Method and equipment for segmenting large intestine effusion and storage medium

Abstract

The invention discloses a method and equipment for segmenting large intestine effusion and a storage medium. The method comprises the following steps: extracting a target skeleton with the maximum skeleton volume from a three-dimensional intestinal image, and taking the target skeleton as an initial segmentation region; based on the initial segmentation region, performing contour extraction by using an active contour model to obtain an intestine segmentation result; calculating a minimum surrounding circle of the intestine segmentation result in the cross section area; fitting the probability distribution of the large intestine region and the probability distribution of the large intestine effusion region in the minimum surrounding circle by using a Gaussian mixture model; and inputting the fitting result into a graph cut algorithm model to carry out effusion segmentation so as to segment a large intestine effusion region from the large intestine region. By means of the scheme, the large intestine effusion area can be accurately extracted, and the complete large intestine area can be segmented by combining the large intestine effusion area with the large intestine area subsequently.

Description

Technical Field

[0001] The present application generally relates to the field of image processing technology. More specifically, the present application relates to a method, device and computer-readable storage medium for segmenting large intestinal effusion. Background Art

[0002] 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 human tomographic data through medical scanning equipment such as computed tomography ("CT") and magnetic resonance imaging ("MRI"), and then uses image processing technology to reconstruct three-dimensional images to form virtual human tissues.

[0003] In actual shooting scenarios, the colon is incompletely inflated or has too much fluid inside when shooting three-dimensional images. The fluid is the colon wall from which water has not been completely removed. This results in an incomplete colon in the reconstructed three-dimensional image, affecting subsequent analysis (such as conformal expansion or three-dimensional navigation, etc.). Currently, there are methods for segmenting the colon such as region growing, watershed, and level set, but these methods themselves cannot automatically identify the features of the large intestine region and cannot segment the complete large intestine region, so they can only be applied to ideal images. In addition, there are also methods for segmenting the large intestine through deep learning, but because the intestinal region is always peristaltic, it is even different between two different detections, unlike other organs that have a fixed shape. Therefore, it is difficult for deep learning methods to learn this fixed pattern of the ever-changing large intestine region, and it is also difficult to segment the complete large intestine region.

[0004] In view of this, it is urgent to provide a solution for segmenting large intestinal effusion, by directly extracting the target skeleton with the largest skeleton volume from the three-dimensional intestinal image as the initial point for segmentation, and combining the prior knowledge of the large intestine shape to calculate the minimum enclosing circle, so as to more finely fit the probability distribution of the large intestine area and the probability distribution of the large intestine effusion area, thereby excluding irrelevant areas and similar effusion areas. Then the fitting results are input into the graph cut algorithm model for effusion segmentation, so as to converge to the global optimal solution through global optimization to accurately extract the large intestine effusion area, and then the large intestine effusion area is combined with the large intestine area to segment the complete large intestine area. Summary of the invention

[0005] In order to at least solve one or more of the technical problems mentioned above, the present application proposes a solution for segmenting large intestinal fluid accumulation in multiple aspects.

[0006] In the first aspect, the present application provides a method for segmenting large intestinal effusion, comprising: extracting a target skeleton with the largest skeleton volume from a three-dimensional intestinal image and using it as an initial segmentation area; based on the initial segmentation area, using an active contour model to perform contour extraction to obtain an intestinal segmentation result; calculating the minimum enclosing circle of the intestinal segmentation result in a cross-sectional area; using a Gaussian mixture model to fit the probability distribution of the large intestine area and the probability distribution of the large intestine effusion area within the minimum enclosing circle; and inputting the fitting result into a graph cut algorithm model for effusion segmentation to segment the large intestine effusion area from the large intestine area.

[0007] In one embodiment, before extracting the target skeleton with the largest skeleton volume from the three-dimensional intestinal image, the method further includes: setting a grayscale threshold; and roughly segmenting the three-dimensional intestinal image based on the grayscale threshold.

[0008] In another embodiment, extracting a target skeleton with the largest skeleton volume from a three-dimensional intestinal image includes: extracting a two-dimensional surface of a large intestine region from the three-dimensional intestinal image; calculating the shortest Euclidean distance from a voxel point in the three-dimensional intestinal image to the two-dimensional surface; and extracting the target skeleton with the largest skeleton volume from the three-dimensional intestinal image according to the shortest Euclidean distance.

