Image processing method, device, readable storage medium and electronic device

Abstract

The present disclosure relates to an image processing method, device, readable storage medium and electronic device. The method comprises: extracting brain tissue in a target SWI image to obtain a SWI brain tissue image; determining a candidate area of ​​microbleeding points in the SWI brain tissue image by using a Hessian matrix; and identifying microbleeding points from the candidate area of ​​microbleeding points. Thus, automatic detection of microbleeding points can be achieved, avoiding the problem of time-consuming and human error in manually searching for microbleeding points. In addition, based on a three-dimensional SWI brain tissue image, extracting a candidate area of ​​microbleeding points in the image by using a Hessian matrix can effectively reduce non-microbleeding point candidate areas that can cause interference, thereby improving the efficiency and accuracy of subsequent precise extraction of microbleeding points.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing technology, and in particular, to an image processing method, device, readable storage medium and electronic device. Background Art

[0002] Cerebral microbleeds (CMBs) are local hemorrhage-containing flavin deposits in the blood vessels caused by damage to small blood vessels in the brain. CMBs usually appear as dot-shaped, round or oval low-signal shadows on the susceptibility weighted imaging (SWI) sequence, with a diameter of 2 mm to 10 mm, and are mainly distributed in the brainstem, basal ganglia, and bilateral cerebral cortex. CMBs have important guiding value in the treatment of acute ischemic stroke. Studies have found that for patients with acute cerebral infarction, when thrombolytic therapy or anticoagulation and anti-aggregation therapy is used, the probability of hemorrhagic transformation in patients with CMBs is higher than that in patients without CMBs. Therefore, clarifying the CMBs of patients with ischemic stroke can help reduce the risk of bleeding after treatment. At present, manual methods are often used to visually find microbleeding points, which is not only time-consuming, but also difficult to ensure accuracy. Summary of the invention

[0003] In order to overcome the problems existing in the related art, the present disclosure provides an image processing method, an apparatus, a readable storage medium and an electronic device.

[0004] In order to achieve the above objectives, in a first aspect, the present disclosure provides an image processing method, comprising:

[0005] Extracting brain tissue in the target SWI image to obtain a SWI brain tissue image;

[0006] Determine the candidate area of ​​microbleeding points in the SWI brain tissue image by using the Hessian matrix;

[0007] Identify microbleeding spots from the microbleeding spot candidate area.

[0008] Optionally, determining the candidate area of ​​microbleeding points in the SWI brain tissue image by using a Hessian matrix includes:

[0009] For each target pixel in the SWI brain tissue image, feature extraction is performed on the target pixel through the Hessian matrix to obtain three eigenvalues ​​of the Hessian matrix corresponding to the target pixel; if the three eigenvalues ​​corresponding to the target pixel represent that the target pixel conforms to a spherical feature, an image area of ​​a preset size centered on the target pixel in the SWI brain tissue image is determined as a candidate area for microbleeding.

[0010] Optionally, the target pixel is determined by:

[0011] Determining whether each pixel in the SWI brain tissue image is a suspected microbleeding point according to the grayscale distribution characteristics in the SWI brain tissue image;

[0012] The suspected microbleeding points in the SWI brain tissue image are determined as target pixels.

[0013] Optionally, the step of identifying microbleeding spots from the microbleeding spot candidate area comprises:

[0014] For each of the microbleeding point candidate areas, a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point are determined, and based on the first probability and the second probability, whether the microbleeding point candidate area is a microbleeding point is determined.

[0015] Optionally, the determining of a first probability that the candidate area for microbleeding spots belongs to a microbleeding spot and a second probability that the candidate area for microbleeding spots does not belong to a microbleeding spot comprises:

[0016] Standardizing the gray value of the SWI brain tissue image;

[0017] The microbleeding point candidate area in the standardized SWI brain tissue image is input into a pre-trained prediction model to obtain a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point.

[0018] Optionally, the prediction model includes a first convolution layer, a second convolution layer, a third convolution layer, a pooling layer, a global average layer, a splicing module and a prediction module;

[0019] The first convolutional layer is used to extract features of the candidate area of ​​microbleeding points in the standardized SWI brain tissue image to obtain a first feature map;

[0020] The second convolutional layer is used to extract features from the first feature map to obtain a second feature map;

[0021] The third convolutional layer is used to perform feature extraction on the second feature map to obtain a third feature map;

[0022] The pooling layer is used to perform pooling processing on the third feature map to obtain a fourth feature map;

[0023] The global average layer is used to compress the fourth feature map;

[0024] The splicing module is used to splice the second feature map, the third feature map and the fourth feature map obtained after compression;

[0025] The prediction module is used to predict a first probability that the candidate area for microbleeding points belongs to a microbleeding point and a second probability that the candidate area for microbleeding points does not belong to a microbleeding point according to the feature map obtained after splicing.

