Method and device for training segmentation model for segmenting aneurysm and storage medium

Abstract

The invention discloses a method and equipment for training a segmentation model for segmenting an aneurysm, and a storage medium. The method comprises the following steps: acquiring multi-modal image data, wherein the multi-modal image data comprises digital subtraction angiography data, CT angiography data and magnetic resonance angiography data; carrying out blood vessel segmentation and aneurysm labeling based on the multi-modal image data to obtain a blood vessel segmentation result and an aneurysm labeling result; constructing label data according to the blood vessel segmentation result and the aneurysm labeling result; and inputting the blood vessel segmentation result and the label data into a segmentation model for performing aneurysm segmentation to obtain an aneurysm segmentation result, so as to train the segmentation model for segmenting the aneurysm. By using the scheme of the invention, the segmentation precision of the segmentation model can be improved so as to obtain a more accurate aneurysm segmentation result.

Description

Technical Field

[0001] The present application generally relates to the field of aneurysm segmentation technology. More specifically, the present application relates to a method for training a segmentation model for segmenting an aneurysm, a method, a device and a computer-readable storage medium for segmenting an aneurysm. Background Art

[0002] Aneurysm is a type of lesion in the blood vessels, which refers to the thickening of the artery diameter and the manifestation of tumor-like expansion. In physical examination screening, imaging examination of the blood vessels is often used to screen for aneurysms. The aforementioned images include digital subtraction angiography ("DSA"), CT vascular imaging ("CTA"), or magnetic resonance angiography ("MRA"). At present, for the segmentation of aneurysms, aneurysms are directly annotated in the image data, and the annotated images are input into the segmentation model for aneurysm segmentation. However, there are differences in the size of aneurysms in different imaging examinations. Direct annotation based on image data cannot balance the characteristics of aneurysms of different sizes, thereby affecting the segmentation accuracy of aneurysms.

[0003] In view of this, there is an urgent need to provide a solution for training a segmentation model for segmenting aneurysms, so as to improve the segmentation accuracy of the segmentation model and obtain more accurate aneurysm segmentation results. Summary of the invention

[0004] In order to at least solve one or more of the technical problems mentioned above, the present application proposes a solution for training a segmentation model for segmenting aneurysms in multiple aspects.

[0005] In a first aspect, the present application provides a method for training a segmentation model for segmenting aneurysms, comprising: acquiring multimodal image data, wherein the multimodal image data comprises digital subtraction angiography data, CT vascular imaging data and magnetic resonance vascular imaging data; performing vascular segmentation and annotating aneurysms based on the multimodal image data to obtain vascular segmentation results and aneurysm annotation results; constructing label data according to the vascular segmentation results and the aneurysm annotation results; and inputting the vascular segmentation results and the label data into a segmentation model to perform aneurysm segmentation to obtain an aneurysm segmentation result, so as to train the segmentation model for segmenting aneurysms.

[0006] In some embodiments, foreground pixels on the blood vessel segmentation result that are not on the aneurysm annotation result are set as first label data; the shortest distance between the foreground pixels on the blood vessel segmentation result and the background pixels is calculated; and second label data or third label data is set according to the shortest distance and the aneurysm annotation result to construct the label data.

[0007] In other embodiments, setting the second label data or the third label data according to the shortest distance and the aneurysm annotation result includes: comparing the shortest distance with a distance threshold and determining whether the corresponding foreground pixel is located within the aneurysm annotation result; and setting the second label data or the third label data according to the comparison result of the shortest distance with the distance threshold and the determination result of whether the corresponding foreground pixel is located within the aneurysm annotation result.

[0008] In some other embodiments, setting the second label data or the third label data according to the comparison result of the shortest distance and the distance threshold and the judgment result of whether the corresponding foreground pixel point is located in the aneurysm annotation result includes: in response to the shortest distance being less than or equal to the distance threshold and the corresponding foreground pixel point being located in the aneurysm annotation result, setting the corresponding foreground pixel point to the second label data; or in response to the shortest distance being greater than the distance threshold and the corresponding foreground pixel point being located in the aneurysm annotation result, setting the corresponding foreground pixel point to the third label data.

