Method and device for performing large intestine segmentation based on three-dimensional image, and storage medium

Abstract

The invention discloses a method and equipment for performing large intestine segmentation based on a three-dimensional image and a storage medium. The method comprises the following steps: acquiring a three-dimensional image of an intestinal tract and performing initial segmentation on the three-dimensional image of the intestinal tract to obtain an initial segmentation result; calculating a first prior probability that each voxel point belongs to a large intestine based on the initial segmentation result; according to the intestinal tract three-dimensional image, the initial segmentation result and the first prior probability, calculating prior distribution of the segmentation result and a second prior probability that each voxel point belongs to the segmentation result by using a Markov random field; calculating a maximum likelihood estimate based on the first prior probability, the prior distribution, and the second prior probability; and determining a large intestine voxel according to a maximum likelihood estimation result so as to realize three-dimensional large intestine segmentation. By utilizing the scheme of the invention, the large intestine can be efficiently and accurately segmented.

Description

Technical Field

[0001] This application generally relates to the field of image segmentation technology. More specifically, this application relates to a method, device, and computer-readable storage medium for three-dimensional large intestine segmentation. Background Art

[0002] The large intestine consists of the cecum, colon, and rectum, and is located at the end of the digestive tract and is an important part of the human digestive system. Together with the small intestine, the large intestine forms the end of the human digestive system. With the development of medical image processing and three-dimensional visualization technology, virtual endoscopy has been widely used in many clinical experiments and various medical diagnoses, such as intestinal examinations, due to its obvious advantages such as non-invasiveness and repeatability. Virtual endoscopy first obtains tomographic scan data of the human body through medical scanning devices such as computed tomography ("CT"), magnetic resonance imaging ("MRI"), etc., and then uses image processing technology to reconstruct three-dimensional images to form virtual human tissues.

[0003] In intestinal examinations, in addition to the large intestine part, other tissue organs may also exist in the reconstructed three-dimensional intestinal image, which will interfere with subsequent research. Therefore, it is necessary to segment the large intestine from the three-dimensional image. Traditional methods often segment through methods such as edge detection and connected component analysis. However, since the brightness of the large intestine is basically the same as that of other tissue organs, the large intestine cannot be accurately segmented using traditional methods. In addition, there are also methods such as deep learning and machine learning to segment the large intestine. However, the large intestine is constantly peristalsing, and even for the same subject, there may be huge differences in the large intestine in the supine and prone positions. This makes the large intestine area have no fixed pattern and is unstable, resulting in difficulty for methods such as deep learning and machine learning that rely on manual marking to fully achieve large intestine segmentation. In addition, there are also methods using level set methods to segment large intestine images. However, the level set method is sensitive to the initial value, and different initial values will produce different results, which does not meet the high requirements for robustness in medical applications. Further, the level set usually requires hundreds or thousands of iterations, so its time cost is extremely high. From an engineering perspective, the level set method can only perform some small-scale local organ segmentations, but it is difficult to segment the large intestine that occupies the entire abdomen.

[0004] In view of this, there is an urgent need to provide a solution for large intestine segmentation based on three-dimensional images. First, an initial segmentation result is obtained through initial segmentation to pre-filter interference noise or interfering tissues, etc. Then, by combining the prior probability of the large intestine and the prior probability of the segmentation result itself, the global optimal solution is guaranteed, so that accurate large intestine voxels can be obtained, and the accuracy of large intestine segmentation can be improved. Further, only one iteration is required through the Markov random field, overcoming the time cost of the level set and improving the efficiency of large intestine segmentation. Summary of the Invention

[0005] To at least solve one or more of the above-mentioned technical problems, the present application proposes a solution for large intestine segmentation based on three-dimensional images in multiple aspects.

