial feature point and the adjacent next initial feature point as the change index corresponding to each initial feature point;

[0021] Take the value obtained by performing negative correlation mapping and normalization on the sum of the change indexes corresponding to all initial feature points as the smoothness coefficient of each movement path;

[0022] Among all movement paths, take the group of matching feature points corresponding to the movement paths with smoothness coefficients less than the preset smoothness threshold as the group of abnormal feature points;

[0023] In the group of abnormal feature points, analyze the proportion of the curvature value of each initial feature point among the curvature values of all initial feature points to obtain the anomaly index of each initial feature point.

[0024] Further, differentiating abnormal feature points from normal feature points includes:

[0025] Take the initial feature points with anomaly indexes greater than the preset anomaly threshold as abnormal feature points, and take the initial feature points with anomaly indexes less than or equal to the preset anomaly threshold as normal feature points.

[0026] Further, the method for obtaining the corrected feature points includes:

[0027] On the movement path of the group of matching feature points with abnormal feature points, for any abnormal feature point, take the two normal feature points with the closest Euclidean distance to this abnormal feature point as the target points, and take the straight line passing through the two target points as the target straight line;

[0028] Take the Euclidean distance between the abnormal feature point and the target straight line as the deviation distance;

[0029] Take the product of the anomaly index of each abnormal feature point and the corresponding deviation distance as the deviation index, and take the difference between the deviation distance of each abnormal feature point and the deviation index as the correction distance between each abnormal feature point and the corresponding target straight line;

[0030] Taking each abnormal feature point as the starting point, draw a perpendicular line segment to the corresponding target straight line. On the perpendicular line segment, starting from the foot of the perpendicular, take a line segment with a length equal to the correction distance corresponding to each feature point, and take the endpoint as the corrected feature point corresponding to each abnormal feature point.

[0031] Further, the method for obtaining the corrected brain magnetic resonance image group includes:

[0032] When there are no abnormal feature points in a certain initial corrected brain magnetic resonance image, the initial corrected brain magnetic resonance image is used as the corrected brain magnetic resonance image;

[0033] When there are abnormal feature points in a certain initial corrected brain magnetic resonance image, the abnormal feature points are replaced with the corresponding corrected feature points of the abnormal feature points to obtain the corrected brain magnetic resonance image;

[0034] All the corrected brain magnetic resonance images together form the corrected brain magnetic resonance image group.

[0035] Further, the method for obtaining the initial feature points includes:

[0036] In each initial brain magnetic resonance image, all the initial feature points are identified based on the SIFT feature detection algorithm.

[0037] Further, the value range of the abnormal threshold is (0.5, 1).

[0038] The present invention has the following beneficial effects:

[0039] First, a series of initial brain magnetic resonance images of patients with mental illnesses are obtained to form a group of initial brain magnetic resonance images, and initial feature points in each initial brain magnetic resonance image are extracted. These initial feature points can represent the structures of specific brain regions. Due to the continuity of the initial brain magnetic resonance images in space and time, there are often a large number of similar initial feature points in two adjacent initial brain magnetic resonance images. Therefore, the similarity of the position distributions of the initial feature points can be analyzed to determine the matching feature point group in the group of initial brain magnetic resonance images. During the magnetic resonance scanning process, as images are continuously acquired, the feature points will form a movement path, and the movement path reflects the position changes of the feature points in different images. When a patient with a mental illness undergoes a magnetic resonance examination, due to factors such as anxiety or discomfort, the patient's head may move involuntarily. This unexpected movement will directly affect the imaging quality, resulting in blurred or ghosted images. Specifically, this movement causes abnormal changes in the positions of the initial feature points in the image, resulting in an uneven movement path of these feature points. This unevenness further causes deviations in the positions between the initial feature points on the path and also changes the bending state of the path. Therefore, based on the foregoing features, an abnormal index of the initial feature points can be calculated and used to distinguish abnormal feature points from normal feature points. In the matching feature point group with abnormal feature points, based on the positional relationship between the abnormal feature points and the normal feature points and the abnormal index, the positional deviation of the abnormal feature points is corrected to determine the corrected feature points corresponding to the abnormal feature points. Finally, in the group of initial brain magnetic resonance images, based on the normal feature points and the corrected feature points, a group of corrected brain magnetic resonance images is determined. The present invention can effectively correct the feature point errors caused by reasons such as patient movement, improve the overall quality of the brain magnetic resonance images to a certain extent, and thus ensure the reliability of the finally constructed three-dimensional brain model. Description of the Drawings

