Aortic Screening Method, Device, Electronic Device and Computer Readable Storage Medium

Abstract

The present application relates to an aortic screening method, apparatus, computer device, and storage medium. The method includes: acquiring an original image of the aorta. Analyzing and processing the original image of the aorta to determine a plurality of segmentation masks of the aorta, a plurality of key points of the aorta, and the continuous cross-sectional diameter of the aorta, where the key points are the contact points between the branches of the aorta and the aortic trunk of the aorta. Determining a plurality of key cross-sections according to the plurality of key points and the plurality of segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta. Determining the aortic risk level according to the plurality of key cross-sections and the continuous cross-sectional diameter. It can improve the accuracy of aortic screening.

Description

Technical Field

[0001] The present application relates to the field of image processing technologies, and particularly to an aorta screening method, apparatus, electronic device, and computer-readable storage medium. Background Art

[0002] The aorta is the main blood vessel that transports blood to various parts of the body, and its health status is very important in the preventive evaluation of many cardiovascular diseases. The aorta can experience lumen dilation, or even rupture, under various pathological factors. Since most cardiovascular diseases such as hypertension, coronary artery, and peripheral artery are closely related to the morphological changes of the aorta, early screening of the aorta is particularly important. The key to aorta screening is the segmentation of the aorta and the positioning of multiple groups of key cross-sections.

[0003] However, the current positioning of key cross-sections needs to be obtained by combining manual measurements. Due to the very limited accuracy of the measurements, objective and accurate measurement parameters cannot be provided, which in turn leads to a relatively low accuracy of aorta screening. Summary of the Invention

[0004] Based on this, in view of the above technical problems, it is necessary to provide an aorta screening method, apparatus, computer device, and storage medium, which can improve the accuracy of aorta screening.

[0005] In a first aspect, the present application provides an aorta screening method, and the method includes:

[0006] Obtain the original image of the aorta.

[0007] Analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-section diameter of the aorta, where the key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0008] Determine multiple key cross-sections according to the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0009] Determine the aorta risk level according to the multiple key cross-sections and the continuous cross-section diameter.

[0010] In one embodiment, the determining the aorta risk level according to the multiple key cross-sections and the continuous cross-section diameter includes:

[0011] Determine key cross-section parameters, where the key cross-section parameters include at least one of the cross-section diameter of each key cross-section in the multiple key cross-sections and the volume and length between two adjacent key cross-sections.

[0012] Determine the aortic risk level based on the multiple key cross-sections, the continuous cross-section diameters, and the key cross-section parameters.

[0013] In one embodiment, the determining the aortic risk level based on the multiple key cross-sections, the continuous cross-section diameters, and the key cross-section parameters includes:

[0014] Determine a global diameter feature vector according to the continuous cross-section diameters; the global diameter feature vector is used to characterize the global diameter size of the aorta.

[0015] Determine the feature vectors of the respective key cross-sections according to the key cross-section parameters; the feature vectors of the respective key cross-sections are used to characterize the parameter features of the respective key cross-sections.

[0016] Determine an image feature vector according to the original image of the aorta and the mask images of the multiple key cross-sections; the image feature vector is used to characterize the relationship feature between the original image of the aorta and the key cross-sections.

[0017] Determine the aortic risk level based on the global diameter feature vector, the feature vectors of the respective key cross-sections, and the image feature vector.

[0018] In one embodiment, the determining the aortic risk level based on the global diameter feature vector, the feature vectors of the respective key cross-sections, and the image feature vector includes:

[0019] Perform feature fusion on the global diameter feature vector, the feature vectors of the respective key cross-sections, and the image feature vector to obtain a fused feature vector.

[0020] Determine the aortic risk level according to the fused feature vector.

[0021] In one embodiment, the analyzing and processing the original image of the aorta to determine the multiple key points includes:

[0022] Determine a probability map of the multiple key points of the aorta and multiple segmentation masks of the aorta according to the original image of the aorta; the probability map is used to characterize the probability of each position where the key points are located in the aorta.

[0023] Determine each of the key points according to the probability map of each of the key points.

[0024] In one embodiment, the determining the coordinates of the key points according to the probability map of the key points includes: determining the position where the maximum value in the probability map of the key points is located as the key point.

[0025] In one embodiment, determining a plurality of key cross-sections according to the plurality of key points and the plurality of segmentation masks includes:

[0026] Determining the centerline of the aorta according to the plurality of segmentation masks.

[0027] Determining the plurality of key cross-sections according to the plurality of key points and the centerline.

[0028] In one embodiment, the plurality of key cross-sections include at least two of the following cross-sections: aortic sinus cross-section, sinoatrial node cross-section, middle ascending aorta cross-section, proximal descending aorta cross-section, aortic arch cross-section, proximal descending aorta, middle descending aorta cross-section, aortic cross-section at the diaphragm, and abdominal aorta cross-section.

[0029] In one embodiment, the plurality of segmentation masks include at least two of the background, aortic trunk, branches of the aorta, and aortic sinus in the original image of the aorta, and the branches of the aorta include at least one of brachiocephalic artery, carotid artery, subclavian artery, and abdominal aorta.

[0030] In a second aspect, the present application further provides an aortic screening device, and the device includes:

[0031] An acquisition module, configured to acquire an original image of the aorta.