[0009] In some other embodiments, the blood vessel segmentation result and the label data are input into a segmentation model to perform aneurysm segmentation to obtain an aneurysm segmentation result for training the segmentation model for segmenting aneurysms, which includes: inputting the blood vessel segmentation result, the first label data, the second label data and the third label data into the segmentation model to perform aneurysm segmentation, and outputting different categories to obtain an aneurysm segmentation result for training the segmentation model for segmenting aneurysms.

[0010] In some other embodiments, it also includes: calculating the proportion of the third label data in the blood vessel segmentation result; and adjusting the training weights of the first label data, the second label data and the third label data in training the segmentation model according to the proportion, so as to optimize the training of the segmentation model used to segment aneurysms.

[0011] In some other embodiments, the segmentation model includes a Unet model, a FPN model or a PSPNet model.

[0012] In a second aspect, the present application provides a method for segmenting an aneurysm, comprising: acquiring image data to be segmented and extracting a binary image of blood vessel segmentation based on the image data to be segmented; and inputting the binary image of blood vessel segmentation into a segmentation model trained by one or more embodiments of the first aspect to perform aneurysm segmentation to obtain an aneurysm segmentation result.

[0013] In a third aspect, the present application provides a device, comprising: a processor; and a memory, on which computer instructions for training a segmentation model for segmenting an aneurysm are stored, and when the computer instructions are executed by the processor, multiple embodiments of the first aspect are implemented; or computer instructions for segmenting an aneurysm are stored, and when the computer instructions are executed by the processor, one embodiment of the second aspect is implemented.

[0014] In a fourth aspect, the present application provides a computer-readable storage medium storing computer program instructions for training a segmentation model for segmenting an aneurysm, which, when executed by one or more processors, implements multiple embodiments of the first aspect; or storing computer instructions for segmenting an aneurysm, which, when executed by a processor, implements one embodiment of the second aspect.

[0015] Through the scheme for training the segmentation model for segmenting aneurysms provided above, the embodiment of the present application uses multimodal image data, and performs blood vessel segmentation and aneurysm annotation on the multimodal image data to obtain blood vessel segmentation results and aneurysm annotation results, and at the same time, combines the blood vessel segmentation results and aneurysm annotation results of different image data to construct label data. Among them, the blood vessel segmentation results are consistent in different modal data, so the embodiment of the present application can more accurately annotate aneurysms through the morphological features of the blood vessels and the aneurysm annotation, avoiding the problem of directly annotating in the image data and being unable to align the image features of different aneurysms, thereby improving the segmentation accuracy of the segmentation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0017] Figure 1 is an exemplary flowchart showing a method for training a segmentation model for segmenting an aneurysm according to an embodiment of the present application;

[0018] Figure 2 is an exemplary flowchart showing the overall process of training a segmentation model for segmenting an aneurysm according to an embodiment of the present application;

[0019] Figure 3 is another exemplary flowchart showing the overall training of a segmentation model for segmenting an aneurysm according to an embodiment of the present application;

[0020] Figure 4is an exemplary flowchart showing a method for segmenting an aneurysm according to an embodiment of the present application;

[0021] Figure 5 is a block diagram showing an exemplary structure of a device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0023] It should be understood that the terms "include" and "comprising" used in the specification and claims of the present application indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0024] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this application specification and claims, unless the context clearly indicates otherwise, the singular forms of "a", "an" and "the" are intended to include plural forms. It should also be further understood that the term "and / or" used in this application specification and claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0025] As used in this specification and claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0026] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0027] Figure 1 FIG. 1 is an exemplary flowchart showing a method 100 for training a segmentation model for segmenting an aneurysm according to an embodiment of the present application. Figure 1As shown in , the method 100 includes: step S101: acquiring multimodal image data, wherein the multimodal image data includes digital subtraction angiography data, CT vascular imaging data and magnetic resonance vascular imaging data; step S102: performing blood vessel segmentation and annotating aneurysms based on the multimodal image data to obtain blood vessel segmentation results and aneurysm annotation results; step S103: constructing label data according to the blood vessel segmentation results and aneurysm annotation results; and step S104: inputting the blood vessel segmentation results and the label data into a segmentation model to perform aneurysm segmentation to obtain an aneurysm segmentation result, so as to train the segmentation model for segmenting aneurysms.

[0028] First, at step S101, multimodal image data is collected. The multimodal image data may include digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data. Among them, digital subtraction angiography data is the gold standard for screening and diagnosis of various vascular diseases, and is also the gold standard for diagnosis and measurement of aneurysms, while CT angiography data and magnetic resonance angiography data can be used for screening and follow-up of aneurysms. The embodiment of the present application can greatly improve the performance of the segmentation model by using the three modal data for training the segmentation model.