[0006] In a first aspect, the present application provides a method for large intestine segmentation based on three-dimensional images, including: acquiring a three-dimensional intestinal image and performing initial segmentation on the three-dimensional intestinal image to obtain an initial segmentation result; calculating a first prior probability that each voxel belongs to the large intestine based on the initial segmentation result; calculating a prior distribution of the segmentation result and a second prior probability that each voxel belongs to the segmentation result by using a Markov random field according to the three-dimensional intestinal image, the initial segmentation result, and the first prior probability; calculating a maximum likelihood estimate based on the first prior probability, the prior distribution, and the second prior probability; and determining large intestine voxels according to the maximum likelihood estimate result to achieve three-dimensional large intestine segmentation.

[0007] In one embodiment, performing initial segmentation on the three-dimensional intestinal image to obtain an initial segmentation result includes: performing global threshold segmentation on the three-dimensional intestinal image to obtain a global threshold segmentation result; and filtering out the target region from the global threshold segmentation result by using a level set method to obtain the initial segmentation result.

[0008] In another embodiment, it further includes: performing a volume filtering and deletion operation on the initial segmentation result.

[0009] In yet another embodiment, the initial segmentation result includes the large intestine and the small intestine, and calculating a first prior probability that each voxel belongs to the large intestine based on the initial segmentation result includes: extracting points on the respective centerlines of the large intestine and the small intestine in the initial segmentation result; calculating a first distance function according to the points on the respective centerlines; and calculating the first prior probability that each voxel belongs to the large intestine based on the first distance function and a second distance function, where the second distance function is calculated based on points in the target region filtered out in the initial segmentation.

[0010] In yet another embodiment, calculating a first distance function according to the points on the respective centerlines includes: calculating a radial length based on the points on the respective centerlines; setting points corresponding to the radial length greater than a preset threshold as the foreground region; and calculating the first distance function based on points in the foreground region.

[0011] In yet another embodiment, calculating the first prior probability that each voxel belongs to the large intestine based on the first distance function and the second distance function includes: normalizing the first distance function and the second distance function to calculate the first prior probability that each voxel belongs to the large intestine.

[0012] In yet another embodiment, calculating the prior distribution of the segmentation result and the second prior probability that each voxel belongs to the segmentation result by using a Markov random field according to the three-dimensional intestinal image and the initial segmentation result includes: constructing an undirected graph of the Markov random field according to the three-dimensional intestinal image and the initial segmentation result; in the undirected graph, calculating the prior distribution based on connecting the nodes in the initial segmentation result with their adjacent target voxels and a first weight; and in the undirected graph, calculating the second prior probability based on connecting the introduced edges between the three-dimensional intestinal image and the nodes in the initial segmentation result and a second weight.

[0013] In yet another embodiment, determining the large intestine voxels according to the maximum likelihood estimation result includes: performing a maximum flow minimum cut operation on the maximum likelihood estimation result; and selecting the voxels with the target voxel value in the segmentation result after the maximum flow minimum cut operation to be determined as the large intestine voxels.

[0014] In a second aspect, the present application provides a device for segmenting the large intestine based on a three-dimensional image, including: a processor; and a memory storing computer instructions for segmenting the large intestine based on a three-dimensional image, which when executed by the processor, enables the implementation of multiple embodiments in the foregoing first aspect.

[0015] In a third aspect, the present application provides a computer-readable storage medium storing computer program instructions for segmenting the large intestine based on a three-dimensional image, which when executed by one or more processors, enables the implementation of multiple embodiments in the foregoing first aspect.