[0026] Optionally, the grayscale value standardization of the SWI brain tissue image includes:

[0027] Obtaining the median value of the grayscale value of each pixel in the layer with the largest cross-sectional area of ​​the SWI brain tissue image;

[0028] The median value is used to normalize the grayscale value of the SWI brain tissue image.

[0029] Optionally, extracting brain tissue in the target SWI image to obtain a SWI brain tissue image includes:

[0030] Performing skull removal processing on the target SWI image to obtain a three-dimensional mask of brain tissue;

[0031] isotropically resampling the target SWI image and the brain tissue three-dimensional mask respectively, so that the three-dimensional pixel spacing of the resampled images is consistent;

[0032] The product of the target SWI image obtained after resampling and the three-dimensional brain tissue mask obtained after resampling is used as the SWI brain tissue image corresponding to the target SWI image.

[0033] Optionally, the method further comprises:

[0034] For each of the micro-bleeding points, three-dimensional region growth is performed on the micro-bleeding point;

[0035] The volume of the microbleeding spot after the three-dimensional region growth is determined and the volume is output.

[0036] In a second aspect, the present disclosure provides an image processing device, comprising:

[0037] An extraction module, used for extracting brain tissue in the target SWI image to obtain a SWI brain tissue image;

[0038] A determination module, used for determining the candidate area of ​​microbleeding points in the SWI brain tissue image extracted by the extraction module through a Hessian matrix;

[0039] The identification module is used to identify the microbleeding points from the microbleeding point candidate area determined by the determination module.

[0040] In a third aspect, the present disclosure provides a non-temporary computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method provided in the first aspect of the present disclosure.

[0041] In a fourth aspect, the present disclosure provides an electronic device, including:

[0042] a memory having a computer program stored thereon;

[0043] A processor is used to execute the computer program in the memory to implement the steps of the method provided in the first aspect of the present disclosure.

[0044] In the above technical solution, the brain tissue in the target SWI image is first extracted to obtain the SWI brain tissue image; then, the candidate area of ​​microbleeding points in the SWI brain tissue image is determined by the Hessian matrix; and the microbleeding points are identified from the candidate area of ​​microbleeding points. In this way, the automatic detection of microbleeding points can be achieved, avoiding the problem of time-consuming and human error in manually searching for microbleeding points. In addition, based on the three-dimensional SWI brain tissue image, the candidate area of ​​microbleeding points in the image is extracted by the Hessian matrix, which can effectively reduce the non-microbleeding point candidate area that can cause interference, thereby improving the efficiency and accuracy of the subsequent precise extraction of microbleeding points.

[0045] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0047] Figure 1 The figure is a flowchart of an image processing method according to an exemplary embodiment.

[0048] Figure 2 It is a schematic diagram of the structure of a prediction model according to an exemplary embodiment.

[0049] Figure 3 is a flowchart of an image processing method according to another exemplary embodiment.

[0050] Figure 4 is a block diagram of an image processing apparatus according to an exemplary embodiment.

[0051] Figure 5 It is a block diagram of an electronic device according to an exemplary embodiment.

[0052] Figure 6It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0053] As discussed in the background technology, currently, the main method for visually searching for microbleeding spots is manual methods, which is not only time-consuming, but also difficult to ensure accuracy. For this reason, it is necessary to automatically search for microbleeding spots with the help of computers. CMBs usually appear as spherical low-signal shadows on the SWI sequence. Therefore, at this stage, a spot detection algorithm is usually used to detect the quasi-circular low-density shadows on a two-dimensional level to obtain candidate areas for microbleeding spots, and then accurately extract microbleeding spots from the candidate areas for microbleeding spots. Since some blood vessels and microbleeding spots are both quasi-circular in two-dimensional terms, there may be a phenomenon that the same microbleeding spot corresponds to multiple two-dimensional candidate areas, and blood vessels may also be used as candidate areas, which will interfere with the subsequent accurate extraction of microbleeding spots, affecting the efficiency and accuracy of microbleeding spot detection.

