Method and apparatus for segmenting inferior mesenteric artery

Abstract

This invention provides a method and apparatus for segmenting the inferior mesenteric artery. The method includes: converting the pixel values ​​of each peritoneal imaging image in a sequence of peritoneal imaging images into CT values ​​to obtain an original tensor matrix; inputting the original tensor matrix into a segmentation model to obtain an image of the inferior mesenteric artery output by the segmentation model. The method and apparatus for segmenting the inferior mesenteric artery provided by this invention can perform coarse-grained segmentation with relatively messy tissue features, and after refining the initially segmented tissue features, perform fine-grained segmentation to extract the vascular morphology of the inferior mesenteric artery separately, thereby improving the accuracy and efficiency of inferior mesenteric artery segmentation.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a method and apparatus for segmenting the inferior mesenteric artery. Background Technology

[0002] The course of the inferior mesenteric artery has long been a focus of attention during abdominal surgery. However, this task has long relied on manual observation in imaging workstations, with manual classification and recording. Moreover, due to limitations in imaging software, it is impossible to extract the model of the inferior mesenteric artery independently, resulting in low accuracy and efficiency in extracting the morphology of the inferior mesenteric artery. Summary of the Invention

[0003] This invention provides a method and apparatus for segmenting the inferior mesenteric artery, which addresses the shortcomings of the prior art in terms of the low accuracy and efficiency of inferior mesenteric artery morphology extraction.

[0004] This invention provides a method for segmenting the inferior mesenteric artery, comprising:

[0005] The pixel values ​​of each laparography image in the laparography image sequence are converted into CT values ​​to obtain the original tensor matrix; wherein the arrangement order of the laparography image sequence is the same as the arrangement order of the slices in the original tensor matrix.

[0006] The original tensor matrix is ​​input into the segmentation model to obtain the inferior mesenteric artery image output by the segmentation model;

[0007] The segmentation model is trained based on the original tensor matrix of the sample and the corresponding arterial region labels labeled on the original tensor matrix of the sample; the segmentation model includes:

[0008] The first segmentation layer is used to perform region growing on the original tensor matrix to obtain the first tensor matrix;

[0009] A feature extraction layer is used to determine a second tensor matrix based on the first tensor matrix and the original tensor matrix;

[0010] The second segmentation layer is used to perform region growing on the second tensor matrix to obtain the image of the inferior mesenteric artery.

[0011] According to a method for segmenting the inferior mesenteric artery provided by the present invention, the step of performing region growing on the original tensor matrix to obtain a first tensor matrix includes:

[0012] The inferior mesenteric artery is located in the first slice of the original tensor matrix to obtain the target coordinates;

[0013] Using the target coordinates as the first original seed, the original tensor matrix is ​​processed using a three-dimensional region growing algorithm to obtain the first tensor matrix.

[0014] According to a method for segmenting the inferior mesenteric artery provided by the present invention, the inferior mesenteric artery is located in the first slice of the original tensor matrix to obtain the target coordinates, including:

[0015] After binarizing the first layer slice, erosion was performed using a circular mask to determine the inferior mesenteric artery region.

[0016] The target coordinates are determined by weighted averaging of the pixel coordinates contained in the inferior mesenteric artery region.

[0017] The size of the circular mask matches the size of the first layer slice.

[0018] According to a method for segmenting the inferior mesenteric artery provided by the present invention, the step of determining a second tensor matrix based on a first tensor matrix and the original tensor matrix includes:

[0019] A morphological opening operation is performed on the original tensor matrix after binarization to obtain a third tensor matrix;

[0020] The second tensor matrix is ​​determined based on the first tensor matrix and the third tensor matrix;

[0021] The third tensor matrix contains physiological tissue features from the original tensor matrix, excluding the inferior mesenteric artery.