[0016] Through the above-provided solution for segmenting the large intestine based on a three-dimensional image, the embodiments of the present application first obtain an initial segmentation result by initially segmenting the three-dimensional intestinal image, and pre-filter interference noises or interfering tissues, etc. Then, based on the initial segmentation result, the first prior probability is calculated, and based on the foregoing three-dimensional intestinal image and the initial segmentation result, a Markov random field is used to calculate the second prior probability and the prior distribution of the segmentation result. Thus, the prior probability that a voxel belongs to the large intestine and the prior probability of the segmentation result itself can be considered simultaneously to improve the segmentation accuracy. Further, the large intestine voxels are determined by jointly performing maximum likelihood estimation of the foregoing probabilities, so as to accurately segment the large intestine. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By reading the following detailed description with reference to the accompanying drawings, the above and other purposes, features, and advantages of the exemplary embodiments of the present application will become readily understood. In the drawings, several embodiments of the present application are shown in an exemplary rather than restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1is an exemplary flowchart showing a method for colon segmentation based on three-dimensional images according to an embodiment of the present application; Figure 2 is an exemplary schematic diagram showing the global threshold segmentation result according to an embodiment of the present application; Figure 3 is an exemplary schematic diagram showing the initial segmentation result according to an embodiment of the present application; Figure 4 is an exemplary schematic diagram showing the colon segmentation result according to an embodiment of the present application; Figure 5 is an exemplary flowchart showing the overall process of colon segmentation based on three-dimensional images according to an embodiment of the present application; Figure 6 is an exemplary structural block diagram of a device for colon segmentation based on three-dimensional images according to an embodiment of the present application Detailed implementation manners Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present application.

[0018] It should be understood that the terms "including" and "comprising" 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 combinations.

[0019] It should also be understood that the terms used in the specification of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. As used in the specification and claims of the present application, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms. It should also be further understood that the term " / and / " used in the specification and claims of the present application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0020] As used in this specification and the claims, the term "if" may be construed, depending on the context, as "when" or "once" or "in response to determining" or "in response to detecting". Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be construed, depending on the context, to mean "once determined" or "in response to determining" or "once [described condition or event] is detected" or "in response to detecting [described condition or event]".

[0021] The following describes in detail the specific embodiments of the present application in conjunction with the accompanying drawings.

[0022] Figure 1 It is an exemplary flowchart showing a method 100 for large intestine segmentation based on three-dimensional images according to an embodiment of the present application. As Figure 1 shown, at step S101, a three-dimensional intestinal image is obtained and the three-dimensional intestinal image is initially segmented to obtain an initial segmentation result. In one implementation scenario, the aforementioned three-dimensional intestinal image can be collected by acquisition devices such as CT, MRI, etc. and obtained through image reconstruction. It can be understood that the three-dimensional intestinal image may contain other interfering tissues (such as lungs, bones, etc.) and interfering noises. Therefore, it is necessary to perform large intestine segmentation based on the three-dimensional intestinal image. First, the three-dimensional intestinal image is initially segmented to filter out noises and irrelevant tissues.

[0023] In one embodiment, global threshold segmentation can be performed on the three-dimensional intestinal image to obtain a global threshold segmentation result, and the level set method is used to filter out the target region from the global threshold segmentation result to obtain an initial segmentation result. It should be understood that from an anatomical perspective, the gray value of the large intestine is relatively low, while the gray values of tissues and bones are relatively high. Therefore, in the embodiments of the present application, global threshold segmentation is first used to extract regions with lower brightness. That is, a threshold is used throughout the image to divide the entire image into target objects and background objects to obtain a full threshold segmentation result (such as Figure 2 shown). In the embodiments of the present application, the target object is the large intestine. In some embodiments, the aforementioned threshold can be determined through the gray histogram of the image.

[0024] Since the brightness of the lungs and the large intestine is almost the same, there will be lungs in the above-mentioned global threshold segmentation result, and the lungs and the large intestine are connected in the image. It is difficult to filter out the lungs through simple connectivity analysis. Therefore, in the embodiments of the present application, the target region (such as the lungs) is filtered out from the global threshold segmentation result by the Level Set Method ("LSM") to obtain an initial segmentation result. In one implementation scenario, the first picture obtained by global threshold segmentation can be set as the initial point of the level set, and an initial point is iterated using, for example, a step function to obtain the contour of the large intestine, thereby filtering out the lungs and obtaining an initial segmentation result. Specifically, in the initial stage, the level set function of the value inside the contour is 1, while the level set function of the value outside the contour is -1. Through iteration, -1 of the level set function gradually increases to a positive number, and the contour of the large intestine is obtained when the iteration stop condition is satisfied.