[0009] In another embodiment, extracting the target skeleton with the largest skeleton volume from the three-dimensional intestinal image according to the shortest Euclidean distance includes: comparing the shortest Euclidean distance with a preset distance threshold; and extracting voxel points whose shortest Euclidean distance is greater than the preset distance threshold from the three-dimensional intestinal image to obtain the target skeleton with the largest skeleton volume.

[0010] In another embodiment, based on the initial segmentation area, using an active contour model to perform contour extraction to obtain the intestinal segmentation result includes: constructing an energy function under the active contour model based on the initial segmentation area; and minimizing the energy function to perform contour extraction to obtain the intestinal segmentation result.

[0011] In yet another embodiment, fitting the probability distribution of the large intestine region and the probability distribution of the large intestine fluid accumulation region within the minimum enclosing circle using a Gaussian mixture model comprises: fitting the probability distribution of the large intestine region and the probability distribution of the large intestine fluid accumulation region by the following formula:

[0012] in, represents the fitting result, represents the gray value within the minimum enclosing circle, represents the probability distribution of the large intestine effusion area, represents the probability distribution of the large intestine region, and represents the mean, and represents the variance, and represents the weight of the corresponding probability distribution.

[0013] In another embodiment, the fitting result is input into the graph cut algorithm model for fluid segmentation to segment the large intestine fluid accumulation area from the large intestine area, including: inputting the fitting result into the graph cut algorithm model for optimization; and taking the area formed by the points where the median value of the optimization result is 1 as the large intestine fluid accumulation area to segment the large intestine fluid accumulation area from the large intestine area.

[0014] In yet another embodiment, the fitting result is input into the graph cut algorithm model for optimization by the following formula:

[0015] in, f represents the gray value within the minimum enclosing circle, represents the fitting result, Y Indicates the value 0 or 1 in the optimization, subscript i represents a voxel point, subscript j Represents voxel points i Neighborhood points.

[0016] In a second aspect, the present application provides a device for segmenting large intestinal effusion, comprising: a processor; and a memory, in which program instructions for segmenting large intestinal effusion are stored, and when the program instructions are executed by the processor, the device implements one or more embodiments of the aforementioned first aspect.

[0017] In a third aspect, the present application provides a computer-readable storage medium having stored thereon computer-readable instructions for segmenting large intestinal effusion, wherein when the computer-readable instructions are executed by one or more processors, one or more embodiments of the aforementioned first aspect are implemented.

[0018] Through the scheme for segmenting large intestinal effusion as provided above, the embodiment of the present application extracts the target skeleton with the largest skeleton volume from the three-dimensional intestinal image and uses it as the initial segmentation area. Then, the points in the initial segmentation area are used as the initial points, and the active contour model is used to extract the contour to obtain the intestinal segmentation result. Furthermore, the minimum enclosing circle is calculated based on the prior knowledge of the shape of the large intestine (that is, the large intestine is circular in each cross-section), and the Gaussian mixture model is used to fit the probability distribution of the large intestine area and the probability distribution of the large intestine effusion area within the minimum enclosing circle. In this way, irrelevant areas and similar effusion areas can be excluded. Then, the fitting results are input into the graph cut algorithm model for effusion segmentation, so as to converge to the global optimal solution through global optimization to accurately extract the large intestine effusion area. Subsequently, the large intestine effusion area is combined with the large intestine area to segment out the complete large intestine area. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] 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: Figure 1 is an exemplary flowchart showing a method for segmenting large intestinal fluid accumulation according to an embodiment of the present application; Figure 2 is an exemplary schematic diagram showing a roughly segmented three-dimensional intestinal image according to an embodiment of the present application; Figure 3 is an exemplary schematic diagram showing a skeleton including a large intestine and a small intestine according to an embodiment of the present application; Figure 4 is an exemplary schematic diagram showing a target skeleton with the largest skeleton volume according to an embodiment of the present application; Figure 5 is an exemplary schematic diagram showing a large intestine region obtained by segmentation according to an embodiment of the present application; Figure 6 is an exemplary schematic diagram showing a cross-sectional area of ​​a large intestine region according to an embodiment of the present application; Figure 7 is an exemplary schematic diagram showing a complete large intestine region according to an embodiment of the present application; Figure 8 is an exemplary flowchart showing acquisition of a complete large intestine region according to an embodiment of the present application; Fig. 9 It is an exemplary structural block diagram showing a device for segmenting large intestinal fluid according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] 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.

