Method, device, storage medium and electronic device for identifying aortic dissection images

Abstract

The present disclosure relates to a method, device, storage medium and electronic device for identifying an aortic dissection image. The method comprises: acquiring an aortic CTA image, and extracting an aortic 3D vascular image from the CTA image; constructing a sequence of layers to be identified based on the aortic 3D vascular image, wherein each layer to be identified in the layer set to be identified is a 2D image; inputting the layer set to be identified into a trained aortic dissection identification model to obtain an identification result characterizing whether there is an aortic dissection in the layer set to be identified; and determining whether the CTA image is an aortic dissection image based on the identification result corresponding to each layer set to be identified in the layer set sequence to be identified. By adopting the method disclosed in the present disclosure, it is possible to quickly and accurately determine whether the aortic CTA image is an aortic dissection image.

Description

Technical Field

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

[0002] Aortic dissection refers to the blood in the aorta cavity entering the aortic media from the tear in the aortic intima, separating the intima and the media, and the blood expands along the long axis of the aorta to form a separation of the true and false cavities of the aortic wall. The diaphragm between the true and false cavities is called the dissecting membrane. Once aortic dissection ruptures, it will cause massive bleeding, which is a clinically critical and severe disease with a very high mortality rate.

[0003] In the related art, due to the heavy workload of doctors and the fact that doctors usually consult patients in the order of their registration, it takes at least 1 to 2 hours to get the image analysis results after the patient undergoes an imaging examination. This method of waiting for manual analysis of images is time-consuming and is not conducive to timely treatment of patients with aortic dissection. Summary of the invention

[0004] The purpose of the present disclosure is to provide a method, device, storage medium and electronic device for identifying aortic dissection images, so as to quickly and accurately identify aortic dissection images, thereby solving the problems existing in the related art.

[0005] In order to achieve the above-mentioned purpose, the first part of the embodiment of the present disclosure provides a method for identifying an aortic dissection image, the method comprising:

[0006] Acquire an aortic CTA image, and extract an aortic 3D vascular image from the CTA image;

[0007] Constructing a sequence of to-be-identified layer sets according to the aorta 3D vascular image, wherein each to-be-identified layer in the to-be-identified layer set is a 2D image;

[0008] Inputting the to-be-identified layer set into the trained aortic dissection recognition model to obtain a recognition result indicating whether aortic dissection exists in the to-be-identified layer set;

[0009] It is determined whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence.

[0010] Optionally, the aorta 3D vascular image includes a plurality of layers, and constructing a sequence of layers to be identified according to the aorta 3D vascular image includes:

[0011] For each target layer except the first and last layers in the aorta 3D vascular image, construct the corresponding set of layers to be identified according to the target layer and the auxiliary layers, wherein the auxiliary layers of the target layer include two layers adjacent to the target layer in the aorta 3D vascular image;

[0012] According to the layer sequence of each target layer in the aorta 3D vascular image, the sequence of each layer set to be identified is determined to obtain a sequence of the layer sets to be identified.

[0013] Optionally, the step of inputting the to-be-identified layer set into a trained aortic dissection recognition model to obtain a recognition result indicating whether aortic dissection exists in the to-be-identified layer set includes:

[0014] The set of layers to be identified is input into the aortic dissection identification model to obtain the identification result indicating whether the target layer in the set of layers to be identified has aortic dissection.

[0015] Optionally, determining whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence includes:

[0016] For each of the to-be-identified layer sets, when the identification result of the to-be-identified layer set indicates that aortic dissection exists in the to-be-identified layer set, determining the to-be-identified layer set as a target to-be-identified layer set;

[0017] Determine the number of the target layer sets to be identified in succession in the sequence of layer sets to be identified;

[0018] When the number exceeds a preset number threshold, the CTA image is determined to be the aortic dissection image.

[0019] Optionally, determining whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence includes:

[0020] For each of the to-be-identified layer sets, when the identification result of the to-be-identified layer set indicates that aortic dissection exists in the to-be-identified layer set, determining the to-be-identified layer set as a target to-be-identified layer set;

[0021] Determine the number of the target layer sets to be identified in succession in the sequence of layer sets to be identified;

[0022] Calculate the axial length of aortic dissection based on the number and interval between layers;

[0023] When the axial length is greater than a preset axial length threshold, the CTA image is determined to be the aortic dissection image.

[0024] Optionally, extracting the aorta 3D vascular image from the CTA image includes:

[0025] Extracting a skeleton image from the CTA image and determining a central axis of the skeleton image;

[0026] Cutting out a cubic image with the central axis as the axis, a preset X-axis axial length, and a preset Y-axis axial length from the CTA image;

[0027] The aorta 3D blood vessel image is extracted from the cube image.

[0028] Optionally, extracting a skeleton image from the CTA image includes:

[0029] Performing image threshold segmentation on the CTA image according to the bone grayscale threshold to obtain a corresponding bone region image, wherein the grayscale value of each voxel in the bone region image is greater than the bone grayscale threshold;

[0030] Performing connected component analysis on the binary image of the bone region image to obtain at least one connected component of the bone region image;

[0031] The connected component with the largest number of voxels in at least one connected component of the skeleton region image is used as a skeleton component, and the image corresponding to the skeleton component is used as the skeleton image.

[0032] The second part of the embodiment of the present disclosure provides a device for identifying an aortic dissection image, the device comprising:

[0033] An acquisition module, used for acquiring an aortic CTA image and extracting an aortic 3D vascular image from the CTA image;

[0034] A construction module, used for constructing a sequence of to-be-identified layer sets according to the aorta 3D vascular image, wherein each to-be-identified layer in the to-be-identified layer set is a 2D image;

[0035] An input module, used for inputting the layer set to be identified into the trained aortic dissection identification model to obtain an identification result indicating whether aortic dissection exists in the layer set to be identified;

[0036] An execution module is used to determine whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence.

[0037] Optionally, the aorta 3D vascular image includes multiple layers, and the construction module includes:

[0038] A construction submodule, for constructing, for each target layer except the first and last layers in the aorta 3D vascular image, the corresponding set of layers to be identified according to the target layer and the auxiliary layers, wherein the auxiliary layers of the target layer include two layers adjacent to the target layer in the aorta 3D vascular image;

[0039] The execution submodule is used to determine the order of each of the to-be-identified layer sets according to the layer order of each of the target layers in the aorta 3D vascular image, so as to obtain a sequence of the to-be-identified layer sets.