[0040] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following described drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0041] Figure 1 It is a flowchart of a method for analyzing and processing magnetic resonance images for mental illness examination provided by an embodiment of the present invention;

[0042] Figure 2 It is a schematic diagram of a movement path provided by an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of a reference point provided by an embodiment of the present invention. Specific implementation manners

[0044] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following specifically describes, in conjunction with the accompanying drawings and preferred embodiments, a nuclear magnetic resonance image analysis and processing method for mental illness examination proposed according to the present invention, including its specific implementation manners, structures, features and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0046] The following specifically describes, in conjunction with the accompanying drawings, the specific solution of a nuclear magnetic resonance image analysis and processing method for mental illness examination provided by the present invention.

[0047] Please refer to Figure 1 , which shows a method flow chart of a nuclear magnetic resonance image analysis and processing method for mental illness examination provided by an embodiment of the present invention. The method includes the following steps:

[0048] Step S1: Obtain an initial group of brain nuclear magnetic resonance images of mental illness patients.

[0049] With the increasing prevalence of mental illness year by year, neuroimaging technology has also developed rapidly and become an important tool in the field of mental illness research and diagnosis. Among them, magnetic resonance imaging (MRI) technology, due to its significant characteristics such as non-invasiveness, simple operation, no radiation and high resolution, has played a key role in the research and diagnosis of mental illness. It allows doctors and researchers to deeply detect and analyze the brain structure and function changes related to mental illness, and has far-reaching significance for the scientific research and clinical practice of mental illness.

[0050] Since patients with mental illnesses may move involuntarily during the brain scan by magnetic resonance imaging (MRI), which can cause the MRI images to be blurred or have ghosting. When constructing a three-dimensional brain model based on the feature points in the MRI images at this time, the accuracy of the final three-dimensional brain model will decrease due to the abnormal movement of some feature points. Therefore, in the embodiments of the present invention, by analyzing the movement of the feature points in the MRI images of the patients, abnormal feature points are screened and corrected to obtain a corrected brain MRI image that can more accurately reflect the brain, thereby effectively improving the accuracy when establishing a three-dimensional brain model subsequently, and helping to accurately reflect the actual brain structure and functional status of patients with mental illnesses.

[0051] Therefore, in the embodiments of the present invention, it is first necessary to obtain an initial set of brain MRI images of patients with mental illnesses. Specifically, a magnetic resonance imaging device is used to scan the brains of patients with mental illnesses to generate a series of high-resolution initial brain MRI images. These initial brain MRI images are obtained through different scanning sequences and slice thicknesses, aiming to capture the subtle structural changes in the brains of patients with mental illnesses. Moreover, every two adjacent initial brain MRI images are continuous in space and time, so they are arranged in a fixed order and jointly form the initial set of brain MRI images.

[0052] In the embodiments of the present invention, the acquisition and obtaining of the initial set of brain MRI images are all authorized by relevant users, and the process does not violate laws and regulations and does not violate public order and good customs.

[0053] Step S2: In each initial brain MRI image, obtain the initial feature points; in the initial set of brain MRI images, analyze the similarity of the position distributions of the initial feature points in every two adjacent initial brain MRI images to determine the matching set of feature points in the initial set of brain MRI images.

[0054] The feature points in the initial brain magnetic resonance images of patients with mental illnesses can represent the structures of specific brain regions. Therefore, it is necessary to extract the initial feature points from each initial brain magnetic resonance image. Since adjacent initial brain magnetic resonance images are continuous in space and time, the differences between adjacent initial brain magnetic resonance images are usually not significant. Therefore, there are often feature points with relatively similar position distributions. Further, when a patient with a mental illness undergoes a magnetic resonance examination, due to factors such as anxiety or discomfort, the patient's head may move involuntarily. This unexpected movement will cause abnormal changes in the positions of the initial feature points in the initial brain magnetic resonance images, affecting the accuracy of subsequent three-dimensional brain model construction. To facilitate the analysis of the movement of the initial feature points in the group of initial brain magnetic resonance images, the initial feature points in the group of initial brain magnetic resonance images can be matched based on the similarity of the position distributions of the initial feature points in adjacent initial brain magnetic resonance images, so as to obtain a group of matching feature points. The initial feature points in a group of matching feature points can be regarded as representing the same brain region.

[0055] First, it is necessary to extract the initial feature points from each initial brain magnetic resonance image.