[0029] Next, at step S102, blood vessel segmentation is performed based on the multimodal imaging data and aneurysms are labeled to obtain blood vessel segmentation results and aneurysm labeling results. In some embodiments, image processing methods or deep learning methods may be used to perform blood vessel segmentation, wherein the aforementioned image processing methods may include but are not limited to image edge detection and image morphology methods. The aforementioned deep learning method may be based on an existing trained segmentation model to perform blood vessel segmentation to obtain a blood vessel segmentation result, and the present application does not impose any restrictions on this. In some implementation scenarios, the aforementioned aneurysm labeling may be performed manually or automatically to label the aneurysm area in the blood vessel segmentation result to obtain an aneurysm labeling result.

[0030] Based on the above-obtained vascular segmentation results and aneurysm labeling results, at step S103, label data is constructed according to the vascular segmentation results and aneurysm labeling results. The vascular segmentation results are consistent in different modal data, which is more conducive to aneurysm morphology analysis, thereby more accurately labeling aneurysms. Therefore, by combining the vascular segmentation results and aneurysm labeling results to construct label data, the accuracy of aneurysm labeling can be improved, so as to improve the accuracy of subsequent segmentation model training.

[0031] In some embodiments, the foreground pixel points on the blood vessel segmentation result that are not on the aneurysm annotation result can be set as the first label data, and then the shortest distance between the foreground pixel points and the background pixel points on the blood vessel segmentation result is calculated, and then the second label data or the third label data is set according to the shortest distance and the aneurysm annotation result to construct the label data. That is, the blood vessel pixel points are first set as the first label data, and then the second and third label data are set in combination with the shortest distance and the aneurysm annotation result. Among them, the first label data corresponds to non-aneurysm. In some implementation scenarios, the aforementioned shortest distance can be, for example, the Euclidean distance between the foreground pixel points and the background pixel points. Preferably, the aforementioned distance threshold can be, for example, 2 mm. It can be understood that by calculating the shortest distance, large aneurysms and small aneurysms can be distinguished more accurately.

[0032] In some embodiments, the shortest distance may be compared with a distance threshold and a determination may be made as to whether the corresponding foreground pixel is located within the aneurysm annotation result, and then the second label data or the third label data may be set according to the comparison result of the shortest distance with the distance threshold and the determination result as to whether the corresponding foreground pixel is located within the aneurysm annotation result. More specifically, in some embodiments, in response to the shortest distance being less than or equal to the distance threshold and the corresponding foreground pixel being located within the aneurysm annotation result, the corresponding foreground pixel is set as the second label data; or in response to the shortest distance being greater than the distance threshold and the corresponding foreground pixel being located within the aneurysm annotation result, the corresponding foreground pixel is set as the third label data.

[0033] That is, first determine whether the shortest distance between the foreground pixel and the background pixel on the blood vessel segmentation result is greater than the distance threshold and determine whether the corresponding foreground pixel is located in the aneurysm annotation result. If the shortest distance is less than or equal to the distance threshold and the corresponding foreground pixel is located in the aneurysm annotation result, set the foreground pixel as the second label data. If the shortest distance between the foreground pixel and the background pixel is greater than the distance threshold and the corresponding foreground pixel is located in the aneurysm annotation result, set the foreground pixel as the third label data. That is, the second and third label data both correspond to aneurysms.

[0034] Further, at step S104, the blood vessel segmentation result and the label data are input into the segmentation model for aneurysm segmentation to obtain the aneurysm segmentation result, so as to train the segmentation model for segmenting aneurysms. In some embodiments, the blood vessel segmentation result, the first label data, the second label data and the third label data can be input into the segmentation model for aneurysm segmentation, and different categories can be output to obtain the aneurysm segmentation result, so as to train the segmentation model for segmenting aneurysms. That is, the blood vessel segmentation result is used as input data, and the first, second and third label data are used as labels to train the segmentation model for segmenting aneurysms. In some implementation scenarios, the aforementioned segmentation model can be, for example, a Unet model, an FPN model or a PSPNet model.