[0040] Optionally, the input module is specifically used to input the set of layers to be identified into the aortic dissection identification model to obtain the identification result characterizing whether the target layer in the set of layers to be identified has aortic dissection.

[0041] Optionally, the execution module includes:

[0042] A first determination submodule is used for, for each of the to-be-identified layer sets, determining the to-be-identified layer set as a target to-be-identified layer set when the recognition result of the to-be-identified layer set indicates that aortic dissection exists in the to-be-identified layer set;

[0043] A second determination submodule is used to determine the number of consecutive target layer sets to be identified in the sequence of layer sets to be identified;

[0044] The third determination submodule is configured to determine that the CTA image is the aortic dissection image when the number exceeds a preset number threshold.

[0045] Optionally, the execution module includes:

[0046] A fourth determination submodule is configured to, for each of the to-be-identified layer sets, determine the to-be-identified layer set as a target to-be-identified layer set when the recognition result of the to-be-identified layer set indicates that aortic dissection exists in the to-be-identified layer set;

[0047] A fifth determination submodule is used to determine the number of the continuous target layer sets to be identified in the sequence of layer sets to be identified;

[0048] A calculation submodule, used for calculating the axial length of the aortic dissection according to the quantity and the interval between the layers;

[0049] The sixth determination submodule is used to determine that the CTA image is the aortic dissection image when the axial length is greater than a preset axial length threshold.

[0050] Optionally, the acquisition module includes:

[0051] a seventh determination submodule, configured to extract a skeleton image from the CTA image and determine a central axis of the skeleton image;

[0052] A cutting submodule, used for cutting out a cubic image with the central axis as the axis, a preset X-axis axial length, and a preset Y-axis axial length from the CTA image;

[0053] The extraction submodule is used to extract the aorta 3D blood vessel image from the cube image.

[0054] Optionally, the seventh determining submodule includes:

[0055] A segmentation submodule, used for performing image threshold segmentation on the CTA image according to a bone grayscale threshold to obtain a corresponding bone region image, wherein the grayscale value of each voxel in the bone region image is greater than the bone grayscale threshold;

[0056] An analysis submodule, configured to perform connected component analysis on the binary image of the bone region image to obtain at least one connected component of the bone region image;

[0057] The eighth determination submodule is used to use the connected component with the largest number of voxels in at least one connected component of the skeleton region image as a skeleton component, and use the image corresponding to the skeleton component as the skeleton image.

[0058] The third part of the embodiment of the present disclosure provides a non-temporary computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods described in the first part are implemented.

[0059] The fourth part of the embodiment of the present disclosure provides an electronic device, including:

[0060] a memory having a computer program stored thereon;

[0061] A processor, configured to execute the computer program in the memory to implement the steps of any one of the methods described in the first part.

[0062] By adopting the above technical solution, at least the following beneficial technical effects can be achieved:

[0063] By acquiring an aortic CTA image, and extracting an aortic 3D vascular image from the CTA image. A sequence of layers to be identified is constructed according to the aortic 3D vascular image, and each layer to be identified in the layer set to be identified is a 2D image. Each layer set to be identified is input into the trained aortic dissection recognition model to obtain a recognition result representing whether there is aortic dissection in each layer set to be identified. According to the recognition result corresponding to each layer set to be identified in the layer set sequence to be identified, it is determined whether the CTA image is an aortic dissection image. In this way disclosed in the present invention, by constructing a sequence of layers to be identified, and using the aortic dissection recognition model to perform aortic dissection recognition on each layer set to be identified, it is possible to quickly determine whether there is aortic dissection in each layer set to be identified. However, since the accuracy of the model cannot reach 100%, in this way disclosed in the present invention, after obtaining the recognition result representing whether there is aortic dissection in each layer set to be identified, it is further determined whether the CTA image is an aortic dissection image according to the recognition result corresponding to each layer set to be identified in the layer set sequence to be identified. This method can improve the accuracy of identifying aortic dissection images. Therefore, by adopting the method disclosed in the present invention, it is possible to quickly and accurately determine whether the aortic CTA image is an aortic dissection image. When the aortic CTA image is quickly and accurately determined to be an aortic dissection image, it is beneficial for patients with aortic dissection to receive timely treatment.

[0064] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] 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:

[0066] Figure 1 The figure is a flowchart of a method for identifying an aortic dissection image according to an exemplary embodiment of the present disclosure.

[0067] Figure 2 An aortic dissection recognition model architecture is shown according to an exemplary embodiment of the present disclosure.

[0068] Figure 3 It is a block diagram of a device for identifying an aortic dissection image according to an exemplary embodiment of the present disclosure.

[0069] Figure 4 It is a block diagram of an electronic device according to an exemplary embodiment.

[0070] Figure 5 is a block diagram of another electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0071] 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.

[0072] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the relevant data protection laws and policies of the country where the device is located and with the authorization given by the owner of the corresponding device.

[0073] In the related art, the research and reports on aortic dissection are focused on the segmentation of the true lumen and false lumen of aortic dissection. The segmentation of the true and false lumens of aortic dissection images of patients with known type B dissection disease can assist doctors in formulating surgical treatment plans. In the related art, the way to identify aortic dissection images is that after the patient undergoes an imaging examination, the patient waits for the doctor to analyze the image to determine whether the image is an aortic dissection image. This method is time-consuming and is not conducive to timely treatment for patients with aortic dissection. Therefore, there is an urgent need for a method that can quickly identify aortic dissection images.

[0074] In view of this, the embodiments of the present disclosure provide a method, device, storage medium and electronic device for identifying aortic dissection images, so as to quickly and accurately determine whether an aortic CTA image is an aortic dissection image, thereby saving the time for patients with aortic dissection to wait for medical imaging analysis and facilitating timely treatment for patients with aortic dissection.

[0075] Figure 1 FIG. 1 is a flow chart showing a method for identifying an aortic dissection image according to an exemplary embodiment of the present disclosure. Figure 1 As shown, the method for identifying aortic dissection images includes the following steps:

[0076] S11 . Acquire an aortic CTA image, and extract an aortic 3D vascular image from the CTA image.

[0077] Among them, the aortic CTA (Computed Tomography Angiography) image refers to the image obtained after the patient undergoes an aortic CTA examination, which can also be called an aortic CT angiography image. The aortic CTA image is a 3D image. Among them, CT (Computed Tomography) refers to electronic computer tomography, which is a technology that uses precisely collimated X-ray beams, gamma rays, ultrasound, etc., together with extremely sensitive detectors to perform cross-sectional scans one after another around a certain part of the human body to obtain a three-dimensional CT image.

