A method, device and computer device for segmenting medical images

Abstract

The present application discloses a method, apparatus, and computer device for segmenting medical images, including: obtaining a click position of a target medical image, taking the click position as the center, obtaining multiple equiangular rays and the distance measurement values of equidistant voxels on each ray outward, based on the distance measurement values, taking the click position as the center, performing multiple segmentations, obtaining the union region of the target connection region, the target segmentation region, and the spherical convex hull region, and further determining the final segmentation result of the tissue organ corresponding to the click position. It solves the problem that due to the lack of existing technologies, there is an urgent need for a new method for segmenting medical images to solve the problems of time-consuming and laborious manual segmentation and low efficiency. On the basis of manual clicks, an unsupervised and untrained intelligent method is used to maximize the automatic recognition and segmentation of the edges of the clicked area, significantly improving the efficiency of manual segmentation and retaining the high compatibility and high accuracy of manual segmentation.

Description

Technical Field

[0001] This application relates to the technical field of medical imaging, and in particular, to a method, device, and computer device for segmenting medical images. Background Art

[0002] The segmentation of medical images is a core task in medical image analysis. The purpose of the medical image segmentation task is to divide different organs or lesion regions in the image, and then extract key information, which is a key step in realizing the visualization of medical images. The segmented image is provided to doctors for different tasks such as quantitative analysis of tissue volume, diagnosis, localization of pathologically changed tissues, depiction of anatomical structures, and treatment planning.

[0003] For the segmentation of medical images, mostly two segmentation methods are adopted: intelligent automatic segmentation and manual segmentation. Among them, the intelligent automatic segmentation method relies on machine learning and deep learning models established through a large amount of data training in advance, and has almost no segmentation ability for data without a model established, and the intelligent automatic segmentation method mostly uses a unified template for modeling, and there are certain errors in the segmentation results; while the manual segmentation method has relatively high accuracy, but usually only supports manual contour drawing, regional smearing, or contour drawing of the Live Wire algorithm based on gray scale gradient detection, etc. Therefore, manual segmentation is a time-consuming and laborious task, which adds a great burden to the daily work of clinicians and has low efficiency.

[0004] Therefore, due to the lack of the prior art, there is an urgent need for a new method for segmenting medical images to solve the problems of time-consuming, laborious, and low efficiency of manual segmentation. Summary of the Invention

[0005] This application provides a method, device, and computer device for segmenting medical images, and solves the technical problem that due to the lack of the prior art, there is an urgent need for a new method for segmenting medical images to solve the problems of time-consuming, laborious, and low efficiency of manual segmentation.

[0006] On the one hand, a method for segmenting medical images is provided, and the method includes:

[0007] Obtain the click position of the target medical image;

[0008] Taking the click position as the center, obtain multiple equiangular rays outward;

[0009] Obtain the distance measurement values of equidistant voxels on each ray, and form a set of measurement value vectors with all the distance measurement values on the same ray;

[0010] Obtain the measure extreme points closest to the click position in each group of the measure value vectors, and connect the convex hulls of the respective measure extreme points to obtain a target connection region;

[0011] Obtain the upper density limit value and the lower density limit value of the target connection region, and based on the upper density limit value and the lower density limit value, segment the target connection region to obtain a target segmentation region;

[0012] With the direction of the connection lines between the click position and the convex hulls of the respective measure extreme points as the extension direction, extend the lengths of each connection line outward to form a spherical region centered on the click position;

[0013] Divide the radius of the spherical region equally to form a plurality of concentric spheres;

[0014] Obtain the spherical measure extreme points of each concentric sphere, and form a spherical convex hull region with the convex hulls of the respective spherical measure extreme points;

[0015] Obtain the union region of the target connection region, the target segmentation region, and the spherical convex hull region, and determine the union region as the target segmentation result of the target medical image.

[0016] On the other hand, a medical image segmentation device is provided, and the device includes:

[0017] A click position acquisition module, configured to acquire the click position of a target medical image;

[0018] A ray acquisition module, configured to acquire a plurality of equiangular rays outward with the click position as the center;

[0019] A measure value vector acquisition module, configured to acquire the distance measure values of equidistant voxels on each ray, and form a group of measure value vectors with all the distance measure values on the same ray;

[0020] A target connection region acquisition module, configured to obtain the measure extreme points closest to the click position in each group of the measure value vectors, and connect the convex hulls of the respective measure extreme points to obtain a target connection region;

[0021] A target segmentation region acquisition module, configured to obtain the upper density limit value and the lower density limit value of the target connection region, and based on the upper density limit value and the lower density limit value, segment the target connection region to obtain a target segmentation region;

[0022] A spherical region acquisition module, configured to extend the lengths of each connection line outward with the direction of the connection lines between the click position and the convex hulls of the respective measure extreme points as the extension direction to form a spherical region centered on the click position;

[0023] A concentric sphere acquisition module, configured to equally divide the radius of the spherical region to form a plurality of concentric spheres;

[0024] A spherical convex hull region acquisition module, configured to obtain the extreme points of the spherical measure of each of the concentric spheres, and form a spherical convex hull region by taking the convex hull of the extreme points of the spherical measure;

[0025] A target segmentation result acquisition module, configured to obtain the union region of the target connection region, the target segmentation region, and the spherical convex hull region, and determine the union region as the target segmentation result of the target medical image.