[0025] After performing the initial segmentation, there may still be a small amount of noise in the initial segmentation result. In this scenario, a more optimal initial segmentation result can be obtained by performing a volume filtering deletion operation on the initial segmentation result. Only the large intestine and the small intestine are included in this initial segmentation result (as shown in, for example Figure 3 ).

[0026] Based on the initial segmentation result obtained above, at step S102, the first prior probability that each voxel point belongs to the large intestine is calculated based on the initial segmentation result. In one embodiment, the points on the respective centerlines of the large intestine and the small intestine in the initial segmentation result can be extracted, and the first distance function is calculated according to the points on the respective centerlines, so as to calculate the first prior probability that each voxel point belongs to the large intestine based on the first distance function and the second distance function. Among them, the aforementioned second distance function is calculated based on the points of the target region filtered out in the initial segmentation.

[0027] In one implementation scenario, the radial length can be calculated based on the points on the respective centerlines, and the points corresponding to the radial length greater than the preset threshold are set as the foreground region, so as to calculate the first distance function based on the points in the foreground region. It can be understood that according to anatomical knowledge, the large intestine always has a larger radial length than the small intestine. Therefore, by using this prior knowledge, the first prior probability that each voxel belongs to the large intestine can be calculated.

[0028] Taking the centerline of the large intestine as an example, assuming that the points on the centerline of the large intestine are denoted as , the radial length can be calculated by the following formula: (1) Among them, represents a point in three-dimensional space, which is an independent variable, represents all points in three-dimensional space, represents the points on the surface of the large intestine; Denotes differentiation, i.e., the tangent vector of the central axis of the large intestine, and inf denotes the minimum value of the elements. The radial length is calculated for each point on the midline using the aforementioned formula (1). Further, for example, a global threshold method can be used to perform binary classification on the radial length, where all points corresponding to radial lengths greater than the threshold belong to the large intestine.

[0029] In an implementation scenario, by setting the points corresponding to radial lengths greater than a preset threshold as the foreground region, for the points within the foreground region, algorithms such as the "Geodesic Distance" algorithm are used to obtain a first distance function . It can be understood that the aforementioned geodesic distance is the distance of the shortest path between points within the foreground region. Since there will be a deviation in judging the probability that a voxel point belongs to the large intestine only using the distance function, the embodiments of the present application also involve a second distance function , which is calculated based on the points in the region filtered out in the initial segmentation. Specifically, the region filtered out in the initial segmentation (such as the lungs, tissues, and bones, etc.) is set as the background region, and similarly, the Geodesic Distance algorithm is used to obtain a second distance function . The first prior probability that each voxel point belongs to the large intestine calculated by combining the first distance function and the second distance function has a relatively high accuracy.

[0030] In an implementation scenario, the first distance function and the second distance function can be normalized to calculate the first prior probability that each voxel point belongs to the large intestine. In an exemplary scenario, normalization can be performed based on the following formula to calculate the first prior probability that a voxel point belongs to the large intestine : (2) Further, at step S103, according to the three-dimensional intestinal image, the initial segmentation result, and the first prior probability, the prior distribution of the segmentation result and the second prior probability that each voxel point belongs to the segmentation result are calculated using a Markov random field. It can be understood that if only the above first prior probability is used for segmentation, when the TV regularization term of the segmentation result is too large, the obtained segmentation result will be poor. Therefore, in addition to considering the first prior probability that a voxel point belongs to the large intestine, the embodiments of the present application also consider the prior probability of the segmentation result itself. By comprehensively considering the two probability distributions, the probability of voxels belonging to the large intestine is optimized, and the segmentation accuracy is improved.