[0078] Since the aorta is the largest artery in the human body, it originates from the left ventricle of the heart, goes up to the right and then down in a slightly arched shape, and then goes down along the spine, and divides into many smaller arteries in the chest and abdominal cavities. Therefore, the aortic CTA examination of patients is mainly performed on the chest and abdominal areas of patients. The aortic CTA image includes the aortic 3D vascular image.

[0079] S12, constructing a sequence of to-be-recognized layer sets according to the aorta 3D vascular image, wherein each to-be-recognized layer in the to-be-recognized layer set is a 2D image.

[0080] Since the aortic 3D vascular image is extracted from the CTA image, which includes multiple scan layers, the aortic 3D vascular image also includes multiple 2D layers, and the axial spacing between each layer is between 0.5 and 1.2 mm. The axial direction refers to the Z-axis direction of the CT scan, which is generally the direction of the human spine.

[0081] In some embodiments, a sequence of layers to be identified may be constructed based on 2D layers in the aorta 3D vascular image, wherein the number of layers in the layers to be identified may be 1, 3, 5, etc.

[0082] S13, inputting the to-be-identified layer set into the trained aortic dissection recognition model to obtain a recognition result indicating whether aortic dissection exists in the to-be-identified layer set.

[0083] In some embodiments, before the set of layers to be identified is input into the trained aortic dissection identification model, each layer in the set of layers to be identified may be preprocessed.

[0084] For example, preprocessing can be to perform window position and window width processing on each layer. Since various tissue structures or lesions have different image pixel values, when it is desired to display the details of a certain tissue structure, a window width and window position suitable for observing the certain tissue structure can be selected to obtain the best display effect. In the embodiment of the present disclosure, when performing window position and window width processing on each layer, the window position pixel value can be set to 200 and the window width to 700. The window position pixel value of 200 and the window width of 700 represent a window position of 200 pixel value and a window width of 350 pixels above and below the pixel value of 200. After the window position and window width processing, a layer with the best display effect of the aorta can be obtained.

[0085] For another example, preprocessing can be normalizing each layer, such as calling the Normalize() function to normalize each layer so that the mean of the image pixel value of each layer is 0 and the standard deviation is 1. Normalizing each layer is beneficial to improving the accuracy of the aortic dissection recognition model (it can improve the accuracy and convergence speed of the model during the training process).

[0086] In a specific implementation, by inputting the to-be-identified layer set (or the to-be-identified layer set after preprocessing) into the trained aortic dissection recognition model, a recognition result indicating whether aortic dissection exists in the to-be-identified layer set can be obtained.

[0087] S14. Determine whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence.

[0088] Since the human heart is in a continuous beating state, the continuous beating of the heart causes the blood vessels to continuously fluctuate. During the CTA examination process, the continuous fluctuation of blood vessels will cause artifacts to appear in the CTA image, and the artifacts are very similar to the imaging characteristics of aortic dissection. Therefore, the recognition result obtained by the aortic dissection recognition model after recognizing a single set of layers to be recognized has a certain error, and it is possible to recognize the vascular fluctuation phenomenon in a single set of layers to be recognized as aortic dissection. Therefore, it is impossible to accurately determine whether the entire CTA image is an aortic dissection image based on the recognition result of whether there is an aortic dissection in a single set of layers to be recognized. In view of this, the embodiment of the present disclosure proposes that after obtaining the recognition results of each set of layers to be recognized based on the aortic dissection recognition model, it is possible to further determine whether the corresponding CTA image is an aortic dissection image based on comprehensive features such as the number of sets of layers to be recognized with aortic dissection in the sequence of the set of layers to be recognized, the sorting position of each set of layers to be recognized with aortic dissection, etc.

[0089] The method disclosed in the present invention is adopted to obtain an aortic CTA image and extract an aortic 3D vascular image from the CTA image. A sequence of layers to be identified is constructed according to the aortic 3D vascular image, and each layer to be identified in the layer set to be identified is a 2D image. Each layer set to be identified is input into the trained aortic dissection recognition model to obtain a recognition result representing whether there is an aortic dissection in each layer set to be identified. According to the recognition result corresponding to each layer set to be identified in the layer set sequence to be identified, it is determined whether the CTA image is an aortic dissection image. In this method disclosed in the present invention, by constructing a sequence of layers to be identified and using the aortic dissection recognition model to perform aortic dissection recognition on each layer set to be identified, it is possible to quickly determine whether there is an aortic dissection in each layer set to be identified. However, since the accuracy of the model cannot reach 100%, in this method disclosed in the present invention, after obtaining the recognition result representing whether there is an aortic dissection in each layer set to be identified, it is further determined whether the CTA image is an aortic dissection image according to the recognition result corresponding to each layer set to be identified in the layer set sequence to be identified. This method can improve the accuracy of identifying aortic dissection images. Therefore, by adopting the method disclosed in the present invention, it is possible to quickly and accurately determine whether the aortic CTA image is an aortic dissection image. When the aortic CTA image is quickly and accurately determined to be an aortic dissection image, the time for patients with aortic dissection to wait for the results of medical imaging analysis can be reduced, which is conducive to timely treatment of patients with aortic dissection.

[0090] Optionally, the aorta 3D vascular image includes a plurality of layers, and constructing a sequence of layers to be identified according to the aorta 3D vascular image includes:

[0091] For each target layer except the first and last layers in the aortic 3D vascular image, a corresponding set of layers to be identified is constructed according to the target layer and the auxiliary layers, wherein the auxiliary layers of the target layer include two layers adjacent to the target layer in the aortic 3D vascular image; according to the layer order of each target layer in the aortic 3D vascular image, the order of each set of layers to be identified is determined to obtain a sequence of the set of layers to be identified.

[0092] For example, assuming that the aorta 3D blood vessel image includes layer 1, layer 2, layer 3, layer 4, layer 5, and layer 6 in sequence (generally, the number of layers is more than 500, and this is just an example). Then, the target layers except the first and last layers include layer 2, layer 3, layer 4, and layer 5.

