MRI-based method for detecting cognitive impairment in diabetic patients

Abstract

The present invention relates to the field of image processing technology, and in particular to a method for detecting cognitive impairment in diabetic patients based on MRI images, comprising: obtaining a T1 weighted image and a T2 weighted image; obtaining brain pixels according to the positions of the pixels in the two weighted images; obtaining the possibility that the brain pixels are brain matter pixels according to the grayscale representation of the brain pixels in the two weighted images, and further obtaining the tissue representation of each pixel in the two weighted images; obtaining the edge features of the brain pixels according to the pixels in the two weighted images and the tissue representation of each pixel in the two weighted images; denoising the two weighted images according to the edge features of the brain pixels and the possibility that they are brain matter pixels. The present invention analyzes the grayscale representation of the pixels in the two weighted images, combines the unique characteristics of the noise, and assigns smoothing parameters to each pixel, thereby avoiding the loss of detail features.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for detecting cognitive impairment in diabetic patients based on MRI images. Background Art

[0002] Diabetes not only affects the patient's blood sugar metabolism, but is also closely related to cognitive dysfunction. Cognitive dysfunction includes symptoms such as memory loss, lack of concentration, and impaired executive function, which seriously affect the patient's quality of life. Since MRI (magnetic resonance imaging) technology can provide high-resolution brain structure and functional images, the cognitive function of diabetic patients is usually detected based on MRI images of the diabetic patient's brain.

[0003] When collecting MRI images of the patient's brain, the thermal motion of electrons in the electronic equipment of the imaging system will generate thermal noise, so the MRI images need to be denoised. However, when the MRI images are denoised using the traditional non-local mean filtering algorithm, the smoothing parameters in the non-local mean filtering algorithm are fixed, and the detail features contained in different pixels in the MRI images are different. If the MRI images are filtered and denoised with a fixed smoothing parameter, the denoised MRI images will lose some detail features, and the white matter and gray matter in the MRI images cannot be accurately obtained. Summary of the invention

[0004] The present invention provides a method for detecting cognitive impairment in diabetic patients based on MRI images to solve the existing problem that if the MRI images are simply denoised using a non-local mean filtering algorithm, the denoised MRI images will lose some detail features.

[0005] The method for detecting cognitive impairment in diabetic patients based on MRI images of the present invention adopts the following technical scheme:

[0006] The following steps are involved:

[0007] Acquiring an MRI image of the patient's brain, wherein the MRI image includes a T1-weighted image and a T2-weighted image;

[0008] According to the positions of the pixels in the T1-weighted image and the T2-weighted image, the brain pixels are obtained; according to the gray values ​​of the pixels corresponding to the brain pixels in the T1-weighted image and the T2-weighted image, the brain pixels are clustered to obtain a number of clusters; according to the positions of the brain pixels in the clusters and the gray values ​​of the brain pixels in the T1-weighted image and the T2-weighted image, the possibility that the brain pixels are brain matter pixels is obtained;

[0009] According to the difference between each pixel and surrounding pixels in the T1-weighted image and the T2-weighted image, the tissue expression of each pixel in the T1-weighted image and the T2-weighted image is obtained;

[0010] According to the corresponding pixel points of the brain pixel points in the T1-weighted image and the T2-weighted image, a one-dimensional pixel point sequence of two corresponding points of the brain pixel point is obtained; according to the one-dimensional pixel point sequence of the two corresponding points of the brain pixel point, the edge features of the brain pixel point are obtained in combination with the tissue expression of each pixel point in the T1-weighted image and the T2-weighted image;

[0011] According to the edge features of the brain pixels and the possibility that the brain pixels are brain matter pixels, the importance of the brain pixels is obtained; according to the importance of the brain pixels, smoothing parameters are assigned to the pixels corresponding to the brain pixels in the T1-weighted image and the T2-weighted image, and the T1-weighted image and the T2-weighted image are denoised;

[0012] The method of assigning smoothing parameters to the pixels corresponding to the brain pixels in the T1-weighted image and the T2-weighted image, and denoising the T1-weighted image and the T2-weighted image includes the following specific methods:

[0013] Preset an initial smoothing parameter; for any brain pixel point, use the product of the inverse normalized result of the importance of the brain pixel point and the initial smoothing parameter as the smoothing parameter of the corresponding point of the brain pixel point;

[0014] The smoothing parameters of each pixel in the T1-weighted image and the T2-weighted image are obtained, and finally the T1-weighted image and the T2-weighted image are denoised using the non-local mean filtering algorithm and the smoothing parameters of each pixel in the T1-weighted image and the T2-weighted image.