[0032] A processing module, configured to analyze and process the original image of the aorta acquired by the acquisition module to determine a plurality of segmentation masks of the aorta, a plurality of key points of the aorta, and the continuous cross-sectional diameter of the aorta, where the key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0033] The processing module is further configured to determine a plurality of key cross-sections according to the plurality of key points and the plurality of segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0034] The processing module is further configured to determine the aortic risk level according to the plurality of key cross-sections and the continuous cross-sectional diameter.

[0035] In one embodiment, the processing module is specifically configured to,

[0036] Determine key cross-section parameters, where the key cross-section parameters include at least one of the cross-sectional diameter of each key cross-section in the plurality of key cross-sections and the volume and length between two adjacent key cross-sections.

[0037] Determine the aortic risk level according to the plurality of key cross-sections, the continuous cross-sectional diameter, and the key cross-section parameters.

[0038] In one embodiment, the processing module is specifically configured to,

[0039] Determine a global diameter feature vector according to the continuous cross-section diameters; the global diameter feature vector is used to characterize the global diameter size of the aorta.

[0040] Determine the feature vectors of the respective key cross-sections according to the key cross-section parameters; the feature vectors of the respective key cross-sections are used to characterize the parameter features of the respective key cross-sections.

[0041] Determine an image feature vector according to the original image of the aorta and the mask images of the multiple key cross-sections; the image feature vector is used to characterize the relationship feature between the original image of the aorta and the key cross-sections.

[0042] Determine the aorta risk level based on the global diameter feature vector, the feature vectors of the respective key cross-sections, and the image feature vector.

[0043] In one embodiment, the processing module is specifically configured to,

[0044] Perform feature fusion on the global diameter feature vector, the feature vectors of the respective key cross-sections, and the image feature vector to obtain a fused feature vector.

[0045] Determine the aorta risk level according to the fused feature vector.

[0046] In one embodiment, the processing module is specifically configured to,

[0047] Determine a probability map of the multiple key points of the aorta and multiple segmentation masks of the aorta according to the original image of the aorta; the probability map is used to characterize the probability of each position where the key points are located in the aorta.

[0048] Determine each of the key points according to the probability maps of the respective key points.

[0049] In one embodiment, the processing module is specifically configured to determine the position where the maximum value is located in the probability map of the key points as the key point.

[0050] In one embodiment, the processing module is specifically configured to,

[0051] Determine the centerline of the aorta according to the multiple segmentation masks.

[0052] Determine the multiple key cross-sections according to the multiple key points and the centerline.

[0053] In one embodiment, the multiple key cross-sections include at least two of the following cross-sections: aortic sinus cross-section, sinoatrial node cross-section, middle ascending aorta cross-section, proximal descending aorta cross-section, aortic arch cross-section, proximal descending aorta, middle descending aorta cross-section, aortic cross-section at the diaphragm, and abdominal aorta cross-section.

[0054] In one embodiment, the multiple segmentation masks include the background in the original image of the aorta, the aortic trunk, the branches of the aorta, and the aortic sinus. The branches of the aorta include the brachiocephalic artery, carotid artery, subclavian artery, and abdominal aorta.

[0055] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the following steps are implemented:

[0056] Obtain the original image of the aorta.

[0057] Analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameters of the aorta. The key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0058] Determine multiple key cross-sections according to the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0059] Determine the aortic risk level according to the multiple key cross-sections and the continuous cross-sectional diameters.

[0060] In a fourth aspect, the present application also provides a computer-readable storage medium. A computer program is stored thereon. When the computer program is executed by a processor, the following steps are implemented:

[0061] Obtain the original image of the aorta.

[0062] Analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameters of the aorta. The key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0063] Determine multiple key cross-sections according to the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0064] Determine the aortic risk level according to the multiple key cross-sections and the continuous cross-sectional diameters.

[0065] Fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program which, when executed by a processor, performs the following steps:

[0066] Obtain the original image of the aorta.

[0067] Analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameter of the aorta. The key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0068] Determine multiple key cross-sections according to the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0069] Determine the aortic risk level according to the multiple key cross-sections and the continuous cross-sectional diameter.

[0070] The above-mentioned aortic screening method, device, computer device and storage medium analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameter of the aorta, and determine multiple key cross-sections based on the multiple key points and multiple segmentation maps, which are more accurate than the key cross-sections obtained by combining manual measurements in the prior art. In addition, the aortic risk level determined by combining the characteristics of the key cross-sections and the continuous cross-sectional diameter can improve the accuracy of aortic screening, thereby providing a more reliable basis for the early screening of the aorta. Description of the Drawings

[0071] Figure 1 It is the architecture diagram of the aortic screening system in an embodiment;

[0072] Figure 2 It is the flowchart of the aortic screening method in an embodiment;

[0073] Figure 3 It is the effect diagram of the aorta in an embodiment;

[0074] Figure 4 It is the structural diagram of the multi-task model in an embodiment;

[0075] Figure 5 It is the structural diagram of the key cross-section of the aorta in an embodiment;

[0076] Figure 6 It is the flowchart of the aortic screening method in an embodiment;

[0077] Figure 7 It is the structural diagram of the key cross-section in an embodiment;

[0078] Figure 8 It is a schematic structural diagram of adjacent key cross-sections in an embodiment;

[0079] Figure 9 It is a schematic flowchart of an aortic screening method in an embodiment;

[0080] Figure 10 It is a schematic structural diagram of a multi-modal feature fusion network in an embodiment;

[0081] Figure 11 It is a schematic structural diagram of an aortic screening device in an embodiment;

[0082] Figure 12 It is an internal structural diagram of a computer device in an embodiment. Detailed implementation manners

[0083] Next, the technical solutions in some embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments provided by the present application, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of the present application.