[0093] The specific implementation method of constructing a set of layers to be identified for each target layer may be that, for any target layer, a corresponding set of layers to be identified is constructed according to the target layer and the auxiliary layers, wherein the auxiliary layers may include two layers adjacent to the target layer. For example, for the target layer, i.e., layer 2, since the two layers adjacent to layer 2 are layer 1 and layer 3, respectively, the auxiliary layers of layer 2 may include layer 1 and layer 3. Then, the set of layers to be identified corresponding to layer 2 includes layer 1, layer 2, and layer 3. For another example, for the target layer, i.e., layer 3, since the two layers adjacent to layer 3 are layer 2 and layer 4, respectively, the auxiliary layers of layer 3 may include layer 2 and layer 4. Then, the set of layers to be identified corresponding to layer 3 includes layer 2, layer 3, and layer 4. Similarly, the set of layers to be identified corresponding to layer 4 includes layer 3, layer 4, and layer 5. The set of layers to be identified corresponding to layer 5 includes layer 4, layer 5, and layer 6.

[0094] After obtaining the to-be-identified layer sets corresponding to each target layer, namely, layer 2, layer 3, layer 4, and layer 5, the order of each to-be-identified layer set can be determined according to the layer order of each target layer in the aorta 3D vascular image to obtain a sequence of to-be-identified layer sets. For example, the order of the to-be-identified layer sets corresponding to layer 2, layer 3, layer 4, and layer 5 is (layer 1, layer 2, and layer 3), (layer 2, layer 3, and layer 4), (layer 3, layer 4, and layer 5), (layer 4, layer 5, and layer 6).

[0095] Since the present disclosure does not limit the number of layers in the to-be-identified layer set, in some possible implementations, for each target layer, a corresponding to-be-identified layer set may be constructed based on the target layer itself.

[0096] In some other possible implementations, for each target layer, a corresponding set of layers to be identified may be constructed based on the target layer and the auxiliary layers, wherein the auxiliary layers may include a plurality of layers randomly selected from the aorta 3D vascular image. Alternatively, the auxiliary layers may include two layers before the target layer and / or two layers after the target layer. For example, for target layer 3, the corresponding auxiliary layers may include layer 1, layer 2, layer 4, and layer 5.

[0097] In addition, in some implementations, the number of layers in each to-be-identified layer set may be the same, for example, the number of layers in each to-be-identified layer set is 3. Similarly, in other implementations, the number of layers in each to-be-identified layer set may be different.

[0098] Optionally, the step of inputting the to-be-identified layer set into a trained aortic dissection recognition model to obtain a recognition result indicating whether aortic dissection exists in the to-be-identified layer set includes:

[0099] The set of layers to be identified is input into the aortic dissection identification model to obtain the identification result indicating whether aortic dissection exists in the target layer in the set of layers to be identified.

[0100] For example, for the set of layers to be identified (layer 1, layer 2, and layer 3) corresponding to the target layer 2 in the aforementioned embodiment, the set of layers to be identified (layer 1, layer 2, layer 3) is input into the aortic dissection recognition model to obtain the recognition result output by the aortic dissection recognition model, which can characterize whether there is an aortic dissection in the target layer 2.

[0101] In some embodiments, the aortic dissection recognition model can be trained in the following manner:

[0102] First, a training data set is constructed. Specifically, an aortic CTA annotated image sample is obtained. The aortic CTA annotated image sample includes an aortic dissection 3D annotated image sample (the location of the aortic dissection is annotated in the aortic dissection 3D annotated image sample) and a normal aorta 3D annotated image sample. For the aortic dissection 3D annotated image sample, extract by layer to obtain multiple 2D layers. And compare the aortic dissection 3D annotated image sample, mark whether there is an aortic dissection in each 2D layer to obtain a layer sample. For example, if there is an aortic dissection in the 2D layer, the annotated label is 1. If there is no aortic dissection in the 2D layer, the annotated label is 0. Similarly, the corresponding layer sample can also be determined for the normal aorta 3D annotated image sample. Then, a layer set sample to be identified is constructed for all layer samples, thereby obtaining an initial training data set.

[0103] It should be noted that, when the layer set sample to be identified includes three layer samples, for example, when the layer set sample to be identified includes a target layer sample and two auxiliary layer samples, the label of the target layer sample can be used as the label of the layer set sample to be identified.

[0104] In some implementations, preprocessing may be performed on the initial training data set, and the preprocessing may be window level and window width processing of each layer sample.

[0105] In addition, the preprocessing may also include random image enhancement processing. Specifically, the random image enhancement processing method may include at least one of the following:

[0106] Layer random rotation processing, preferably, the parameters are set as the minimum rotation angle min=-10 and the maximum rotation angle max=10;

[0107] Layer contrast enhancement processing, preferably, setting the parameters to a saturation factor minimum value min=0.1 and a maximum value max=1.9;

[0108] Cutout enhancement processing (i.e. randomly cutting out part of the sample area and filling it with pixel values), preferably, the parameters are set such that the cutting and filling area is a rectangle, and the ratio coefficient of the width / height of the rectangle to the width / height of the layer has a minimum value of min=0 and a maximum value of max=0.4, and the filling pixel value is the average pixel value of the layer.

[0109] For the layer zoom-in / zoom-out enhancement processing, preferably, the parameters are set to a minimum zoom factor of min = 0.8 and a maximum zoom factor of max = 1.2. After the layer zoom-in / zoom-out enhancement processing, the layer is restored to its initial size and normalized so that the mean of the layer pixel value is 0 and the standard deviation is 1.

[0110] After the initial training data set is preprocessed, the training data set can be obtained.

[0111] The aortic dissection recognition model to be trained is trained according to the training data set, so as to obtain a trained aortic dissection recognition model.

[0112] In some embodiments, the architecture of the aortic dissection recognition model can adopt the EfficientNet-b7 model architecture, and the parameters trained on the Imagenet dataset can be used as the initialization parameters of the aortic dissection recognition model to be trained. Then, the aortic dissection recognition model is transferred and trained according to the training dataset constructed according to the embodiment of the present disclosure, so as to obtain the trained aortic dissection recognition model.

[0113] In some embodiments, the aortic dissection recognition model architecture may also include only a partial module architecture for extracting image features in the EfficientNet-b7 model, and then combine the adaptive average pooling layer, the dropout layer, and the fully connected layer to build a complete model architecture. For example, Figure 2 shown. Figure 2 In the aortic dissection recognition model architecture in the EfficientNet-b7 model, the partial module architecture for extracting image features in the EfficientNet-b7 model is used to extract the image features of each layer sample in the layer set to be identified. After the image features are processed by an adaptive average pooling layer (such as nn.AdaptiveAvgPool2d(1) of Pytorch in the related technology), a dropout layer, and a fully connected layer, an identification result can be obtained to characterize whether aortic dissection exists in the layer set sample to be identified (target layer sample).