[0022] According to a method for segmenting the inferior mesenteric artery provided by the present invention, the step of determining the second tensor matrix based on the first tensor matrix and the third tensor matrix includes:

[0023] Perform a matrix inner product operation on the first tensor matrix and the third tensor matrix to determine the fourth tensor matrix;

[0024] Perform matrix subtraction on the first tensor matrix and the fourth tensor matrix to determine the second tensor matrix;

[0025] The fourth tensor matrix contains the physiological tissue features shared by the first tensor matrix and the third tensor matrix.

[0026] According to a method for segmenting the inferior mesenteric artery provided by the present invention, the step of performing region growing on the second tensor matrix to obtain an image of the inferior mesenteric artery includes:

[0027] Based on the second original seed, the second tensor matrix is ​​processed using a three-dimensional region growing algorithm to obtain the image of the inferior mesenteric artery;

[0028] The second original seed is determined based on the second tensor matrix.

[0029] The present invention also provides a device for segmenting the inferior mesenteric artery, comprising:

[0030] The preprocessing module is used to convert the pixel values ​​of each laparography image in the laparography image sequence into CT values ​​to obtain the original tensor matrix; wherein the arrangement order of the laparography image sequence is the same as the arrangement order of the slices in the original tensor matrix.

[0031] The segmentation module is used to input the original tensor matrix into the segmentation model to obtain the inferior mesenteric artery image output by the segmentation model;

[0032] The segmentation model is trained based on the original tensor matrix of the sample and the corresponding arterial region labels labeled on the original tensor matrix of the sample; the segmentation model includes:

[0033] The first segmentation layer is used to perform region growing on the original tensor matrix to obtain the first tensor matrix;

[0034] A feature extraction layer is used to determine a second tensor matrix based on the first tensor matrix and the original tensor matrix;

[0035] The second segmentation layer is used to perform region growing on the second tensor matrix to obtain the image of the inferior mesenteric artery.

[0036] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for segmenting the inferior mesenteric artery as described above.

[0037] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for segmenting the inferior mesenteric artery as described above.

[0038] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the method for segmenting the inferior mesenteric artery as described above.

[0039] The present invention provides a method and apparatus for segmenting the inferior mesenteric artery. Based on a sequence of peritoneal angiography images at different torsions, the image is converted into a raw tensor matrix using medical image processing techniques. This raw tensor matrix is ​​then used as input to a segmentation model. A first segmentation layer aggregates and grows pixels with inferior mesenteric artery features within the raw tensor matrix. A feature extraction layer then performs digital subtraction on the initially segmented first tensor matrix. Finally, a second segmentation layer performs secondary segmentation on the subtracted and refined second tensor matrix, extracting an image of the inferior mesenteric artery containing only its features. This method enables coarse-grained segmentation of complex tissue features, followed by fine-grained segmentation after refining the initially segmented tissue features, thus extracting the vascular morphology of the inferior mesenteric artery and improving the accuracy and efficiency of inferior mesenteric artery segmentation. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0041] Figure 1 This is one of the flowcharts illustrating the method for segmenting the inferior mesenteric artery provided by the present invention;

[0042] Figure 2 This is a schematic diagram illustrating the principle of the three-dimensional region growing algorithm provided by this invention;

[0043] Figure 3 This is a flowchart of the target coordinate positioning process provided by the present invention;

[0044] Figure 4 This is the second flowchart of the method for segmenting the inferior mesenteric artery provided by the present invention;

[0045] Figure 5 This is a schematic diagram of the structure of the inferior mesenteric artery segmentation device provided by the present invention;

[0046] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0048] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more.

[0049] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used herein, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” are intended to include the plural forms.

[0050] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0051] Figure 1 This is one of the flowcharts illustrating the method for segmenting the inferior mesenteric artery provided by this invention. For example... Figure 1 As shown, the method for segmenting the inferior mesenteric artery provided in this embodiment of the invention includes: step 101, converting the pixel values ​​of each peritoneal imaging image in the peritoneal imaging image sequence into CT values ​​to obtain the original tensor matrix.

[0052] The order of the laparoscopic imaging sequence is the same as the order of the slices in the original tensor matrix.

[0053] It should be noted that the execution subject of the inferior mesenteric artery segmentation method provided in this embodiment of the invention is the inferior mesenteric artery segmentation device.