[0084] Unless otherwise required by the context, throughout the specification and claims, the term "comprise" and its other forms, such as the third-person singular form "comprises" and the present participle form "comprising", are interpreted as open and inclusive meanings, that is, "including, but not limited to". In the description of the specification, the terms "one embodiment", "some embodiments", "exemplary embodiments", "example", "specific example" or "some examples", etc. are intended to indicate that the specific features, structures, materials or characteristics related to the embodiment or example are included in at least one embodiment or example of the present application. The schematic representations of the above terms do not necessarily refer to the same embodiment or example. In addition, the described specific features, structures, materials or characteristics can be included in any one or more embodiments or examples in any appropriate manner.

[0085] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0086] When describing some embodiments, the expressions "coupled" and "connected" and their derivatives may be used. For example, when describing some embodiments, the term "connected" may be used to indicate that two or more components have direct physical or electrical contact with each other. Also, for example, when describing some embodiments, the term "coupled" may be used to indicate that two or more components have direct physical or electrical contact. However, the term "coupled" or "communicatively coupled" may also mean that two or more components do not have direct contact with each other, but still cooperate or interact with each other. The embodiments disclosed herein are not necessarily limited to the content herein.

[0087] "At least one of A, B, and C" has the same meaning as "at least one of A, B, or C", and both include the following combinations of A, B, and C: only A, only B, only C, the combination of A and B, the combination of A and C, the combination of B and C, and the combination of A, B, and C.

[0088] As used herein, depending on the context, the term "if" is optionally construed to mean "when" or "at the time of" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "if it is determined that..." or "if [the stated condition or event] is detected" is optionally construed to mean "when it is determined that..." or "in response to determining..." or "at the time of detecting [the stated condition or event]" or "in response to detecting [the stated condition or event]".

[0089] The use of "suitable for" or "configured to" herein means open and inclusive language, which does not exclude devices suitable for or configured to perform additional tasks or steps.

[0090] In addition, the use of "based on" or "in accordance with" means open and inclusive, because a process, step, calculation, or other action "based on" or "in accordance with" one or more of the stated conditions or values may, in practice, be based on additional conditions or values beyond those stated.

[0091] Since most cardiovascular diseases are closely related to the morphological changes of the aorta, early screening of the aorta is particularly important. Plain CT is very suitable as a popular early screening tool because of its low cost and low radiation. For the plain CT images of the aorta, the current hospital film reading system usually relies on some traditional image processing methods.

[0092] During the automated reconstruction process of the aorta in plain CT, abnormal situations such as noise and fracture cannot be effectively solved; in the observation of the reconstructed images, key cross-sections need to be manually measured, and diameter information of multiple groups of cross-sections is rarely provided, and volume and length information cannot be measured. With the development of computer technology, most of the existing related algorithms segment the aorta in a single-model manner, and then use the segmentation results and post-processing techniques to measure key cross-sections. The accuracy of the key cross-sections obtained by such measurement is not high.

[0093] To address the above technical problems, an embodiment of the present application provides an aorta screening method. By utilizing the complementarity and constraint relationship between two features, namely the segmentation mask and key points, after segmenting the aorta, multiple more accurate key cross-sections can be obtained. And key cross-section parameters including the cross-section diameter of each key cross-section among the multiple key cross-sections and at least one of the volume and length between two adjacent key cross-sections are determined through the multiple key cross-sections. Thus, the aorta risk level determined based on at least two of the original image of the aorta, multiple key cross-sections, continuous cross-section diameters, and key cross-section parameters is more referential in aorta screening. Moreover, in the process of using the aorta screening method provided by the embodiment of the present application to determine the aorta risk level, only the original image of the aorta needs to be obtained. Therefore, it has the beneficial effects of simple operation and easy implementation.

[0094] For the convenience of using this embodiment, refer to Figure 1 the architecture of the aorta screening system 10 shown in the figure. In the aorta screening system 10, it includes a terminal device 11 and an imaging device 12. The imaging device 12 can photograph the aorta and obtain the original image of the aorta. The medical device 12 can be any one of a digital imaging device, an X-ray computed tomography device, a magnetic resonance imaging device, an ultrasonic imaging device, a nuclear medicine imaging device.

[0095] In an exemplary solution, the terminal device 1 can have a multi-general or dedicated computing device environment or configuration. For example: a personal computer, a server computer, a handheld device or a portable device, a tablet device, a multi-processor device, a distributed computing environment including any of the above devices or equipment, and so on.

[0096] Combined with the above Figure 1 , the aorta screening method provided by the embodiment of the present application will be introduced in detail. Refer to Figure 2, the method includes:

[0097] S21. Obtain the original image of the aorta.