[0114] In some embodiments, the specific training method for training the aortic dissection recognition model to be trained may be as follows:

[0115] The data input of the aortic dissection recognition model to be trained is set to 3 channels, which are used to simultaneously input 3 layer samples in the layer set samples to be recognized.

[0116] Set the data loading mode to distributed sampling collection loading mode.

[0117] The loss function is set to a class-weighted binary cross entropy loss function (such as the BCEWithLogitsLoss function of Pytorch in the related art), wherein the class weight class_weight is preferably set to [0.2, 0.8] to adapt to the scenario of using an unbalanced dataset to train the model.

[0118] The optimizer uses the Adam optimizer, uses the default parameter settings (that is, the initial parameters use the default values), and the model optimization method uses the mixed precision training method. The mixed precision training method is a model training method in the relevant technology and is not explained here.

[0119] Optionally, determining whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence includes:

[0120] For each of the layer sets to be identified, when the identification result of the layer set to be identified indicates that aortic dissection exists in the layer set to be identified, the layer set to be identified is determined as a target layer set to be identified; the number of consecutive target layer sets to be identified in the sequence of layer sets to be identified is determined; and when the number exceeds a preset number threshold, the CTA image is determined to be the aortic dissection image.

[0121] For example, it is assumed that the layers to be identified are A (layer 1, layer 2, layer 3), B (layer 2, layer 3, layer 4), C (layer 3, layer 4, layer 5), and D (layer 4, layer 5, layer 6). When the identification result of the layer set B to be identified indicates that aortic dissection exists in the layer set B to be identified (target layer 2), the layer set B to be identified can be determined as the target layer set to be identified.

[0122] If the layers to be identified B, C, and D are all target layers to be identified, then it can be determined that the number of consecutive target layers to be identified in the layers to be identified sequence A, B, C, and D is 3 (i.e., B, C, and D). When the number 3 exceeds the preset number threshold N, it can be determined that the corresponding CTA image is an aortic dissection image. On the contrary, when the number 3 does not exceed the preset number threshold N, it can be determined that the corresponding CTA image is not an aortic dissection image.

[0123] The preset number threshold N may be 9, 10, etc. The specific value may be determined according to the parameter setting of the CT detection instrument.

[0124] Since the human heart is in a continuous beating state, the continuous beating of the heart causes the blood vessels to continuously fluctuate. During the CTA examination, the continuous fluctuation of blood vessels will cause artifacts to appear in the CTA image. The artifacts are very similar to the imaging characteristics of aortic dissection, and it is easy to misjudge the artifacts as aortic dissection. Therefore, in order to eliminate the interference caused by artifacts, the embodiment of the present disclosure also provides another implementation method for determining whether the CTA image is an aortic dissection image based on the recognition results corresponding to each layer set to be identified in the layer set sequence to be identified, which is as follows:

[0125] Determining whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence includes:

[0126] For each of the layer sets to be identified, when the identification result of the layer set to be identified indicates that aortic dissection exists in the layer set to be identified, the layer set to be identified is determined as a target layer set to be identified; the number of consecutive target layer sets to be identified in the sequence of layer sets to be identified is determined; the axial length of the aortic dissection is calculated based on the number and the layer spacing; and when the axial length is greater than a preset axial length threshold, the CTA image is determined to be the aortic dissection image.

[0127] The interval between the layers is [0.5, 1.2], in units of mm (millimeter), which is the same as the axial interval between the layers mentioned above.

[0128] During the experiment, it was found that aortic dissection appears to continuously cross layers, and the axial length of the dissection is usually greater than 30 mm. The axial length of the artifact caused by the undulation of the blood vessels is generally less than 30 mm. Therefore, the disclosed embodiment proposes that the axial length of the aortic dissection can be calculated based on the number of continuous target layer sets to be identified in the layer set sequence to be identified, and the layer spacing. When the axial length is greater than a preset axial length threshold, such as 30 mm, it can be accurately determined that the CTA image is an aortic dissection image. On the contrary, when the axial length is less than or equal to a preset axial length threshold, such as 30 mm, it can be determined that the CTA image is not an aortic dissection image.

[0129] It should be noted that, in some embodiments, since the set of layers to be identified is input into the trained aortic dissection recognition model, an identification result characterizing whether aortic dissection exists in the set of layers to be identified is obtained. Specifically, the set of layers to be identified can be input into the aortic dissection recognition model to obtain the identification result characterizing whether the target layer in the set of layers to be identified has aortic dissection.

[0130] Therefore, the layer set to be identified is input into the trained aortic dissection identification model to obtain a recognition result, which indicates whether there is aortic dissection in the layer set to be identified. Specifically, the recognition result indicates whether there is aortic dissection in the target layer in the layer set to be identified. Then, the number of continuous target layer sets to be identified in the sequence of layer sets to be identified is determined, and the number can indicate the number of continuous target images with aortic dissection.

[0131] Therefore, two other implementations that are substantially the same as the above implementation of determining whether a CTA image is an aortic dissection image based on the recognition results corresponding to each to-be-recognized layer set in the to-be-recognized layer set sequence are:

[0132] Embodiment 1: Determining whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence includes:

[0133] For each of the to-be-identified layer sets, when the identification result of the to-be-identified layer set indicates that aortic dissection exists in the target layer in the to-be-identified layer set, determining the target layer in the to-be-identified layer set as a valid target layer;

[0134] Determine the number of continuous valid target layers in the aorta 3D vascular image;

[0135] When the number exceeds a preset number threshold, the CTA image is determined to be the aortic dissection image.

[0136] Embodiment 2: Determining whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence comprises:

[0137] For each of the to-be-identified layer sets, when the identification result of the to-be-identified layer set indicates that aortic dissection exists in the target layer in the to-be-identified layer set, determining the target layer in the to-be-identified layer set as a valid target layer;

[0138] Determine the number of continuous valid target layers in the aorta 3D vascular image;

[0139] Calculate the axial length of aortic dissection based on the number and interval between layers;

[0140] When the axial length is greater than a preset axial length threshold, the CTA image is determined to be the aortic dissection image.

[0141] Optionally, extracting the aorta 3D vascular image from the CTA image includes:

[0142] An image region including an aortic vessel is determined from the CTA image, where the size of the image region is smaller than the size of the CTA image; and an aortic 3D vessel image is extracted from the image region.