[0054] The submesenteric artery segmentation method provided in this application embodiment is applicable to users processing abdominal angiography images using electronic devices to extract the submesenteric artery vascular model separately.

[0055] The aforementioned electronic devices can be implemented in various forms. For example, the electronic devices described in the embodiments of this application may include mobile terminals such as mobile phones, smartphones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), navigation devices, smart bracelets, smartwatches, etc., as well as fixed terminals such as digital TVs, desktop computers, etc. Hereinafter, it is assumed that the electronic device is a mobile terminal. However, those skilled in the art will understand that, in addition to elements specifically designed for mobile purposes, the construction according to the embodiments of this application can also be applied to fixed-type terminals.

[0056] It should be noted that a laparoscopic image sequence refers to a collection of images of multiple tomographic planes formed by reconstructing images based on the attenuation law of X-rays after scanning a part of the human body with different tissue densities (such as iterative method or back projection method).

[0057] The laparography images included in the laparography image sequence include, but are not limited to, X-ray films, computed tomography (CT) images, and magnetic resonance imaging (MRI) images.

[0058] Preferably, the abdominal imaging sequence can be composed of enhanced CT images formed by injecting a contrast agent into a vein during scanning, and the contrast agent enhancing areas rich in blood supply through the blood circulation system.

[0059] Specifically, in step 101, the inferior mesenteric artery segmentation device unifies the size of individual peritoneal images in the peritoneal imaging image sequence, converts the pixel values ​​in each peritoneal imaging image into raw CT values, and converts them into slices in the raw tensor matrix in the order of the images in the sequence.

[0060] The CT value can be characterized by the Hounsfield Unit (HU) value, which reflects the degree of absorption of X-rays emitted by the imaging equipment by the abdominal tissue. Water's absorption is used as a reference; water has a HU value of 0, values ​​with attenuation coefficients greater than water are positive, and values ​​with coefficients less than water are negative. The upper and lower limits for the HU values ​​are defined as those for bone cortex and air, respectively. The formula for calculating the HU value is as follows:

[0061] HU=pixel_val*slope+intercept

[0062] Where pixel_val is the pixel value of a certain pixel in the laparoscopic image, and HU is the HU value (i.e., CT value) corresponding to that pixel. slope is the rescaling slope involved in the mapping process, and intercept is the rescaling intercept involved in the mapping process. slope and intercept are determined by the parameters of the imaging equipment.

[0063] Step 102: Input the original tensor matrix into the segmentation model to obtain the inferior mesenteric artery image output by the segmentation model.

[0064] The segmentation model is trained based on the original tensor matrix of the samples and the corresponding arterial region labels. The segmentation model includes:

[0065] The first segmentation layer is used to perform region growing on the original tensor matrix to obtain the first tensor matrix.

[0066] The feature extraction layer is used to determine the second tensor matrix based on the first tensor matrix and the original tensor matrix.

[0067] The second segmentation layer is used to perform region growing on the second tensor matrix to obtain images of the inferior mesenteric artery.

[0068] It should be noted that the segmentation model can be an artificial intelligence model, and the embodiments of the present invention do not specifically limit the model type.

[0069] For example, the segmentation model can be a neural network. The structure and parameters of the neural network include, but are not limited to, the number of input layers, hidden layers, and output layers, as well as the weight parameters of each layer. This invention does not specifically limit the type and structure of the neural network.

[0070] For example, a segmentation model can be a feedforward neural network, which consists of an input layer, hidden layers, and an output layer, wherein:

[0071] The input layer is at the very front of the entire network, directly receiving the original tensor matrix corresponding to the laparoscopic imaging image sequence.

[0072] Hidden layers can have one or more layers. They perform weighted summations on the input vector using their own neurons. The calculation formula can be expressed as:

[0073] z = b + w1*x1 + w2*x2 + ... + wm*xm

[0074] Where z is the sum of weights of the hidden layer output, x1, x2, x3...xm are the m feature vectors of each sample, b is the bias, and w1, w2...wm are the weights corresponding to each feature vector.