[0098] Exemplarily, the original image of the aorta can be a CT angiography image. Specifically, the original image of the aorta is an angiography image that includes the aorta. The angiography image may also include other body structures in addition to the aorta. Here, the other image structures in the angiography image except for the image of the aorta can be referred to as the background.

[0099] S22. Analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameter of the aorta. The key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0100] In one embodiment, the multiple segmentation masks include at least two of the background, aortic trunk, branches of the aorta, and aortic sinuses in the original image of the aorta. The branches of the aorta include at least one of the brachiocephalic artery, carotid artery, subclavian artery, and abdominal aorta.

[0101] Optionally, analyzing and processing the original image of the aorta to determine multiple key points includes: determining the probability map of multiple key points of the aorta and multiple segmentation masks of the aorta according to the original image of the aorta; wherein, the probability map is used to represent the probability of the key points at each position in the aorta. According to the probability map of each key point, each key point is determined.

[0102] In one implementation, when analyzing and processing the original image of the aorta, the aorta in the original image can be first identified, and then, based on the identified aorta, multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameter of the aorta are determined. A possible identification method is: after obtaining the original image, considering that the size of each pixel in the original image is different in the directions of each coordinate axis of the coordinate system, in order to ensure the accuracy of aorta identification, the original image is adjusted so that the size of each pixel in the original image is the same in the directions corresponding to each coordinate axis in the coordinate system, and the original image is input into the aorta identification model, and the aorta part in the original image can be identified to obtain the identification result of the aorta, such as Figure 3 shown, Figure 3 the irregular cylindrical figure in is the identified aorta.

[0103] In another possible implementation, an image segmentation and key point localization model is constructed to obtain multiple segmentation masks of the aorta and the probability heat map of multiple key points of the aorta. Refer to Figure 4, this module is a multi-task model for image segmentation and key point positioning, and its basic structure is based on V-NET. Among them, image segmentation can obtain multiple segmentation masks of the aorta, and key point positioning can obtain probability heat maps of multiple key points of the aorta. It should be noted that one implementation method of the probability map in the embodiment of the present application is a probability heat map.

[0104] Specifically, refer to Figure 4 In the V-NET structure, the two tasks of image segmentation and key point localization share the downsampling path, and are divided into two parallel network branches in the upsampling path. The specific training process of the image segmentation and key point localization model is as follows: obtain training samples, which contain multiple original images of aortas and the segmentation labels corresponding to the original images of each aorta and the probability heat map of the coordinate points; first perform the downsampling task on the original image of the aorta to obtain the sampling block, use the sampling block to perform the first upsampling task, and learn the background, aortic trunk, aortic branches (including brachiocephalic artery, carotid artery, subclavian artery and abdominal aorta) in the original image of the aorta based on the four types of segmentation masks of the aortic sinus. Use the sampling block to perform the second upsampling task and learn the probability heat map of the four coordinate points. The probability heat map of the four coordinate points represents the heat map of the contact coordinate points of the four aortic branches and the aortic trunk. The testing process is as follows: the original data is preprocessed and input into the multi-task network model, and the complete aortic segmentation map and the probability heat map of the four coordinate points can be obtained at the same time. It should be noted that the types of segmentation masks mentioned above are only exemplary descriptions of the four types of background, aortic trunk, aortic branches (including brachiocephalic artery, carotid artery, subclavian artery and abdominal aorta) and aortic sinus. The embodiments of the present application do not limit the types of segmentation masks included. For example, only the three types of segmentation masks of background, aortic trunk and aortic branches may be included. In addition, the probability heat maps of the four coordinate points mentioned above are also only exemplary descriptions. The embodiments of the present application do not limit the probability heat maps of the coordinate points. For example, three of the probability heat maps of the four coordinate points may be included.

[0105] S23. Determine multiple key sections based on the multiple key points and the multiple segmentation masks; the key sections are used to characterize the physiological characteristics of the aorta.

[0106] In one embodiment, the multiple key sections include at least two of the following sections: aortic sinus section, sinoatrial node section, middle ascending aorta section, proximal descending aorta section, aortic arch section, proximal descending aorta, middle descending aorta section, aortic section at the diaphragm, and abdominal aorta section.