[0021] 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.

[0022] 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.

[0023] As used in this specification and claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" can 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.

[0024] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0025] Figure 1 FIG. 1 is an exemplary flowchart of a method 100 for segmenting large intestinal fluid according to an embodiment of the present application. Figure 1 As shown in , at step S101, a target skeleton with the largest skeleton volume is extracted from the three-dimensional intestinal image and used as the initial segmentation region. In an implementation scenario, the aforementioned three-dimensional intestinal image can be acquired by an acquisition device such as CT, MRI, etc., and obtained through image reconstruction. For the reconstructed original three-dimensional intestinal image, before extracting the target skeleton with the largest skeleton volume from the three-dimensional intestinal image, the three-dimensional intestinal image can be roughly segmented to filter out interference information such as air and fat in the image.

[0026] In one embodiment, a grayscale threshold can be set to roughly segment the three-dimensional intestinal image based on the grayscale threshold. According to the Hounsfield scale, the HU value (also known as the CT value, or grayscale value) of air is -1000, and the HU value of fat is -120. Therefore, in some embodiments, the aforementioned grayscale threshold can be set to -1000 to -120. Based on the grayscale threshold, the three-dimensional intestinal image can be roughly segmented, and after the rough segmentation, it also includes, for example, the lungs, small intestine, large intestine, and interference noise areas. Furthermore, a subsequent segmentation operation is performed on the result after the rough segmentation to extract the large intestine area.

[0027] It can be understood that the biggest difference between the intestinal area and other areas is that the intestinal area is a long and narrow area, so theoretically the intestinal area can be judged according to the length of the central axis of the intestine. However, the large intestine and the small intestine cannot be distinguished based on the length of the central axis alone, and because the large intestine is thicker than the small intestine, the small intestine can be further excluded based on the thickness. For example, in some embodiments, the large intestine and the small intestine can be obtained based on the length of the central axis, and then the large intestine can be obtained based on the thickness. In an embodiment of the present application, the large intestine area is segmented by multiplying the length by the thickness, that is, by the volume of the skeleton. Specifically, the target skeleton with the largest skeleton volume is extracted from the three-dimensional intestinal image and used as the initial segmentation area.

[0028] In one embodiment, the two-dimensional surface of the large intestine region can be first extracted from the three-dimensional intestinal image, and then the shortest Euclidean distance from the voxel point in the three-dimensional intestinal image to the two-dimensional surface is calculated to extract the target skeleton with the largest skeleton volume from the three-dimensional intestinal image according to the shortest Euclidean distance. In the implementation scenario, for a voxel point in the three-dimensional intestinal image, if there is both a large intestine and other tissues or organs (such as small intestine, lungs, etc.) in its neighborhood (for example, 6 neighborhoods), then the voxel point is a two-dimensional surface point of the large intestine region. By judging all voxel points, the two-dimensional surface of the large intestine region can be extracted.

[0029] Based on the extracted two-dimensional surface of the large intestine region, for the remaining voxel points in the three-dimensional intestinal image (i.e., the voxel points other than the two-dimensional surface points), the shortest Euclidean distance from the voxel point to the two-dimensional surface is calculated. In some embodiments, for example, a binary image Euclidean distance transformation acceleration algorithm can be used to implement the aforementioned calculation of the shortest Euclidean distance from the voxel point to the two-dimensional surface. It can be understood that the binary image Euclidean distance transformation is a process of representing the distance between spatial points (target points and background points), and finally converting the binary image into a grayscale image. Specifically, it can be calculated by the formula Implement the Euclidean distance transform of binary images, where Represents voxel points To 2D surface point The shortest Euclidean distance, (*) represents the Euclidean distance function, represents the remaining voxel point sets (i.e., target sets) in the three-dimensional intestinal image, Represents a two-dimensional surface (i.e., the background set). Based on the calculated shortest Euclidean distance, the target skeleton with the largest skeleton volume can be extracted from the three-dimensional intestinal image.