[0075] The output layer is the last layer, used to decode the vector obtained after weighted summation and output the segmented image containing the inferior mesenteric artery.

[0076] It should be noted that the sample data includes the original sample tensor matrix corresponding to the sample data, as well as artery region labels pre-annotated on the coordinates corresponding to the artery regions in the original sample tensor matrix. The sample data is divided into training and test sets according to a certain ratio.

[0077] For example, the ratio of training set to test set in sample data includes, but is not limited to, 9:1, 8:2, etc., and the embodiments of the present invention do not specifically limit this.

[0078] Specifically, in step 102, the inferior mesenteric artery segmentation device initializes the weight coefficients between each layer of the constructed segmentation model, then inputs a set of sample data from the training set into the neural network with the current weight coefficients, and sequentially calculates the output of each node in the input layer, hidden layer, and output layer. The cumulative error between the final output result of the output layer and its actual connection position state type is used to correct the weight coefficients between each node in the input layer and hidden layer according to the gradient descent method. Following the above process, until all samples in the training set are traversed, the weight coefficients of the input layer and hidden layer can be obtained.

[0079] The segmentation device for the inferior mesenteric artery restores the segmentation model in step 102 based on the weight coefficients of the input layer and hidden layer of the neural network, and inputs each set of original tensor matrices in the test set into the trained segmentation model to obtain the corresponding segmentation results.

[0080] The embodiments of the present invention do not specifically limit the form of the segmentation result.

[0081] For example, the segmentation result can be a three-dimensional matrix with the same width and height as the original tensor matrix, and any value in the array can take the range [0, 1]. Here, 0 means that the coordinate point is not in the inferior mesenteric artery region, and 1 means that the coordinate point is in the inferior mesenteric artery region. Furthermore, the location information of the coordinate point in the three-dimensional array can be used to determine which areas in the angiography are inferior mesenteric arteries.

[0082] For example, the segmentation result can be a statistical value, which can be used to explain whether the inferior mesenteric artery exists in the angiography and which areas are inferior mesenteric artery vessels by statistically analyzing the values ​​of all coordinate points.

[0083] The embodiments of the present invention do not impose specific limitations on the segmentation model.

[0084] For example, the segmentation model consists of an input layer, a hidden layer, and an output layer. The hidden layer's function is to extract features from the input 3D angiography image using its own neurons, extracting feature information that is helpful in identifying the inferior mesenteric artery.

[0085] Preferably, the hidden layer comprises at least three layers: a first segmentation layer, a feature extraction layer, and a second segmentation layer, wherein:

[0086] The first segmentation layer, based on predefined criteria, starts from the growth point of the original tensor matrix (the growth point can be a single pixel or a small region formed by multiple pixels), and merges adjacent pixels or regions with similar properties with the growth point. These new pixels are then treated as new growth points, and the above process is repeated until no new growth points are generated. The set of all growth points is then taken as the result of coarse-grained segmentation, i.e., the first tensor matrix.

[0087] The first tensor matrix is ​​separated from the original tensor matrix and mainly includes the physiological tissue characteristics of the aorta, inferior mesenteric artery, and even some vertebrae.

[0088] The feature extraction layer can extract and fuse features from the original tensor matrix containing complete physiological tissue features and the first tensor matrix containing partial physiological tissue features to obtain the second tensor matrix.

[0089] The second tensor matrix includes only the physiological and tissue features of the inferior mesenteric artery and other minor defective areas.

[0090] The second segmentation layer can optimize the segmentation of the second tensor matrix, shielding the image features from the interference of flawed regions. After performing a region growing operation, it segments the image to include only the physiological tissue features corresponding to the inferior mesenteric artery.

[0091] Among them, the image of the inferior mesenteric artery can be used to perform three-dimensional reconstruction of the inferior mesenteric artery, so as to guide users of the inferior mesenteric artery segmentation device or medical workers to classify it into its corresponding vascular type, and provide targeted data support for subsequent treatment.