[0054] In view of this, the present disclosure provides an image processing method, an apparatus, a readable storage medium, and an electronic device.

[0055] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0056] Figure 1 FIG. 1 is a flowchart of an image processing method according to an exemplary embodiment. Figure 1 As shown, the method includes S101 to S103.

[0057] In S101 , brain tissue in the target SWI image is extracted to obtain a SWI brain tissue image.

[0058] In the present disclosure, since the target SWI image contains not only brain tissue but also skull, in order to avoid interference of the skull in subsequent microbleed point identification, the brain tissue in the target SWI image needs to be extracted to obtain a three-dimensional SWI brain tissue image.

[0059] In S102, the candidate regions of microbleeding points in the SWI brain tissue image are determined by using the Hessian matrix.

[0060] In one implementation, the Hessian matrix may be the second-order partial derivative of the SWI brain tissue image.

[0061] In another embodiment, the Hessian matrix can be the convolution of the second-order partial derivative of the SWI brain tissue image and the Gaussian function. The Gaussian function is used to remove Gaussian noise, and it has a good low-pass filtering effect. Moreover, the convolution does not affect the overall structure of the SWI brain tissue image, and to a certain extent, the SWI brain tissue image is denoised, thereby ensuring the accuracy of the features extracted from the SWI brain tissue image, thereby improving the accuracy of the candidate area of ​​the microbleed point.

[0062] In S103, microbleeding spots are identified from the microbleeding spot candidate area.

[0063] In the above technical solution, the brain tissue in the target SWI image is first extracted to obtain the SWI brain tissue image; then, the candidate area of ​​microbleeding points in the SWI brain tissue image is determined by the Hessian matrix; and the microbleeding points are identified from the candidate area of ​​microbleeding points. In this way, the automatic detection of microbleeding points can be achieved, avoiding the problem of time-consuming and human error in manually searching for microbleeding points. In addition, based on the three-dimensional SWI brain tissue image, the candidate area of ​​microbleeding points in the image is extracted by the Hessian matrix, which can effectively reduce the non-microbleeding point candidate area that can cause interference, thereby improving the efficiency and accuracy of the subsequent precise extraction of microbleeding points.

[0064] The following is a detailed description of the specific implementation method of extracting the brain tissue in the target SWI image to obtain the SWI brain tissue image in the above S101. Specifically, it can be achieved by the following steps 1) to 3):

[0065] 1) Remove the skull from the target SWI image to obtain a 3D mask of the brain tissue.

[0066] In the present disclosure, the target SWI image can be processed by removing the skull using the active contour method to obtain a three-dimensional mask of the brain tissue, wherein the three-dimensional mask of the brain tissue is a two-category image, one of which is the brain tissue and the other is other structures except the brain tissue.

[0067] 2) The target SWI image and the brain tissue 3D mask are isotropically resampled respectively so that the 3D pixel spacing of the resampled images is consistent.

[0068] In the present disclosure, the target SWI image and the three-dimensional brain tissue mask can be isotropically resampled respectively by linear interpolation and nearest neighbor interpolation so that the three-dimensional pixel spacing of the resampled images is consistent, that is, the pixel spacing of the three dimensions of the target SWI image is unified to a preset spacing (for example, 0.5 mm), and the pixel spacing of the three dimensions of the three-dimensional brain tissue mask is unified to the above preset spacing.

[0069] 3) The product of the target SWI image obtained after resampling and the three-dimensional brain tissue mask obtained after resampling is used as the SWI brain tissue image corresponding to the target SWI image.

[0070] In the above embodiment, the target SWI image and the brain tissue three-dimensional mask are respectively isotropically resampled so that the three-dimensional pixel spacing of the resampled images is consistent, thereby ensuring the consistency of the three-dimensional pixel spacing of the SWI brain tissue image, thereby ensuring the accuracy of the features extracted from the SWI brain tissue image (for example, the three eigenvalues ​​of the Hessian matrix), and further ensuring the accuracy of the candidate area of ​​microbleeding points.

[0071] The following is a detailed description of the specific implementation method of determining the candidate areas of microbleeding points in the SWI brain tissue image by using the Hessian matrix in the above S102. Specifically, it can be achieved by the following steps (1) and (2):

[0072] (1) For each target pixel in the SWI brain tissue image, the Hessian matrix is ​​used to extract the features of the target pixel, and the three eigenvalues ​​of the Hessian matrix corresponding to the target pixel are obtained.