[0030] In one embodiment, the shortest Euclidean distance can be compared with a preset distance threshold to extract voxel points whose shortest Euclidean distance is greater than the preset distance threshold from the three-dimensional intestinal image to obtain the target skeleton with the largest skeleton volume. It should be understood that inside the large intestine, the closer to the central axis, the greater the distance. Therefore, a preset distance threshold can be used so that when all points greater than the preset distance threshold are extracted, the small intestine area can be filtered out. As can be seen from the foregoing, the small intestine is thinner than the large intestine, so the area in the small intestine that is greater than the preset distance threshold is far less than that in the large intestine. Preferably, the aforementioned preset distance threshold can be, for example, 10. In combination with the anatomical features of the large and small intestines, by setting the preset distance threshold to 10, the small intestine can be filtered out while avoiding exceeding the thickness of the large intestine, causing the large intestine area to disappear, thereby being able to extract the target skeleton with the largest volume.

[0031] Based on the target skeleton with the largest skeleton volume obtained above, it is used as the initial segmentation region, and at step S102, based on the initial segmentation region, contour extraction is performed using the active contour model to obtain the intestinal segmentation result. In one embodiment, an energy function can be constructed under the active contour model based on the initial segmentation region, and then the energy function can be minimized to perform contour extraction to obtain the intestinal segmentation result. Specifically, by taking the initial segmentation region as the initial value and optimizing the initial segmentation region to minimize the energy function constructed under the active contour model, the intestinal segmentation result is obtained. In the embodiment of the present application, the intestinal segmentation result is the large intestine region.

[0032] In some embodiments, the above-mentioned active contour model can be, for example, an edge-based active contour model (such as a snake algorithm model), or can be, for example, a region-based active contour model (such as a Chan-Vese algorithm model). Among them, the edge-based active contour model is to evolve the curve to a specific edge of the image, that is, to find the target curve. The region-based active contour model is to evolve a specific region with the target curve as the boundary, that is, to find the specific region. Taking the region-based active contour model as an example, an energy function is constructed by, for example, the Chan-Vese algorithm model. Preferably, the embodiment of the present application also uses, for example, a level set method to optimize the initial segmentation area, so as to minimize the energy function for contour extraction, thereby obtaining the large intestine region.

[0033] Specifically, the Chan-Vese algorithm model constructs an energy function by using the gray value of the three-dimensional intestinal image as energy to evolve the curve to the target area. In the embodiment of the present application, the initial segmented area is used as the level set function (function of the surface), and the level set function is used to replace the curve evolution in the Chan-Vese algorithm model. For example, in an implementation scenario, the energy function can be constructed by the following formula: (1) in, Represents the initial segmentation area (that is, the gray value of the level set function), Represents the grayscale value of the original three-dimensional intestinal image, represents the energy of the area within the curve, represents the energy of the area outside the curve, and Represent the pixel mean values ​​in the area inside the curve and outside the curve respectively, and denote the weights of the area inside the curve and the area outside the curve respectively. Further, and denote the regularization terms for length and surface area, respectively. represents a step function, controlling the area inside the curve ( ) and the area outside the curve ( ), represents the Diktonic function.

[0034] In some embodiments, the initial segmentation region can be optimized by, for example, a variational method and a gradient descent flow. , to minimize the above energy function, and when optimized to When >0, the intestinal segmentation result is obtained, that is, the large intestine area is obtained.

[0035] Next, at step S103, the minimum enclosing circle of the intestinal segmentation result in the cross-sectional area is calculated. It can be understood that the grayscale distribution of the large intestine fluid effusion area is significantly different from that of the large intestine area. Therefore, based on the prior knowledge of the shape of the large intestine, the large intestine is circular in each cross-section. Based on this, the embodiment of the present application calculates the minimum enclosing circle of the large intestine area in each cross-sectional area, and the minimum enclosing circle can contain the effusion. In some embodiments, for example, a minimum circle covering algorithm can be used to calculate the minimum enclosing circle of the large intestine area in each cross-sectional area. The minimum circle covering algorithm can find the minimum circle covering n points within a linear time complexity.

