A white matter high signal detection system and method

Abstract

A white matter high signal detection system and method of the present invention comprises a database; an image preprocessing module; a computing module; and a processor. The system performs three matching judgments through a first method, a second method, and a third method, compares first image data to be detected with a plurality of reference image data, gradually screens reference image data that best matches the first image data to be detected, and then outputs corresponding reference score data in combination with known information such as reference image data and a mapping relationship table, and obtains a white matter high signal load condition. Through the above-mentioned method, the system can accurately and quickly find the reference image data that matches the first image data to be detected, and output the white matter high signal score and load condition in the first image data to be detected, thereby facilitating subsequent diagnosis and treatment by doctors, improving the efficiency and accuracy of judgment, and avoiding misjudgment caused by human factors.

Description

Technical Field

[0001] The present invention relates to the field of medical image technology, and in particular to a white matter high signal detection system and method capable of assisting doctors in diagnosis. Background Art

[0002] White matter hyperintensity (WMH) is also called leukoaraiosis (LA). This imaging term was first proposed by neurologist Hachinski. The disease often manifests as high signal on T2WI or T2 fluid attenuated inversion recovery sequence (FLAIR), and equal or low signal on T1WI.

[0003] With the continuous deepening of research, WMH is now generally considered to be a manifestation of cerebral small vessel disease (CSVD). The pathogenesis of WMH is mainly related to impaired cerebral blood flow autoregulation, collagen vascular disease, blood brain barrier (BBB) ​​damage and genetic factors. The incidence of the disease is positively correlated with age, and the incidence rate in people over 60 years old exceeds 85%. There are also a large number of studies showing that WMH is closely related to cognitive impairment and the occurrence of other cerebrovascular diseases. With the rapid development of imaging technology, the research on WMH has made some breakthrough progress in recent years.

[0004] The existing technology only achieves image segmentation of white matter hyperintensities, but does not further evaluate them on this basis, so it cannot help doctors make preliminary diagnoses based on the evaluation results. Doctors usually need to identify images based on the segmented images, that is, to manually judge white matter hyperintensities, which is time-consuming and laborious, and often leads to misjudgments, and the accuracy is easily affected by human factors. Therefore, there is an urgent need for a white matter hyperintensity detection system and method that can assist doctors in diagnosis. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a white matter high signal detection system and method that can assist doctors in diagnosis.

[0006] The present invention provides a white matter high signal detection system, comprising:

[0007] A database pre-stored with at least one reference image data and reference area data of different white matter high signals in the reference image data;

[0008] An image preprocessing module, which is used to preprocess the MRI image data and output first image data;

[0009] A calculation module, used for calculating first area data of white matter high signals in the first image data;

[0010] A processor, which determines whether the first image data and at least one reference image data are preliminarily matched by a first method; if they are preliminarily matched, determines whether the positions of white matter high signals in the first image data and different white matter high signals in the reference image data are matched by a second method; if they are matched, determines whether the first area data in the first image data and the reference area data of different white matter high signals in the reference image data are matched by a third method; if so, outputs corresponding reference score data and obtains the white matter high signal load.

[0011] The present invention provides a white matter high signal detection system, wherein the first method is: the processor scales at least one of the reference image data to a preset resolution and generates a reference scaled image, scales the first image data to a preset resolution and generates a comparative scaled image, the processor calculates the grayscale averages of all pixels of the reference scaled image and the comparative scaled image, and compares the first grayscale value of each pixel in at least one of the reference scaled images with the grayscale average, and the second grayscale value of each pixel in the comparative scaled image with the grayscale average;

[0012] The processor calculates the absolute values ​​of the differences between multiple first pixel values ​​a and second pixel values ​​b based on a first pixel value a whose first grayscale value is greater than or equal to the grayscale average value in at least one of the reference scaled images, and a second pixel value b whose second grayscale value is greater than or equal to the grayscale average value in the contrast scaled image, and determines that at least one of the reference scaled images corresponding to the minimum value of the absolute values ​​of the differences between the contrast scaled image and the first pixel values ​​a and the second pixel values ​​b matches, and the first image data corresponding to the contrast scaled image preliminarily matches the corresponding at least one reference image data.

[0013] The present invention provides a white matter high signal detection system, wherein the second method is: the processor delineates the contours of white matter high signals on first image data and at least one reference image data that preliminarily matches the first image data, respectively obtaining a first signal image containing only the contours of white matter high signals in the first image data and a reference signal image containing the contours of white matter high signals in at least one reference image data that preliminarily matches the first image data, the processor scaling the first signal image and at least one reference signal image to match an m×n grid size, and placing them in an m×n grid according to corresponding positions;

[0014] The processor determines whether the distance between the geometric center g of the first signal image and the geometric center o of at least one reference signal image is less than or equal to the first variable, and if not, removes the reference scaled image corresponding to the minimum value of the absolute value of the difference between the first pixel value a and the second pixel value b, and repeats the first method;

[0015] If so, it is determined that the positions of the white matter hyperintensities in the first image data match those of the white matter hyperintensities in at least one reference image data.

[0016] The present invention provides a white matter high signal detection system, wherein the third method is: the processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image;

[0017] The processor calculates an intersection area J of an area S1 of the first signal image and an area S2 of the reference signal image as S1∩S2;

[0018] The processor determines that when the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than a first preset threshold, it is determined that the first signal image matches only one of the reference signal images, and the first area data of the white matter high signal in the first image data matches the reference area data of the white matter high signal in only one reference image data;

[0019] The processor determines that when the absolute value of the difference between the intersection area J and the area S2 of at least one reference signal image is less than a first preset threshold, the first step is repeated k times until the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than the first preset threshold;

[0020] The first step is: adjusting the m×n grid to a (m+k)×(n+k) grid with the same total area, the processor determining whether the distances between the geometric center of the first signal image and the geometric center of at least one reference signal image are both less than or equal to the second variable, if so, determining that the white matter high signal in the first image data matches the position of the white matter high signal in at least one reference signal image, if not, filtering out the reference signal image whose geometric center is greater than the second variable from the geometric center of the first signal image, and determining that the white matter high signal in the first image data matches the position of the white matter high signal in the remaining reference image data; the processor determining whether the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than a first preset threshold, if so, ending the first step, if not, repeating the first step and adding 1 to the value of k.