[0073] In the present disclosure, the three eigenvalues ​​of the Hessian matrix reflect the grayscale change degree of the target pixel.

[0074] In addition, there are multiple ways to determine the target pixel point. In one embodiment, each pixel point in the SWI brain tissue image can be used as a target pixel point. That is, at this time, the Hessian matrix is ​​used to determine whether each pixel point in the SWI brain tissue image is a candidate area for microbleeding.

[0075] In another embodiment, the target pixel may be determined by: determining whether each pixel in the SWI brain tissue image is a suspected microbleeding point according to the grayscale distribution characteristics in the SWI brain tissue image; and determining the suspected microbleeding point in the SWI brain tissue image as the target pixel.

[0076] In the present disclosure, the grayscale of the microbleeding points in the brain on the SWI image is darker than the grayscale of the brain tissue part, that is, the grayscale value of the microbleeding points in the brain on the SWI image is lower than the grayscale value of the brain tissue part. Therefore, the grayscale distribution characteristics in the SWI brain tissue image can be used to determine whether each pixel in the SWI brain tissue image is a suspected microbleeding point. Specifically, for each pixel in the SWI brain tissue image, if the grayscale value of the pixel is less than or equal to the preset grayscale value threshold, the pixel is determined to be a suspected microbleeding point; if the grayscale value of the pixel is greater than the above-mentioned preset grayscale value threshold, the pixel is determined to be a non-suspected microbleeding point.

[0077] In this way, feature extraction is performed only on suspected microbleeding points in the SWI brain tissue image, rather than on every pixel in the SWI brain tissue image, thereby improving the extraction efficiency of microbleeding point candidate areas, and effectively reducing non-microbleeding point candidate areas such as blood vessels and calcification points that may cause interference, thereby improving the accuracy of microbleeding point candidate areas.

[0078] (2) If the three eigenvalues ​​corresponding to the target pixel point indicate that the target pixel point conforms to the spherical feature, an image region of a preset size centered on the target pixel point in the SWI brain tissue image is determined as a candidate region for microbleeding.

[0079] In the present disclosure, since microbleeding points usually present a quasi-spherical shape on the SWI sequence, the target pixel point is judged whether it meets the spherical feature according to the three characteristic values ​​corresponding to the target pixel point to determine whether the target pixel point belongs to a microbleeding point. If the target pixel point meets the spherical feature, the target pixel point is determined to be a microbleeding point. At this time, an image area of ​​a preset size centered on the target pixel point in the SWI brain tissue image can be determined as a candidate area for microbleeding points; if the target pixel point does not meet the spherical feature, the target pixel point is determined not to be a microbleeding point.

[0080] For example, the following equation (1) can be used to determine whether the target pixel point meets the spherical feature according to the three eigenvalues ​​corresponding to the target pixel point:

[0081]

[0082] Among them, the intermediate variable Intermediate variables c. d. t is a constant, for example, c = 0.7, d t =0.5; when d=1, it indicates that the target pixel point meets the spherical feature. At this time, the image area of ​​a preset size centered on the target pixel point in the SWI brain tissue image can be determined as the candidate area for microbleeding points; when d=0, it indicates that the target pixel point does not meet the spherical feature.

[0083] For example, a cubic area of ​​25*25*25 (ie, a preset size) centered at the target pixel in the SWI brain tissue image may be determined as a candidate area for microbleeding points.

[0084] The specific implementation method of identifying microbleeding spots from the microbleeding spot candidate area in S103 is described in detail below. Specifically, for each microbleeding spot candidate area determined in S102, a first probability that the microbleeding spot candidate area belongs to a microbleeding spot and a second probability that the microbleeding spot candidate area does not belong to a microbleeding spot can be determined, and whether the microbleeding spot candidate area is a microbleeding spot can be determined based on the first probability and the second probability.

[0085] If the first probability is greater than or equal to the second probability, the candidate area for microbleeding is determined to be a microbleeding point; if the first probability is less than the second probability, the candidate area for microbleeding is determined to be a non-microbleeding point.

[0086] The following is a detailed description of the specific implementation method for determining the first probability that the candidate area of ​​microbleeding spots belongs to a microbleeding spot and the second probability that the candidate area of ​​microbleeding spots does not belong to a microbleeding spot. Specifically, in one implementation, the following steps [1] and [2] can be used to achieve the above:

[0087] [1] Grayscale values ​​of SWI brain tissue images were normalized.