[0031] In an embodiment, an undirected graph of a Markov random field is constructed according to the three-dimensional intestinal image and the initial segmentation result. In the undirected graph, the prior distribution is calculated based on the connection between the nodes in the initial segmentation result and their adjacent target voxels and the first weight, and in the undirected graph, the second prior probability is calculated based on the introduced edges connecting the three-dimensional intestinal image and the nodes in the initial segmentation result and the second weight.

[0032] Specifically, a Markov random field is an undirected graph that contains two types of nodes, one of which represents the original three-dimensional data (such as a three-dimensional intestinal image), denoted as ; the other represents the segmentation result, denoted as table. This undirected graph also contains two types of edges. The first type of edge represents the prior distribution of the segmentation result, which connects and its adjacent target voxels (such as 6), and the corresponding first weight is . Thus, the prior distribution of the segmentation result can be expressed as . The second type of edge represents the second prior probability that each voxel belongs to the segmentation result. By introducing an edge to connect and , the corresponding second weight is . Accordingly, the second prior probability that each voxel point belongs to the segmentation result can be expressed as , represents the first prior probability.

[0033] After obtaining the above prior probabilities, at step S104, the maximum likelihood estimation is calculated based on the first prior probability, the prior distribution, and the second prior probability. In one implementation scenario, based on the Hammersley-Clifford theorem, the maximum likelihood estimation of the joint probability distribution can be performed, which is specifically expressed as follows: (3) which can be equivalent to: (4) Furthermore, at step S105, the large intestine voxels are determined according to the maximum likelihood estimation result to achieve three-dimensional large intestine segmentation. In one embodiment, a maximum flow minimum cut operation can be performed on the maximum likelihood estimation result, and the voxels with the value of the target voxel value in the segmentation result after the maximum flow minimum cut operation are selected to be determined as the large intestine voxels. In some embodiments, when takes the value of , the possibility that all voxels belong to the large intestine is , where represents the total number of all voxels of the three-dimensional voxels, and it is difficult to enumerate all combinations. Therefore, in the embodiments of the present application, through the maximum flow minimum cut theorem, a maximum flow minimum cut operation is performed, and the voxels with the value of the target voxel value are selected from the segmentation result after the maximum flow minimum cut operation to be determined as the large intestine voxels. As an example, by selecting to be for all voxels, they are the large intestine voxels (as shown in Figure 4 for example).

[0034] As described above, in the embodiment of the present application, an initial segmentation result is obtained by initially segmenting the three-dimensional intestinal image, and the first prior probability is calculated based on the initial segmentation result. Based on the aforementioned three-dimensional intestinal image and the initial segmentation result, the second prior probability and the prior distribution of the segmentation result are calculated using a Markov random field. Based on this, by simultaneously considering the prior probability that a voxel belongs to the large intestine and the prior probability of the segmentation result itself, the global optimal solution is ensured, so that accurate large intestine voxels can be obtained, improving the accuracy of large intestine segmentation. Further, the Markov random field only requires one iteration, overcoming the time cost of the level set and improving the efficiency of large intestine segmentation.

[0035] Figure 2 is an exemplary schematic diagram showing the global threshold segmentation result according to the embodiment of the present application. As Figure 2 shown, it is the global threshold segmentation result obtained by performing global threshold segmentation on the three-dimensional intestinal image. According to the foregoing, the entire image can be divided into a large intestine region and a background region by using a threshold. Since the gray value of the large intestine is relatively low, while the gray values of tissues and bones are relatively high, the pixels exceeding the threshold can be set to 0, which is the background region, and the pixel values lower than the threshold are retained to obtain the global threshold segmentation result. As can be seen from the figure, the global threshold segmentation result also includes the lungs (such as indicated by the arrow in the figure), so further segmentation is required to obtain the initial segmentation result, such as Figure 3 shown.