[0056] Preferably, in an embodiment of the present invention, the method for obtaining the initial feature points includes:

[0057] The Scale Invariant Feature Transform (SIFT) algorithm is an algorithm for extracting and describing local features of images. This algorithm detects feature points in different scale spaces and calculates the main direction of the gradient in the neighborhood of the feature points as the direction feature of the point. Therefore, this algorithm has scale invariance and rotation invariance, and at the same time has a certain stability and robustness to illumination changes, affine transformations, etc. Therefore, in the embodiments of the present invention, the SIFT algorithm can be used to extract the feature points in each initial brain magnetic resonance image as the initial feature points.

[0058] It should be noted that the process of extracting feature points by SIFT is a well-known technical means to those skilled in the art, and the specific process will not be elaborated here.

[0059] After obtaining the initial feature points in each initial brain magnetic resonance image, the similarity of the position distributions of the initial feature points in every two adjacent initial brain magnetic resonance images can be analyzed to determine the group of matching feature points in the group of initial brain magnetic resonance images.

[0060] Preferably, in an embodiment of the present invention, the method for obtaining the group of matching feature points includes:

[0061] Based on the initial group of brain magnetic resonance images, a three-dimensional model of the patient's brain is initially constructed, and a three-dimensional coordinate system is established with the centroid of the three-dimensional brain model as the origin, so as to obtain the three-dimensional coordinates of each initial feature point in each initial brain magnetic resonance image. Each initial brain magnetic resonance image represents a slice of the brain. These slices are usually obtained from different depths and angles of the brain and are connected to each other in space, jointly depicting the overall picture of the brain structure. Therefore, by superimposing and reconstructing these slices, a three-dimensional image of the brain can be generated. The three-dimensional coordinate system ensures that the spatial positions of the initial feature points in different initial brain magnetic resonance images can be uniformly measured, providing a basis for subsequent precise matching.

[0062] Select an initial feature point in each initial brain magnetic resonance image as the target feature point, and match the target feature point with each initial feature point in the next adjacent initial brain magnetic resonance image to obtain all the feature point pairs of the target feature point. For example, for the target feature point in the first initial brain magnetic resonance image, if there are 3 initial feature points in the second initial brain magnetic resonance image, then there are three pairs of feature point pairs for this target point.

[0063] The Euclidean distance is a measure of the straight-line distance between two points, which reflects the degree of proximity of the initial feature points in terms of spatial position. In the three-dimensional brain model, the smaller the Euclidean distance, the more likely it means that the initial feature points correspond to the same structure, and the higher the inter-layer correspondence evaluation, that is, the higher the matching degree. Therefore, in each group of feature point pairs of the target feature point, the Euclidean distance between the three-dimensional coordinates of the two initial feature points is used as the distance factor. The smaller the distance factor between the two initial feature points in a certain group of feature point pairs, the higher the matching degree of these two feature points.

[0064] During the process of performing magnetic resonance imaging (MRI) on a patient's brain, when the scanning direction is determined, the coordinates of the initial feature points representing the same anatomical structure in adjacent initial brain MRI images taken are basically unchanged in the plane perpendicular to the scanning plane. That is, there is no obvious angular deviation between the direction between two initial feature points with a relatively high degree of matching and the scanning direction. Therefore, in this embodiment of the present invention, the deviation angle between two initial feature points can be determined according to the relative position relationship between the two initial feature points: In each pair of feature points of the target feature point, taking the position of the target feature point as the starting point and the position of the other initial feature point as the ending point, a first direction vector is obtained; in the scanning direction of the initial brain MRI image group, a unit vector is obtained as the second direction vector; the angle value between the first direction vector and the second direction vector is calculated as the deviation angle. If the deviation angle corresponding to a certain pair of feature points is smaller, it indicates that there is a smaller angular deviation between the line connecting the two initial feature points in this pair of feature points and the scanning direction, and the two initial feature points in this pair of feature points can be regarded as having a higher degree of matching. The calculation method of the deviation angle can be based on the dot product formula and the inverse cosine function, and the specific process will not be elaborated here.