[0035] In combination with the above description, it can be seen that the embodiment of the present application performs blood vessel segmentation and aneurysm labeling on digital subtraction angiography data, CT vascular imaging data, and magnetic resonance vascular imaging data to obtain blood vessel segmentation results and aneurysm labeling results, and at the same time, the label data constructed by combining the blood vessel segmentation results and aneurysm labeling results is used to train the segmentation model. Based on this, the problem of image feature alignment when directly segmenting aneurysms from images can be avoided, thereby improving the accuracy of aneurysm recognition. Furthermore, by calculating the shortest distance between foreground points and background points, the embodiment of the present application can more accurately distinguish large aneurysms from small aneurysms, thereby improving the recognition accuracy of aneurysms.

[0036] Figure 2 FIG. 1 is an exemplary flowchart showing the overall training of a segmentation model for segmenting an aneurysm according to an embodiment of the present application. It should be understood that Figure 2 is the above Figure 1 A specific implementation of method 100, so the above Figure 1 The description also applies to Figure 2 .

[0037] like Figure 2 As shown in , at step S201, digital subtraction angiography data, CT angiography data and magnetic resonance angiography data are collected. Then, at step S202, blood vessel segmentation is performed based on the aforementioned three multimodal imaging data to obtain a blood vessel segmentation result. In some embodiments, blood vessel segmentation can be performed by, for example, edge detection, morphology or an existing segmentation model to obtain a blood vessel segmentation result. At step S203, aneurysm annotation is performed in the blood vessel segmentation result to obtain an aneurysm annotation result. Among them, for a foreground pixel point in the blood vessel segmentation result that is not on the aneurysm annotation result, at step S204, the foreground pixel point is set as the first label data.

[0038] Further, at step S205, the shortest distance between the foreground pixel and the background pixel on the blood vessel segmentation result is calculated. As mentioned above, the shortest distance may be, for example, the Euclidean distance. At step S206 and step S207, it is determined whether the shortest distance is greater than a distance threshold. Preferably, the aforementioned distance threshold may be, for example, 2 mm. When the shortest distance is not greater than the distance threshold and when the foreground pixel is located within the aneurysm annotation result, at step S208, the corresponding foreground pixel is set to the second label data. When the shortest distance is greater than the distance threshold and the foreground pixel is located within the aneurysm annotation result, at step S209, the foreground pixel is set to the third label data to construct the label data. As an example, the aforementioned first, second, and third label data may be represented by 1, 2, and 3, respectively, where 2 and 3 both correspond to aneurysms.

[0039] Based on the blood vessel segmentation result and the first, second and third label data obtained above, at step S210, the blood vessel segmentation result, the first label data, the second label data and the third label data are input into the segmentation model for training to obtain a trained segmentation model.

[0040] In some embodiments, the embodiments of the present application also involve calculating the proportion of the first label data in the blood vessel segmentation result, and adjusting the training weights of the first label data, the second label data, and the third label data in the training segmentation model according to the proportion, so as to optimize the training of the segmentation model for segmenting aneurysms. Based on this, the training process can be further optimized to further improve the accuracy of the segmentation model. Specifically, in some implementation scenarios, the training weight of each label data can be set by the data ratio between the preset range values ​​of the proportion.

[0041] As an example, assume that the number of pixel points of the third label data is N3, the number of pixel points of the second label data is N2, and the number of pixel points of the first label data is N1. The proportion of the third label data can be obtained as N3 / (N1 + N2 + N3). Further, the total number of all data is denoted as num, and the aforementioned preset range value can be, for example, (0.1, 0.4, 1). In this scenario, when the number of data where N3 / (N1 + N2 + N3)>1 is num1, the data ratio ratio1 of this interval = num1 / num; when the number of data where 0.4 < N3 / (N1 + N2 + N3) <= 1 is num2, the data ratio ratio2 of this interval = num2 / num; when the number of data where 0.1 < N3 / (N1 + N2 + N3) <= 0.4 is num3, the data ratio ratio3 of this interval = num3 / num. In training the segmentation model, the training weights corresponding to the first label data, the second label data, and the third label data can be set to 1 / ratio1, 1 / ratio2, and 1 / ratio3 respectively. Based on this, the weight of the aneurysm with a larger proportion in the blood vessel can be reduced, while the weight of the aneurysm with a smaller proportion can be increased, so as to balance the volume difference between large aneurysms and small aneurysms, and further improve the accuracy of the segmentation model.