[0021] The present invention provides a white matter high signal detection system, wherein the third method is: the processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image;

[0022] The processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image as a whole set Q;

[0023] The complementary area B1 of the area S1 of the first signal image with respect to the area S2 of at least one reference signal image is C Q S1;

[0024] The area S2 of the reference signal image is the complementary area B2 of the area S1 of the first signal image = C Q S2;

[0025] The processor determines that when the area S2 of only one reference signal image with respect to the complementary area B2 of the area S1 of the first signal image is less than a second preset threshold, it is determined that the first signal image matches only one of the reference signal images, and the first area data of the white matter high signal in the first image data matches the reference area data of the white matter high signal in only one reference image data;

[0026] The processor determines that when the area S2 of at least one reference signal image and the complementary area B2 of the area S1 of the first signal image are smaller than the second preset threshold, the first step is repeated k times until only the complementary area B2 of the area S2 of the reference signal image and the complementary area S1 of the first signal image are smaller than the second preset threshold:

[0027] The first step is: adjusting the m×n grid to a (m+k)×(n+k) grid with the same total area, the processor determining whether the distances between the geometric center of the first signal image and the geometric center of at least one reference signal image are both less than or equal to a second variable, if so, determining that the white matter hyperintensity signal in the first image data matches the position of the white matter hyperintensity signal in at least one reference signal image, if not, filtering out the reference signal image whose distance between the geometric center and the geometric center of the first signal image is greater than the second variable, and determining that the white matter hyperintensity signal in the first image data matches the position of the white matter hyperintensity signal in the remaining reference image data; the processor determining whether the area S2 of only one reference signal image with respect to the complement area B2 of the area S1 of the first signal image is less than a second preset threshold, if so, ending the first step, if not, repeating the first step and increasing the value of k by 1.

[0028] The present invention provides a white matter high signal detection system, wherein the step of "outputting corresponding reference score data and obtaining white matter high signal load conditions" comprises:

[0029] The database pre-stores a white matter high signal mapping relationship table;

[0030] The processor outputs the corresponding white matter high signal score and load condition according to the white matter high signal mapping relationship table and in combination with the area S2 of the reference signal image.

[0031] The present invention provides a white matter high signal detection system, wherein the processor is electrically connected to the image preprocessing module and the calculation module.

[0032] The present invention provides a white matter high signal detection system, wherein the first grayscale value and the second grayscale value range from 0 to 63.

[0033] The present invention provides a white matter high signal detection system, wherein the white matter high signal mapping relationship table adopts a modified Scheltens scale.

[0034] The present invention provides a method for detecting white matter high signal, comprising the following steps:

[0035] storing at least one reference image data and reference area data of different white matter high signals in the reference image data;

[0036] Preprocessing the MRI image data and outputting first image data;

[0037] Calculating first area data of white matter high signals in the first image data;

[0038] Determine whether the first image data is preliminarily matched with at least one reference image data; if so, determine whether the positions of white matter high signals in the first image data match those of different white matter high signals in the reference image data; if so, determine whether the first area data in the first image data matches the reference area data of different white matter high signals in the reference image data; if so, output the corresponding reference score data and obtain the white matter high signal load situation.

[0039] The difference between the white matter high signal detection system and method of the present invention and the prior art is that the white matter high signal detection system and method of the present invention can accurately and quickly find the reference image data that matches the first image data to be detected, and output the white matter high signal score and load condition in the first image data to be detected, thereby facilitating doctors to conduct subsequent diagnosis and treatment, improving the efficiency and accuracy of judgment, and avoiding misjudgment caused by human factors.

[0040] The following is a further description of a white matter high signal detection system and method of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of a system and method for detecting white matter hyperintensities;

[0042] Figure 2 This is a diagram of the skull removal and segmentation process;

[0043] Figure 3 It is a schematic diagram of outline drawing;

[0044] Figure 4 It is a schematic diagram for placing the grid;

[0045] Figure 5 yes Figure 4 A schematic diagram after scaling of the first position;

[0046] Figure 6 yes Figure 4 A schematic diagram after scaling the second position of ;

[0047] Figure 7 yes Figure 4 A schematic diagram after scaling the third position of ;

[0048] Figure 8 yes Figure 4 A schematic diagram after scaling the fourth position of ;

[0049] Fig. 9 yes Figure 4 A schematic diagram after scaling the fifth position;

[0050] Fig.10 yes Figure 5 , Figure 6 Overlapping schematics;

[0051] Fig.11 yes Figure 5 , Figure 7 Overlapping schematics;

[0052] Fig.12 yes Fig.10 Grid refinement diagram of ;

[0053] Fig.13 yes Fig.11 Grid refinement diagram of ;

[0054] Fig.14 yes Fig.12 Grid refinement diagram of ;

[0055] Fig.15 yes Fig.13 Grid refinement diagram. DETAILED DESCRIPTION

[0056] like Figures 1 to 15 As shown, see Figure 1 , 2 The present invention provides a white matter high signal detection system and method, comprising:

[0057] A database pre-stored with at least one reference image data and reference area data of different white matter high signals in the reference image data;

[0058] An image preprocessing module, which is used to preprocess the MRI image data and output first image data;

[0059] A calculation module, used for calculating first area data of white matter high signals in the first image data;

[0060] A processor, which determines whether the first image data and at least one reference image data are preliminarily matched by a first method, and if not, repeats the first method; if they are preliminarily matched, determines whether the positions of white matter high signals in the first image data and different white matter high signals in the reference image data match by a second method; if they do not match, repeats the first method and the second method; if they match, determines whether the first area data in the first image data and the reference area data of different white matter high signals in the reference image data match by a third method; if not, roughly repeats the first method, the second method, and the third method; if so, outputs the corresponding reference score data and obtains the white matter high signal load situation.