[0065] Based on the foregoing analysis, it can be seen that in a certain pair of feature points corresponding to the target feature point, if the distance factor between the two initial feature points is smaller, it indicates that the inter-layer correspondence evaluation of the two initial feature points is higher and they are more likely to represent the same brain structure; if the deviation angle corresponding to the two initial feature points is smaller, it also indicates that the matching degree between the two initial feature points is higher. Therefore, calculate the product of the distance factor and the deviation angle between the two initial feature points, and perform a negative correlation mapping and normalization processing on this product to achieve logical relationship correction, thereby obtaining the matching coefficient between the two initial feature points. At this time, the larger the matching coefficient, the higher the matching degree between the two initial feature points. The negative correlation mapping and normalization processing here can be carried out using the formula , where represents the exponential function with the natural constant e as the base, and x represents the independent variable.

[0066] Then, among all pairs of feature points of the target feature point, the pair of feature points corresponding to the maximum matching coefficient can be used as a pair of matching feature points. At this time, through calculation and screening, several pairs of matching feature points can be determined in each pair of adjacent initial brain MRI images, and the inter-layer correspondence evaluation of the two initial feature points in each pair of matching feature points is relatively high and is very likely to represent the same brain region.

[0067] Finally, in the initial nuclear magnetic resonance (NMR) image group, all matching feature point groups are determined by synthesizing the correspondence relationships of all pairs of matching feature points in every two adjacent initial brain NMR images. For example, in the initial NMR image group, there are three initial NMR images, denoted as Image 1, Image 2, and Image 3 respectively. If between Image 1 and Image 2, the initial feature point a (existing in Image 1) and the initial feature point b (existing in Image 2) form a pair of matching feature points, and between Image 2 and Image 3, the initial feature point b (existing in Image 2) and the initial feature point c (existing in Image 3) form a pair of matching feature points, then the initial feature points a, b, and c form a matching feature point group.

[0068] Thus far, the initial feature points between the initial brain NMR images have been matched, and the matching feature point groups in the initial brain NMR image group have been obtained.

[0069] Step S3: In the initial brain NMR image group, obtain the movement paths of the initial feature points in the matching feature point groups; on each movement path, analyze the position change situation and the path bending state of the initial feature points, and calculate the abnormality index of the initial feature points to distinguish abnormal feature points from normal feature points; in the matching feature point groups with abnormal feature points, based on the positional relationship between the abnormal feature points and the normal feature points and the abnormality index of the abnormal feature points, determine the corrected feature points corresponding to the abnormal feature points.

[0070] When a mental illness patient undergoes an NMR examination, the patient's head may move involuntarily. This accidental movement will cause abnormal changes in the positions of the initial feature points in the initial brain NMR images. The abnormal changes will make the movement paths of the initial feature points in different initial brain NMR images become uneven. This uneven characteristic will cause fluctuations in the position change situation of the initial feature points, and the movement paths will also undergo significant bending changes. Therefore, on the movement paths corresponding to each group of matching feature point groups, based on the foregoing features, the abnormality index of the initial feature points can be calculated to distinguish abnormal feature points from normal feature points. The abnormal feature points can be regarded as the initial feature points affected by the patient's own head movement. In order to obtain a brain NMR image that can accurately represent the brain state and structure, and thus ensure the accuracy of the subsequent constructed three-dimensional brain model, in the embodiments of the present invention, in the matching feature point groups with abnormal feature points, based on the relative positional relationship between the normal feature points and the abnormal feature points and the abnormality index of the abnormal feature points, the position deviation situation of the abnormal feature points is corrected to obtain the corrected feature points.

[0071] First, it is necessary to determine the movement paths of the initial feature points in each matching feature point group. Preferably, in an embodiment of the present invention, the method for obtaining the movement path includes:

[0072] In each group of matched feature points, the initial feature points are sequentially and smoothly connected according to the order of the initial brain magnetic resonance images in the group of initial brain magnetic resonance images, so as to obtain the movement paths of the initial feature points in each group of matched feature points.

[0073] In other embodiments of the present invention, in each group of matched feature points, the least squares method can also be used to perform curve fitting on all the initial feature points, so as to obtain a curve, and the curve is used as the movement path, and the trend of the movement path is the same as the arrangement order of the initial brain magnetic resonance images.

[0074] For details, please refer to Figure 2 , which shows a schematic diagram of the movement path in an embodiment of the present invention.

[0075] After obtaining the movement path, the abnormal index of the initial feature points can be further determined based on the bending condition of the movement path and the position change condition of the initial feature points on the movement path. Preferably, in an embodiment of the present invention, the method for obtaining the abnormal index includes:

[0076] The curvature value is an important index for measuring the bending degree of the path. In the initial brain magnetic resonance image, the movement path of the initial feature points is usually relatively smooth. If there is a large bend, it is considered that the initial feature points are affected and have abnormal movement. Therefore, on each movement path, the curvature value at each initial feature point is calculated to quantify the bending degree at each initial feature point.