[0088] [2] The candidate area of ​​the microbleeding spot in the standardized SWI brain tissue image is input into a pre-trained prediction model to obtain a first probability that the candidate area of ​​the microbleeding spot belongs to a microbleeding spot and a second probability that the candidate area of ​​the microbleeding spot does not belong to a microbleeding spot.

[0089] Specifically, if Figure 2 As shown, the above prediction model includes a first convolutional layer, a second convolutional layer, a third convolutional layer, a pooling layer, a global average layer, a splicing module and a prediction module.

[0090] like Figure 2 As shown, the first convolution layer is used to extract features of the microbleeding point candidate area in the SWI brain tissue image obtained after standardization to obtain a first feature map; the second convolution layer is used to extract features of the first feature map to obtain a second feature map; the third convolution layer is used to extract features of the second feature map to obtain a third feature map; the pooling layer is used to perform pooling processing on the third feature map to obtain a fourth feature map; the global average layer is used to compress the fourth feature map; the splicing module is used to splice the second feature map, the third feature map and the compressed fourth feature map; the prediction module is used to predict a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point according to the feature map obtained after splicing by the splicing module.

[0091] For example, the prediction module may include multiple fully connected layers connected in series, such as Figure 2 As shown, the prediction module may include three fully connected layers in series.

[0092] In the above prediction model, multi-level features are fused (i.e., the second feature map, the third feature map, and the fourth feature map obtained after compression), so that by utilizing multi-level feature information and global feature information, the first probability that the candidate area of ​​the microbleeding point belongs to a microbleeding point and the second probability that the candidate area of ​​the microbleeding point does not belong to a microbleeding point can be more accurately predicted, thereby being able to more accurately distinguish between microbleeding points and non-microbleeding points, for example, distinguishing between spherical blood vessels and irregularly shaped bleeding points.

[0093] The specific implementation method of grayscale standardization of SWI brain tissue images in the above step [1] is described in detail below. Specifically, it can be achieved by the following steps:

[0094] First, the median grayscale value of each pixel in the layer with the largest cross-sectional area of ​​the SWI brain tissue image was obtained;

[0095] Then, the gray value of the SWI brain tissue image was normalized using the median value.

[0096] For example, the above median value can be used to normalize the gray value of the SWI brain tissue image by the following equation (2):

[0097]

[0098] Where v′ is the gray value of the pixel in the SWI brain tissue image after standardization; v is the gray value of the pixel in the SWI brain tissue image; v r is the median value mentioned above.

[0099] Since the layer with the largest cross-sectional area of ​​the SWI brain tissue image contains more gray and white matter than other layers, the median of the grayscale value of each pixel in the layer with the largest cross-sectional area of ​​the SWI brain tissue image can filter out noise, thereby well estimating the mean of the grayscale value of the normal gray and white matter part, so that the grayscale value of the gray and white matter part in the SWI brain tissue image obtained after standardization is close to 1, and thus the gray and white matter and other tissues in the SWI brain tissue image obtained after standardization can be well distinguished. Therefore, the first probability that the candidate area of ​​the microbleeding point belongs to the microbleeding point and the second probability that the candidate area of ​​the microbleeding point does not belong to the microbleeding point can be more accurately predicted, further improving the accuracy of subsequent precise extraction of microbleeding points.

[0100] In addition, in order to further improve the accuracy of subsequent microbleeding point extraction, the quantitative susceptibility mapping (QSM) image or phase (Phase, Pha) image corresponding to the above-mentioned target SWI image can also be used as the basis for identifying microbleeding points. Specifically, in another embodiment, the first probability that the microbleeding point candidate area belongs to the microbleeding point and the second probability that the microbleeding point candidate area does not belong to the microbleeding point can be determined by the following steps 1] and 2].

[0101] 1] grayscale values ​​of the SWI brain tissue image and the additional image are respectively normalized, wherein the additional image is one of the QSM image corresponding to the target SWI image and the Pha image corresponding to the target SWI image.

[0102] 2] The microbleeding point candidate area in the standardized SWI brain tissue image and the pixel points corresponding to the microbleeding point candidate area in the additional image after standardization are input into a pre-trained prediction model to obtain a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point.

[0103] In the above embodiment, because the additional image has quantitative information, that is, the characteristics of microbleeding points and calcification points in the additional image are bright spots and dark spots respectively, and the additional image has different image characteristics from the SWI brain tissue image, the calcification points and microbleeding points can be distinguished more accurately, that is, the accuracy of subsequent precise extraction of microbleeding points is further improved.