[0026] In a possible implementation manner, the measure value vector acquisition module is further configured to:

[0027] Determine the equally spaced voxels on each ray as sampling points, and obtain the density values corresponding to the respective sampling points;

[0028] Obtain an operator for the target size, and obtain the distance measure values of the respective sampling points based on the operator and the density values.

[0029] In a possible implementation manner, the target connection region acquisition module is further configured to:

[0030] Obtain the curve extreme values of each group of the measure value vectors;

[0031] Based on the curve extreme values, obtain the measure extreme points closest to the click position in each group of the measure value vectors;

[0032] Obtain the convex hulls of the respective measure extreme points, and connect the respective convex hulls to obtain a target connection region.

[0033] In a possible implementation manner, the target segmentation region acquisition module is further configured to:

[0034] Obtain the voxel mean value and voxel standard deviation of the target connection region;

[0035] Based on the voxel mean value and voxel standard deviation, obtain the density upper limit value and density lower limit value of the target connection region;

[0036] Centered on the click position, based on the density upper limit value and density lower limit value, segment the target connection region to obtain a target segmentation region.

[0037] In a possible implementation manner, the spherical region acquisition module is further configured to:

[0038] Taking the direction of the line connecting the click position and the convex hull of each measurement extreme point as the extension direction, extend the length of each line outward by a target multiple to form a spherical region centered at the click position.

[0039] In a possible implementation manner, the spherical convex hull region obtaining module is further configured to:

[0040] Obtain the spherical measurement extreme points of each of the concentric spheres;

[0041] Obtain the extreme value direction and average intensity of each of the spherical measurement extreme points;

[0042] Select, from all the spherical measurement extreme points, the spherical measurement maximum value points whose extreme value direction is the maximum value and whose intensity is greater than or equal to the average intensity;

[0043] Select, from all the spherical measurement extreme points, the spherical measurement minimum value points whose extreme value direction is the minimum value and whose intensity is less than or equal to the average intensity;

[0044] Form a spherical convex hull region by the convex hull of each of the spherical measurement maximum value points and each of the spherical measurement minimum value points.

[0045] In a possible implementation manner, the apparatus is further configured to:

[0046] Construct the convex hull of the union region;

[0047] Obtain the vertices of each of the convex hulls and the target number of adjacent points corresponding to each vertex;

[0048] Based on the target number of adjacent points, perform curve fitting on each vertex to fit the outer contour of the union region into a smooth contour.

[0049] In another aspect, a computer device is provided. The computer device includes a processor and a memory. At least one instruction is stored in the memory, and the at least one instruction is loaded and executed by the processor to implement a medical image segmentation method as described above.

[0050] In yet another aspect, a computer-readable storage medium is provided. At least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement a medical image segmentation method as described above.

[0051] The technical solution provided by this application may include the following beneficial effects:

[0052] The present application discloses a method for segmenting medical images. After obtaining the manually clicked positions corresponding to the target medical images, multiple equally angled rays are obtained outward with the clicked positions as the centers, and the distance measurement values of equally spaced voxels on each ray are obtained. Based on the distance measurement values, multiple segmentations are performed with the manually clicked positions as the centers. Based on simply manually clicking positions, the present application uses an unsupervised and untrained intelligent method to maximize the automatic recognition and segmentation of the edges of the clicked regions, significantly improving the efficiency of manual segmentation and retaining the high compatibility and high accuracy of manual segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0054] Figure 1 is a schematic structural diagram of a medical image segmentation system shown according to an exemplary embodiment.

[0055] Figure 2 is a method flowchart of a medical image segmentation method shown according to an exemplary embodiment.

[0056] Figure 3 is a method flowchart of a medical image segmentation method shown according to an exemplary embodiment.

[0057] Figure 4 shows a schematic diagram of concentric sphere segmentation involved in an embodiment of the present application.

[0058] Figure 5 is a block diagram of the structure of a medical image segmentation device shown according to an exemplary embodiment.

[0059] Figure 6 shows a block diagram of the structure of a computer device shown according to an exemplary embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] The following will clearly and completely describe the technical solutions of the present application with reference to the drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts fall within the scope of protection of the present application.