[0077] Since the unevenness of the movement path will also cause changes in the positions of the initial feature points, which will be intuitively reflected in the distance between the initial feature points. The unevenness will cause an increase in the length of the movement path between the initial feature points. Therefore, the curvature value corresponding to each initial feature point is combined with the Euclidean distance between each initial feature point and the next adjacent initial feature point, and the product of the two is used as the change index corresponding to each initial feature point. At this time, on the movement path, the larger the Euclidean distance between an initial feature point and the next adjacent initial feature point, and the larger the curvature value at this initial feature point, it indicates that the movement path is bent and the smoothness is reduced, and the position of this initial feature point has shifted. Therefore, the larger the change index, the higher the abnormal degree of the initial feature point is considered.

[0078] So far, the change index of each initial feature point on each movement path can be calculated. Then, on each movement path, the value obtained by performing negative correlation mapping and normalization on the sum of the change indices corresponding to all initial feature points is used as the smoothing coefficient of each movement path. The larger the smoothing coefficient corresponding to a certain movement path, the lower the degree of abnormal arrangement of the initial feature points on that movement path. Conversely, the smaller the smoothing coefficient, the more initial feature points with abnormal arrangement exist on that movement path. The negative correlation mapping and normalization method here can adopt the formula , where represents the exponential function with the natural constant e as the base, and x represents the independent variable; or adopt the formula , where x represents the independent variable, represents a preset parameter, whose function is to prevent the denominator from being 0, represents the normalization function, and the preset parameter can take the value of 0.001. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0079] Therefore, among all movement paths, the set of matching feature points corresponding to the movement paths with smoothing coefficients less than the preset smoothing threshold is used as the set of abnormal feature points.

[0080] Finally, in the set of abnormal feature points, evaluate the abnormal degree of each feature point relative to other feature points: The proportion of the curvature value of each initial feature point in the curvature values of all initial feature points can be analyzed. That is, in each set of abnormal feature points, the ratio of the curvature value of each initial feature point to the sum of the curvature values of all initial feature points is used as the abnormal index of each initial feature point. The larger the abnormal index, the higher the abnormal degree of the initial feature point, that is, the more serious the position deviation.

[0081] It should be noted that the preset smoothing threshold is set to 0.7. The specific value can be adjusted according to the actual scenario and is not limited here.

[0082] After obtaining the abnormal indices of all initial feature points, the normal feature points and abnormal feature points can be distinguished based on this index.

[0083] Preferably, in an embodiment of the present invention, the method for distinguishing normal feature points and abnormal feature points includes:

[0084] Based on the foregoing analysis, it can be seen that when the abnormal index of the initial feature point is larger, the higher the abnormal degree of the initial feature point, that is, the more serious the position deviation. Therefore, the initial feature points with abnormal indices greater than the preset abnormal threshold are used as abnormal feature points, and then the initial feature points with abnormal indices less than or equal to the preset abnormal threshold are used as normal feature points.

[0085] Based on the calculation method of the abnormal index, its value ranges from 0 to 1. Therefore, in the embodiments of the present invention, the value range of the preset abnormal threshold is set to (0.5, 1). In this embodiment, the value is 0.65, and the value can be adjusted according to the implementation scenario, which is not limited herein.

[0086] So far, all the initial feature points can be classified into normal feature points and abnormal feature points. The abnormal feature points will deviate from the normal structural positions, thus affecting the accuracy of the initial brain magnetic resonance image. Therefore, by correcting the deviation positions of the abnormal feature points, the errors caused by them can be reduced, making the subsequent analysis process more accurate.

[0087] Since the initial feature points in the matching feature point group have consistency and continuity, which highly likely represent the same structure of the brain, the deviation of the abnormal feature points can be corrected by using the positional relationship between the normal feature points and the abnormal feature points and the abnormal index of the abnormal feature points in the matching feature point group with abnormal feature points, so as to obtain the corrected feature points corresponding to the abnormal feature points.

[0088] Preferably, in an embodiment of the present invention, the method for obtaining the corrected feature points includes:

[0089] On the moving path of the matching feature point group with abnormal feature points, for any abnormal feature point, the two normal feature points with the closest Euclidean distance to the abnormal feature point are used as target points, and the straight line passing through the two target points is used as the target line. The target line can be used as a reference benchmark to evaluate the deviation degree of the abnormal feature point. The Euclidean distance between the abnormal feature point and the target line is used as the deviation distance. The larger the deviation distance, the greater the possible deviation degree of the abnormal feature point in space.