[0036] Figure 3 is an exemplary schematic diagram showing the initial segmentation result according to the embodiment of the present application. As Figure 3 shown, it is the initial segmentation result. In some embodiments, the lungs can be filtered out from the global threshold segmentation result by using, for example, LSM. In addition, after the initial segmentation is performed, there may still be a small amount of noise, and operations such as volume filtering deletion can be used to obtain the initial segmentation result. As can be seen from the figure, the lungs and a small amount of noise are filtered out in the initial segmentation result after the initial segmentation, and the remaining parts are the large intestine and the small intestine.

[0037] Figure 4 is an exemplary schematic diagram showing the large intestine segmentation result according to the embodiment of the present application. As Figure 4 shown, it is the schematic diagram of the finally segmented large intestine. As described above, based on the above initial segmentation result and the three-dimensional intestinal image, the first prior probability that each voxel point belongs to the large intestine, the prior distribution of the segmentation result, and the second prior probability that each voxel point belongs to the segmentation result can be calculated. Further, the large intestine voxels are determined by the maximum likelihood estimation of the joint probability distribution to obtain the large intestine segmentation result as shown in the figure. Among them, for more details on determining the large intestine voxels, reference can be made to the above Figure 1 description, which will not be elaborated in the present application.

[0038] Figure 5 It is an exemplary flowchart showing the overall process of large intestine segmentation based on three-dimensional images according to an embodiment of the present application. As Figure 5 shown, at step S501, three-dimensional images of the intestine are acquired. In one implementation scenario, the aforementioned three-dimensional images of the intestine can be acquired by acquisition devices such as CT, MRI, etc. and obtained through image reconstruction. Then, at step S502, global threshold segmentation is performed on the three-dimensional images of the intestine to obtain a global threshold segmentation result. Based on the global threshold segmentation, the large intestine region and the background region can be obtained, but it also includes the lungs and a small amount of noise. In order to filter out the lungs and a small amount of noise, at steps S503 and S504, the lungs are filtered out from the global threshold segmentation result by the level set method, and the noise is filtered out by operations such as volume filtering deletion operation to obtain a better initial segmentation result.

[0039] Next, at step S505, the first prior probability that each voxel point belongs to the large intestine, the prior distribution of the segmentation result, and the second prior probability that each voxel point belongs to the segmentation result are calculated. Specifically, the first prior probability that each voxel point belongs to the large intestine can be calculated based on the above formula (2). The prior distribution of the segmentation result and the second prior probability that each voxel point belongs to the segmentation result are calculated using the Markov random field. The calculation of each prior probability can refer to the above Figure 1 description, which will not be elaborated in the present application. Further, at step S506, the maximum likelihood estimation is calculated (see the above formula (3) and formula (4)), and further segmentation is performed based on the maximum flow minimum cut theorem to obtain the large intestine segmentation result at step S507.

[0040] Figure 6 It is an exemplary structural block diagram of a device 600 for large intestine segmentation based on three-dimensional images according to an embodiment of the present application. It can be understood that the device 600 may include the device 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 multifunctional device including various peripheral devices.

[0041] As Figure 6 shown, the device of the present application may further include a central processing unit or central processing unit (“CPU”) 611, which may be a general-purpose CPU, a dedicated CPU, or other information processing and program execution units. Further, the device 600 may further include a large-capacity memory 612 and a read-only memory (“ROM”) 613, where the large-capacity memory 612 may be configured to store various types of data, including various three-dimensional images of the intestine, various segmentation results, algorithm data, intermediate results, and various programs required to run the device 600. The ROM 613 may be configured to store data and instructions for power-on self-test of the device 600, initialization of each functional module in the system, basic input / output drivers of the system, and data and instructions required to boot the operating system.

[0042] Optionally, device 600 may further include other hardware platforms or components, such as the illustrated tensor processing unit (“TPU”) 614, graphics processing unit (“GPU”) 615, field programmable gate array (“FPGA”) 616, and machine learning unit (“MLU”) 617. It can be understood that although a variety of hardware platforms or components are illustrated in device 600, these are merely exemplary and not restrictive, and those skilled in the art may add or remove corresponding hardware according to actual needs. For example, device 600 may include only a CPU, associated storage devices, and interface devices to implement the method for large intestine segmentation based on three-dimensional images of the present application.