[0042] Figure 3 is another exemplary flowchart showing the overall process of training a segmentation model for aneurysm segmentation according to an embodiment of the present application. It should be understood that Figure 3 is the above Figure 1 Another specific implementation of method 100 in, so the above description about Figure 1 also applies to Figure 3 .

[0043] As Figure 3 shown in, at step S301, digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data are collected. Then, at step S302, based on the above three multimodal image data, blood vessel segmentation is performed to obtain a blood vessel segmentation result. In some embodiments, blood vessel segmentation can be performed, for example, through edge detection, morphology, or an existing segmentation model to obtain a blood vessel segmentation result. At step S303, aneurysm annotation is performed in the blood vessel segmentation result to obtain an aneurysm annotation result. According to the obtained blood vessel segmentation result and aneurysm annotation result, at step S304, the first label data, the second label data, and the third label data are constructed based on the blood vessel segmentation result and the aneurysm annotation result. Among them, more details about constructing the first label data, the second label data, and the third label data can be referred to the above Figure 1 , Figure 2 The description made will not be repeated herein.

[0044] After obtaining the above-mentioned blood vessel segmentation results and label data, at step S305, the blood vessel segmentation results, the first label data, the second label data and the third label data are input into the segmentation model for training, and at step S306, the proportion of the third label data in the blood vessel segmentation results is calculated. Then, at step S307, the proportion of data whose proportion is within a preset range is calculated, such as the ratio1, ratio2 and ratio3 obtained above. Further, at step S308, the training weight of each label data is set to train the segmentation model to obtain a trained segmentation model. As an example, the training weight settings corresponding to the first label data, the second label data and the third label data are set to 1 / ratio1, 1 / ratio2 and 1 / ratio3 respectively.

[0045] Figure 4 FIG. 4 is an exemplary flowchart showing a method 400 for segmenting an aneurysm according to an embodiment of the present application. Figure 4 As shown in , the method 400 includes: step S401: acquiring image data to be segmented and extracting a binary image of blood vessel segmentation based on the image data to be segmented; step S402: inputting the binary image of blood vessel segmentation into a trained segmentation model to perform aneurysm segmentation to obtain an aneurysm segmentation result.

[0046] In step S401, image data to be segmented is obtained and a binary image of blood vessel segmentation is extracted based on the image data to be segmented. As can be seen from the foregoing, a binary image of blood vessel segmentation can be extracted by methods such as image edge detection, image morphology or deep learning.

[0047] Next, at step S402, the binary image of the blood vessel segmentation is input into the trained segmentation model for aneurysm segmentation to obtain an aneurysm segmentation result. That is, the binary image of the blood vessel segmentation is directly input into the trained segmentation model, and the aneurysm segmentation results of different categories are obtained by output. As an example, when the third and second labels are output corresponding to categories 3 and 2, it is an aneurysm. When the first label is output corresponding to category 1, it is a non-aneurysm.

[0048] Based on the above, it can be seen that the embodiment of the present application extracts the vascular segmentation results and annotates the aneurysms on the multimodal data to simultaneously combine the vascular related features and the aneurysm annotation results to construct the label data, and then train the segmentation model. Based on this, it is possible to overcome the problem that the image features cannot be aligned when the aneurysm is directly segmented from the image, thereby improving the accuracy of aneurysm segmentation. Furthermore, the embodiment of the present application also trains the segmentation model by adjusting the training weights of each data to balance the volume difference between large aneurysms and small aneurysms, thereby further optimizing the performance of the segmentation model.

[0049] Figure 5is an exemplary structural block diagram of a device 500 according to an embodiment of the present application. It is understood that the device 500 may include the apparatus of the embodiment of the present application, and the device implementing the solution of the present application may be a single device (such as a computing device) or a multi-function device including various peripheral devices.

[0050] like Figure 5 As shown in , the device of the present application may also include a central processing unit or central processing unit ("CPU") 511, which may be a general-purpose CPU, a dedicated CPU, or other information processing and program execution unit. Further, the device 500 may also include a large-capacity memory 512 and a read-only memory ("ROM") 513, wherein the large-capacity memory 512 may be configured to store various types of data, including various multimodal data, blood vessel segmentation results and aneurysm annotation results, algorithm data, intermediate results, and various programs required to run the device 500. The ROM 513 may be configured to store data and instructions required for power-on self-test of the device 500, initialization of various functional modules in the system, basic input / output drivers of the system, and booting the operating system.