[0104] Figure 3 FIG. 1 is a flowchart of an image processing method according to another exemplary embodiment. Figure 3 As shown, the above method also includes S104 and S105.

[0105] In S104, for each microbleeding point, three-dimensional region growth is performed on the microbleeding point.

[0106] In the present disclosure, three-dimensional region growing is a process of gradually aggregating a pixel or a sub-region into a complete independent connected region according to a predefined growth criterion. Since three-dimensional region growing is a well-known technical means for those skilled in the art, it will not be described in detail in the present disclosure.

[0107] In S105 , the volume of the microbleeding spot after the three-dimensional region growth is determined and the volume is output.

[0108] In this implementation, three-dimensional region growth is performed on the micro-bleeding spots, and the edge regions of the micro-bleeding spots can be included, thereby improving the calculation accuracy of the volume of the micro-bleeding spots.

[0109] In addition, after the volume of the microbleeding spot is determined, the volume is output, so that the user can directly know the volume of each microbleeding spot, which is convenient for guiding the medication of acute ischemic stroke according to the volume.

[0110] Based on the same inventive concept, the present disclosure also provides an image processing device. Figure 4 As shown, the device 400 includes:

[0111] An extraction module 401 is used to extract brain tissue in a target SWI image to obtain a SWI brain tissue image;

[0112] A determination module 402 is used to determine the candidate area of ​​microbleeding points in the SWI brain tissue image extracted by the extraction module 401 through a Hessian matrix;

[0113] The identification module 403 is used to identify microbleeding spots from the microbleeding spot candidate areas determined by the determination module 402 .

[0114] In the above technical solution, the brain tissue in the target SWI image is first extracted to obtain the SWI brain tissue image; then, the candidate area of ​​microbleeding points in the SWI brain tissue image is determined by the Hessian matrix; and the microbleeding points are identified from the candidate area of ​​microbleeding points. In this way, the automatic detection of microbleeding points can be achieved, avoiding the problem of time-consuming and human error in manually searching for microbleeding points. In addition, based on the three-dimensional SWI brain tissue image, the candidate area of ​​microbleeding points in the image is extracted by the Hessian matrix, which can effectively reduce the non-microbleeding point candidate area that can cause interference, thereby improving the efficiency and accuracy of the subsequent precise extraction of microbleeding points.

[0115] Optionally, the determination module 402 includes: a first determination submodule, used to perform feature extraction on each target pixel in the SWI brain tissue image through the Hessian matrix to obtain three eigenvalues ​​of the Hessian matrix corresponding to the target pixel; and a second determination submodule, used to determine an image area of ​​a preset size centered on the target pixel in the SWI brain tissue image as a candidate area for microbleeding if the three eigenvalues ​​corresponding to the target pixel indicate that the target pixel conforms to a spherical feature.

[0116] Optionally, the determination module 402 further includes: a third determination submodule, configured to determine whether each pixel in the SWI brain tissue image is a suspected microbleeding point according to the grayscale distribution characteristics in the SWI brain tissue image; and determine the suspected microbleeding point in the SWI brain tissue image as a target pixel point.

[0117] Optionally, the identification module 403 is used to determine, for each of the microbleeding point candidate areas, a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point, and determine whether the microbleeding point candidate area is a microbleeding point based on the first probability and the second probability.

[0118] Optionally, the identification module 403 includes: a first standardization submodule, used to standardize the grayscale value of the SWI brain tissue image; a prediction submodule, used to input the microbleeding point candidate area in the SWI brain tissue image obtained after standardization into a pre-trained prediction model to obtain a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point.

[0119] Optionally, the prediction model includes a first convolutional layer, a second convolutional layer, a third convolutional layer, a pooling layer, a global average layer, a splicing module and a prediction module; wherein the first convolutional layer is used to perform feature extraction on the microbleeding point candidate area in the SWI brain tissue image obtained after standardization to obtain a first feature map; the second convolutional layer is used to perform feature extraction on the first feature map to obtain a second feature map; the third convolutional layer is used to perform feature extraction on the second feature map to obtain a third feature map; the pooling layer is used to perform pooling processing on the third feature map to obtain a fourth feature map; the global average layer is used to compress the fourth feature map; the splicing module is used to splice the second feature map, the third feature map and the compressed fourth feature map; the prediction module is used to predict a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point based on the feature map obtained after splicing.