[0036] Further, at step S104, the probability distribution of the large intestine region and the probability distribution of the large intestine effusion region are fitted within the minimum enclosing circle using a Gaussian mixture model. According to the prior knowledge of the large intestine, there are only large intestine distribution and effusion distribution within the minimum enclosing circle. Therefore, the embodiment of the present application uses two Gaussian distributions in the Gaussian mixture model, one of which fits the large intestine distribution and the other fits the effusion distribution. For example, in an exemplary scenario, the probability distribution of the large intestine region and the probability distribution of the large intestine effusion region can be fitted by the following formula: (2) in, represents the fitting result, Represents the gray value within the minimum enclosing circle, represents the probability distribution of the large intestine fluid accumulation area, represents the probability distribution of the large intestine region, and represents the mean, and represents the variance, and represents the weight of the corresponding probability distribution.

[0037] Based on the fitting result of the above fitting, at step S105, the fitting result is input into the graph cut algorithm model for fluid segmentation, so as to segment the large intestine fluid area from the large intestine area. In one embodiment, the fitting result is input into the graph cut algorithm model for optimization, and the area formed by the point where the median value of the optimization result is 1 is used as the large intestine fluid area, so as to segment the large intestine fluid area from the large intestine area.

[0038] In some embodiments, the distribution corresponding to the larger mean in the above fitting process is used as the probability distribution of the large intestinal effusion area, and then the probability distribution of the large intestinal effusion area after fitting is used as the initial value of the graph cut algorithm model to optimize through the graph cut algorithm model. In an implementation scenario, the fitting result can be input into the graph cut algorithm model for optimization through the following formula: (3) in, f Represents the gray value within the minimum enclosing circle, represents the fitting result (i.e., the probability distribution of the large intestinal effusion area), Y Indicates the value 0 or 1 in the optimization, subscript i represents a voxel point, subscript j Represents voxel points i Further, the point where the median value of the optimization result is 1 (that is, ) is used as the large intestine fluid accumulation area. In some embodiments, a complete large intestine area can be obtained by combining the large intestine fluid accumulation area with the large intestine area.

[0039] In combination with the above description, it can be known that the embodiment of the present application extracts the target skeleton with the largest skeleton volume from the three-dimensional intestinal image and uses it as the initial segmentation area. Then, the points in the initial segmentation area are used as the initial points, and the active contour model is used to extract the contour to obtain the intestinal segmentation result. Furthermore, the minimum enclosing circle is calculated based on the prior knowledge of the shape of the large intestine (that is, the large intestine is a circle in each cross section), and the Gaussian mixture model is used to fit the probability distribution of the large intestine area and the probability distribution of the large intestine fluid accumulation area within the minimum enclosing circle. In this way, irrelevant areas and similar fluid accumulation areas can be excluded. Then, the fitting results are input into the graph cut algorithm model for fluid accumulation segmentation, so as to converge to the global optimal solution through global optimization to accurately extract the large intestine fluid accumulation area. Subsequently, the large intestine fluid accumulation area is combined with the large intestine area to segment out the complete large intestine area.

[0040] Figure 2 is an exemplary schematic diagram showing a roughly segmented three-dimensional intestinal image according to an embodiment of the present application. In some embodiments, a grayscale threshold can be set to roughly segment the originally acquired three-dimensional intestinal image by using the grayscale threshold to obtain a Figure 2 The roughly segmented three-dimensional intestinal image is shown in . As an example, the grayscale threshold can be set to -1000 to -120. As can be seen from the figure, after the roughly segmented, the image also includes, for example, lung 201, small intestine 202, large intestine 203 and interference noise areas.

[0041] Figure 3 is an exemplary schematic diagram showing a skeleton including a large intestine and a small intestine according to an embodiment of the present application. As can be seen from the foregoing, the biggest difference between the intestinal region and other regions is that the intestinal region is a narrow and long region, so that, for example, the lung or noise region can be filtered out by, for example, the length of the central axis, and the image can be obtained as follows: Figure 3 The skeleton shown in FIG. 1 includes the large intestine and the small intestine. In addition, since the large intestine is thicker than the small intestine, the small intestine can be further filtered out according to the thickness to obtain the large intestine. In an embodiment of the present application, the large intestine region is segmented by the volume of the skeleton. In an implementation scenario, the two-dimensional surface of the large intestine region can be first extracted from the three-dimensional intestinal image, and then the voxel points whose shortest Euclidean distance from the remaining voxel points in the three-dimensional intestinal image to the two-dimensional surface are extracted that are greater than a preset threshold, so as to obtain a target skeleton with the largest skeleton volume.