[0051] Optionally, the device 500 may also include other hardware platforms or components, such as the tensor processing unit ("TPU") 514, graphics processing unit ("GPU") 515, field programmable gate array ("FPGA") 516, and machine learning unit ("MLU") 517 shown. It is understood that although a variety of hardware platforms or components are shown in the device 500, they are merely exemplary and not restrictive, and those skilled in the art may add or remove corresponding hardware according to actual needs. For example, the device 500 may only include a CPU, related storage devices, and interface devices to implement the method for training a segmentation model for segmenting an aneurysm or the method for segmenting an aneurysm of the present application.

[0052] In some embodiments, in order to facilitate the transmission and interaction of data with an external network, the device 500 of the present application further includes a communication interface 518, so that it can be connected to a local area network / wireless local area network ("LAN / WLAN") 505 through the communication interface 518, and then connected to a local server 506 or to the Internet ("Internet") 507 through the LAN / WLAN. Alternatively or additionally, the device 500 of the present application can also be directly connected to the Internet or a cellular network through the communication interface 518 based on wireless communication technology, such as wireless communication technology based on the third generation ("3G"), the fourth generation ("4G") or the fifth generation ("5G") of the generation. In some application scenarios, the device 500 of the present application can also access a server 508 and a database 509 of an external network as needed to obtain various known algorithms, data and modules, and can remotely store various data, such as various types of data or instructions for presenting, for example, multimodal data, vascular segmentation results and aneurysm annotation results.

[0053] The peripheral devices of the device 500 may include a display device 502, an input device 503, and a data transmission interface 504. In one embodiment, the display device 502 may include, for example, one or more speakers and / or one or more visual displays, which are configured to train the segmentation model for segmenting aneurysms of the present application or to perform voice prompts and / or image video display for segmenting aneurysms. The input device 503 may include other input buttons or controls such as a keyboard, a mouse, a microphone, a gesture capture camera, etc., which are configured to receive input of audio data and / or user instructions. The data transmission interface 504 may include, for example, a serial interface, a parallel interface or a universal serial bus interface ("USB"), a small computer system interface ("SCSI"), a serial ATA, a FireWire ("FireWire"), a PCI Express, and a high-definition multimedia interface ("HDMI"), etc., which are configured for data transmission and interaction with other devices or systems. According to the solution of the present application, the data transmission interface 504 can receive multimodal data collected from CT and MRI equipment, and transmit data or results including multimodal data or various other types to the device 500.

[0054] The CPU 511, mass storage 512, ROM 513, TPU 514, GPU 515, FPGA 516, MLU 517 and communication interface 518 of the device 500 of the present application can be interconnected through a bus 519, and data interaction with peripheral devices can be achieved through the bus. In one embodiment, through the bus 519, the CPU 511 can control other hardware components in the device 500 and its peripheral devices.

[0055] Combination of the above Figure 5The device that can be used to perform the present application is described. It should be understood that the device structure or architecture here is only exemplary, and the implementation method and implementation entity of the present application are not limited thereto, but can be changed without departing from the spirit of the present application.

[0056] According to the above description in combination with the accompanying drawings, those skilled in the art can also understand that the embodiments of the present application can also be implemented by software programs. Therefore, the present application also provides a computer-readable storage medium, on which computer-readable instructions for training a segmentation model for segmenting an aneurysm or for segmenting an aneurysm are stored. When the computer-readable instructions are executed by one or more processors, they can be used to implement the present application in combination with the accompanying drawings. Figure 1 The method described in the present invention for training a segmentation model for segmenting an aneurysm; or implementing the present invention in combination with the attached Figure 4 A method for segmenting an aneurysm is described.

[0057] It should be noted that although the operations of the method of the present application are described in a specific order in the accompanying drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. On the contrary, the steps depicted in the flow chart can be performed in a different order. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step, and / or one step can be decomposed into multiple steps.

[0058] It should be understood that when the terms "first", "second", "third" and "fourth" are used in the claims, the specification and the drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0059] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this application specification and claims, unless the context clearly indicates otherwise, the singular forms of "a", "an" and "the" are intended to include plural forms. It should also be further understood that the term "and / or" used in this application specification and claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0060] Although the implementation methods of the present application are as above, the contents described are only examples adopted to facilitate the understanding of the present application, and are not intended to limit the scope and application scenarios of the present application. Any technician in the technical field described in the present application can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present application, but the scope of patent protection of the present application shall still be subject to the scope defined in the attached claims.