[0061] It should be understood that the "indication" mentioned in the embodiments of the present application can be a direct indication, an indirect indication, or a representation of an associated relationship. For example, A indicates B, which can mean that A directly indicates B, for example, B can be obtained through A; it can also mean that A indirectly indicates B, for example, A indicates C and B can be obtained through C; it can also mean that there is an associated relationship between A and B.

[0062] In the description of the embodiments of the present application, the term "corresponding" can indicate a direct or indirect corresponding relationship between two parties, can also indicate an associated relationship between the two parties, or can be relationships such as indication and being indicated, configuration and being configured.

[0063] Figure 1 It is a schematic structural diagram of a medical image segmentation system shown according to an exemplary embodiment. The segmentation system includes a server 110 and a medical image acquisition device 120.

[0064] Optionally, the medical image acquisition device 120 can be Computed Tomography (CT), Nuclear magnetic resonance, and can be used to acquire three-dimensional medical images or two-dimensional medical images.

[0065] Optionally, the medical image acquisition device 120 is communicatively connected to the server 110 through a transmission network (such as a wireless communication network). The medical image acquisition device 120 can upload the acquired medical images to the server 110 through the wireless communication network so that the server 110 can process the acquired medical images.

[0066] Optionally, the server 110 can also establish a wireless communication connection with the medical image acquisition device 120 through the wireless communication network. The server 110 can be a server cluster or a distributed system composed of multiple physical servers, or can also be a cloud server that provides technical computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0067] Optionally, the system can also include a management device, which is used to manage the system (such as managing the connection status between each module in the medical image acquisition device 120 and the server), and the management device is connected to the server through a communication network. Optionally, the communication network is a wired network or a wireless network.

[0068] Optionally, the above-mentioned wireless network or wired network uses standard communication technologies and / or protocols. The network is usually the Internet, but it can also be any other network, including but not limited to any combination of local area networks, metropolitan area networks, wide area networks, mobile, limited or wireless networks, private networks or virtual private networks. In some embodiments, technologies and / or formats including Hypertext Markup Language, Extensible Markup Language, etc. are used to represent the data exchanged through the network. In addition, conventional encryption technologies such as Secure Sockets Layer, Transport Layer Security, Virtual Private Network, Internet Protocol Security, etc. can be used to encrypt all or some of the links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above data communication technologies.

[0069] Figure 2 is a method flowchart of a method for segmenting medical images shown according to an exemplary embodiment. This method is executed by a computer device, and the computer device can be a server 110 as shown in Figure 1 as shown in Figure 2 As shown, the method for segmenting medical images may include the following steps:

[0070] Step S201, obtain the click position of the target medical image.

[0071] In a possible implementation manner, first obtain the target medical image, which can be two-dimensional or three-dimensional, and then manually click on the internal area of the lesion of the target medical image. Correspondingly, when the computer device receives the manual click operation, it determines the click position corresponding to the manual click operation in the target medical image..

[0072] Wherein, the target medical image is any medical image to be segmented.

[0073] Step S202, taking the click position as the center, obtain multiple equally angled rays outward.

[0074] In a possible implementation manner, after obtaining the click position, take the click position as the center and emit multiple rays outward at a uniform angle (i.e., equally angled).

[0075] Step S203, obtain the distance measurement values of equally spaced voxels on each ray, and form a set of measurement value vectors with all the distance measurement values on the same ray.

[0076] In a possible implementation manner, after emitting each ray, collect equally spaced voxels on each ray, obtain the distance measurement values corresponding to each equally spaced voxel on each ray, form a set of measurement value vectors with all the distance measurement values on the same ray, and obtain multiple sets of measurement value vectors. The number of sets of measurement value vectors is equal to the number of emitted rays.

[0077] Step S204: Obtain the measure extreme points in each group that are closest to the click position among the measure value vectors, and connect the convex hulls of these measure extreme points to obtain the target connection region.

[0078] In a possible implementation manner, after obtaining the measure value vectors of each group, select the distance measure value in each group of measure value vectors that is closest to the click position, and determine it as the measure extreme point. Then obtain the convex hulls of each measure extreme point, connect the vertices of each convex hull, and obtain the target connection region.

[0079] Step S205: Obtain the upper density limit value and the lower density limit value of the target connection region, and based on the upper density limit value and the lower density limit value, segment the target connection region to obtain the target segmentation region.

[0080] In a possible implementation manner, after obtaining the target connection region, calculate the upper density limit value and the lower density limit value of the target connection region based on the average voxel intensity and the standard deviation of voxel intensity of the target connection region. Taking the click position as the center and using the upper density limit value and the lower density limit value as parameters, segment the target connection region to obtain the target segmentation region.

[0081] Step S206: Taking the direction of the connection lines between the click position and the convex hulls of the measure extreme points as the extension direction, extend the lengths of each connection line outward to form a spherical region centered on the click position.