[0120] Optionally, the first standardization submodule includes: an acquisition submodule, used to acquire the median of the grayscale values ​​of each pixel in the layer with the largest cross-sectional area of ​​the SWI brain tissue image; and a second standardization submodule, used to standardize the grayscale values ​​of the SWI brain tissue image using the median.

[0121] Optionally, the extraction module 401 includes: a processing submodule, used to perform skull removal processing on the target SWI image to obtain a three-dimensional brain tissue mask; a resampling submodule, used to perform isotropic resampling on the target SWI image and the three-dimensional brain tissue mask respectively, so that the three-dimensional pixel spacing of the resampled images is consistent; and a fourth determination submodule, used to use the product of the resampled target SWI image and the resampled three-dimensional brain tissue mask as the SWI brain tissue image corresponding to the target SWI image.

[0122] Optionally, the device 400 further includes: a region growth module, for performing three-dimensional region growth on each microbleeding point; and an output module, for determining the volume of the microbleeding point after the three-dimensional region growth, and outputting the volume.

[0123] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0124] The present disclosure also provides a non-temporary computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the above-mentioned image processing method provided by the present disclosure are implemented.

[0125] Figure 5 FIG. 5 is a block diagram of an electronic device 500 according to an exemplary embodiment. Figure 5 As shown, the electronic device 500 may include: a processor 501 and a memory 502. The electronic device 500 may also include one or more of a multimedia component 503, an input / output (I / O) interface 504, and a communication component 505.

[0126] The processor 501 is used to control the overall operation of the electronic device 500 to complete all or part of the steps in the above-mentioned image processing method. The memory 502 is used to store various types of data to support the operation of the electronic device 500, and these data may include instructions for any application or method used to operate on the electronic device 500, and application-related data, such as contact data, messages sent and received, pictures, audio, video, etc. The memory 502 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Read-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 503 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 502 or sent through the communication component 505. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 504 provides an interface between the processor 501 and other interface modules, and the other interface modules may be keyboards, mice, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 505 is used for wired or wireless communication between the electronic device 500 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC, or other 5G, etc., or a combination of one or more of them, is not limited here. Therefore, the corresponding communication component 505 may include: Wi-Fi module, Bluetooth module, NFC module, etc.

[0127] In an exemplary embodiment, the electronic device 500 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned image processing method.

[0128] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned image processing method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 502 including program instructions, and the above-mentioned program instructions can be executed by the processor 501 of the electronic device 500 to complete the above-mentioned image processing method.

[0129] Figure 6 6 is a block diagram of an electronic device 600 according to an exemplary embodiment. For example, the electronic device 600 may be provided as a server. Figure 6 The electronic device 600 includes a processor 622, which may be one or more, and a memory 632 for storing a computer program executable by the processor 622. The computer program stored in the memory 632 may include one or more modules each corresponding to a set of instructions. In addition, the processor 622 may be configured to execute the computer program to perform the above-mentioned image processing method.

[0130] In addition, the electronic device 600 may further include a power supply component 626 and a communication component 650. The power supply component 626 may be configured to perform power management of the electronic device 600, and the communication component 650 may be configured to implement communication of the electronic device 600, for example, wired or wireless communication. In addition, the electronic device 600 may further include an input / output (I / O) interface 658. The electronic device 600 may operate based on an operating system stored in the memory 632, such as Windows Server 2000. TM , Mac OSX TM , Unix TM , Linux TM etc.

[0131] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned image processing method are implemented. For example, the non-transitory computer-readable storage medium can be the above-mentioned memory 632 including program instructions, and the above-mentioned program instructions can be executed by the processor 622 of the electronic device 600 to complete the above-mentioned image processing method.

[0132] In another exemplary embodiment, a computer program product is also provided. The computer program product includes a computer program executable by a programmable device, and the computer program has a code portion for executing the above-mentioned image processing method when executed by the programmable device.

[0133] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, a variety of simple modifications can be made to the technical solution of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0134] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0135] In addition, various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

Claims

1. An image processing method, characterized in that: include: Extracting brain tissue in the target SWI image to obtain a SWI brain tissue image; Determine the candidate area of ​​microbleeding points in the SWI brain tissue image by using the Hessian matrix; identifying microbleeding spots from the microbleeding spot candidate area; The step of extracting brain tissue from the target SWI image to obtain a SWI brain tissue image includes: Performing skull removal processing on the target SWI image to obtain a three-dimensional mask of brain tissue; isotropically resampling the target SWI image and the brain tissue three-dimensional mask respectively, so that the three-dimensional pixel spacing of the resampled images is consistent; The product of the target SWI image obtained after resampling and the three-dimensional brain tissue mask obtained after resampling is used as the SWI brain tissue image corresponding to the target SWI image.