[0143] Since the image resolution of the CTA image obtained after the patient undergoes aortic CTA examination is very large, the image resolution of the CTA image is generally (512*512), and the number of layers is generally more than 500 layers. Therefore, the amount of calculation required to directly extract the aortic 3D vascular image from the CTA image is large, and the performance requirements of the computing device are high. In addition, research has found that the aorta is only located in a small area in the center of the CTA image. Therefore, a preferred implementation is to first determine an image area including the aortic blood vessels from the CTA image, and the size of the image area is smaller than the size of the CTA image. Then extract the aortic 3D vascular image from the image area. This can avoid a large amount of calculations to directly extract the aortic 3D vascular image from the CTA image.

[0144] In some embodiments, the image center point of the CTA image can be used as a cutting reference to cut out an image region of a preset size, and then the aorta 3D vascular image is extracted from the image region. The preset size can be an empirical value determined based on a large number of aorta 3D vascular images.

[0145] Considering individual differences and differences in the settings of CT equipment, the aorta may not be completely located at the center of the CTA image. Therefore, if the above method of cutting out an image area of ​​a preset size using the image center point of the CTA image as a cutting reference is adopted, part of the aortic blood vessels located outside the image area may not be extracted. As a result, the aortic 3D blood vessel image extracted from the image area is not a complete aortic image.

[0146] In view of this, preferably, extracting the aorta 3D vascular image from the CTA image includes:

[0147] A skeleton image is extracted from the CTA image, and a central axis of the skeleton image is determined; a cube image with the central axis as the axis, a preset X-axis axial length, and a preset Y-axis axial length is cut out from the CTA image; and the aorta 3D vascular image is extracted from the cube image.

[0148] In view of the physiological characteristics of the aorta being contained in the sternal skeleton and the high density of bone images, which makes it easy to extract, the disclosed embodiment proposes that the image area where the skeleton is located can be first determined from the CTA image, and then the complete aortic 3D vascular image can be further extracted from the image area where the skeleton is located.

[0149] For example, a skeleton image is extracted from a CTA image, and the central axis of the skeleton image is determined. A cube image with the central axis as the axis, a preset X-axis axial length, and a preset Y-axis axial length is cut out from the CTA image. A 3D vascular image of the aorta is extracted from the cube image. The direction of the central axis is the Z axis. The preset X-axis axial length and the preset Y-axis axial length are empirical values. In detail, an image processing tool can be used to analyze the shape characteristics of the skeleton image, and determine the initial vertex coordinates (X0, Y0, Z0) of the circumscribed cube of the skeleton, and calculate the center point coordinates (X0, Y0, Z0) of each layer according to the preset X-axis axial length a and the preset Y-axis axial length b. c ,Y c ), where X c =X0+a / 2,Y c =Y0+b / 2. Next, when the CTA image has 512*512 pixels and 512 layers, the area [X c -128,X c +128], [Y c -128, Y c+128], [0, 512] to cut out a cube image from the CTA image. In this way, a high-resolution CTA image (512*512*512) can be cut into a cube image including the aorta with a resolution of (256*256*512). Not only can a complete aorta 3D vascular image be extracted from a cube image with a lower resolution, but the amount of calculation is also reduced.

[0150] Optionally, extracting a skeleton image from the CTA image includes:

[0151] The CTA image is subjected to image threshold segmentation according to a bone grayscale threshold to obtain a corresponding bone region image, wherein the grayscale value of each voxel in the bone region image is greater than the bone grayscale threshold; a connected component analysis is performed on the binary image of the bone region image to obtain at least one connected component of the bone region image; the connected component with the largest number of voxels in at least one connected component of the bone region image is used as a bone component, and the image corresponding to the bone component is used as the skeleton image.

[0152] The bone grayscale threshold can be set to 200. Performing image threshold segmentation on the CTA image according to the bone grayscale threshold of 200 can obtain the corresponding bone region image. The grayscale value of each voxel in the bone region image is greater than (or equal to) the bone grayscale threshold of 200. The bone grayscale threshold of 200 may not be an accurate bone extraction value because the embodiment of the present disclosure is only for determining the bone region image where the bone is located. It is not required that each voxel in the bone region image represents the bone, and the bone region image may include other voxels representing non-bone.

[0153] After obtaining the image of the bone region where the bone is located, a connected component analysis can be performed on the binary image of the bone region image to obtain at least one connected component of the bone region image. The connected component with the largest number of voxels in at least one connected component of the bone region image is used as a bone component, and the image corresponding to the bone component is used as a skeleton image. The connected component represents a 3D image region composed of foreground voxel points with the same voxel value (or satisfying a specific similarity criterion) and adjacent positions in the 3D image.

[0154] Optionally, extracting the aorta 3D vascular image from the cube image includes:

[0155] Each layer in the cube image is preprocessed, including window level and window width processing (the window level pixel value is 200 and the window width is 700, which can ensure that the aorta area is located in the layer and the display effect is better).

[0156] Preprocessing may also include image enhancement processing, such as scaling the pixel values ​​of each layer to [0, 255].

[0157] Preprocessing can also include normalization, so that the pixel mean of each layer after enhancement is 0 and the standard deviation is 1, so as to improve the accuracy of the model and the convergence speed of the model during training.

[0158] Each preprocessed layer is input into the trained segmentation model to obtain a binary layer output by the segmentation model. The voxel value in the binary layer is 0 or 1, where 1 represents the aortic voxel and 0 represents the non-aortic voxel. The segmentation model can specifically adopt the 2D UNet model in the related technology and the Pytorch framework. The training method of the segmentation model is similar to the training method of the 2D UNet model in the related technology, which will not be repeated here.

[0159] Each binary layer is multiplied with the corresponding layer in the preprocessed cubic image (the voxel values ​​at the same position are multiplied) to obtain the aorta 3D vascular image.

[0160] After experimental verification, the above-mentioned method for identifying aortic dissection images disclosed in the present invention can be used to determine whether the aortic CTA image is an aortic dissection image within 2 minutes. Compared with the method of waiting for manual and manual judgment of whether the aortic CTA image is an aortic dissection image in the prior art, the judgment time is shortened. Moreover, the present invention has also been experimentally verified, and the recognition results of the test set show that the recall rate reached 100% and the accuracy rate reached 98.3%. Therefore, the method for identifying aortic dissection images disclosed in the present invention can quickly and accurately determine whether the aortic CTA image is an aortic dissection image, thereby saving the time for patients with aortic dissection to wait for medical imaging analysis, and facilitating timely treatment of patients with aortic dissection. For example, when it is determined that a patient's aortic CTA image is an aortic dissection image, medical staff can be prompted so that the patient with aortic dissection can receive timely treatment.