[0090] Based on the acquisition process of the abnormal index, when the abnormal index of the abnormal feature point is larger, it is considered that the abnormal degree of the abnormal feature point is higher and the deviation situation is more serious. Therefore, considering that the abnormal degrees of the abnormal feature points may be different, the abnormal index of the abnormal feature point is introduced as a weight to weight the deviation distance of the abnormal feature point: the product of the abnormal index of each abnormal feature point and the corresponding deviation distance is used as the deviation index. The larger the deviation index, the more serious the position deviation of the abnormal feature point.

[0091] Then, the difference between the deviation distance and the deviation index of each abnormal feature point is used as the correction distance between each abnormal feature point and the corresponding target line. At this time, the larger the deviation index, the more serious the position deviation of the abnormal feature point, so the degree of correction required is higher, and the correction distance is smaller, that is, the distance from the reference benchmark of the target line should be closer, so that the abnormal feature point is closer to the distribution of the normal feature point after correction.

[0092] Finally, taking each abnormal feature point as a starting point, draw a perpendicular line segment to the corresponding target line. On the perpendicular line segment, starting from the foot of the perpendicular, take a line segment with a length equal to the correction distance corresponding to each feature point, and use the endpoint as the corrected feature point corresponding to each abnormal feature point. Thus, the position coordinates of the corrected feature points can be obtained, and therefore their positions in the corresponding initial group of brain magnetic resonance images can be determined. In this way, the deviation of the positions of the abnormal feature points can be corrected to make them closer to the distribution characteristics of the normal feature points, while retaining the relevance of the remaining original positions.

[0093] In other embodiments of the present invention, the method for determining the corrected feature points may also be:

[0094] On the movement path of the matching feature point group with abnormal feature points, if the abnormal feature point is the first point or the last point on the movement path, then it can be processed according to the aforementioned method for determining the corrected feature point. If the abnormal feature point is not the first point or the last point on the movement path, then the corrected feature point is determined by the following method:

[0095] On the movement path, take the two normal feature points adjacent to the abnormal feature point as endpoints to obtain a line segment, and take the midpoint of this line segment as the reference point. Please refer to Figure 3 , which shows a schematic diagram of the reference point in an embodiment of the present invention. The position of the reference point can be regarded as the possible position where the abnormal feature point might be without being affected. Therefore, the Euclidean distance between the abnormal feature point and the reference point is used as the deviation distance of the abnormal feature point. The larger the deviation distance, the greater the possible deviation degree of the abnormal feature point in space.

[0096] Similarly, considering that the abnormal degrees of the abnormal feature points may be different, an abnormal index of the abnormal feature point is introduced as a weight to weight the deviation distance of the abnormal feature point: the product of the abnormal index of the abnormal feature point and the corresponding deviation distance is used as the deviation index. The larger the deviation index, the more serious the position deviation of the abnormal feature point.

[0097] Then, the difference between the deviation distance of the abnormal feature point and the deviation index is used as the correction distance between the abnormal feature point and the corresponding reference point. At this time, the larger the deviation index, the more serious the position deviation of the abnormal feature point, so the degree of correction required is higher, and thus the correction distance is smaller, that is, the distance from the reference point should be closer, so that the abnormal feature point is closer to the distribution of the normal feature points after correction.

[0098] Finally, taking the abnormal feature point as the starting point, draw a line segment with the reference point as the midpoint. On this line segment, take the point with a length distance from the reference point equal to the correction distance as the corrected feature point corresponding to the abnormal feature point.

[0099] Step S4: In the initial brain magnetic resonance image group, based on the normal feature points and the corrected feature points, determine the corrected brain magnetic resonance image group.

[0100] Based on the above steps, the corrected feature points corresponding to the abnormal feature points in each initial brain magnetic resonance image can be obtained. Therefore, the corrected brain magnetic resonance image group can be determined according to the normal feature points and the corrected feature points.

[0101] Preferably, in an embodiment of the present invention, the method for obtaining the corrected brain magnetic resonance image group includes:

[0102] When there are no abnormal feature points in a certain initial corrected brain magnetic resonance image, then use this initial corrected brain magnetic resonance image as the corrected brain magnetic resonance image.