[0015] Preferably, the obtaining of brain pixel points includes the following specific methods:

[0016] The pixel points with the same coordinate position in the T1-weighted image and the T2-weighted image are recorded as the same brain pixel point, and a number of brain pixel points are obtained.

[0017] Preferably, the clustering of brain pixels to obtain a plurality of clusters includes the following specific methods:

[0018] A clustering space was constructed with the grayscale value of brain pixels in T1-weighted images as the horizontal coordinate and the grayscale value in T2-weighted images as the vertical coordinate. All brain pixels were placed in the clustering space, and the distance measurement adopted the Euclidean distance of different brain pixels in the clustering space. The k-means clustering algorithm was used to cluster all brain pixels in the clustering space into several clusters.

[0019] Preferably, the possibility of obtaining brain pixels as brain matter pixels includes the following specific methods:

[0020] For any brain pixel in any cluster, the distance between the brain pixel and the cluster center is obtained, and the possibility that the brain pixel is a brain matter pixel is obtained by combining the grayscale value of the brain pixel in the T1-weighted image and the T2-weighted image. The specific calculation formula is:

[0021]

[0022] In the formula, Indicates the possibility that a brain pixel is a brain-matter pixel; Represents the distance between the brain pixel and the cluster center; Represents the gray value of brain pixels in T1-weighted images; Represents the gray value of brain pixels in T2-weighted images; It represents the absolute value function; Represents an exponential function with a natural constant as base.

[0023] Preferably, the specific method of obtaining the tissue expression of each pixel in the T1-weighted image and the T2-weighted image includes:

[0024] For any pixel in a T1-weighted image or a T2-weighted image, a A local window of size is recorded as the local window of the pixel point; and the tissue expression degree of the pixel point is obtained according to the grayscale value of each pixel point in the local window of the pixel point.

[0025] Preferably, the specific calculation formula for obtaining the tissue expression degree of the pixel point is:

[0026]

[0027] In the formula, Indicates The organizational representation of each pixel; Indicates The grayscale mean of all pixels in the local window of pixels; Indicated in The grayscale mean of all pixels except the central pixel in the local window of pixels; Indicated in In the local window of pixels, except for the The grayscale mean of all pixels except pixels; Indicates The gray value of each pixel; Indicates The local window of the pixel The gray value of each pixel; Indicates The local window of the pixel The gray value of each pixel; Indicates The number of pixels in a local window of pixels; It represents the absolute value function; Expressing the Traverse all pixels in the local window of pixels and take the maximum value; Represents an exponential function with a natural constant as base.

[0028] Preferably, the one-dimensional pixel point sequence of two corresponding points of brain pixel points is obtained by the specific method of:

[0029] For any brain pixel point, the pixel point corresponding to the brain pixel point in the T1-weighted image and the T2-weighted image is recorded as the corresponding point of the brain pixel point; the local windows of the two corresponding points are respectively obtained, and the local windows of the two corresponding points are respectively scanned by the serpentine scanning method to obtain a one-dimensional pixel point sequence of the two corresponding points of the brain pixel point.

[0030] Preferably, the method of obtaining edge features of brain pixels includes:

[0031] For any brain pixel, the edge feature of the brain pixel is obtained according to the one-dimensional pixel sequence of the two corresponding points of the brain pixel and the tissue expression of each pixel in the one-dimensional pixel sequence of the two corresponding points of the brain pixel. The specific calculation formula is:

[0032]

[0033] In the formula, Represents the edge features of brain pixels; The number of pixels in the one-dimensional pixel sequence representing the corresponding points of the brain pixels; The first corresponding point in the one-dimensional pixel sequence representing the brain pixel The gray value of each pixel; The grayscale mean of all pixels in the one-dimensional pixel sequence representing the first corresponding point of the brain pixel; The first corresponding point in the one-dimensional pixel sequence representing the brain pixel The organizational representation of each pixel; The first pixel in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel The gray value of each pixel; The grayscale mean of all pixels in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel; The first pixel in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel The organizational representation of each pixel; The standard deviation of the grayscale of all pixels in the one-dimensional pixel sequence representing the first corresponding point of the brain pixel; The standard deviation of the grayscale of all pixels in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel; Represents the absolute value function.