[0082] In a possible implementation manner, after obtaining the target segmentation region, taking the click position as the center and using the direction of the connection lines between the click position and the convex hulls of the measure extreme points as the extension direction, extend the lengths of each connection line by a target length multiple to form a spherical region centered on the click position.

[0083] Among them, the target length multiple can be preset according to specific needs.

[0084] Step S207: Divide the radius of the spherical region equally to form multiple concentric spheres.

[0085] In a possible implementation manner, after obtaining the spherical region, taking the click position as the center, divide the radius of the spherical region equally to form multiple concentric spheres.

[0086] Step S208: Obtain the spherical measure extreme points of each concentric sphere, and form a spherical convex hull region with the convex hulls of these spherical measure extreme points.

[0087] In a possible implementation, after dividing multiple concentric spheres, calculate the distance measure values of each concentric spherical surface along each concentric spherical surface, obtain the spherical measure extreme points of each concentric sphere based on the distance measure values of each concentric spherical surface, and then obtain the convex hulls of each spherical measure extreme point. The vertices of each convex hull form a spherical convex hull region.

[0088] Step S209, obtain the union region of the target connection region, the target segmentation region, and the spherical convex hull region, and determine the union region as the target segmentation result of the target medical image.

[0089] In a possible implementation, after obtaining the spherical convex hull region, obtain the union region of the target connection region, the target segmentation region, and the spherical convex hull region. The union region is the segmentation result of the internal region of the selected lesion in the target medical image (i.e., the tissue region corresponding to the clicked position).

[0090] In summary, the present application discloses a method for segmenting medical images. After obtaining the artificial click position corresponding to the target medical image, taking the click position as the center, obtaining multiple equiangular rays outward, and obtaining the distance measure values of equidistant voxels on each ray. Based on the distance measure values, taking the artificial click position as the center, perform multiple segmentations. Based on a simple manual click position, the present application uses an unsupervised and non-training intelligent method to maximize the automatic recognition and segmentation of the edge of the clicked area, significantly improving the efficiency of manual segmentation and retaining the high compatibility and high accuracy of manual segmentation.

[0091] Figure 3 is a flowchart of a method for segmenting a medical image shown according to an exemplary embodiment. This method is executed by a computer device, and the computer device can be a data processing device in a target vehicle as shown in Figure 1 shown. As shown in Figure 3 shown, the image data processing method may include the following steps:

[0092] Step S301, obtain the click position of the target medical image.

[0093] In a possible implementation, first obtain the target medical image to be segmented. The target medical image can be Computed Tomography (CT), or it can be Nuclear Magnetic Resonance. It can be a two-dimensional medical image or a three-dimensional medical image. When the above target medical image is a two-dimensional image, the corresponding image element is a pixel. When the above target medical image is a three-dimensional image, the corresponding image element is a voxel.

[0094] After obtaining the target medical image to be segmented, manually click on the internal area of the lesion (or the tissue or organ to be segmented) of the target medical image, and obtain the click position.

[0095] Optionally, when manually clicking on the internal area of the lesion of the target medical image, in order to facilitate segmentation, try to click on the center position of the internal area of the lesion.

[0096] Step S302: Taking the click position as the center, obtain multiple equiangular rays outward.

[0097] In a possible implementation, after obtaining the click position, taking the click position as the center, emit multiple rays outward at a uniform angle (i.e., equiangular). Exemplarily, the uniform angle can be 30°.

[0098] Step S303: Obtain the distance measure values of the equidistant voxels on each ray, and form all the distance measure values on the same ray into a set of measure value vectors.

[0099] In a possible implementation, determine the equidistant voxels on each ray as sampling points, and obtain the density values corresponding to each of these sampling points;

[0100] Obtain an operator of the target size, and based on the operator and the density value, obtain the distance measure values of each sampling point.

[0101] Further, after emitting each ray, collect the equidistant voxels on each ray, determine the equidistant voxels on each ray as sampling points, and based on the voxel value and standard deviation corresponding to each ray, obtain the density values corresponding to each equidistant voxel (i.e., the above sampling points) on each ray.

[0102] After obtaining each density value, set an operator of the target size (the core position of each operator corresponds to each sampling point), and use a distance measure related to the density value (such as the Euclidean distance or Chebyshev distance of the density) to calculate the distance measure value of the sampling point corresponding to each operator. Among them, the distance measure values of the sampling points on each ray form a set of vectors, obtain multiple sets of measure value vectors, and the number of sets of measure value vectors is equal to the number of emitted rays.

[0103] Optionally, exemplarily, the equidistant voxels can be 5 voxels, and the target size can be preset as needed, such as 3×3; 5×5.

[0104] Optionally, the operator has two parameters: size and calculation direction. The operator can be linear. For example, it calculates the density difference between every two adjacent sampling points on each ray starting from the click position and extending in the distal direction. The operator can also be a pattern type. For example, it calculates in the same direction as above, but calculates the mean value of adjacent points on adjacent rays. The combination form of the sampling points covered by the operator in the calculation can be diverse.