[0061] The present invention performs three matching judgments through the first method, the second method, and the third method, compares the first image data to be detected with multiple reference image data, gradually screens the reference image data that best matches the first image data to be detected, and then combines the reference image data, the mapping relationship table and other known information to output the corresponding reference score data, and obtains the white matter high signal load situation.

[0062] Through the above-mentioned method, the present invention can accurately and quickly find the reference image data that matches the first image data to be detected, and output the white matter high signal score and load condition in the first image data to be detected, thereby facilitating doctors to conduct subsequent diagnosis and treatment, improving the efficiency and accuracy of judgment, and avoiding misjudgment caused by human factors.

[0063] The reference image data are multiple MRI image data stored in a database and set according to different areas of white matter high signals.

[0064] The reference image data and the acquired MRI images are obtained by taking pictures at the same distance using the same device.

[0065] Through the above-mentioned arrangement, the present invention realizes that the MRI images and reference image data taken by the same person are the most matched, or the user can match the reference image data of other people with similar brain structures without being affected by the background area and brain size.

[0066] The MRI image data is the MRI image data to be detected, including but not limited to MRI T1FLAIR sequence head scan images and MRI T2FLAIR sequence head scan images.

[0067] The step of "the image preprocessing module preprocesses the MRI image data" includes: using SPM to perform head motion correction processing on the MRI image data, the head motion correction function in SPM can determine the parameters of the time series image and its reference registration, and on this basis, align each frame image in an experimental sequence with the reference image of this sequence to correct the head motion;

[0068] Use N4BiasFieldCorrection bias field correction and Gaussian filter-based methods to perform bias field correction on MRI image data to reduce misdiagnosis and improve diagnostic accuracy;

[0069] Use Python to normalize MRI image data to scale the data to the same order of magnitude, improving the training speed and accuracy of the segmentation model;

[0070] The skull removal algorithm is used to remove the skull, eyes, skin edge tissue, etc. in the MRI image data to remove non-target images and obtain the target image to further remove the noise interference caused by non-brain tissue to ensure segmentation accuracy. Figure 2 Middle picture;

[0071] The MRI image is input into the pre-trained white matter hyperintensity segmentation model to output the white matter hyperintensity segmentation result in MRI. The segmentation result can be found in Figure 2 The figure on the right is prior art and will not be described in detail here.

[0072] Among them, the step of the "calculation module, which is used to calculate the first area data of white matter high signals in the first image data" is: placing the first signal image of white matter high signals in the first image data in a grid of equivalent size, calculating the number of grids occupied by the first signal image, and thus estimating the first area data; or constructing a rectangular coordinate system for the first signal image of white matter high signals in the first image data, obtaining the coordinates of the outer contour of the first signal image, and obtaining the first area data according to the area calculation method of the Euclidean plane; of course, the first area data can also be obtained by other methods.

[0073] The “whether the first image data is preliminarily matched with at least one reference image data” refers to whether the overall image structure and shape of the first image data are similar to those of at least one reference image data.

[0074] Among them, the "whether the positions of the white matter high signals in the first image data match the positions of the different white matter high signals in the reference image data" means: whether the positions of the white matter high signals in the first image data and the different white matter high signals in the reference image data are in very close positions.

[0075] Among them, the "whether the first area data in the first image data matches the reference area data of different white matter high signals in the reference image data" means: whether the area of ​​the white matter high signals in the first image data and the area of ​​the different white matter high signals in the reference image data are in a very close area size.

[0076] Among them, the step of "outputting the corresponding reference score data and obtaining the white matter high signal load situation" can use a white matter high signal mapping relationship table to obtain the corresponding white matter high signal score and load situation. The white matter high signal mapping relationship table can use the Fazekas scale (0 to 6 points), the modified Scheltens scale (0 to 30 points), the Ylikoski scale (0 to 48 points) or the Manolio scale (0 to 9 points), etc., which are all evaluation scales based on MRI imaging examinations. They are existing technologies and will not be repeated here.

[0077] Among them, the manner of roughly repeating the first method, the second method, and the third method can refer to the first step.

[0078] For further explanation of the present invention, see Figure 1 , 2 , the first method is: the processor scales at least one of the reference image data to a preset resolution and generates a reference scaled image, scales the first image data to a preset resolution and generates a comparison scaled image, the processor calculates the grayscale averages of all pixels of the reference scaled image and the comparison scaled image, and compares the first grayscale value of each pixel in at least one of the reference scaled images with the grayscale average, and the second grayscale value of each pixel in the comparison scaled image with the grayscale average;

[0079] The processor calculates the absolute values ​​of the differences between multiple first pixel values ​​a and second pixel values ​​b based on a first pixel value a whose first grayscale value is greater than or equal to the grayscale average value in at least one of the reference scaled images, and a second pixel value b whose second grayscale value is greater than or equal to the grayscale average value in the contrast scaled image, and determines that at least one of the reference scaled images corresponding to the minimum value of the absolute values ​​of the differences between the contrast scaled image and the first pixel values ​​a and the second pixel values ​​b matches, and the first image data corresponding to the contrast scaled image preliminarily matches the corresponding at least one reference image data.