[0043] In some embodiments, for the convenience of data transfer and interaction with an external network, device 600 of the present application further includes a communication interface 618, so that it can be connected to a local area network / wireless local area network (“LAN / WLAN”) 605 through this communication interface 618, and further can be connected to a local server 606 or connected to the Internet (“Internet”) 607 through this LAN / WLAN. Alternatively or additionally, device 600 of the present application may also be directly connected to the Internet or a cellular network based on wireless communication technology through communication interface 618, such as based on the 3rd generation (“3G”), 4th generation (“4G”), or 5th generation (“5G”) wireless communication technology. In some application scenarios, device 600 of the present application may also access server 608 and database 609 of an external network as needed, so as to obtain various known algorithms, data, and modules, and may remotely store various data, such as various types of data or instructions for presenting, for example, three-dimensional images of the intestine, various segmentation results, etc.

[0044] The peripherals of device 600 may include a display device 602, an input device 603, and a data transmission interface 604. In one embodiment, the display device 602 may include, for example, one or more speakers and / or one or more visual displays, which are configured to provide voice prompts and / or display image videos for the large intestine segmentation based on three-dimensional images in this application. The input device 603 may include, for example, a keyboard, a mouse, a microphone, a gesture capture camera, and other input buttons or controls, which are configured to receive the input of audio data and / or user instructions. The data transmission interface 604 may include, for example, a serial interface, a parallel interface, or a Universal Serial Bus interface ("USB"), a Small Computer System Interface ("SCSI"), Serial ATA, FireWire, 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 this application, the data transmission interface 604 may receive the three-dimensional images of the intestine collected by a CT device and transmit to device 600 various types of data or results including the three-dimensional images of the intestine.

[0045] The above CPU 611, mass storage 612, ROM 613, TPU 614, GPU 615, FPGA 616, MLU 617, and communication interface 618 of device 600 in this application may be interconnected with each other through a bus 619 and achieve data interaction with the peripherals through this bus. In one embodiment, through this bus 619, the CPU 611 may control other hardware components and their peripherals in device 600.

[0046] The above combination Figure 6 has described the device that can be used to perform the large intestine segmentation based on three-dimensional images in this application. It should be understood that the device structure or architecture here is only exemplary, and the implementation manner and implementation entity of this application are not limited by it, but can be changed without departing from the spirit of this application.

[0047] According to the above description with reference to the drawings, those skilled in the art can also understand that the embodiments of this application can also be implemented through software programs. Thus, this application also provides a computer-readable storage medium, on which computer-readable instructions for large intestine segmentation based on three-dimensional images are stored. When the computer-readable instructions are executed by one or more processors, they can be used to implement the method for large intestine segmentation based on three-dimensional images described in this application in combination with the attached Figure 1 drawings.

[0048] 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 these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the order of execution of the steps depicted in the flowchart can be changed. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step for execution, and / or one step may be decomposed into multiple steps for execution.

[0049] It should be understood that when terms such as "first", "second", "third", and "fourth" are used in the claims, the specification, and the accompanying drawings of the present application, they are only used to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" 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 combinations.

[0050] It should also be understood that the terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification and claims of the present application, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms. It should be further understood that the term "and / or" used in the specification and claims of the present application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0051] Although the embodiments of the present application are as described above, the above content is only an example adopted for the convenience of understanding the present application and is not intended to limit the scope and application scenarios of the present application. Any person skilled in the art within the technical field of the present application can make any modifications and changes in the form of implementation and details without departing from the spirit and scope disclosed by the present application. However, the scope of patent protection of the present application shall still be subject to the scope defined by the appended claims.