[0105] Optionally, for different tissues and organs, different operators can be used. When clicking, the operator combination can be automatically selected according to the clicked position (for example, when the medical image is of the head, a pattern-type operator is used; when it is of the liver, a linear operator is used, and it can be automatically adjusted according to the medical image information).

[0106] Step S304: Obtain the measure extreme points in each group of the measure value vectors that are closest to the click position, and connect the convex hulls of each of these measure extreme points to obtain the target connection region.

[0107] In a possible implementation, obtain the curve extreme values of each group of the measure value vectors;

[0108] Based on the curve extreme values, obtain the measure extreme points in each group of the measure value vectors that are closest to the click position;

[0109] Obtain the convex hulls of each of these measure extreme points, and connect each of these convex hulls to obtain the target connection region.

[0110] Furthermore, after obtaining each group of measure value vectors, obtain the curve extreme values of each group of the measure value vectors. Based on the curve extreme values, select the distance measure value in each group of measure value vectors that is closest to the click position, and determine it as the measure extreme point (please refer to part d in Figure 4 ), and form a set of measure extreme points with all the measure extreme points. Use the convex hull algorithm to obtain the convex hulls of each measure extreme point in the set of measure extreme points, and connect the vertices of each convex hull, thereby obtaining the target connection region (please refer to part e in Figure 4 ).

[0111] Among them, the measure extreme point can be a maximum value or a minimum value, and its typical form is the tip point of the curve extreme value drawn by the measure value vector.

[0112] Step S305: Obtain the density upper limit value and the density lower limit value of the target connection region, and based on the density upper limit value and the density lower limit value, segment the target connection region to obtain the target segmentation region.

[0113] In a possible implementation, obtain the voxel mean value and the voxel standard deviation of the target connection region;

[0114] Based on the voxel mean value and the voxel standard deviation, obtain the upper density limit value and the lower density limit value of the target connection region;

[0115] Taking the click position as the center, based on the upper density limit value and the lower density limit value, segment the target connection region to obtain the target segmentation region.

[0116] Further, after obtaining the target connection region, obtain the average voxel intensity and the voxel intensity standard deviation of the target connection region. Taking the click position as the center and using statistical information such as the average voxel intensity and the voxel intensity standard deviation as parameters, calculate the upper density limit value and the lower density limit value of the target connection region according to a specific distribution function (such as Gaussian distribution, uniform distribution).

[0117] Taking the click position as the center and using the upper density limit value and the lower density limit value as parameters, use the region growing algorithm to segment the target connection region to obtain the target segmentation region.

[0118] When calculating the upper density limit value and the lower density limit value, first use the distribution function combined with the statistical parameters to obtain the distribution of the density values, and then take the value range with a relatively large probability (such as 95%) as the upper density limit value and the lower density limit value.

[0119] Step S306: Taking the direction of the connection line between the click position and the convex hull of each measure extreme point as the extension direction, extend the length of each connection line outward to form a spherical region centered on the click position.

[0120] In a possible implementation manner, taking the direction of the connection line between the click position and the convex hull of each measure extreme point as the extension direction, extend the length of each connection line outward by a target multiple to form a spherical region centered on the click position.

[0121] Further, after obtaining the target segmentation region, taking the click position as the center and taking the direction of the connection line between the click position and the convex hull of each measure extreme point as the extension direction, extend the length of each connection line outward by a target length multiple to form a spherical region centered on the click position.

[0122] Among them, the target length multiple can be preset according to specific needs. Exemplarily, the target length multiple can be 0.5 times.

[0123] Step S307: Divide the radius of the spherical region into equal parts to form a plurality of concentric spheres.

[0124] In a possible implementation manner, after obtaining the spherical region, taking the click position as the center, divide the radius of the spherical region into equal parts to form a plurality of concentric spheres (please refer to Figure 4 the h part in).

[0125] Step S308: Obtain the spherical measure extreme points of each of the concentric spheres, and form a spherical convex hull region with the convex hull of these spherical measure extreme points.

[0126] In a possible implementation, obtain the spherical measure extreme points of each of the concentric spheres;

[0127] Obtain the extreme value directions and average intensities of each of these spherical measure extreme points;

[0128] From all these spherical measure extreme points, select the spherical measure maximum value points whose extreme value directions are maximum values and whose intensities are greater than or equal to the average intensity;

[0129] From all these spherical measure extreme points, select the spherical measure minimum value points whose extreme value directions are minimum values and whose intensities are less than or equal to the average intensity;

[0130] Form a spherical convex hull region with the convex hull of each of these spherical measure maximum value points and each of these spherical measure minimum value points.