[0161] Figure 3 FIG. 1 is a block diagram of a device for identifying an aortic dissection image according to an exemplary embodiment of the present disclosure. Figure 3 As shown, the device 300 includes:

[0162] An acquisition module 310 is used to acquire an aortic CTA image and extract an aortic 3D vascular image from the CTA image;

[0163] A construction module 320 is used to construct a sequence of to-be-identified layer sets according to the aorta 3D vascular image, wherein each to-be-identified layer in the to-be-identified layer set is a 2D image;

[0164] An input module 330 is used to input the layer set to be identified into the trained aortic dissection identification model to obtain an identification result indicating whether aortic dissection exists in the layer set to be identified;

[0165] The execution module 340 is used to determine whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence.

[0166] By adopting this device, an aortic CTA image is acquired, and an aortic 3D vascular image is extracted from the CTA image. A sequence of layers to be identified is constructed according to the aortic 3D vascular image, and the layers to be identified in each layer set to be identified are 2D images. Each layer set to be identified is input into the trained aortic dissection recognition model to obtain a recognition result representing whether there is aortic dissection in each layer set to be identified. According to the recognition result corresponding to each layer set to be identified in the sequence of layers to be identified, it is determined whether the CTA image is an aortic dissection image. In this way of the present disclosure, by constructing a sequence of layers to be identified, and using the aortic dissection recognition model to perform aortic dissection recognition on each layer set to be identified, it is possible to quickly determine whether there is aortic dissection in each layer set to be identified. However, since the accuracy of the model cannot reach 100%, in this way of the present disclosure, after obtaining the recognition result representing whether there is aortic dissection in each layer set to be identified, it is further determined whether the CTA image is an aortic dissection image according to the recognition result corresponding to each layer set to be identified in the sequence of layers to be identified. This method can improve the accuracy of identifying aortic dissection images. Therefore, by adopting the method disclosed in the present invention, it is possible to quickly and accurately determine whether the aortic CTA image is an aortic dissection image. When the aortic CTA image is quickly and accurately determined to be an aortic dissection image, it is beneficial for patients with aortic dissection to receive timely treatment.

[0167] Optionally, the aorta 3D blood vessel image includes multiple layers, and the construction module 320 includes:

[0168] A construction submodule, for constructing, for each target layer except the first and last layers in the aorta 3D vascular image, the corresponding set of layers to be identified according to the target layer and the auxiliary layers, wherein the auxiliary layers of the target layer include two layers adjacent to the target layer in the aorta 3D vascular image;

[0169] The execution submodule is used to determine the order of each of the to-be-identified layer sets according to the layer order of each of the target layers in the aorta 3D vascular image, so as to obtain a sequence of the to-be-identified layer sets.

[0170] Optionally, the input module 330 is specifically configured to input the set of layers to be identified into the aortic dissection identification model to obtain the identification result indicating whether the target layer in the set of layers to be identified has aortic dissection.

[0171] Optionally, the execution module 340 includes:

[0172] A first determination submodule is used for, for each of the to-be-identified layer sets, determining the to-be-identified layer set as a target to-be-identified layer set when the recognition result of the to-be-identified layer set indicates that aortic dissection exists in the to-be-identified layer set;

[0173] A second determination submodule is used to determine the number of consecutive target layer sets to be identified in the sequence of layer sets to be identified;

[0174] The third determination submodule is configured to determine that the CTA image is the aortic dissection image when the number exceeds a preset number threshold.

[0175] Optionally, the execution module 340 includes:

[0176] A fourth determination submodule is configured to, for each of the to-be-identified layer sets, determine the to-be-identified layer set as a target to-be-identified layer set when the recognition result of the to-be-identified layer set indicates that aortic dissection exists in the to-be-identified layer set;

[0177] A fifth determination submodule is used to determine the number of the continuous target layer sets to be identified in the sequence of layer sets to be identified;

[0178] A calculation submodule, used for calculating the axial length of the aortic dissection according to the quantity and the interval between the layers;

[0179] The sixth determination submodule is used to determine that the CTA image is the aortic dissection image when the axial length is greater than a preset axial length threshold.

[0180] Optionally, the acquisition module 310 includes:

[0181] a seventh determination submodule, configured to extract a skeleton image from the CTA image and determine a central axis of the skeleton image;

[0182] A cutting submodule, used for cutting out a cubic image with the central axis as the axis, a preset X-axis axial length, and a preset Y-axis axial length from the CTA image;

[0183] The extraction submodule is used to extract the aorta 3D blood vessel image from the cube image.

[0184] Optionally, the seventh determining submodule includes:

[0185] A segmentation submodule, used for performing image threshold segmentation on the CTA image according to a bone grayscale threshold to obtain a corresponding bone region image, wherein the grayscale value of each voxel in the bone region image is greater than the bone grayscale threshold;

[0186] An analysis submodule, configured to perform connected component analysis on the binary image of the bone region image to obtain at least one connected component of the bone region image;

[0187] The eighth determination submodule is used to use the connected component with the largest number of voxels in at least one connected component of the skeleton region image as a skeleton component, and use the image corresponding to the skeleton component as the skeleton image.

[0188] 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.

[0189] Figure 4 FIG. 7 is a block diagram of an electronic device 700 according to an exemplary embodiment. Figure 4 As shown, the electronic device 700 may include: a processor 701 and a memory 702. The electronic device 700 may also include one or more of a multimedia component 703, an input / output (I / O) interface 704, and a communication component 705.

[0190] The processor 701 is used to control the overall operation of the electronic device 700 to complete all or part of the steps in the above-mentioned method for aortic dissection images. The memory 702 is used to store various types of data to support the operation of the electronic device 700, and these data may include, for example, instructions for any application or method operated on the electronic device 700, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 702 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, magnetic disk or optical disk. The multimedia component 703 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 702 or sent through the communication component 705. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 704 provides an interface between the processor 701 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 705 is used for wired or wireless communication between the electronic device 700 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 705 may include: Wi-Fi module, Bluetooth module, NFC module, etc.

[0191] In an exemplary embodiment, the electronic device 700 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 method for imaging aortic dissection.

[0192] 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 method for aortic dissection imaging are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 702 including program instructions, and the above-mentioned program instructions can be executed by the processor 701 of the electronic device 700 to complete the above-mentioned method for aortic dissection imaging.

[0193] Figure 5 1 is a block diagram of an electronic device 1900 according to an exemplary embodiment. For example, the electronic device 1900 may be provided as a server. Figure 5 The electronic device 1900 includes a processor 1922, which may be one or more, and a memory 1932 for storing a computer program executable by the processor 1922. The computer program stored in the memory 1932 may include one or more modules each corresponding to a set of instructions. In addition, the processor 1922 may be configured to execute the computer program to perform the above-mentioned method for imaging aortic dissection.