[0061] In addition, the collection and acquisition of various data in this application complies with relevant laws and regulations and is authorized by the data provider. Any organization or individual that needs to obtain external data must obtain authorization in accordance with the law and ensure data security. It is not allowed to illegally collect, use, process, or transmit unauthorized or unprotected data, or to illegally buy, sell, provide, or disclose unauthorized or unprotected data.

Claims

1. A method for training a segmentation model for segmenting an aneurysm, comprising: Acquiring multimodal imaging data, wherein the multimodal imaging data includes digital subtraction angiography data, CT angiography data, and magnetic resonance angiography data; Performing blood vessel segmentation and annotating aneurysms based on the multimodal image data to obtain blood vessel segmentation results and aneurysm annotation results; Constructing label data according to the blood vessel segmentation result and the aneurysm annotation result; and The blood vessel segmentation result and the label data are input into a segmentation model to perform aneurysm segmentation to obtain an aneurysm segmentation result, so as to train the segmentation model used for segmenting aneurysms.

2. The method according to claim 1, wherein constructing label data according to the blood vessel segmentation result and the aneurysm annotation result comprises: Setting the foreground pixel points on the blood vessel segmentation result and not on the aneurysm annotation result as first label data; Calculating the shortest distance between a foreground pixel and a background pixel on the blood vessel segmentation result; as well as The second label data or the third label data is set according to the shortest distance and the aneurysm marking result to construct the label data.

3. The method according to claim 2, wherein setting the second label data or the third label data according to the shortest distance and the aneurysm marking result comprises: Comparing the shortest distance with a distance threshold and determining whether the corresponding foreground pixel is located within the aneurysm annotation result; as well as The second label data or the third label data is set according to the comparison result of the shortest distance and the distance threshold and the judgment result of whether the corresponding foreground pixel point is located in the aneurysm marking result.

4. The method according to claim 3, wherein setting the second label data or the third label data according to the comparison result of the shortest distance and the distance threshold and the judgment result of whether the corresponding foreground pixel point is located within the aneurysm annotation result comprises: In response to the shortest distance being less than or equal to the distance threshold, and the corresponding foreground pixel being located within the aneurysm annotation result, setting the corresponding foreground pixel as the second label data; or In response to the shortest distance being greater than the distance threshold and the corresponding foreground pixel being located within the aneurysm annotation result, the corresponding foreground pixel is set as the third label data.

5. The method according to claim 3, wherein inputting the blood vessel segmentation result and the label data into a segmentation model to perform aneurysm segmentation to obtain an aneurysm segmentation result, so as to train the segmentation model for segmenting aneurysms comprises: The blood vessel segmentation result, the first label data, the second label data and the third label data are input into the segmentation model to perform aneurysm segmentation, and different categories are output to obtain aneurysm segmentation results to train the segmentation model for segmenting aneurysms.

6. The method according to claim 5, further comprising: Calculating the proportion of the third label data in the blood vessel segmentation result; as well as The training weights of the first label data, the second label data, and the third label data in training the segmentation model are adjusted according to the proportions to optimize the training of the segmentation model for segmenting aneurysms.

7. The method according to claim 1, wherein the segmentation model comprises a Unet model, a FPN model or a PSPNet model.

8. A method for segmenting an aneurysm, comprising: Acquiring image data to be segmented and extracting a binary image of blood vessel segmentation based on the image data to be segmented; as well as The binary image of the blood vessel segmentation is input into the segmentation model trained according to the method described in any one of claims 1 to 7 to perform aneurysm segmentation to obtain an aneurysm segmentation result.

9. A device comprising: processor; as well as A memory having computer instructions for training a segmentation model for segmenting an aneurysm stored thereon, which, when executed by a processor, implement the method according to any one of claims 1 to 7; or having computer instructions for segmenting an aneurysm stored thereon, which, when executed by a processor, implement the method according to claim 8.

10. A computer-readable storage medium storing computer program instructions for training a segmentation model for segmenting an aneurysm, wherein the computer program instructions, when executed by one or more processors, implement the method according to any one of claims 1-7; or storing computer program instructions for segmenting an aneurysm, wherein the computer program instructions, when executed by one or more processors, implement the method according to claim 8.

Images