[0107] In a possible implementation manner, based on the multiple key points and multiple segmentation masks obtained above, here, taking four key points and seven segmentation masks including background, aortic trunk, brachiocephalic artery, carotid artery, subclavian artery, abdominal aorta, and aortic sinus as an example for illustration. First, the central line is extracted using the seven segmentation masks of the aorta, and at the same time, the coordinates of each key point are determined using the maximum value of all probability values in the probability heat map of each key point. Then, using the anatomical information provided by medical guidelines, the central line, and the coordinates of the four key points, nine key cross-sections required are located. After determining the nine key cross-sections, the diameters of the nine key cross-sections are further obtained, and the volume and length between two adjacent key cross-sections are calculated. Refer to Figure 5 The nine key cross-sections of the aorta shown in are: aortic sinus cross-section 1, sinoatrial node cross-section 2, middle ascending aorta cross-section 3, proximal descending aorta cross-section 4, aortic arch cross-section 5, proximal descending aorta 6, middle descending aorta cross-section 7, aortic cross-section 8 at the diaphragm, and abdominal aorta cross-section 9. Among them, when determining the aortic sinus cross-section 1, the image of the aortic sinus is determined based on the pixel values in the image, and then the aortic sinus cross-section 1 is determined. Generally, in the original image of the aorta, the pixel values of the aortic sinus and the aortic trunk are different. For example, in the original image of the aorta, the pixel value of the segmentation mask of the aortic sinus is 2, and the pixel value of the segmentation mask of the aortic trunk is 1; therefore, the segmentation mask of the aortic sinus in the original image of the aorta can be determined according to the pixel value of the segmentation mask of the aortic sinus, and then the aortic sinus cross-section 1 is obtained. The sinoatrial node cross-section 2 is the cross-section at the connection of the aortic sinus and the aortic trunk. The middle ascending aorta cross-section 3 is the cross-section in the aortic trunk at the midpoint of the distance between the sinoatrial node cross-section 2 and the proximal descending aorta cross-section 4. The proximal descending aorta cross-section 4 is the aortic artery cross-section at the origin of the innominate artery. The aortic arch cross-section 5 is the cross-section in the aortic trunk between the left common carotid artery and the left subclavian artery. The proximal descending aorta 6 is the cross-section in the aortic trunk about 2 cm below the left subclavian artery. The middle descending aorta cross-section 7 is the cross-section in the aortic trunk at the midpoint of the distance between the proximal descending aorta 6 and the aortic cross-section 8 at the diaphragm. The aortic cross-section 8 at the diaphragm is the cross-section in the aortic trunk 2 cm above the origin of the celiac trunk.

[0108] S24. Determine the aortic risk level according to multiple key cross-sections and the diameters of consecutive cross-sections.

[0109] In one implementation manner, a first sub-network for extracting feature parameters of multiple key cross-sections is constructed based on a convolutional neural network, and a second sub-network for extracting feature parameters of consecutive cross-section diameters is constructed based on a convolutional neural network. The feature parameters obtained by the first sub-network and the second sub-network are fused, and the feature recognition of a multi-layer neural network is performed on the fused feature parameters, so as to obtain the aortic risk level.

[0110] The above aortic screening method determines multiple segmentation masks, multiple key points, and the continuous cross-sectional diameters of the aorta by analyzing and processing the original images of the aorta, and determines multiple key cross-sections based on the multiple key points and multiple segmentation maps, which is more accurate than the key cross-sections obtained by combining manual measurements in the prior art. In addition, the aortic risk level determined by combining the features of the key cross-sections and the continuous cross-sectional diameters can improve the accuracy of aortic screening, thereby providing a more reliable basis for the early screening of the aorta.

[0111] In one embodiment, referring to Figure 6 , S24 specifically includes:

[0112] S241. Determine the key cross-section parameters, where the key cross-section parameters include the cross-sectional diameters of each key cross-section among the multiple key cross-sections and at least one of the volume and length between two adjacent key cross-sections.

[0113] Specifically, the cross-sectional diameters of each key cross-section and the volume and length between two adjacent key cross-sections in this step are described as follows: According to the multiple segmentation masks, the centerline of the aorta is determined, and then the aorta is straightened using the straightened curved planar reformation (CPR) algorithm based on the centerline to obtain a quasi-cylindrical body. Based on this quasi-cylindrical body, the cross-sectional diameters of each key cross-section and the volume and length between two adjacent key cross-sections are determined.

[0114] It can be understood that the aorta is curved, and straightening the aorta according to the centerline means straightening the aorta along the centerline. In one embodiment, the specific straightening process of straightening the aorta using the straightened curved planar reformation (CPR) algorithm based on the centerline may include: starting from the aortic inlet, sequentially selecting the center points on the centerline in the order from proximal to distal, and straightening a section of the aorta centered on the center point to obtain a straightened image of a section of the aorta corresponding to the center point.

[0115] Specifically, 1. Calculation method of the cross-sectional diameter of each key cross-section

[0116] After determining the key cross-sections, assuming one of the key cross-sections is Figure 7The first cross-section shown in [the figure], the intersection of the first cross-section and the center line is point A, the pixel value of the first cross-section is 1. Count the number of pixels with a value of 1 in the first cross-section, and then combine the actual physical scale of each pixel in the medical image to determine the area of the first cross-section. According to the circular area formula, the corresponding diameter r can be obtained.

[0117] 2. Determination method of the volume between two adjacent key cross-sections

[0118] Refer to Figure 8 , count the number of pixels with a value of 1 in the aorta between the first cross-section and the second cross-section, and then combine the actual physical scale of each pixel in the medical image to calculate the length and volume of the aorta between the first cross-section and the second cross-section.

[0119] S242. Determine the aortic risk level according to multiple key cross-sections, continuous cross-section diameters, and key cross-section parameters.

[0120] In one implementation, a first sub-network for extracting the feature parameters of multiple key cross-sections is constructed based on a convolutional neural network, a second sub-network for extracting the feature parameters of continuous cross-section diameters is constructed based on a convolutional neural network, and a third sub-network for extracting the feature parameters of key cross-section parameters is constructed based on a convolutional neural network. The feature parameters obtained by the first sub-network, the second sub-network, and the third sub-network are fused, and the feature recognition of a multi-layer neural network is performed on the fused feature parameters, so as to obtain the aortic risk level.