[0103] When there are abnormal feature points in a certain initial corrected brain magnetic resonance image, then replace the abnormal feature points with the corrected feature points corresponding to the abnormal feature points to obtain the corrected brain magnetic resonance image.

[0104] Finally, all the corrected brain magnetic resonance images together form the corrected brain magnetic resonance image group.

[0105] After obtaining the corrected brain magnetic resonance image group, since the feature point offset caused by the slight movement of the patient's head is eliminated in the magnetic resonance image at this time, the three-dimensional brain model constructed based on the brain magnetic resonance images in the corrected brain magnetic resonance image group can more accurately reflect the structure and morphology of the patient's brain, etc., thereby providing a more accurate basis for the doctor's judgment and analysis.

[0106] In summary, a series of initial brain magnetic resonance images of patients with mental disorders are first obtained to form a group of initial brain magnetic resonance images, and initial feature points in each initial brain magnetic resonance image are extracted. These initial feature points can represent the structures of specific brain regions. Due to the continuity of the initial brain magnetic resonance images in space and time, there are often a large number of similar initial feature points in two adjacent initial brain magnetic resonance images. Therefore, the similarity of the position distributions of the initial feature points can be analyzed to determine a group of matching feature points in the group of initial brain magnetic resonance images. During the magnetic resonance scanning process, as images are continuously acquired, the feature points will form a movement path, and the movement path reflects the position changes of the feature points in different images. When a patient with a mental disorder undergoes a magnetic resonance examination, due to factors such as anxiety or discomfort, the patient's head may move involuntarily. This unexpected movement will directly affect the imaging quality, resulting in blurred or ghosted images. Specifically, this movement causes abnormal changes in the positions of the initial feature points in the image, resulting in the movement path of these feature points becoming uneven. This unevenness further causes deviations in the positions between the initial feature points on the path, and the bending state of the path also changes. Therefore, based on the foregoing features, an abnormality index of the initial feature points can be calculated and used to distinguish abnormal feature points from normal feature points. In the group of matching feature points with abnormal feature points, based on the positional relationship between the abnormal feature points and the normal feature points and the abnormality index, the deviation of the positions of the abnormal feature points is corrected to determine the corrected feature points corresponding to the abnormal feature points. Finally, in the group of initial brain magnetic resonance images, based on the normal feature points and the corrected feature points, a group of corrected brain magnetic resonance images is determined. The embodiments of the present invention can effectively correct the errors of feature points caused by reasons such as patient movement, and to a certain extent improve the overall quality of brain magnetic resonance images, thereby ensuring the reliability of the finally constructed three-dimensional brain model.

[0107] It should be noted that the above sequence of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0108] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for analyzing and processing nuclear magnetic resonance images for mental illness examination, characterized in that, The method includes: Obtaining an initial group of brain magnetic resonance images of patients with mental illnesses; In each initial brain magnetic resonance image, obtaining initial feature points; in the initial group of brain magnetic resonance images, analyzing the similarity of the position distributions of the initial feature points in every two adjacent initial brain magnetic resonance images to determine a group of matching feature points in the initial group of brain magnetic resonance images; In the initial group of brain magnetic resonance images, obtaining the movement paths of the initial feature points in the group of matching feature points; on each movement path, analyzing the position change situation and the path bending state of the initial feature points, calculating an abnormal index of the initial feature points to distinguish abnormal feature points from normal feature points; in the group of matching feature points with abnormal feature points, determining corrected feature points corresponding to the abnormal feature points based on the position relationship between the abnormal feature points and the normal feature points and the abnormal index of the abnormal feature points; In the initial group of brain magnetic resonance images, determining a corrected group of brain magnetic resonance images based on the normal feature points and the corrected feature points; The method for obtaining the group of matching feature points includes: Based on the initial group of brain magnetic resonance images, constructing a three-dimensional model of the patient's brain, and constructing a three-dimensional coordinate system with the centroid of the three-dimensional model of the brain as the origin, so as to obtain the three-dimensional coordinates of each initial feature point in each initial brain magnetic resonance image; In each initial brain magnetic resonance image, randomly selecting an initial feature point as a target feature point, and matching the target feature point with each initial feature point in the next adjacent initial brain magnetic resonance image to obtain all feature point pairs of the target feature point; In each group of feature point pairs of the target feature point, taking the Euclidean distance between the three-dimensional coordinates of the two initial feature points as a distance factor, determining the deviation angle between the two initial feature points according to the relative position relationship between the two initial feature points, and taking the value obtained by performing negative correlation mapping and normalization on the product of the distance factor and the deviation angle between the two initial feature points as the matching coefficient between the two initial feature points; In all feature point pairs of the target feature point, taking the feature point pair corresponding to the maximum matching coefficient as a group of matching feature point pairs; In the initial group of brain magnetic resonance images, determining all groups of matching feature points by synthesizing the corresponding relationships of all matching feature point pairs in every two adjacent initial brain magnetic resonance images.