[0080] The present invention uses the first method to first scale the reference image data and the first image data to the same resolution, and then calculate and compare the size relationship between the grayscale value and the average grayscale value of the scaled reference image data, the first image data, so that after the preliminary screening by the first method, the reference image data that does not satisfy the relevant grayscale value relationship with the first image data is screened out, and the reference image data that satisfies the relevant grayscale value relationship with the first image data is retained, so as to carry out the next step of screening and matching.

[0081] The first grayscale value and the second grayscale value may be defined as follows: due to the different colors and brightness of each point in the scene, each point on the black-and-white photo or the black-and-white image reproduced by a television receiver presents a different degree of gray.

[0082] The first grayscale value and the second grayscale value are to divide white and black into several levels according to a logarithmic relationship, which are called "grayscale levels". The first grayscale value and the second grayscale value generally range from 0 to 255, with white being 255 and black being 0.

[0083] Of course, the first grayscale value and the second grayscale value preferably range from 0 to 63, with white being 63 and black being 0. That is, the preferred range used in the present invention is the combination of the four levels of the above range.

[0084] Wherein, at least one of the reference image data and the first image data has a grayscale level corresponding to its resolution, and for the convenience of calculation, its resolution is made the same as the grayscale level.

[0085] Wherein, the preset resolution is 6×6 to 720×720, preferably 256×256. For example, the resolution of at least one of the reference image data is 720×720, and the resolution of the first image data is 800×800. The at least one of the reference image data is converted into a reference scaled image with a resolution of 256×256 by an image processing tool, and the first image data is converted into a comparison scaled image with a resolution of 256×256. At this time, the reference scaled image and the comparison scaled image both have 65536 pixels and 64 gray levels. The processor calculates the gray average of 131072 pixels in the reference scaled image and the comparison scaled image, and the calculation method is: add the first gray value of 65536 pixels in the reference scaled image and the second gray value of 65536 pixels in the comparison scaled image and divide by 131072. For example, the calculated gray average is 30. The processor calculates that in at least one of the reference scaled images, there are 23655, 32120, 35330, 52123, and 61300 first pixel values ​​a whose first grayscale value is greater than or equal to the grayscale average value 30, respectively; in the contrast scaled image, there are 35120 second pixel values ​​b whose second grayscale value is greater than or equal to the grayscale average value 30, and the absolute values ​​of the differences with the multiple first pixel values ​​a are 11465, 3000, 210, 17003, and 26180, respectively, and the minimum value is 35330-35120=210. The processor determines that the contrast scaled image matches the reference scaled image whose second pixel value b is 35330, that is, the first image data corresponding to the contrast scaled image and the reference image data corresponding to the reference scaled image match.

[0086] For further explanation of the present invention, see Figures 1 to 9 , the second method is: the processor delineates the contours of white matter high signals on the first image data and at least one reference image data that preliminarily matches the first image data, and obtains a first signal image containing only the contours of white matter high signals in the first image data and a reference signal image containing the contours of white matter high signals in at least one reference image data that preliminarily matches the first image data, respectively; the processor scales the first signal image and at least one reference signal image to match the size of an m×n grid, and places them in the m×n grid according to corresponding positions;

[0087] The processor determines whether the distance between the geometric center g of the first signal image and the geometric center o of at least one reference signal image is less than or equal to the first variable, and if not, removes the reference scaled image corresponding to the minimum value of the absolute value of the difference between the first pixel value a and the second pixel value b, and repeats the first method;

[0088] If so, it is determined that the positions of the white matter hyperintensities in the first image data match those of the white matter hyperintensities in at least one reference image data.

[0089] The present invention uses the second method to first outline the white matter high signal contours of the first image data and the reference image data remaining after being screened out by the first method, generate a first signal image and multiple reference signal images, and then place them in a grid that matches their sizes to determine whether the geometric center of the first signal image is sufficiently close to the geometric center of the reference signal image, and screen out the reference signal images whose geometric centers are far from the geometric center of the first signal image, leaving the reference signal images whose geometric centers are sufficiently close to the geometric center of the first signal image, that is, determine that their positions match, so as to make further subsequent judgments.

[0090] Among them, the processor can automatically outline the white matter high signal and obtain the geometric center of the first signal image and the geometric center of at least one reference signal image. This is a prior art and will not be described in detail here.

[0091] The side length of each grid in the m×n grid is x, in mm, and the total area of ​​the entire grid is mnx 2 , unit is mm 2 , the first variable is the maximum distance between two points in each grid of m×n grid, that is

[0092] Wherein, m and n are both integers greater than or equal to 3, and the sizes of m and n can be adjusted according to the contour shape of the white matter high signal and the required degree of precision. The initial value of m×n is preferably 4×4, then the total area of ​​the m×n grid is mnx 2 The initial value is preferably 16x 2 .

[0093] The process of "the processor scaling the first signal image and at least one reference signal image to match the size of the m×n grid and placing them in the m×n grid according to corresponding positions" refers to Figures 3 to 9 , Figure 3 The upper left corner is the first signal image "containing only the white matter high signal contour in the first image data" and Figure 3 The other four reference signal images except the upper left corner are the “contours of white matter high signals in the reference image data” that are preliminarily matched with the first image data, Figure 4 To place the above images in a 4×4 grid according to their relative positions, Figure 5 is the amplified first signal image, Figures 6 to 9 These are 4 magnified reference signal images.

[0094] The geometric center of the first signal image is represented by the letter g, see Figure 5 The geometric center of at least one reference signal image is represented by the letter o, see Figures 6 to 9 .