[0121] In this embodiment, determining the key cross-section parameters including at least one of the cross-section diameters of each key cross-section in multiple key cross-sections and the volume and length between two adjacent key cross-sections, combined with multiple key cross-sections and continuous cross-section diameters, can improve the accuracy when determining the aortic risk level, making the aortic risk level more in line with the actual situation.

[0122] In one of the embodiments, refer to Figure 9 , S242 specifically includes:

[0123] S2421. Determine the global diameter feature vector according to the continuous cross-section diameters; the global diameter feature vector is used to characterize the global diameter size of the aorta.

[0124] Specifically, determine the global diameter feature vector according to the continuous cross-section diameters and the global diameter feature network. Among them, the global diameter network is trained and generated based on a multi-layer convolutional neural network. The length of the global feature vector is N×1.

[0125] It can be understood that the continuous cross-sections are specifically composed of multiple cross-sections of the aorta intercepted according to a preset rule, and the diameter of each cross-section is determined, and then the continuous cross-section diameter of the aorta is obtained. Among them, the determination method of the diameter of each cross-section can refer to the determination process of the cross-section diameter of the above-mentioned key cross-sections, which will not be elaborated here.

[0126] S2422. Determine the eigenvectors of each key cross-section according to the key cross-section parameters; the eigenvectors of each key cross-section are used to characterize the parameter characteristics of each key cross-section.

[0127] Specifically, determine the eigenvectors of each key cross-section according to the key cross-section parameters and the key cross-section feature network. Among them, the key cross-section network is trained and generated based on a multi-layer convolutional neural network. The length of the eigenvector of each key cross-section is N×1.

[0128] S2423. Determine the image eigenvector according to the original image of the aorta and the mask images of multiple key cross-sections; the image eigenvector is used to characterize the relationship characteristics between the original image of the aorta and the key cross-sections.

[0129] Specifically, determine the image eigenvector according to the original image of the aorta, the mask images of multiple key cross-sections, and the image feature network. Among them, the image feature network is trained and generated based on a multi-layer convolutional neural network. The length of each image eigenvector is N×1.

[0130] S2424. Determine the aortic risk level based on the global diameter eigenvector, the eigenvectors of each key cross-section, and the image eigenvector.

[0131] In one of the embodiments, determining the aortic risk level based on the global diameter eigenvector, the eigenvectors of each key cross-section, and the image eigenvector includes: performing feature fusion on the global diameter eigenvector, the eigenvectors of each key cross-section, and the image eigenvector to obtain a fused eigenvector. Determine the aortic risk level according to the fused eigenvector. In this embodiment, by using the multi-feature fusion method combining the global diameter eigenvector, the eigenvectors of each key cross-section, and the image eigenvector to predict the aortic risk level, the complementary advantages of multiple modalities of information can be fully utilized, making the risk classification result more accurate.

[0132] It can be understood that in combination with the above embodiments, the eigenvectors output by the global diameter feature network, the key cross-section feature network, and the image feature network are concatenated to obtain a fully connected layer of N×3, and then through multi-layer feature extraction, the four-classification result of the risk level is output.

[0133] It should be noted that with reference to Figure 10, the four steps S2421 - S2424 actually involve a multi - modal feature fusion network. This multi - modal feature fusion network includes three parallel sub - networks for extracting relevant features, namely the global diameter feature network, the key cross - section feature network, and the image feature network, as well as a splicing network (also called the fusion network) for splicing the relevant features of the three sub - networks at the end. This fusion network is also obtained based on training with a convolutional neural network. The output layers of these three sub - networks, namely the global diameter feature network, the key cross - section feature network, and the image feature network, can be understood as the fully - connected layers of the multi - modal feature fusion network. The length of the feature vector of the fully - connected layer output by each sub - network is N×1. The final output result of this multi - modal feature fusion network is the aortic risk level. Risk level classification: Class A (75% - 100%), Class B (50% - 75%), Class C (25% - 50%), Class D (0 - 25%).

[0134] In this embodiment, by using a multi - modal method that combines the global diameter feature vector, the feature vectors of each key cross - section, and the image feature vector to predict the aortic risk level, it is possible to make full use of the features of multiple modalities, broaden the feature coverage range for risk classification prediction, improve the accuracy of the prediction result, and enhance the robustness of the risk classification result.

[0135] In one of the embodiments, according to the probability map of the key points, determining the coordinates of the key points includes: determining the position where the maximum value in the probability map of the key points is located as the key point.

[0136] Specifically, combined with Figure 4 , in the second up - sampling task, a neural network is used to process the sampling blocks obtained in the down - sampling task to obtain the probability map of the key points (i.e., the probability heat map); specifically, based on the neural network processing the sampling blocks, a probability feature map corresponding to the sampling blocks will be obtained. Each pixel point in this probability feature map corresponds to a probability value, and each probability value represents the probability that this pixel point is a key point. Connect the probabilities that are the same or similar in the probability feature map (similar to contour lines), and in the probability feature map, it will form a probability heat map around the key points. The probability heat map means that the positions with different probabilities in the figure are distinguished by different colors to highlight the areas or points with high probability values.

[0137] Obtain the information of the key points based on the probability heat map, and determine the key points based on the information of the key points.

[0138] Specifically, based on the characteristics of the probability heat map, find the point with the largest probability value in this probability heat map, that is, obtain the information of the key points in the segmentation mask. The information of the key points can include: information such as the coordinates of the key points. Based on the information of the key points, the key points can be determined.