[0042] Figure 4 is an exemplary schematic diagram showing a target skeleton with the largest skeleton volume according to an embodiment of the present application. Figure 4The target skeleton is shown in . As can be seen from the figure, the small intestine skeleton is filtered out of the target skeleton, and the large intestine skeleton remains. In an implementation scenario, the target skeleton is used as the initial segmentation region, an energy function is constructed under the active contour model, and contour extraction is performed by minimizing the energy function to obtain the intestinal segmentation result, that is, the large intestine region. Preferably, the energy function can be constructed by, for example, the Chan-Vese algorithm model, and combined with, for example, the level set method to optimize the initial segmentation region, so as to minimize the energy function for contour extraction, thereby obtaining the large intestine region (for example Figure 5 For more details on constructing the energy function and its optimization, please refer to the above Figure 1 The described contents will not be repeated in this application.

[0043] Figure 5 FIG. 2 is an exemplary schematic diagram showing a large intestine region obtained by segmentation according to an embodiment of the present application. Figure 5 Shown in the figure is the large intestine region obtained by segmentation using the Chan-Vese algorithm combined with the level set method. It can be seen from the figure that the large intestine region where fluid accumulation exists is missing, such as the area indicated by arrow A in the figure. Therefore, the embodiment of the present application further extracts the fluid accumulation region, and subsequently combines the fluid accumulation region with the large intestine region to obtain the completed large intestine region. In an implementation scenario, the minimum enclosing circle of the large intestine region in each cross-sectional area is first calculated, and then the probability distribution of the large intestine region and the probability distribution of the large intestine fluid accumulation region are fitted within the minimum enclosing circle to extract the probability distribution of the large intestine fluid accumulation region.

[0044] Figure 6 FIG. 2 is an exemplary schematic diagram showing a cross-sectional area of ​​a large intestine region according to an embodiment of the present application. Figure 6 Figure (a) shows the original image of the large intestine in the cross-section area. Figure 6 Figure (b) shows the large intestine regions S1, S2 and S3 in the original image of the cross-sectional region. As mentioned above, the minimum enclosing circle of the large intestine region in each cross-sectional region is first calculated, for example Figure 6 As shown in c1, c2 and c3 in Figure (c). In some embodiments, for example, a minimum circle covering algorithm can be used to calculate the minimum enclosing circle of the large intestine region in each cross-sectional area. Then, based on the above formula (2), a Gaussian mixture model is used to fit the probability distribution of the large intestine region and the probability distribution of the large intestine fluid accumulation region within the minimum enclosing circle, and the distribution corresponding to the larger mean in the fitting process is used as the probability distribution of the large intestine fluid accumulation region. Furthermore, using the probability distribution of the large intestine fluid accumulation region as the initial value, the above formula (3) is used to perform fluid accumulation segmentation through a graph cut algorithm to obtain the fluid accumulation region, for example Figure 6The partial fluid accumulation is shown in region E of (d) of FIG. Similarly, by obtaining the fluid accumulation region in each cross-sectional region of the large intestine region in the embodiment of the present application, all fluid accumulation regions can be obtained. Subsequently, the fluid accumulation region is combined with the large intestine region to obtain the complete large intestine region, for example Figure 7 shown.

[0045] Figure 7 is an exemplary schematic diagram showing a complete large intestine region according to an embodiment of the present application. Figure 7 The figure shows a complete large intestine region obtained by combining the fluid accumulation region with the large intestine region. The areas indicated by arrows in the figure are all fluid accumulation regions. In the implementation scenario, by performing conformal unfolding or three-dimensional navigation analysis on the complete large intestine region, intestinal diseases, such as polyp detection, can be checked.