[0095] Among them, for example, in the example of the first method, there are 35120 second pixel values ​​b in the contrast scaled image whose second grayscale value is greater than or equal to the grayscale average value 30, and the absolute values ​​of the differences with multiple first pixel values ​​a are 11465, 3000, 210, 17003, and 26180, respectively, and the minimum value is 35330-35120=210. If the processor determines that the distance between the geometric center g of the first signal image and the geometric center o of at least one reference signal image is greater than the first variable, the reference scaled image corresponding to the minimum value of the absolute value of the difference between the first pixel value a and the second pixel value b is eliminated, that is, the reference scaled image corresponding to the minimum value 210 of the absolute value of the difference with the first pixel value a is eliminated, and the first method is repeated.

[0096] For further explanation of the present invention, see Figure 1 , 10 , 11, 12, 13, 14, 15, the third method is: the processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image;

[0097] The processor calculates an intersection area J of an area S1 of the first signal image and an area S2 of the reference signal image as S1∩S2;

[0098] The processor determines that when the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than a first preset threshold, it is determined that the first signal image matches only one of the reference signal images, and the first area data of the white matter high signal in the first image data matches the reference area data of the white matter high signal in only one reference image data;

[0099] The processor determines that when the absolute value of the difference between the intersection area J and the area S2 of at least one reference signal image is less than a first preset threshold, the first step is repeated k times until the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than the first preset threshold;

[0100] The first step is: adjusting the m×n grid to a (m+k)×(n+k) grid with the same total area, the processor determining whether the distances between the geometric center of the first signal image and the geometric center of at least one reference signal image are both less than or equal to the second variable, if so, determining that the white matter high signal in the first image data matches the position of the white matter high signal in at least one reference signal image, if not, filtering out the reference signal image whose geometric center is greater than the second variable from the geometric center of the first signal image, and determining that the white matter high signal in the first image data matches the position of the white matter high signal in the remaining reference image data; the processor determining whether the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image that matches the position of the white matter high signal in the first image data is less than a first preset threshold, if so, ending the first step, if not, repeating the first step and adding 1 to the value of k.

[0101] The present invention uses the third method to first calculate the area of ​​the first signal image and the area of ​​the reference image data remaining after being screened by the first method and the second method, and take the intersection, and judge that when the area of ​​the first signal image only satisfies the relevant intersection relationship with the area of ​​one of the above reference image data, it means that only one reference image data is left, and then it is judged that the first signal image and the area of ​​this reference image data also match. If the above judgment is not satisfied, it means that the area of ​​at least one reference image data satisfies the relevant intersection relationship, then the grid is refined and returned to the second method, and then multiple reference image data are screened again by judging whether the geometric center is close enough, until only one reference image data is left with an area that satisfies the relevant intersection relationship, thereby screening out the final reference image data that best matches the first signal image.

[0102] The side length of each grid in the m×n grid is x, and the total area of ​​the entire grid is mnx 2 , then after the m×n grid is divided into (m+k)×(n+k) grids, the side length of each grid is The second variable is the maximum distance between two points in each grid of the (m+k)×(n+k) grid, that is, the second variable is

[0103] Wherein, k is: when the first step is cycled for the kth time, the number of grids changes from m×n to (m+k)×(n+k), for example, when the first step is cycled for the first time, the number of grids changes from m×n to (m+1)×(n+1); when the first step is cycled for the second time, the number of grids changes from m×n to (m+2)×(n+2); when the first step is cycled for the third time, the number of grids changes from m×n to (m+3)×(n+3); ......

[0105] And so on.

[0106] The area S1 of the first signal image may be first area data.

[0107] The units of the area S1 of the first signal image, the area S2 of the reference signal image, and the intersection area J are mm. 2 .

[0108] The first preset threshold is any real number in [0, 0.5], preferably 0.25, and the unit is mm 2 .

[0109] Among them, the area of ​​each reference signal image in the “area S2 of at least one reference signal image” is S2.

[0110] If there is no reference image data matching the white matter high signal position in the first image data, the reference signal image whose geometric center is at a distance greater than the second variable from the geometric center of the first signal image is screened out from the database, and the first method and the second method are repeated. If the number of repetitions k is greater than the number of reference image data, a signal that the query result is 0 is output to the user.

[0111] For example, Figure 5 is a first signal image that has been placed in a 4×4 grid and contains only the outlines of the white matter high signals in the first image data, Figures 6 to 9 are four reference signal images placed in a 4×4 grid that contain only the outlines of white matter high signals in the reference image data that matches the first image data. The grid coordinates of the geometric center of the first signal image are (2, 3), and the grid coordinates of the geometric centers of the four reference signal images are (2, 3), (2, 3), (3, 2), and (1, 1), respectively. Then, the geometric center of the first signal image is Figure 6 , Figure 7 If the distance between the geometric center of the reference image data is smaller than the first variable, then it is determined that the white matter high signal in the first signal image is Figure 6 , Figure 7 Matching the location of white matter high signals in the reference image data;

[0112] See also Fig.10 , the processor calculates Figure 5 The area S1 of the first signal image in Figure 6 The area S2 of the reference signal image is calculated Fig.10 The intersection area J in is S1∩S2;

[0113] See also Fig.11 , the processor calculates Figure 5 The area S1 of the first signal image in Figure 7The area S2 of the reference signal image is calculated Fig.11 The intersection area J in is S1∩S2;

[0114] The processor determines the intersection area J and Figure 6 , Figure 7 If the absolute value of the difference between the area S2 of the reference signal image and the image S3 is less than the first preset threshold, the first step is performed for the first time, and k is 1;

[0115] The first signal image is Figure 6 The reference signal images are placed in multiple (m+1)×(n+1) grids according to the corresponding positions, that is, in 5×5 grids, see Fig.12 ;