[0139] In this embodiment, by determining the position of the maximum value in the probability map of the key points and determining them as key points, it can provide a data basis for subsequent determination of the key cross-sections, eliminating the need for manual measurement to obtain the key cross-sections.

[0140] In one embodiment, multiple key cross-sections are determined based on multiple key points and multiple segmentation masks, including: determining the centerline of the aorta according to the multiple segmentation masks. Determining multiple key cross-sections according to the multiple key points and the centerline.

[0141] Specifically, a skeleton extraction algorithm can be used to extract the segmented centerlines of the segmented blood vessels between adjacent segmentation masks. The segmented centerline can include one starting point or multiple starting points. When the segmented centerline includes multiple starting points, it indicates that the segmented blood vessels have multiple unconnected blood vessels. The segmented centerline can include one junction point and one segmented line, or multiple junction points and multiple segmented lines. The starting point is the center point at the lowest end of the segmented blood vessel. The junction point is the center point at the blood vessel junction. The segmented line is the connection line of the center points of the segmented blood vessels. Multiple connection lines form the centerline of the aorta.

[0142] In this embodiment, the centerline of the aorta is determined according to the multiple segmentation masks, and multiple key cross-sections are determined according to the multiple key points and the centerline. By utilizing the complementarity and constraint relationship between the two features of the segmentation mask and the key points after the segmentation of the aorta, the required key cross-sections can be automatically obtained without manual measurement. The embodiment of the present application has better automation performance and higher efficiency in the technology of obtaining key cross-sections.

[0143] It should be understood that although Figure 2 、 6 、the steps in the flowcharts of 9 are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, Figure 2 、 6 、at least a part of the steps in 9 can include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same moment, but can be executed at different moments. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of the steps or stages in other steps or other steps.

[0144] In one embodiment, referring to Figure 11 , an aortic screening device 110 is provided. The device 110 includes:

[0145] An acquisition module 111, configured to acquire the original image of the aorta.

[0146] A processing module 112 is configured to analyze and process the original image of the aorta acquired by the acquisition module 111 to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameter of the aorta. The key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0147] The processing module 112 is further configured to determine multiple key cross-sections according to the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0148] The processing module 112 is further configured to determine the aortic risk level according to the multiple key cross-sections and the continuous cross-sectional diameter.

[0149] In one embodiment, the processing module 112 is specifically configured to,

[0150] Determine key cross-section parameters, where the key cross-section parameters include at least one of the cross-sectional diameter of each key cross-section among the multiple key cross-sections and the volume and length between two adjacent key cross-sections.

[0151] Determine the aortic risk level according to the multiple key cross-sections, the continuous cross-sectional diameter, and the key cross-section parameters.

[0152] In one embodiment, the processing module 112 is specifically configured to,

[0153] Determine a global diameter feature vector according to the continuous cross-sectional diameter; the global diameter feature vector is used to characterize the global diameter size of the aorta.

[0154] Determine the feature vector of each key cross-section according to the key cross-section parameters; the feature vector of each key cross-section is used to characterize the parameter features of each key cross-section.

[0155] Determine an image feature vector according to the original image of the aorta and the mask images of the multiple key cross-sections; the image feature vector is used to characterize the relationship features between the original image of the aorta and the key cross-sections.

[0156] Determine the aortic risk level based on the global diameter feature vector, the feature vectors of each key cross-section, and the image feature vector.

[0157] In one embodiment, the processing module 112 is specifically configured to,

[0158] Perform feature fusion on the global diameter feature vector, the feature vectors of each key cross-section, and the image feature vector to obtain a fused feature vector.

[0159] Determine the aortic risk level according to the fused feature vector.

[0160] In one embodiment, the processing module 112 is specifically configured to,

[0161] Based on the original image of the aorta, a probability map of multiple key points of the aorta and multiple segmentation masks of the aorta are determined; the probability map is used to characterize the probability of the key points at each position in the aorta.

[0162] Based on the probability map of each key point, each key point is determined.

[0163] In one embodiment, the processing module 112 is specifically configured to determine the position where the maximum value in the probability map of the key point as the key point.

[0164] In one embodiment, the processing module 112 is specifically configured to

[0165] Based on the multiple segmentation masks, the center line of the aorta is determined.

[0166] Based on the multiple key points and the center line, multiple key cross-sections are determined.

[0167] In one embodiment, the multiple key cross-sections include at least two of the following cross-sections: aortic sinus cross-section, sinoatrial node cross-section, middle ascending aorta cross-section, proximal descending aorta cross-section, aortic arch cross-section, proximal descending aorta, middle descending aorta cross-section, aortic cross-section at the diaphragm, and abdominal aorta cross-section.

[0168] In one embodiment, the multiple segmentation masks include the background, aortic trunk, branches of the aorta, and aortic sinus in the original image of the aorta. The branches of the aorta include brachiocephalic artery, carotid artery, subclavian artery, and abdominal aorta.

[0169] For the specific limitations of the aortic screening device 110, reference can be made to the limitations of the aortic screening method in the above text, which will not be elaborated here. Each module in the above aortic screening device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0170] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 12As shown. The computer device includes a processor, a memory, and a network interface connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store initial data, and the network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements an aortic screening method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the casing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0171] Those skilled in the art can understand that Figure 12 the structure shown is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0172] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the following steps are implemented:

[0173] Obtain the original image of the aorta.