[0046] Figure 8 FIG. 1 is an exemplary flowchart showing the acquisition of a complete large intestine region according to an embodiment of the present application. Figure 8 As shown in , at step S801, a three-dimensional intestinal image is acquired. In an implementation scenario, the three-dimensional intestinal image can be acquired by acquisition equipment such as CT, MRI, etc. At step S802, a grayscale threshold is set for rough segmentation. In some embodiments, the aforementioned grayscale threshold can be set to -1000 to -120. After rough segmentation, it also includes, for example, lungs, small intestine, large intestine, and interference noise areas. Then, at step S803, the target skeleton with the largest skeleton volume is extracted and used as the initial segmentation area. Specifically, the voxel points whose shortest Euclidean distance is greater than the preset distance threshold can be extracted from the three-dimensional intestinal image to obtain the target skeleton with the largest skeleton volume, and use it as the initial segmentation area.

[0047] Based on the initial segmented area, at step S804, an active contour model is used to perform contour extraction to obtain an intestinal segmentation result. In some embodiments, the intestinal segmentation result, that is, the large intestine region, can be extracted by, for example, the Chan-Vese algorithm combined with the level set method. Next, at step S805, the minimum enclosing circle of the intestinal segmentation result in the cross-sectional area is calculated, and at step S806, the probability distribution of the large intestine fluid accumulation area is fitted. By using the probability distribution of the large intestine fluid accumulation area as the initial value, at step S807, it is input into the graph cut algorithm model for fluid accumulation segmentation to obtain the large intestine fluid accumulation area. Furthermore, at step S808, the segmented large intestine fluid accumulation area is combined with the large intestine area to obtain a complete large intestine area.

[0048] Fig. 9is an exemplary structural block diagram of a device 900 for segmenting large intestinal effusion according to an embodiment of the present application. It is understood that the device 900 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 multifunctional device including various peripheral devices.

[0049] like Fig. 9 As shown in , the device of the present application may also include a central processing unit or central processing unit ("CPU") 911, which may be a general-purpose CPU, a dedicated CPU, or other information processing and program execution unit. Further, the device 900 may also include a large-capacity memory 912 and a read-only memory ("ROM") 913, wherein the large-capacity memory 912 may be configured to store various types of data, including various programs required for three-dimensional intestinal images, initial segmentation areas, intestinal segmentation results, effusion areas, algorithm data, intermediate results, and the device 900. ROM 913 may be configured to store data and instructions required for power-on self-test of the device 900, initialization of various functional modules in the system, basic input / output drivers of the system, and booting the operating system.

[0050] Optionally, the device 900 may also include other hardware platforms or components, such as the tensor processing unit ("TPU") 914, the graphics processing unit ("GPU") 915, the field programmable gate array ("FPGA") 916, and the machine learning unit ("MLU") 917 shown. It is understood that although a variety of hardware platforms or components are shown in the device 900, they are only exemplary and not restrictive, and those skilled in the art can add or remove corresponding hardware according to actual needs. For example, the device 900 may only include a CPU, related storage devices, and interface devices to implement the method for segmenting large intestinal effusion of the present application.

[0051] In some embodiments, in order to facilitate the transmission and interaction of data with an external network, the device 900 of the present application further includes a communication interface 918, so that it can be connected to a local area network / wireless local area network ("LAN / WLAN") 905 through the communication interface 918, and then connected to a local server 906 or to the Internet ("Internet") 907 through the LAN / WLAN. Alternatively or additionally, the device 900 of the present application can also be directly connected to the Internet or a cellular network through the communication interface 918 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") . In some application scenarios, the device 900 of the present application can also access a server 908 and a database 909 of an external network as needed to obtain various known algorithms, data and modules, and can remotely store various data, such as various data or instructions for presenting, for example, three-dimensional intestinal images, initial segmentation areas, intestinal segmentation results, effusion areas, etc.

[0052] The peripheral devices of the device 900 may include a display device 902, an input device 903, and a data transmission interface 904. In one embodiment, the display device 902 may include, for example, one or more speakers and / or one or more visual displays, which are configured to perform voice prompts and / or image video displays for segmenting large intestinal effusions of the present application. The input device 903 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 904 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 904 can receive a three-dimensional intestinal image collected from a CT device, and transmit data or results including a three-dimensional intestinal image or various other types to the device 900.