[0034] Preferably, the obtaining of the importance of brain pixels includes the following specific methods:

[0035] For any brain pixel, the product of the edge feature of the brain pixel and the possibility that the brain pixel is a brain matter pixel is taken as the importance of the brain pixel.

[0036] The beneficial effects of the technical solution of the present invention are as follows: the present invention records the pixels at the same position in two weighted images as the same brain pixel, and obtains the possibility that the brain pixel is a brain substance pixel according to the grayscale expression of the brain pixel in the two weighted images. Since the pixels at the same position in the two weighted images are different expressions of the same position in the brain in the two weighted images, and each tissue in the brain has a stable expression in the two weighted images, the possibility that the brain pixel is a brain substance pixel can be obtained based on this; since the grayscale value of the noise pixel in the MRI image has the characteristic of Gaussian distribution, the tissue expression of the pixel can be obtained by the degree of influence of each pixel on its surroundings;

[0037] Furthermore, according to the pixel points of the brain in the two weighted images, the edge features of the brain pixels are obtained in combination with the tissue expression of each pixel in the two weighted images. Since the texture features existing in the local area of ​​the pixel are obvious when the pixel is located at the edge, the similarity of the textures of two pixels at the same position in different weighted images in the local area can be used as the edge features of the brain pixels. Since the most important detail features in the present invention are the edge features of gray matter and white matter, it is necessary to further combine the possibility that the brain pixel is a brain matter pixel, and assign a smoothing parameter to each pixel, so as to denoise the two weighted images, and finally achieve denoising of the MRI image while retaining the detail features in the MRI image. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 The present invention is a flowchart of the steps of the method for detecting cognitive impairment in diabetic patients based on MRI images;

[0040] Figure 2 Schematic diagram of clustering space. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method for detecting cognitive impairment in diabetic patients based on MRI images proposed by the present invention, its specific implementation, structure, characteristics and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The specific scheme of the method for detecting cognitive impairment in diabetic patients based on MRI images provided by the present invention is described in detail below with reference to the accompanying drawings.

[0044] See also Figure 1 , which shows a flowchart of a method for detecting cognitive impairment in diabetic patients based on MRI images provided by an embodiment of the present invention, the method comprising the following steps:

[0045] Step S001: Obtain MRI images of the patient's brain.

[0046] It should be noted that when collecting MRI images of the patient's brain, thermal noise will be generated due to the thermal motion of electrons in the electronic equipment of the imaging system. In order to avoid the influence of noise in the MRI image on the gray matter and white matter in the MRI image, the MRI image needs to be denoised before the white matter and gray matter in the MRI image are obtained. Therefore, the specific application scenario of this embodiment is: denoising the MRI image. Therefore, it is necessary to collect the MRI image of the patient's brain first.

[0047] Specifically, a nuclear magnetic resonance imaging device is used to collect MRI images of the patient's brain, and the MRI images of the patient's brain include a T1-weighted image and a T2-weighted image.

[0048] Step S002: Obtain brain pixels according to the positions of the pixels in the T1-weighted image and the T2-weighted image; cluster the brain pixels to obtain a number of clusters according to the gray values ​​of the pixels corresponding to the brain pixels in the T1-weighted image and the T2-weighted image; obtain the possibility that the brain pixels are brain matter pixels according to the positions of the brain pixels in the clusters and the gray values ​​of the brain pixels in the T1-weighted image and the T2-weighted image.