[0194] In addition, the electronic device 1900 may further include a power supply component 1926 and a communication component 1950. The power supply component 1926 may be configured to perform power management of the electronic device 1900, and the communication component 1950 may be configured to implement communication of the electronic device 1900, for example, wired or wireless communication. In addition, the electronic device 1900 may further include an input / output (I / O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server 2003. TM , Mac OS X TM , Unix TM , Linux TM etc.

[0195] 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 method for imaging aortic dissection are implemented. For example, the non-transitory computer-readable storage medium may be the above-mentioned memory 1932 including program instructions, and the above-mentioned program instructions may be executed by the processor 1922 of the electronic device 1900 to complete the above-mentioned method for imaging aortic dissection.

[0196] In another exemplary embodiment, a computer program product is also provided, which includes a computer program executable by a programmable device, and has code portions for performing the above-mentioned method for imaging aortic dissection when the computer program is executed by the programmable device.

[0197] 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.

[0198] 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.

[0199] 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. A method for identifying an aortic dissection image, characterized in that: The method comprises: Acquire aortic CTA images; Extracting a sternum skeleton image from the CTA image, and determining a central axis of the sternum skeleton image; The step of extracting a skeleton image from the CTA image comprises: Performing image threshold segmentation on the CTA image according to the bone grayscale threshold to obtain a corresponding bone region image, wherein the grayscale value of each voxel in the bone region image is greater than the bone grayscale threshold; Performing connected component analysis on the binary image of the bone region image to obtain at least one connected component of the bone region image; Using the connected component with the largest number of voxels in at least one connected component of the skeleton region image as a skeleton component, and using the image corresponding to the skeleton component as the skeleton image; Cutting out a cubic image with the central axis as the axis, a preset X-axis axial length, and a preset Y-axis axial length from the CTA image; Extracting an aorta 3D vascular image from the cube image; Constructing a sequence of to-be-identified layer sets according to the aorta 3D vascular image, wherein each to-be-identified layer in the to-be-identified layer set is a 2D image; Inputting the layer set to be identified into a trained aortic dissection identification model to obtain an identification result indicating whether there is aortic dissection in the layer set to be identified, wherein the aortic dissection identification model is obtained by training a training data set, wherein the training data set includes a plurality of layer set samples to be identified, each of the layer set samples to be identified includes a plurality of layer samples, and the plurality of layer samples include a target layer sample and an auxiliary layer sample; Determine whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence; Determining whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence includes: For each of the to-be-identified layer sets, when the identification result of the to-be-identified layer set indicates that aortic dissection exists in the to-be-identified layer set, determining the to-be-identified layer set as a target to-be-identified layer set; Determine the number of the target layer sets to be identified in succession in the sequence of layer sets to be identified; When the number exceeds a preset number threshold, the CTA image is determined to be the aortic dissection image.

2. The method according to claim 1, characterized in that The aorta 3D blood vessel image includes a plurality of layers, and constructing a sequence of layers to be identified according to the aorta 3D blood vessel image includes: For each target layer except the first and last layers in the aorta 3D vascular image, construct the corresponding set of layers to be identified according to the target layer and the auxiliary layers, wherein the auxiliary layers of the target layer include two layers adjacent to the target layer in the aorta 3D vascular image; According to the layer sequence of each target layer in the aorta 3D vascular image, the sequence of each layer set to be identified is determined to obtain a sequence of the layer sets to be identified.

3. The method according to claim 2, characterized in that The step of inputting the to-be-identified layer set into the trained aortic dissection identification model to obtain an identification result indicating whether aortic dissection exists in the to-be-identified layer set comprises: The set of layers to be identified is input into the aortic dissection identification model to obtain the identification result indicating whether the target layer in the set of layers to be identified has aortic dissection.

4. The method according to any one of claims 1 to 3, characterized in that Determining whether the CTA image is an aortic dissection image according to the recognition results corresponding to each of the to-be-recognized layer sets in the to-be-recognized layer set sequence includes: For each of the to-be-identified layer sets, when the identification result of the to-be-identified layer set indicates that aortic dissection exists in the to-be-identified layer set, determining the to-be-identified layer set as a target to-be-identified layer set; Determine the number of the target layer sets to be identified in succession in the sequence of layer sets to be identified; Calculate the axial length of aortic dissection based on the number and interval between layers; When the axial length is greater than a preset axial length threshold, the CTA image is determined to be the aortic dissection image.

5. A device for identifying an aortic dissection image, characterized in that: The device comprises: An acquisition module is used to acquire an aortic CTA image; extract a sternum skeleton image from the CTA image and determine the central axis of the sternum skeleton image; cut out a cubic image with the central axis as the axis, a preset X-axis axial length, and a preset Y-axis axial length from the CTA image; extract an aortic 3D vascular image from the cubic image; perform image threshold segmentation on the CTA image according to a bone grayscale threshold to obtain a corresponding bone region image, wherein the grayscale value of each voxel in the bone region image is greater than the bone grayscale threshold; perform connected component analysis on the binary image of the bone region image to obtain at least one connected component of the bone region image; use the connected component with the largest number of voxels in at least one connected component of the bone region image as a bone component, and use the image corresponding to the bone component as the skeleton image; A construction module, used for constructing a sequence of to-be-identified layer sets according to the aorta 3D vascular image, wherein each to-be-identified layer in the to-be-identified layer set is a 2D image; An input module is used to input the layer set to be identified into a trained aortic dissection recognition model to obtain a recognition result indicating whether there is aortic dissection in the layer set to be identified, wherein the aortic dissection recognition model is obtained by training a training data set, wherein the training data set includes a plurality of layer set samples to be identified, each of the layer set samples to be identified includes a plurality of layer samples, and the plurality of layer samples include a target layer sample and an auxiliary layer sample; An execution module is used to determine whether the CTA image is an aortic dissection image based on the recognition results corresponding to each of the layer sets to be identified in the sequence of layer sets to be identified; for each of the layer sets to be identified, if the recognition result of the layer set to be identified indicates that aortic dissection exists in the layer set to be identified, determine the layer set to be identified as a target layer set to be identified; determine the number of consecutive target layer sets to be identified in the sequence of layer sets to be identified; and determine that the CTA image is an aortic dissection image if the number exceeds a preset number threshold.

6. 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 4 are implemented.

7. 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 4.

Images