[0053] The CPU 911, mass storage 912, ROM 913, TPU 914, GPU 915, FPGA 916, MLU 917 and communication interface 918 of the device 900 of the present application can be interconnected through a bus 919, and data interaction with peripheral devices can be achieved through the bus. In one embodiment, through the bus 919, the CPU 911 can control other hardware components in the device 900 and its peripheral devices.

[0054] Combination of the above Fig. 9The device for segmenting large intestinal fluid that can be used to implement 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.

[0055] 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 segmenting large intestinal effusion 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 described method for segmenting large bowel fluid collections.

[0056] 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.

[0057] 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.

[0058] 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.

[0059] 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.

[0060] 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 dividing large intestinal fluid, characterized in that: include: Extract the target skeleton with the largest skeleton volume from the three-dimensional intestinal image and use it as the initial segmentation area; Based on the initial segmented area, an active contour model is used to perform contour extraction to obtain an intestinal segmentation result; Calculate the minimum enclosing circle of the intestinal segmentation result in the cross-sectional area; Use a Gaussian mixture model to fit the probability distribution of the large intestine area and the probability distribution of the large intestine fluid accumulation area within the minimum enclosing circle; as well as The fitting result is input into the graph cut algorithm model for fluid accumulation segmentation, so as to segment the large intestine fluid accumulation area from the large intestine area; Based on the initial segmented area, using the active contour model to extract contours to obtain the intestinal segmentation result includes: Based on the initial segmented region, constructing an energy function under the active contour model; and Minimizing the energy function to perform contour extraction to obtain the intestinal segmentation result; The method of fitting the probability distribution of the large intestine region and the probability distribution of the large intestine effusion region within the minimum enclosing circle using the Gaussian mixture model includes: fitting the probability distribution of the large intestine region and the probability distribution of the large intestine effusion region by the following formula: in, represents the fitting result, represents the gray value within the minimum enclosing circle, represents the probability distribution of the large intestine effusion area, represents the probability distribution of the large intestine region, and represents the mean, and represents the variance, and represents the weight of the corresponding probability distribution; The fitting results are input into the graph cut algorithm model for fluid segmentation, so as to segment the large intestine fluid area from the large intestine area, including: Inputting the fitting results into a graph cut algorithm model for optimization; and The area formed by the points whose median value of the optimization result is 1 is used as the large intestine fluid accumulation area, so as to segment the large intestine fluid accumulation area from the large intestine area.

2. The method according to claim 1, characterized in that Before extracting the target skeleton with the largest skeleton volume from the three-dimensional intestinal image, the following steps are also included: Setting grayscale thresholds; and The three-dimensional intestinal image is roughly segmented based on the grayscale threshold.

3. The method according to claim 1 or 2, characterized in that: The target skeletons with the largest skeleton volume extracted from the three-dimensional intestinal images include: extracting a two-dimensional surface of a large intestine region from the three-dimensional intestinal image; Calculating the shortest Euclidean distance from a voxel point in the three-dimensional intestinal image to the two-dimensional surface; and The target skeleton with the largest skeleton volume is extracted from the three-dimensional intestinal image according to the shortest Euclidean distance.

4. The method according to claim 3, characterized in that The target skeleton with the largest skeleton volume extracted from the three-dimensional intestinal image according to the shortest Euclidean distance comprises: Comparing the shortest Euclidean distance with a preset distance threshold; and Voxel points whose shortest Euclidean distance is greater than the preset distance threshold are extracted from the three-dimensional intestinal image to obtain the target skeleton with the largest skeleton volume.

5. The method according to claim 1, characterized in that: The fitting result is input into the graph cut algorithm model for optimization by the following formula: in, f represents the gray value within the minimum enclosing circle, represents the fitting result, Y Indicates the value 0 or 1 in the optimization, subscript i represents a voxel point, subscript j Represents voxel points i Neighborhood points.

6. A device for dividing large intestinal fluid, characterized in that: include: processor; as well as A memory, in which program instructions for segmenting large intestinal effusion are stored. When the program instructions are executed by the processor, the device for segmenting large intestinal effusion implements the method according to any one of claims 1-5.

7. A computer-readable storage medium, characterized in that: Computer-readable instructions for segmenting large intestinal fluid are stored thereon, and when the computer-readable instructions are executed by one or more processors, the method according to any one of claims 1-5 is implemented.

Images