[0116] The first signal image is Figure 7 The reference signal images are placed in multiple (m+1)×(n+1) grids according to the corresponding positions, that is, in 5×5 grids, see Fig.13 ;

[0117] The processor determines Figure 5 The geometric center of the first signal image in Figure 6 , Figure 7 The distances between the geometric centers of the reference signal images are all less than or equal to the second variable, that is, less than Then it is determined that the position of the white matter high signal in the first image data matches the position of the white matter high signal in the reference image data; the processor determines that the two intersection areas J and Figure 6 , Figure 7 If the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than the first preset threshold, the first step is repeated for the second time, k is 2..., until the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than the first preset threshold, for example, see Fig.14 , 15 , when the first step is repeated for the 12th time, that is, when the grid is divided into (m+12)×(n+12), that is, 16×16 grids, Figure 5 The geometric center of the first signal image in Figure 6 The distance between the geometric center of the reference signal image is less than the first variable Figure 5 The geometric center of the first signal image in Figure 7 The distance between the geometric center of the reference signal image is greater than the first variable Then filter out Figure 7 Here we only assume that the comparison Figure 6 , Figure 7In actual operation, there may be multiple reference information images to be compared with the first signal image.

[0118] As a variation of the present invention, see Figure 1 , 10 , 11, 12, 13, 14, 15, the third method is: the processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image;

[0119] The processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image as a whole set Q;

[0120] The complementary area B1 of the area S1 of the first signal image with respect to the area S2 of at least one reference signal image is C Q S1;

[0121] The area S2 of the reference signal image is the complementary area B2 of the area S1 of the first signal image = C Q S2;

[0122] The processor determines that when the area S2 of only one reference signal image with respect to the complementary area B2 of the area S1 of the first signal image is less than a second preset threshold, it is determined that the first signal image matches only one of the reference signal images, and the first area data of the white matter high signal in the first image data matches the reference area data of the white matter high signal in only one reference image data;

[0123] The processor determines that when the area S2 of at least one reference signal image and the complementary area B2 of the area S1 of the first signal image are smaller than the second preset threshold, the first step is repeated k times until only the complementary area B2 of the area S2 of the reference signal image and the complementary area S1 of the first signal image are smaller than the second preset threshold:

[0124] The first step is: adjusting the m×n grid to a (m+k)×(n+k) grid with the same total area, the processor determining whether the distances between the geometric center of the first signal image and the geometric center of at least one reference signal image are both less than or equal to the second variable, if so, determining that the white matter hyperintensity in the first image data matches the position of the white matter hyperintensity in at least one reference signal image, if not, filtering out the reference signal images whose distances between the geometric center and the geometric center of the first signal image are greater than the second variable, and determining that the white matter hyperintensity in the first image data matches the position of the white matter hyperintensity in the remaining reference image data; the processor determining whether the area S2 of only one reference signal image that matches the position of the white matter hyperintensity in the first image data is less than the second preset threshold value in terms of the complement area B2 of the area S1 of the first signal image, if so, ending the first step, if not, repeating the first step and increasing the value of k by 1.

[0125] The present invention uses the third method to first calculate the area of ​​the first signal image and the area of ​​the reference image data remaining after being screened by the first method and the second method, and take the complement, and determine that when the area of ​​the first signal image only satisfies the relevant complement relationship with the area of ​​one of the above reference image data, it means that only one reference image data is left, and then it is determined that the first signal image and the reference image data area also match. If the above judgment is not satisfied, it means that the areas of multiple reference image data satisfy the relevant complement relationship, then the grid is refined and returned to the second method, and then multiple reference image data are screened again by judging whether the geometric center is close enough, until only one reference image data area satisfies the relevant complement relationship, thereby screening out the final reference image data that best matches the first signal image.

[0126] The unit of the complementary area B2 is mm 2 .

[0127] The second preset threshold is any real number in [0, 0.5], preferably 0.25, and the unit is mm 2 .

[0128] For further explanation of the present invention, see Figure 1 The step of "outputting corresponding reference score data and obtaining white matter high signal load" includes:

[0129] The database pre-stores a white matter high signal mapping relationship table;

[0130] The processor outputs the corresponding white matter high signal score and load condition according to the white matter high signal mapping relationship table and in combination with the area S2 of the reference signal image.

[0131] The present invention can output the white matter high signal score and load condition in the first image data to be detected in the above manner, so as to facilitate doctors to perform subsequent diagnosis and treatment, improve the efficiency and accuracy of judgment, and avoid misjudgment caused by human factors.

[0132] Among them, the white matter high signal mapping relationship table can evaluate the degree and range of white matter looseness, which contains the correspondence between the area of ​​white matter high signal and the white matter high signal score. In the third method, the area S1 of the first signal image and the area S2 of the reference signal image have been matched, and the area S2 of the reference signal image that has been screened and matched is known, and the corresponding white matter high signal score and load condition can be found through the white matter high signal mapping relationship table.

[0133] Among them, the white matter high signal mapping relationship table can adopt the Fazekas scale (0 to 6 points), the modified Scheltens scale (0 to 30 points), the Ylikoski scale (0 to 48 points) or the Manolio scale (0 to 9 points), etc., which are all evaluation scales based on MRI imaging examinations, which are existing technologies and will not be repeated here.

[0134] For example, a modified Scheltens scale (0 to 30 points) as shown in Table 1 may be used, which scores the periventricular and deep lesions (white matter hyperintensities) separately. Since the first signal image of the white matter hyperintensity contour in the first image data to be detected has been matched with the reference signal image of the white matter hyperintensity contour in a certain reference image data, and the shape, position, and area of ​​the latter reference signal image are all known, the corresponding white matter hyperintensity score and load can be found out based on the above known information in combination with the modified Scheltens scale.