2. The magnetic resonance image analysis and processing method for mental illness examination according to claim 1, wherein The method for obtaining the deviation angle includes: In each group of feature point pairs of the target feature point, taking the position of the target feature point as the starting point and the position of the other initial feature point as the ending point to obtain a first direction vector; In the scanning direction of the initial group of brain magnetic resonance images, obtaining a unit vector as a second direction vector; Calculating the angle value between the first direction vector and the second direction vector as the deviation angle.

3. A magnetic resonance imaging analysis and processing method for mental illness examination according to claim 1, characterized in that, The method for obtaining the movement path includes: In each group of matching feature points, sequentially and smoothly connecting the initial feature points in the order of the initial brain magnetic resonance images, so as to obtain the movement paths of the initial feature points in each group of matching feature points.

4. A method for analyzing and processing magnetic resonance images for mental illness examination according to claim 1, characterized in that, The method for obtaining the abnormal index includes: On each moving path, calculate the curvature value at each initial feature point, and take the product of the curvature value corresponding to each initial feature point and the Euclidean distance between each initial feature point and the adjacent next initial feature point as the change index corresponding to each initial feature point; Take the value obtained by performing negative correlation mapping and normalization on the sum of the change indices corresponding to all initial feature points as the smoothing coefficient of each moving path; Among all moving paths, take the matching feature point groups corresponding to the moving paths with smoothing coefficients less than the preset smoothing threshold as abnormal feature point groups; In the abnormal feature point group, analyze the proportion of the curvature value of each initial feature point among the curvature values of all initial feature points to obtain the abnormality index of each initial feature point.

5. A method for analyzing and processing nuclear magnetic resonance images for mental illness examination according to claim 1, characterized in that, The distinguishing between abnormal feature points and normal feature points includes: Take the initial feature points with abnormality indices greater than the preset abnormality threshold as abnormal feature points, and take the initial feature points with abnormality indices less than or equal to the preset abnormality threshold as normal feature points.

6. A method for analyzing and processing nuclear magnetic resonance images for mental illness examination according to claim 1, characterized in that, The method for obtaining the corrected feature points includes: On the moving path of the matching feature point group where there are abnormal feature points, for any abnormal feature point, take the two normal feature points with the closest Euclidean distance to this abnormal feature point as target points, and take the straight line passing through the two target points as the target straight line; Take the Euclidean distance between the abnormal feature point and the target straight line as the deviation distance; Take the product of the abnormality index of each abnormal feature point and the corresponding deviation distance as the deviation index, and take the difference between the deviation distance of each abnormal feature point and the deviation index as the correction distance between each abnormal feature point and the corresponding target straight line; Taking each abnormal feature point as the starting point, draw a perpendicular line segment to the corresponding target straight line. On the perpendicular line segment, starting from the foot of the perpendicular, take a line segment with a length equal to the correction distance corresponding to each feature point, and take the end point as the corrected feature point corresponding to each abnormal feature point.

7. A method for analyzing and processing nuclear magnetic resonance images for mental illness examination according to claim 1, characterized in that The method for obtaining the corrected brain magnetic resonance image group includes: When there are no abnormal feature points in a certain initial corrected brain magnetic resonance image, take this initial corrected brain magnetic resonance image as the corrected brain magnetic resonance image; When there are abnormal feature points in a certain initial corrected brain magnetic resonance image, use the corrected feature points corresponding to the abnormal feature points to replace the abnormal feature points, thereby obtaining the corrected brain magnetic resonance image; All the corrected brain magnetic resonance images together form the corrected brain magnetic resonance image group.

8. A method for analyzing and processing nuclear magnetic resonance images for mental illness examination according to claim 1, characterized in that, The method for obtaining the initial feature points includes: In each initial brain magnetic resonance image, identify all initial feature points based on the SIFT feature detection algorithm.

9. A method for analyzing and processing nuclear magnetic resonance images for mental illness examination according to claim 5, characterized in that, The value range of the abnormality threshold is (0.5, 1).

Images