[0131] Furthermore, after dividing into multiple concentric spheres, calculate the distance measure values of each concentric spherical surface along each concentric spherical surface, obtain the spherical measure extreme points of each concentric sphere based on the distance measure values of each concentric spherical surface, and then obtain the convex hull of each spherical measure extreme point. The vertices of each convex hull form a spherical convex hull region. Among them, there are multiple local spherical measure extreme points on the spherical surface or circle. For all points that form local spherical measure extreme points, calculate the extreme value directions and corresponding average intensities of all points, and select the top 50% of them for each direction (maximum value or minimum value) as the output spherical measure extreme points.

[0132] Step S309: Obtain the union region of the target connection region, the target segmentation region, and the spherical convex hull region, and determine this union region as the target segmentation result of the target medical image.

[0133] Furthermore, after obtaining the spherical convex hull region, obtain the union region of the target connection region, the target segmentation region, and the spherical convex hull region. This union region is the segmentation result of the internal region of the lesion selected in the target medical image (i.e., the tissue region corresponding to the clicked position). Based on the manually simple clicked position, an unsupervised and non-training intelligent method is used to maximally realize the automatic recognition and segmentation of the edge of the selected region.

[0134] Step S310: Construct the convex hull of this union region, and obtain the vertices of each convex hull and the adjacent points corresponding to the target quantity of each vertex.

[0135] Step S311: Based on the adjacent points corresponding to the target quantity, perform curve fitting on each vertex to fit the outer contour of this union region into a smooth contour.

[0136] In a possible implementation, after obtaining the union region, first construct a fine convex hull of the union region. Then, for the points on the convex hull surface (i.e., the vertices of the convex hull), use the method of curve fitting (such as B-spline interpolation), combined with the density distribution map between adjacent points, to fit the adjacent points with at least three levels of distance around each vertex to form a smooth contour, and obtain the extraction result.

[0137] Among them, the level refers to the nearest 1 point, 2 points, and 3 points of each vertex, corresponding to level one, two, and three respectively.

[0138] In summary, the present application discloses a method for segmenting medical images. After obtaining the manually clicked positions corresponding to the target medical images, taking the clicked positions as the center, obtain multiple equiangular rays outward, and obtain the distance measurement values of the equidistant voxels on each ray. Based on the distance measurement values, take the manually clicked positions as the center and perform multiple segmentations. Based on a simple manual click position, the present application uses an unsupervised and non-training intelligent method to maximize the automatic recognition and segmentation of the edges of the clicked area, significantly improving the efficiency of manual segmentation and retaining the high compatibility and high accuracy of manual segmentation.

[0139] Figure 5 It is a structural block diagram of a medical image segmentation device shown according to an exemplary embodiment. The segmentation device includes:

[0140] A click position acquisition module 501, configured to acquire the click positions of the target medical images;

[0141] A ray acquisition module 502, configured to acquire multiple equiangular rays outward with the click position as the center;

[0142] A measurement value vector acquisition module 503, configured to acquire the distance measurement values of the equidistant voxels on each ray, and form a group of measurement value vectors with all the distance measurement values on the same ray;

[0143] A target connection region acquisition module 504, configured to acquire the measurement extreme points closest to the click position in each group of the measurement value vectors, and connect the convex hulls of the respective measurement extreme points to acquire a target connection region;

[0144] A target segmentation region acquisition module 505, configured to acquire the density upper limit value and the density lower limit value of the target connection region, and segment the target connection region based on the density upper limit value and the density lower limit value to acquire a target segmentation region;

[0145] A spherical region acquisition module 506, configured to extend each connection line length outward in the extension direction of the connection line between the click position and the convex hull of each measure extreme point, so as to form a spherical region centered on the click position;

[0146] A concentric sphere acquisition module 507, configured to equally divide the radius of the spherical region to form a plurality of concentric spheres;

[0147] A spherical convex hull region acquisition module 508, configured to acquire the spherical measure extreme points of each concentric sphere, and form a spherical convex hull region by the convex hull of each spherical measure extreme point;

[0148] A target segmentation result acquisition module 509, configured to acquire the union region of the target connection region, the target segmentation region and the spherical convex hull region, and determine the union region as the target segmentation result of the target medical image.

[0149] In a possible implementation manner, the measure value vector acquisition module 503 is further configured to:

[0150] Determine the equally spaced voxels on each ray as sampling points, and acquire the density values corresponding to each sampling point;

[0151] Acquire an operator for the target size, and acquire the distance measure value of each sampling point based on the operator and the density value.

[0152] In a possible implementation manner, the target connection region acquisition module 504 is further configured to:

[0153] Acquire the curve extreme values of each group of the measure value vectors;

[0154] Based on the curve extreme values, acquire the measure extreme points closest to the click position in each group of the measure value vectors;

[0155] Acquire the convex hull of each measure extreme point, and connect each convex hull to acquire a target connection region.