[0135] It should be noted that what is found in the above method is only the corresponding white matter high signal score and load condition, which is only used as a reference for doctors to conduct subsequent diagnosis and treatment, rather than the result of disease diagnosis and treatment.

[0136]

[0137] Table 1

[0138] As a further explanation of the present invention, the processor is electrically connected to the image preprocessing module and the computing module.

[0139] Through the above arrangement, the present invention enables the processor to be electrically connected to the image preprocessing module and the computing module to achieve signal transmission.

[0140] As a further explanation of the present invention, the first grayscale value and the second grayscale value range from 0 to 63.

[0141] Through the above arrangement, the present invention achieves more convenient calculation and improves calculation efficiency while achieving the same effect.

[0142] As a further explanation of the present invention, the white matter high signal mapping relationship table adopts a modified Scheltens scale.

[0143] The present invention uses the modified Scheltens scale as the white matter high signal mapping relationship table to achieve the following advantages: good reliability and validity, including the size and location of the lesions when evaluating white matter lesions, and also simultaneously evaluating the number of deep white matter lesions, which has a good correlation with the area of ​​white matter lesions, and is more sensitive to longitudinal observation of the progression of white matter lesions and their relationship with clinical manifestations.

[0144] The present invention provides a method for detecting white matter high signal, see Figure 1 , storing at least one reference image data and reference area data of different white matter high signals in the reference image data;

[0145] Preprocessing the MRI image data and outputting first image data;

[0146] Calculating first area data of white matter high signals in the first image data;

[0147] Determine whether the first image data is preliminarily matched with at least one reference image data; if so, determine whether the positions of white matter high signals in the first image data match those of different white matter high signals in the reference image data; if so, determine whether the first area data in the first image data matches the reference area data of different white matter high signals in the reference image data; if so, output the corresponding reference score data and obtain the white matter high signal load situation.

[0148] The embodiments described above are merely descriptions of preferred implementation modes of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A white matter high signal detection system, characterized by: include A database pre-stored with at least one reference image data and reference area data of different white matter high signals in the reference image data; An image preprocessing module, which is used to preprocess the MRI image data and output first image data; A calculation module, used for calculating first area data of white matter high signals in the first image data; a processor, which determines whether the first image data and at least one reference image data are preliminarily matched by a first method, and if they are preliminarily matched, determines whether the positions of white matter hypersignals in the first image data and different white matter hypersignals in the reference image data are matched by a second method, and if they are matched, determines whether first area data in the first image data and reference area data of different white matter hypersignals in the reference image data are matched by a third method, and if so, outputs corresponding reference score data and obtains a white matter hypersignal load; The first method is: the processor scales at least one of the reference image data to a preset resolution and generates a reference scaled image, scales the first image data to a preset resolution and generates a comparative scaled image, the processor calculates the grayscale averages of all pixels of the reference scaled image and the comparative scaled image, and compares the first grayscale value of each pixel in at least one of the reference scaled images with the grayscale average, and the second grayscale value of each pixel in the comparative scaled image with the grayscale average; The processor calculates the absolute values ​​of the differences between the first pixel values ​​a and the second pixel values ​​b according to the first pixel value a whose first grayscale value is greater than or equal to the grayscale average value in at least one of the reference zoomed images, and the second pixel value b whose second grayscale value is greater than or equal to the grayscale average value in the contrast zoomed image, and determines that at least one of the reference zoomed images corresponding to the minimum absolute value of the difference between the contrast zoomed image and the first pixel value a and the second pixel value b matches, and the first image data corresponding to the contrast zoomed image preliminarily matches the corresponding at least one of the reference image data; The second method is: the processor delineates the contours of white matter high signals on the first image data and at least one reference image data that preliminarily matches the first image data, respectively obtaining a first signal image containing only the contours of white matter high signals in the first image data and a reference signal image containing the contours of white matter high signals in at least one reference image data that preliminarily matches the first image data, the processor scaling the first signal image and at least one reference signal image to match an m×n grid size, and placing them in an m×n grid according to corresponding positions; The processor determines whether the distance between the geometric center g of the first signal image and the geometric center o of at least one reference signal image is less than or equal to the first variable, and if not, removes the reference scaled image corresponding to the minimum value of the absolute value of the difference between the first pixel value a and the second pixel value b, and repeats the first method; If yes, determining that the location of the white matter hyperintensity signal in the first image data matches the location of the white matter hyperintensity signal in at least one reference image data; The third method is: the processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image; The processor calculates an intersection area J of an area S1 of the first signal image and an area S2 of the reference signal image as S1∩S2; The processor determines that when the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than a first preset threshold, it is determined that the first signal image matches only one of the reference signal images, and the first area data of the white matter high signal in the first image data matches the reference area data of the white matter high signal in only one reference image data; The processor determines that when the absolute value of the difference between the intersection area J and the area S2 of at least one reference signal image is less than a first preset threshold, the first step is repeated k times until the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image is less than the first preset threshold; The first step is: adjusting the m×n grid to a (m+k)×(n+k) grid with the same total area, the processor determining whether the distances between the geometric center of the first signal image and the geometric center of at least one reference signal image are both less than or equal to the second variable, if so, determining that the white matter high signal in the first image data matches the position of the white matter high signal in at least one reference signal image, if not, filtering out the reference signal image whose geometric center is greater than the second variable from the geometric center of the first signal image, and determining that the white matter high signal in the first image data matches the position of the white matter high signal in the remaining reference image data; the processor determining whether the absolute value of the difference between the intersection area J and the area S2 of only one reference signal image that matches the position of the white matter high signal in the first image data is less than a first preset threshold, if so, ending the first step, if not, repeating the first step and adding 1 to the value of k.