[0049] It should be noted that brain MRI images contain not only gray matter and white matter but also cerebrospinal fluid. Doctors will detect the patient's cognitive function through the gray matter and white matter in the patient's brain MRI images. Therefore, gray matter and white matter are more important than cerebrospinal fluid in MRI images. In addition, MRI images include T1-weighted images and T2-weighted images. In T1 images, the brightness of white matter is high, the brightness of gray matter is medium, and the brightness of cerebrospinal fluid is low; in T2 images, the brightness of white matter is medium, and the brightness of gray matter and cerebrospinal fluid is high.

[0050] It should be further explained that the pixel points at the same coordinate position in the T1-weighted image and the T2-weighted image are different representations of the same position in the brain in the T1-weighted image and the T2-weighted image. Therefore, the pixel points with the same coordinates in the T1-weighted image and the T2-weighted image can be recorded as the same brain pixel point. Since gray matter, white matter, and cerebrospinal fluid have stable grayscale representations in both the T1-weighted image and the T2-weighted image, the possibility that the brain pixel point is a brain matter pixel point can be obtained based on the respective representations of the brain pixel point in the T1-weighted image and the T2-weighted image. The possibility that the brain pixel point is a brain matter pixel point refers to the possibility that the brain pixel point is gray matter or white matter.

[0051] Preferably, in one embodiment of the present invention, pixel points with the same coordinate position in the T1-weighted image and the T2-weighted image are recorded as the same brain pixel point, so as to obtain a plurality of brain pixel points.

[0052] A clustering space is constructed with the grayscale value of the brain pixel in the T1-weighted image as the horizontal coordinate and the grayscale value in the T2-weighted image as the vertical coordinate; all brain pixels are placed in the clustering space, and the distance measurement adopts the Euclidean distance of different brain pixels in the clustering space, and the k-means clustering algorithm is used. In this embodiment, k=3 is used for description, and all brain pixels in the clustering space are clustered into 3 clusters, such as Figure 2 As shown, Figure 2 Schematic diagram of clustering space.

[0053] It should be noted that since noise pixels do not have the characteristic of having a stable grayscale representation in both T1-weighted images and T2-weighted images, for any brain pixel in any cluster, the greater the distance between the brain pixel and the cluster center, the more likely the brain pixel is a noise pixel, that is, the less likely it is a brain matter pixel; the brain matter pixel is a gray matter pixel or a white matter pixel.

[0054] It should be further explained that in T1 images, the brightness of white matter is high, the brightness of gray matter is medium, and the brightness of cerebrospinal fluid is low; in T2 images, the brightness of white matter is medium, and the brightness of gray matter and cerebrospinal fluid is high, that is, the grayscale performance of brain matter pixels in T1-weighted images and T2-weighted images is similar, while the grayscale performance of cerebrospinal fluid pixels in T1-weighted images and T2-weighted images is not similar. Therefore, the possibility that brain pixels are brain matter pixels can be obtained through the position of brain pixels in clusters and the grayscale values ​​of brain pixels in T1-weighted images and T2-weighted images.

[0055] Preferably, in one embodiment of the present invention, for any brain pixel in any cluster, the distance between the brain pixel and the cluster center is obtained, and the possibility that the brain pixel is a brain matter pixel is obtained by combining the grayscale value of the brain pixel in the T1-weighted image and the T2-weighted image. The specific calculation formula is:

[0056]

[0057] In the formula, Indicates the possibility that a brain pixel is a brain-matter pixel; Represents the distance between the brain pixel and the cluster center; Represents the gray value of brain pixels in T1-weighted images; Represents the gray value of brain pixels in T2-weighted images; It represents the absolute value function; represents an exponential function with a natural constant as the base. In this embodiment, Model to present inverse proportional relationship and normalization, As the input of the model, the implementer can set the inverse proportional function and normalization function according to the actual situation.

[0058] It should be noted that The larger the value of , the greater the distance between the brain pixel and its cluster center, and the more likely the brain pixel is to be a noise pixel, that is, the less likely the brain pixel is to be a brain matter pixel; The larger the value of indicates that the grayscale performance of cerebrospinal fluid pixels in T1-weighted images and T2-weighted images is different, that is, the less likely the brain pixels are brain matter pixels. The smaller the value is, the less likely the brain pixel is to be a brain pixel.

[0059] At this point, the possibility that the brain pixel is a brain matter pixel is obtained.