[0156] In a possible implementation manner, the target segmentation region acquisition module 505 is further configured to:

[0157] Acquire the voxel mean value and voxel standard deviation of the target connection region;

[0158] Based on the voxel mean value and voxel standard deviation, acquire the density upper limit value and density lower limit value of the target connection region;

[0159] Centered on the click position, based on the density upper limit value and density lower limit value, segment the target connection region to acquire a target segmentation region.

[0160] In a possible implementation, the spherical region acquisition module 506 is further configured to:

[0161] Taking the direction of the connection line between the click position and the convex hull of each measurement extreme point as the extension direction, extend the length of each connection line outward by a target multiple to form a spherical region centered on the click position.

[0162] In a possible implementation, the spherical convex hull region acquisition module 508 is further configured to:

[0163] Obtain the spherical measurement extreme points of each concentric sphere;

[0164] Obtain the extreme direction and average intensity of each spherical measurement extreme point;

[0165] From all the spherical measurement extreme points, select the spherical measurement maximum points whose extreme direction is the maximum value and whose intensity is greater than or equal to the average intensity;

[0166] From all the spherical measurement extreme points, select the spherical measurement minimum points whose extreme direction is the minimum value and whose intensity is less than or equal to the average intensity;

[0167] Construct a spherical convex hull region with the convex hull of each spherical measurement maximum point and each spherical measurement minimum point.

[0168] In a possible implementation, the device is further configured to:

[0169] Construct the convex hull of the union region;

[0170] Obtain the vertices of each convex hull and the target number of adjacent points corresponding to each vertex;

[0171] Based on the target number of adjacent points, perform curve fitting on each vertex to fit the outer contour of the union region into a smooth contour.

[0172] In summary, the present application discloses a medical image segmentation device. After obtaining the manually clicked position corresponding to the target medical image, taking the clicked position as the center, obtaining multiple equiangular rays outward, and obtaining the distance measurement values of equidistant voxels on each ray, and based on the distance measurement values, performing multiple segmentations with the manually clicked position as the center. Based on a simple manual click position, the present application uses an unsupervised and non-training intelligent method to maximize the automatic recognition and segmentation of the edge of the clicked area, significantly improving the efficiency of manual segmentation and retaining the high compatibility and high accuracy of manual segmentation.

[0173] Figure 6The block diagram of a computer device shown in an exemplary embodiment of the present application is presented. The computer device includes a memory and a processor. The memory is used to store a computer program, and when the computer program is executed by the processor, the above-mentioned method for segmenting medical images is implemented.

[0174] One embodiment of the present application further provides a computer storage medium, which is used to store a computer program, and when the computer program is executed by a processor, the above-mentioned method for segmenting medical images is implemented.

[0175] Among them, the processor can be a Central Processing Unit (CPU). The processor can also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. chips, or a combination of the above types of chips.

[0176] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the method in the embodiment of the present invention. The processor executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions, and modules stored in the memory, that is, the method in the above method embodiment is implemented.

[0177] The memory can include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created by the processor, etc. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory optionally includes a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0178] Those skilled in the art can understand that to implement all or part of the processes in the above-described implementation methods, it can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the implementation methods of the above-mentioned various methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memories.

[0179] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for segmenting medical images, characterized in that, The method includes: Obtaining the click position of the target medical image; Taking the click position as the center and obtaining multiple equally angled rays outward; Obtaining the distance measure values of equally spaced voxels on each ray, and forming a set of measure value vectors with all the distance measure values on the same ray; Obtaining the measure extreme points closest to the click position for each group of the measure value vectors, and connecting the convex hulls of the respective measure extreme points to obtain a target connection region; Obtaining the density upper limit value and the density lower limit value of the target connection region, and segmenting the target connection region based on the density upper limit value and the density lower limit value to obtain a target segmentation region; Taking the connection direction between the click position and the convex hull of the respective measure extreme points as the extension direction, and extending the lengths of the respective connections outward to form a spherical region centered on the click position; Dividing the radius of the spherical region equally to form multiple concentric spheres; Obtaining the spherical measure extreme points of the respective concentric spheres, and forming a spherical convex hull region with the convex hulls of the respective spherical measure extreme points; Obtaining the union region of the target connection region, the target segmentation region, and the spherical convex hull region, and determining the union region as the target segmentation result of the target medical image; The obtaining the distance measure values of equally spaced voxels on each ray includes: Determining the equally spaced voxels on each ray as sampling points, and obtaining the density values corresponding to the respective sampling points; Obtaining an operator of a target size, and obtaining the distance measure values of the respective sampling points based on the operator and the density values; The taking the connection direction between the click position and the convex hull of the respective measure extreme points as the extension direction, and extending the lengths of the respective connections outward to form a spherical region centered on the click position includes: Taking the connection direction between the click position and the convex hull of the respective measure extreme points as the extension direction, and extending the lengths of the respective connections outward by a target multiple to form a spherical region centered on the click position; The obtaining the spherical measure extreme points of the respective concentric spheres, and forming a spherical convex hull region with the convex hulls of the respective spherical measure extreme points includes: Obtaining the spherical measure extreme points of the respective concentric spheres; Obtaining the extreme direction and the average intensity of the respective spherical measure extreme points; Selecting, from all the spherical measure extreme points, the spherical measure maximum points whose extreme direction is the maximum value and whose intensity is greater than or equal to the average intensity; Selecting, from all the spherical measure extreme points, the spherical measure minimum points whose extreme direction is the minimum value and whose intensity is less than or equal to the average intensity; Forming a spherical convex hull region with the convex hulls of the respective spherical measure maximum points and the respective spherical measure minimum points.