[0174] Analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameter of the aorta. The key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0175] Determine multiple key cross-sections according to the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0176] Determine the aortic risk level according to the multiple key cross-sections and the continuous cross-sectional diameter.

[0177] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the following steps are implemented:

[0178] Obtain the original image of the aorta.

[0179] Analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameter of the aorta, where the key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0180] Determine multiple key cross-sections based on the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0181] Determine the aortic risk level based on the multiple key cross-sections and the continuous cross-sectional diameter.

[0182] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the following steps:

[0183] Obtain the original image of the aorta.

[0184] Analyze and process the original image of the aorta to determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameter of the aorta, where the key points are the contact points between the branches of the aorta and the aortic trunk of the aorta.

[0185] Determine multiple key cross-sections based on the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta.

[0186] Determine the aortic risk level based on the multiple key cross-sections and the continuous cross-sectional diameter.

[0187] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include Read-Only Memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0188] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0189] The above-described embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. An aortic screening method, characterized in that, The method includes: Obtaining an original image of the aorta; Analyzing and processing the original image of the aorta to determine a plurality of segmentation masks of the aorta, a plurality of key points of the aorta, and the continuous cross-sectional diameters of the aorta, where the key points are the contact points between the branches of the aorta and the aortic trunk of the aorta; Determining a plurality of key cross-sections according to the plurality of key points and the plurality of segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta; Determining key cross-section parameters, where the key cross-section parameters include at least one of the cross-sectional diameter of each key cross-section in the plurality of key cross-sections and the volume and length between two adjacent key cross-sections; Determining the aortic risk level according to the plurality of key cross-sections, the continuous cross-sectional diameters, and the key cross-section parameters.

2. The aortic screening method according to claim 1, wherein The determining the aortic risk level according to the plurality of key cross-sections, the continuous cross-sectional diameters, and the key cross-section parameters includes: Determining a global diameter feature vector according to the continuous cross-sectional diameters; the global diameter feature vector is used to characterize the global diameter size of the aorta; Determining the feature vectors of the respective key cross-sections according to the key cross-section parameters; the feature vectors of the respective key cross-sections are used to characterize the parameter features of the respective key cross-sections; Determining an image feature vector according to the original image of the aorta and the mask images of the plurality of key cross-sections; the image feature vector is used to characterize the relationship feature between the original image of the aorta and the key cross-sections; Determining the aortic risk level based on the global diameter feature vector, the feature vectors of the respective key cross-sections, and the image feature vector.

3. The aortic screening method according to claim 2, wherein The determining the aortic risk level based on the global diameter feature vector, the feature vectors of the respective key cross-sections, and the image feature vector includes: Performing feature fusion on the global diameter feature vector, the feature vectors of the respective key cross-sections, and the image feature vector to obtain a fused feature vector; Determining the aortic risk level according to the fused feature vector.

4. The aortic screening method according to claim 1, wherein The analyzing and processing the original image of the aorta to determine the plurality of key points includes: Determining a probability map of the plurality of key points of the aorta and a plurality of segmentation masks of the aorta according to the original image of the aorta; the probability map is used to characterize the probability of the key points at each position in the aorta; Determining each of the key points according to the probability map of each of the key points.

5. The aortic screening method according to claim 4, wherein The determining the coordinates of the key points according to the probability map of the key points includes: Determining the position where the maximum value in the probability map of the key points is located as the key point.

6. The aortic screening method according to claim 1, wherein The determining a plurality of key cross-sections according to the plurality of key points and the plurality of segmentation masks includes: Determining the centerline of the aorta according to the plurality of segmentation masks; Determining the plurality of key cross-sections according to the plurality of key points and the centerline.

7. The aortic screening method according to claim 1, wherein The determining the key cross-section parameters includes: Determining the centerline of the aorta according to the plurality of segmentation masks; Using a straight-curved surface reconstruction algorithm, the aorta is straightened according to the centerline to obtain a quasi-cylindrical body; Based on the quasi-cylindrical body, at least one of the cross-sectional diameters of each key cross-section among the multiple key cross-sections, the volume and the length between two adjacent key cross-sections is determined.

8. An aortic screening device, characterized in that, The device includes: An acquisition module, configured to acquire an original image of the aorta; A processing module, configured to analyze and process the original image of the aorta acquired by the acquisition module, determine multiple segmentation masks of the aorta, multiple key points of the aorta, and the continuous cross-sectional diameters of the aorta, where the key points are the contact points between the branches of the aorta and the aortic trunk of the aorta; The processing module is further configured to determine multiple key cross-sections according to the multiple key points and the multiple segmentation masks; the key cross-sections are used to characterize the physiological characteristics of the aorta; The processing module is further configured to determine key cross-section parameters, where the key cross-section parameters include at least one of the cross-sectional diameters of each key cross-section among the multiple key cross-sections, the volume and the length between two adjacent key cross-sections; according to the multiple key cross-sections, the continuous cross-sectional diameters, and the key cross-section parameters, determine the aorta risk level.

9. An electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, characterized in that, When the computer program is executed by the processor, the processor is caused to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images