[0060] Step S003: Obtaining the tissue expression of each pixel in the T1-weighted image and the T2-weighted image according to the difference between each pixel and surrounding pixels in the T1-weighted image and the T2-weighted image.

[0061] It should be noted that the noise in MRI images is caused by the thermal motion of electrons in the electronic equipment of the imaging system, which in turn generates thermal noise; and the grayscale value of thermal noise in any region will be Gaussian distributed around the mean of the region; and since the grayscale changes uniformly within each region in the MRI image, the tissue expression of the pixel point can be quantified by controlling the variable, that is, by controlling the degree of influence of different pixels on the surrounding area after removal, the tissue expression of the pixel point can be obtained; the tissue expression of the pixel point is the possibility that the pixel point is not a noise pixel point, and the greater the tissue expression of the pixel point, the less likely the pixel point is to be a noise pixel point.

[0062] Preferably, in one embodiment of the present invention, for any pixel point in the T1-weighted image or the T2-weighted image, a The local window of size is recorded as the local window of the pixel point. is the preset local window side length, The specific value of can be set according to the actual situation, and this embodiment does not limit it. It is particularly noted that if the local window exceeds the image boundary range, the local window is filled with pixels by quadratic linear interpolation; according to the gray value of each pixel in the local window of the pixel, the tissue expression of the pixel is obtained, and the specific calculation formula is:

[0063]

[0064] In the formula, Indicates The organizational representation of each pixel; Indicates The grayscale mean of all pixels in the local window of pixels; Indicated in The grayscale mean of all pixels except the central pixel in the local window of pixels; Indicated in In the local window of pixels, except for the The grayscale mean of all pixels except pixels; Indicates The gray value of each pixel; Indicates The local window of the pixel The gray value of each pixel; Indicates The local window of the pixel The gray value of each pixel; Indicates The number of pixels in a local window of pixels; It represents the absolute value function; Expressing the Traverse all pixels in the local window of pixels and take the maximum value; represents an exponential function with a natural constant as the base. In this embodiment, Model to present inverse proportional relationship and normalization, As the input of the model, the implementer can set the inverse proportional function and normalization function according to the actual situation.

[0065] It should be noted that It represents the grayscale change in the area after removing the central pixel, so The larger the value is, the greater the impact on the region after removing the central pixel, and the more the gray value of the central pixel has the characteristics of Gaussian distribution; It represents the difference between the central pixel and all the pixels in the area, so The larger the value of is, the greater the difference between the central pixel and all the pixels in the region is, and the gray value of the central pixel has the characteristics of Gaussian distribution. The smaller the value of The more pixels there are, the more likely they are noise pixels.

[0066] At this point, the tissue expression of the pixels in the T1-weighted image and the T2-weighted image is obtained.

[0067] Step S004: According to the corresponding pixel points of the brain pixel points in the T1-weighted image and the T2-weighted image, a one-dimensional pixel point sequence of two corresponding points of the brain pixel points is obtained; according to the one-dimensional pixel point sequence of two corresponding points of the brain pixel points, combined with the tissue expression of each pixel point in the T1-weighted image and the T2-weighted image, the edge features of the brain pixel points are obtained.

[0068] It should be noted that since the boundary between gray matter and white matter in MRI images is blurred and there is an area with slow transition, the boundary between gray matter and white matter cannot be obtained by the gradient value of the pixel point. The pixel points at the same coordinate position in the T1-weighted image and the T2-weighted image are different representations of the same position in the brain in the T1-weighted image and the T2-weighted image. Therefore, the textures of the two pixels at the same position in the T1-weighted image and the T2-weighted image are similar in the local area. When the pixel point is located at the edge, the texture features in the local area of ​​the pixel point are obvious. Therefore, the similarity of the textures of the two pixels at the same position in the T1-weighted image and the T2-weighted image in the local area can be used as the edge features of the brain pixels.

[0069] Preferably, in one embodiment of the present invention, for any brain pixel point, the pixel point corresponding to the brain pixel point in the T1-weighted image and the T2-weighted image is recorded as the corresponding point of the brain pixel point; local windows of two corresponding points are respectively obtained, and the local windows of the two corresponding points are respectively scanned using the serpentine scanning method to obtain a one-dimensional pixel point sequence of the two corresponding points of the brain pixel point. Since the serpentine scanning method is a well-known prior art, it will not be described in detail in this embodiment.