2. A white matter high signal detection system, characterized in that: include A database pre-stored with at least one reference image data and reference area data of different white matter high signals in the reference image data; An image preprocessing module, which is used to preprocess the MRI image data and output first image data; A calculation module, used for calculating first area data of white matter high signals in the first image data; a processor, which determines whether the first image data and at least one reference image data are preliminarily matched by a first method, and if they are preliminarily matched, determines whether the positions of white matter hypersignals in the first image data and different white matter hypersignals in the reference image data are matched by a second method, and if they are matched, determines whether first area data in the first image data and reference area data of different white matter hypersignals in the reference image data are matched by a third method, and if so, outputs corresponding reference score data and obtains a white matter hypersignal load; The first method is: the processor scales at least one of the reference image data to a preset resolution and generates a reference scaled image, scales the first image data to a preset resolution and generates a comparative scaled image, the processor calculates the grayscale averages of all pixels of the reference scaled image and the comparative scaled image, and compares the first grayscale value of each pixel in at least one of the reference scaled images with the grayscale average, and the second grayscale value of each pixel in the comparative scaled image with the grayscale average; The processor calculates the absolute values ​​of the differences between the first pixel values ​​a and the second pixel values ​​b according to the first pixel value a whose first grayscale value is greater than or equal to the grayscale average value in at least one of the reference zoomed images, and the second pixel value b whose second grayscale value is greater than or equal to the grayscale average value in the contrast zoomed image, and determines that at least one of the reference zoomed images corresponding to the minimum absolute value of the difference between the contrast zoomed image and the first pixel value a and the second pixel value b matches, and the first image data corresponding to the contrast zoomed image preliminarily matches the corresponding at least one of the reference image data; The second method is: the processor delineates the contours of white matter high signals on the first image data and at least one reference image data that preliminarily matches the first image data, respectively obtaining a first signal image containing only the contours of white matter high signals in the first image data and a reference signal image containing the contours of white matter high signals in at least one reference image data that preliminarily matches the first image data, the processor scaling the first signal image and at least one reference signal image to match an m×n grid size, and placing them in an m×n grid according to corresponding positions; The processor determines whether the distance between the geometric center g of the first signal image and the geometric center o of at least one reference signal image is less than or equal to the first variable, and if not, removes the reference scaled image corresponding to the minimum value of the absolute value of the difference between the first pixel value a and the second pixel value b, and repeats the first method; If yes, determining that the location of the white matter hyperintensity signal in the first image data matches the location of the white matter hyperintensity signal in at least one reference image data; The third method is: the processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image; The processor calculates the area S1 of the first signal image and the area S2 of at least one reference signal image as a whole set Q; The complementary area B1 of the area S1 of the first signal image with respect to the area S2 of at least one reference signal image is C Q S1; The area S2 of the reference signal image is the complementary area B2 of the area S1 of the first signal image = C Q S2; The processor determines that when the area S2 of only one reference signal image with respect to the complementary area B2 of the area S1 of the first signal image is less than a second preset threshold, it is determined that the first signal image matches only one of the reference signal images, and the first area data of the white matter high signal in the first image data matches the reference area data of the white matter high signal in only one reference image data; The processor determines that when the area S2 of at least one reference signal image and the complementary area B2 of the area S1 of the first signal image are smaller than the second preset threshold, the first step is repeated k times until only the complementary area B2 of the area S2 of the reference signal image and the complementary area S1 of the first signal image are smaller than the second preset threshold: The first step is: adjusting the m×n grid to a (m+k)×(n+k) grid with the same total area, the processor determining whether the distances between the geometric center of the first signal image and the geometric center of at least one reference signal image are both less than or equal to the second variable, if so, determining that the white matter hyperintensity in the first image data matches the position of the white matter hyperintensity in at least one reference signal image, if not, filtering out the reference signal images whose distances between the geometric center and the geometric center of the first signal image are greater than the second variable, and determining that the white matter hyperintensity in the first image data matches the position of the white matter hyperintensity in the remaining reference image data; the processor determining whether the area S2 of only one reference signal image that matches the position of the white matter hyperintensity in the first image data is less than the second preset threshold value in terms of the complement area B2 of the area S1 of the first signal image, if so, ending the first step, if not, repeating the first step and increasing the value of k by 1.

3. A white matter high signal detection system according to claim 1 or 2, characterized in that: The step of "outputting corresponding reference score data and obtaining white matter high signal load" includes: The database pre-stores a white matter high signal mapping relationship table; The processor outputs the corresponding white matter high signal score and load condition according to the white matter high signal mapping relationship table and in combination with the area S2 of the reference signal image.

4. A white matter high signal detection system according to claim 3, characterized in that: The processor is electrically connected to the image preprocessing module and the computing module.

5. A white matter high signal detection system according to claim 4, characterized in that: The first grayscale value and the second grayscale value range from 0 to 63.

6. A white matter high signal detection system according to claim 5, characterized in that: The white matter high signal mapping relationship table adopts the modified Scheltens scale.

7. An analysis method based on a white matter high signal detection system according to any one of claims 1 or 2, characterized in that The following steps are involved: storing at least one reference image data and reference area data of different white matter high signals in the reference image data; Preprocessing the MRI image data and outputting first image data; Calculating first area data of white matter high signals in the first image data; Determine whether the first image data is preliminarily matched with at least one reference image data; if so, determine whether the positions of white matter high signals in the first image data match those of different white matter high signals in the reference image data; if so, determine whether the first area data in the first image data matches the reference area data of different white matter high signals in the reference image data; if so, output the corresponding reference score data and obtain the white matter high signal load situation.

Images