2. The method according to claim 1, characterized in that: The method of determining the candidate area of ​​microbleeding points in the SWI brain tissue image by using the Hessian matrix includes: For each target pixel in the SWI brain tissue image, feature extraction is performed on the target pixel through the Hessian matrix to obtain three eigenvalues ​​of the Hessian matrix corresponding to the target pixel; if the three eigenvalues ​​corresponding to the target pixel represent that the target pixel conforms to a spherical feature, an image area of ​​a preset size centered on the target pixel in the SWI brain tissue image is determined as a candidate area for microbleeding.

3. The method according to claim 2, characterized in that The target pixel is determined by: Determining whether each pixel in the SWI brain tissue image is a suspected microbleeding point according to the grayscale distribution characteristics in the SWI brain tissue image; The suspected microbleeding points in the SWI brain tissue image are determined as target pixels.

4. The method according to claim 1, characterized in that: The step of identifying microbleeding spots from the microbleeding spot candidate area comprises: For each of the microbleeding point candidate areas, a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point are determined, and based on the first probability and the second probability, whether the microbleeding point candidate area is a microbleeding point is determined.

5. The method according to claim 4, characterized in that The determining of the first probability that the candidate area for microbleeding spots belongs to the microbleeding spots and the second probability that the candidate area for microbleeding spots does not belong to the microbleeding spots includes: Standardizing the gray value of the SWI brain tissue image; The microbleeding point candidate area in the standardized SWI brain tissue image is input into a pre-trained prediction model to obtain a first probability that the microbleeding point candidate area belongs to a microbleeding point and a second probability that the microbleeding point candidate area does not belong to a microbleeding point.

6. The method according to claim 5, characterized in that The prediction model includes a first convolution layer, a second convolution layer, a third convolution layer, a pooling layer, a global average layer, a splicing module and a prediction module; The first convolutional layer is used to extract features of the candidate area of ​​microbleeding points in the standardized SWI brain tissue image to obtain a first feature map; The second convolutional layer is used to extract features from the first feature map to obtain a second feature map; The third convolutional layer is used to perform feature extraction on the second feature map to obtain a third feature map; The pooling layer is used to perform pooling processing on the third feature map to obtain a fourth feature map; The global average layer is used to compress the fourth feature map; The splicing module is used to splice the second feature map, the third feature map and the fourth feature map obtained after compression; The prediction module is used to predict a first probability that the candidate area for microbleeding points belongs to a microbleeding point and a second probability that the candidate area for microbleeding points does not belong to a microbleeding point according to the feature map obtained after splicing.

7. The method according to claim 5, characterized in that The grayscale value standardization of the SWI brain tissue image comprises: Obtaining the median value of the grayscale value of each pixel in the layer with the largest cross-sectional area of ​​the SWI brain tissue image; The median value is used to normalize the gray value of the SWI brain tissue image.

8. The method according to any one of claims 1 to 7, characterized in that The method further comprises: For each of the micro-bleeding points, three-dimensional region growth is performed on the micro-bleeding point; The volume of the microbleeding spot after the three-dimensional region growth is determined and the volume is output.

9. An image processing device, characterized in that: include: An extraction module, used for extracting brain tissue in the target SWI image to obtain a SWI brain tissue image; A determination module, used for determining a candidate area of ​​microbleeding points in the SWI brain tissue image extracted by the extraction module through a Hessian matrix; an identification module, used for identifying microbleeding points from the microbleeding point candidate area determined by the determination module; Wherein, the extraction module comprises: A processing submodule, used for performing skull removal processing on the target SWI image to obtain a three-dimensional mask of brain tissue; A resampling submodule, used for isotropically resampling the target SWI image and the brain tissue three-dimensional mask, respectively, so that the three-dimensional pixel spacing of the resampled images is consistent; The fourth determination submodule is configured to use the product of the target SWI image obtained after resampling and the three-dimensional brain tissue mask obtained after resampling as the SWI brain tissue image corresponding to the target SWI image.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 8 are implemented.

11. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 8.

Images