[0070] It should be noted that, since both the T1-weighted image and the T2-weighted image have noise, in order to avoid the influence of the noise in the one-dimensional pixel sequence of two corresponding points of the brain pixel point on the calculation of the similarity, it is necessary to assign a weight to each pixel in the one-dimensional pixel sequence of two corresponding points of the brain pixel point according to the tissue expression of each pixel in the one-dimensional pixel sequence of two corresponding points of the brain pixel point when calculating the similarity, so that pixels with large tissue expression are given greater participation when participating in the similarity calculation, thereby reducing the interference of noise and making the calculation result of the similarity more accurate.

[0071] Preferably, in one embodiment of the present invention, for any brain pixel, the edge feature of the brain pixel is obtained according to the one-dimensional pixel sequence of two corresponding points of the brain pixel and the tissue expression of each pixel in the one-dimensional pixel sequence of two corresponding points of the brain pixel, and the specific calculation formula is:

[0072]

[0073] In the formula, Represents the edge features of brain pixels; The number of pixels in the one-dimensional pixel sequence representing the corresponding points of the brain pixels; The first corresponding point in the one-dimensional pixel sequence representing the brain pixel The gray value of each pixel; The grayscale mean of all pixels in the one-dimensional pixel sequence representing the first corresponding point of the brain pixel; The first corresponding point in the one-dimensional pixel sequence representing the brain pixel The organizational representation of each pixel; The first pixel in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel The gray value of each pixel; The grayscale mean of all pixels in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel; The first pixel in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel The organizational representation of each pixel; The standard deviation of the grayscale of all pixels in the one-dimensional pixel sequence representing the first corresponding point of the brain pixel; The standard deviation of the grayscale of all pixels in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel; Represents the absolute value function.

[0074] It should be noted that, For the existing correlation coefficient calculation formula, when calculating the similarity between the one-dimensional pixel sequences of two corresponding points of brain pixels, the tissue expression of the pixels in the one-dimensional pixel sequence is used as a weight to reduce the participation of the pixels with low tissue expression in the one-dimensional pixel sequence, so as to avoid the interference of noise pixels.

[0075] At this point, the edge features of brain pixels are obtained.

[0076] Step S005: Obtain the importance of brain pixels according to the edge features of brain pixels and the possibility that the brain pixels are brain parenchyma pixels; assign smoothing parameters to the pixels corresponding to the brain pixels in T1-weighted images and T2-weighted images according to the importance of the brain pixels, and denoise the T1-weighted images and T2-weighted images.

[0077] It should be noted that the specific application scenario of this embodiment is: the MRI image is denoised before the gray matter and white matter in the MRI image are obtained, so the most important detail features in the MRI image are the edge features of the gray matter and the white matter; therefore, after the edge features of the brain pixels are obtained through step S004, it is also necessary to combine the possibility that the brain pixels obtained in step S002 are brain matter pixels to obtain the importance of the brain pixels, and further assign smoothing parameters to each pixel in the T1-weighted image and the T2-weighted image based on the importance of the brain pixels, so as to finally achieve denoising of the MRI image while retaining the detail features in the MRI image.

[0078] Preferably, in a specific embodiment of the present invention, for any brain pixel, the product of the edge feature of the brain pixel and the possibility that the brain pixel is a brain matter pixel is used as the importance of the brain pixel, and the specific calculation formula is:

[0079]

[0080] In the formula, Indicates the importance of brain pixels; Represents the edge features of brain pixels; Indicates the possibility that a brain pixel is a brain-matter pixel.

[0081] It should be noted that in the non-local mean filtering algorithm, the smaller the smoothing parameter is, the more detailed features in the image can be retained. Therefore, the more important pixels in the MRI image are assigned smaller smoothing parameters to avoid the loss of detailed features of the pixels.