[0052] In addition, the collection and acquisition of various data in the present application comply with relevant laws and regulations and are authorized by the data provider. Any organization or individual that needs to obtain external data should obtain authorization in accordance with the law and ensure data security, and shall not illegally collect, use, process, or transmit unauthorized or unprotected data, nor illegally buy, sell, provide, or disclose unauthorized or unprotected data.

Claims

1. A method for colon segmentation based on three-dimensional images, characterized in that, including: obtaining a three-dimensional image of the intestine and performing initial segmentation on the three-dimensional image of the intestine to obtain an initial segmentation result; calculating a first prior probability that each voxel belongs to the large intestine based on the initial segmentation result; calculating a prior distribution of the segmentation result and a second prior probability that each voxel belongs to the segmentation result by using a Markov random field according to the three-dimensional image of the intestine, the initial segmentation result, and the first prior probability; calculating a maximum likelihood estimate based on the first prior probability, the prior distribution, and the second prior probability; determining large intestine voxels according to the maximum likelihood estimation result to achieve three-dimensional large intestine segmentation.

2. The method according to claim 1, wherein Wherein performing initial segmentation on the three-dimensional image of the intestine to obtain an initial segmentation result includes: performing global threshold segmentation on the three-dimensional image of the intestine to obtain a global threshold segmentation result; using a level set method to filter out a target region from the global threshold segmentation result to obtain the initial segmentation result.

3. The method according to claim 1 or 2, characterized in that, Wherein it further includes: performing a volume filtering and deletion operation on the initial segmentation result.

4. The method according to claim 3, characterized in that Wherein the initial segmentation result includes the large intestine and the small intestine, and calculating a first prior probability that each voxel belongs to the large intestine based on the initial segmentation result includes: extracting points on the respective centerlines of the large intestine and the small intestine in the initial segmentation result; calculating a first distance function according to the points on the respective centerlines; calculating the first prior probability that each voxel belongs to the large intestine based on the first distance function and a second distance function, wherein the second distance function is calculated based on points of a target region filtered out in the initial segmentation.

5. The method according to claim 4, wherein Wherein calculating a first distance function according to the points on the respective centerlines includes: calculating a radial length based on the points on the respective centerlines; setting points corresponding to the radial length greater than a preset threshold as a foreground region; calculating the first distance function based on points in the foreground region.

6. The method according to claim 4, characterized in that Wherein calculating the first prior probability that each voxel belongs to the large intestine based on the first distance function and the second distance function includes: normalizing the first distance function and the second distance function to calculate the first prior probability that each voxel belongs to the large intestine.

7. The method according to claim 1, wherein Wherein calculating a prior distribution of the segmentation result and a second prior probability that each voxel belongs to the segmentation result by using a Markov random field according to the three-dimensional image of the intestine and the initial segmentation result includes: constructing an undirected graph of a Markov random field according to the three-dimensional image of the intestine and the initial segmentation result; in the undirected graph, calculating the prior distribution based on connecting nodes in the initial segmentation result with their adjacent target voxels and a first weight; in the undirected graph, calculating the second prior probability based on incoming edges connecting the three-dimensional image of the intestine and nodes in the initial segmentation result and a second weight.

8. The method according to claim 1, characterized in that Wherein determining large intestine voxels according to the maximum likelihood estimation result includes: performing a maximum flow minimum cut operation on the maximum likelihood estimation result; selecting voxels with a value of the target voxel value in the segmentation result after the maximum flow minimum cut operation to be determined as the large intestine voxels.

9. An apparatus for colon segmentation based on three-dimensional images, characterized in that, including: a processor; A memory storing computer instructions for performing large intestine segmentation based on three-dimensional images, which, when executed by a processor, enable the implementation of the method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, A computer program instruction for performing large intestine segmentation based on three-dimensional images is stored thereon, and when the computer program instruction is executed by one or more processors, it enables the implementation of the method according to any one of claims 1-8.

Images