2. The method according to claim 1, wherein The obtaining the measure extreme points closest to the click position for each group of the measure value vectors, and connecting the convex hulls of the respective measure extreme points to obtain a target connection region includes: Obtaining the curve extremes of each group of the measure value vectors; Based on the curve extremes, obtaining the measure extreme points closest to the click position for each group of the measure value vectors; Obtain the convex hulls of the respective measure extreme points, and connect the respective convex hulls to obtain a target connection region.

3. The method according to claim 1, characterized in that, The obtaining of the upper density limit value and the lower density limit value of the target connection region, and based on the upper density limit value and the lower density limit value, segmenting the target connection region to obtain a target segmentation region, includes: Obtain the voxel mean value and the voxel standard deviation of the target connection region; Based on the voxel mean value and the voxel standard deviation, obtain the upper density limit value and the lower density limit value of the target connection region; With the click position as the center, based on the upper density limit value and the lower density limit value, segment the target connection region to obtain a target segmentation region.

4. The method according to claim 1, characterized in that, Before determining the union region as the target segmentation result of the target medical image, the method further includes: Construct the convex hull of the union region; Obtain the vertices of each convex hull and the corresponding target number of adjacent points of each vertex; Based on the target number of adjacent points, perform curve fitting on each vertex to fit the outer contour of the union region into a smooth contour.

5. A segmentation device for medical images, characterized in that, The device includes: A click position acquisition module, configured to acquire the click position of a target medical image; A ray acquisition module, configured to outwardly acquire a plurality of equiangular rays with the click position as the center; A measure value vector acquisition module, configured to acquire the distance measure values of equidistant voxels on each ray, and form a group of measure value vectors with all the distance measure values on the same ray; A target connection region acquisition module, configured to acquire the measure extreme points closest to the click position in each group of measure value vectors, and connect the convex hulls of the respective measure extreme points to obtain a target connection region; A target segmentation region acquisition module, configured to acquire the upper density limit value and the lower density limit value of the target connection region, and based on the upper density limit value and the lower density limit value, segment the target connection region to obtain a target segmentation region; A spherical region acquisition module, configured to extend each connection line length outward in the extension direction of the connection line between the click position and the convex hull of the respective measure extreme points to form a spherical region centered on the click position; A concentric sphere acquisition module, configured to equally divide the radius of the spherical region to form a plurality of concentric spheres; A spherical convex hull region acquisition module, configured to acquire the spherical measure extreme points of each concentric sphere, and form a spherical convex hull region with the convex hulls of the respective spherical measure extreme points; A target segmentation result acquisition module, configured to acquire the union region of the target connection region, the target segmentation region, and the spherical convex hull region, and determine the union region as the target segmentation result of the target medical image; The measure value vector acquisition module is further configured to: Determine the equidistant voxels on each ray as sampling points, and acquire the density values corresponding to the respective sampling points; Acquire an operator of a target size, and based on the operator and the density value, acquire the distance measure values of the respective sampling points; The spherical region acquisition module is further configured to: Taking the direction of the connection line between the click position and the convex hull of each measurement extreme point as the extension direction, extend the length of each connection line outward by a target multiple to form a spherical region centered on the click position; The spherical convex hull region acquisition module is further configured to: Obtain the spherical measurement extreme points of each concentric sphere; Obtain the extreme value direction and average intensity of each of the spherical measurement extreme points; From all the spherical measurement extreme points, select the spherical measurement maximum value points whose extreme value direction is the maximum value and whose intensity is greater than or equal to the average intensity; From all the spherical measurement extreme points, select the spherical measurement minimum value points whose extreme value direction is the minimum value and whose intensity is less than or equal to the average intensity; Form a spherical convex hull region by the convex hull of each of the spherical measurement maximum value points and each of the spherical measurement minimum value points.

6. A computer device, characterized in that, The computer device includes a processor and a memory, and at least one instruction is stored in the memory, and the at least one instruction is loaded and executed by the processor to implement a method for segmenting a medical image according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, At least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement a method for segmenting a medical image according to any one of claims 1 to 4.

Images