[0082] Preferably, in one embodiment of the present invention, an initial smoothing parameter is preset , The specific value of can be set according to the actual situation, and this embodiment does not limit it. The initial smoothing parameter is a parameter in the non-local mean filtering. For any brain pixel, the product of the inverse normalization result of the importance of the brain pixel and the initial smoothing parameter is used as the smoothing parameter of the corresponding point of the brain pixel. The specific calculation formula is:

[0083]

[0084] In the formula, Smoothing parameters of corresponding points of brain pixels; Indicates the importance of brain pixels; Represents the preset initial smoothing parameter; represents an exponential function with a natural constant as the base. In this embodiment, Model to present inverse proportional relationship and normalization, As the input of the model, the implementer can set the inverse proportional function and normalization function according to the actual situation.

[0085] The smoothing parameter of each pixel in the T1-weighted image and the T2-weighted image is obtained, and finally the T1-weighted image and the T2-weighted image are denoised using the non-local mean filtering algorithm and the smoothing parameter of each pixel in the T1-weighted image and the T2-weighted image. Since the non-local mean filtering algorithm is a well-known prior art, it will not be described in detail in this embodiment.

[0086] At this point, this embodiment is completed.

[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting cognitive impairment in diabetic patients based on MRI images, characterized in that: The method comprises the following steps: Acquiring an MRI image of the patient's brain, wherein the MRI image includes a T1-weighted image and a T2-weighted image; According to the positions of the pixels in the T1-weighted image and the T2-weighted image, the brain pixels are obtained; according to the gray values ​​of the pixels corresponding to the brain pixels in the T1-weighted image and the T2-weighted image, the brain pixels are clustered to obtain a number of clusters; according to the positions of the brain pixels in the clusters and the gray values ​​of the brain pixels in the T1-weighted image and the T2-weighted image, the possibility that the brain pixels are brain matter pixels is obtained; According to the difference between each pixel and surrounding pixels in the T1-weighted image and the T2-weighted image, the tissue expression of each pixel in the T1-weighted image and the T2-weighted image is obtained; According to the corresponding pixel points of the brain pixel points in the T1-weighted image and the T2-weighted image, a one-dimensional pixel point sequence of two corresponding points of the brain pixel point is obtained; according to the one-dimensional pixel point sequence of the two corresponding points of the brain pixel point, the edge features of the brain pixel point are obtained in combination with the tissue expression of each pixel point in the T1-weighted image and the T2-weighted image; According to the edge features of the brain pixels and the possibility that the brain pixels are brain matter pixels, the importance of the brain pixels is obtained; according to the importance of the brain pixels, smoothing parameters are assigned to the pixels corresponding to the brain pixels in the T1-weighted image and the T2-weighted image, and the T1-weighted image and the T2-weighted image are denoised; The method of assigning smoothing parameters to the pixels corresponding to the brain pixels in the T1-weighted image and the T2-weighted image, and denoising the T1-weighted image and the T2-weighted image includes the following specific methods: Preset an initial smoothing parameter; for any brain pixel point, use the product of the inverse normalized result of the importance of the brain pixel point and the initial smoothing parameter as the smoothing parameter of the corresponding point of the brain pixel point; The smoothing parameters of each pixel in the T1-weighted image and the T2-weighted image are obtained, and finally the T1-weighted image and the T2-weighted image are denoised using the non-local mean filtering algorithm and the smoothing parameters of each pixel in the T1-weighted image and the T2-weighted image.

2. The method for detecting cognitive impairment in diabetic patients based on MRI images according to claim 1, characterized in that: The specific method of obtaining brain pixel points includes: The pixel points with the same coordinate position in the T1-weighted image and the T2-weighted image are recorded as the same brain pixel point, and a number of brain pixel points are obtained.

3. The method for detecting cognitive impairment in diabetic patients based on MRI images according to claim 1, characterized in that: The specific method of clustering brain pixels to obtain a number of clusters includes: A clustering space was constructed with the grayscale value of brain pixels in T1-weighted images as the horizontal coordinate and the grayscale value in T2-weighted images as the vertical coordinate. All brain pixels were placed in the clustering space, and the distance measurement adopted the Euclidean distance of different brain pixels in the clustering space. The k-means clustering algorithm was used to cluster all brain pixels in the clustering space into several clusters.

4. The method for detecting cognitive impairment in diabetic patients based on MRI images according to claim 1, characterized in that: The possibility of obtaining brain pixels as brain matter pixels includes the following specific methods: For any brain pixel in any cluster, the distance between the brain pixel and the cluster center is obtained, and the possibility that the brain pixel is a brain matter pixel is obtained by combining the grayscale value of the brain pixel in the T1-weighted image and the T2-weighted image. The specific calculation formula is: In the formula, Indicates the possibility that a brain pixel is a brain-matter pixel; Represents the distance between the brain pixel and the cluster center; Represents the gray value of brain pixels in T1-weighted images; Represents the gray value of brain pixels in T2-weighted images; It represents the absolute value function; Represents an exponential function with a natural constant as base.

5. The method for detecting cognitive impairment in diabetic patients based on MRI images according to claim 1, characterized in that: The specific method of obtaining the tissue expression of each pixel in the T1-weighted image and the T2-weighted image includes: For any pixel in a T1-weighted image or a T2-weighted image, a A local window of size is recorded as the local window of the pixel point; and the tissue expression degree of the pixel point is obtained according to the grayscale value of each pixel point in the local window of the pixel point.

6. The method for detecting cognitive impairment in diabetic patients based on MRI images according to claim 5, characterized in that: The specific calculation formula for obtaining the tissue expression of the pixel point is: In the formula, Indicates The organizational representation of each pixel; Indicates The grayscale mean of all pixels in the local window of pixels; Indicated in The grayscale mean of all pixels except the central pixel in the local window of pixels; Indicated in In the local window of pixels, except for the The grayscale mean of all pixels except pixels; Indicates The gray value of each pixel; Indicates The local window of the pixel The gray value of each pixel; Indicates The local window of the pixel The gray value of each pixel; Indicates The number of pixels in a local window of pixels; It represents the absolute value function; Expressing the Traverse all pixels in the local window of pixels and take the maximum value; Represents an exponential function with a natural constant as base.

7. The method for detecting cognitive impairment in diabetic patients based on MRI images according to claim 1, characterized in that: The specific method of obtaining a one-dimensional pixel point sequence of two corresponding points of brain pixel points includes: For any brain pixel point, the pixel point corresponding to the brain pixel point in the T1-weighted image and the T2-weighted image is recorded as the corresponding point of the brain pixel point; the local windows of the two corresponding points are respectively obtained, and the local windows of the two corresponding points are respectively scanned by the serpentine scanning method to obtain a one-dimensional pixel point sequence of the two corresponding points of the brain pixel point.

8. The method for detecting cognitive impairment in diabetic patients based on MRI images according to claim 1, characterized in that: The specific method of obtaining edge features of brain pixels is as follows: For any brain pixel, the edge feature of the brain pixel is obtained according to the one-dimensional pixel sequence of the two corresponding points of the brain pixel and the tissue expression of each pixel in the one-dimensional pixel sequence of the two corresponding points of the brain pixel. The specific calculation formula is: In the formula, Represents the edge features of brain pixels; The number of pixels in the one-dimensional pixel sequence representing the corresponding points of the brain pixels; The first corresponding point in the one-dimensional pixel sequence representing the brain pixel The gray value of each pixel; The grayscale mean of all pixels in the one-dimensional pixel sequence representing the first corresponding point of the brain pixel; The first corresponding point in the one-dimensional pixel sequence representing the brain pixel The organizational representation of each pixel; The first pixel in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel The gray value of each pixel; The grayscale mean of all pixels in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel; The first pixel in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel The organizational representation of each pixel; The standard deviation of the grayscale of all pixels in the one-dimensional pixel sequence representing the first corresponding point of the brain pixel; The standard deviation of the grayscale of all pixels in the one-dimensional pixel sequence representing the second corresponding point of the brain pixel; Represents the absolute value function.

9. The method for detecting cognitive impairment in diabetic patients based on MRI images according to claim 1, characterized in that: The specific method of obtaining the importance of brain pixels is as follows: For any brain pixel, the product of the edge feature of the brain pixel and the possibility that the brain pixel is a brain matter pixel is taken as the importance of the brain pixel.

Images