[0092] This invention utilizes laparoscopic image sequences from different torsions, converting them into raw tensor matrices using medical image processing techniques. These raw tensor matrices are then used as input to a segmentation model. A first segmentation layer aggregates and grows pixels with inferior mesenteric artery features within the raw tensor matrix. A feature extraction layer then performs digital subtraction on the initially segmented tensor matrix. Finally, a second segmentation layer refines the subtracted tensor matrix, resulting in a secondary segmentation that extracts only the inferior mesenteric artery features. This approach enables coarse-grained segmentation of complex tissue features, followed by fine-grained segmentation after refining the initially segmented tissue features. This allows for the separate extraction of the inferior mesenteric artery's vascular morphology, improving the accuracy and efficiency of inferior mesenteric artery segmentation.

[0093] Based on any of the above embodiments, a region growing process is performed on the original tensor matrix to obtain a first tensor matrix, including: locating the inferior mesenteric artery in the first slice of the original tensor matrix to obtain the target coordinates.

[0094] Specifically, in step 102, the segmentation device for the inferior mesenteric artery can process the first slice in the original tensor matrix corresponding to the first abdominal angiography image, using the theory that the abdominal aorta usually shows a high HU value in medical angiography as a positioning condition, and outputting the coordinates of the points that meet the condition as the target coordinates.

[0095] The target coordinates can be one or more, and the target coordinates are used to define the inferior mesenteric artery region.

[0096] Using the target coordinates as the first original seed, the original tensor matrix is ​​processed using a 3D region growing algorithm to obtain the first tensor matrix.

[0097] It should be noted that the first original seed refers to the original seed set in the first segmentation layer regarding the regional growth strategy.

[0098] Specifically, the segmentation device for the inferior mesenteric artery sets the target coordinates as the first original seed for the initial segmentation of the original tensor matrix, and triggers the 3D-Region Grow Algorithm (3D-RGA) to calculate the HU gradient value between the neighboring pixels of the first original seed and the first original seed. Pixels whose HU gradient values ​​meet the requirements are taken as new growth points. The process of discovering new growth points is iterated until no new growth points are found. The pixel value corresponding to the discovered growth point is set to 1, and the pixel value corresponding to the non-growth point is set to 0, so as to form the first tensor matrix.

[0099] The embodiments of the present invention do not specifically limit the execution process of the three-dimensional region growing algorithm.

[0100] For example, Figure 2 This is a schematic diagram illustrating the principle of the three-dimensional region growing algorithm provided by this invention. (For example...) Figure 2 As shown, taking the process of performing one iteration starting from the first original seed as an example, the implementation process of a three-dimensional region growing algorithm is given:

[0101] The segmentation device for the inferior mesenteric artery can select the remaining 26 pixels (excluding the first original seed) in a unit cubic region centered on the first original seed, and then evaluate each pixel based on iterative conditions:

[0102] (1) The HU value of the pixel is required to be less than 800 and greater than 100.

[0103] (2) The HU difference between adjacent pixels is required to be less than 75.

[0104] If a pixel in an adjacent cubic region satisfies both of the above conditions, then that pixel is set as a new growth point. Subsequent iterations follow the same principle and will not be elaborated here. Furthermore, pixels with a value of 1 in the resulting first tensor matrix are growth points, and pixels with a value of 0 are non-growth points.

[0105] This invention utilizes the absorption characteristics of the inferior mesenteric artery to contrast radiation. Based on the CT values ​​recorded in the first slice of the original tensor matrix, the inferior mesenteric artery region is initially located. The target coordinates obtained from this location are then used as the initial seed for three-dimensional region growing of the original tensor matrix, iteratively obtaining a first tensor matrix containing the inferior mesenteric artery region. This method enables the localization of the inferior mesenteric artery using contrast features and allows region growing to begin from pixels belonging to the inferior mesenteric artery region. This ensures that the physiological tissue features of the inferior mesenteric artery are not missed during coarse-grained segmentation, providing an accurate range for subsequent segmentation optimization.

[0106] Based on any of the above embodiments, the inferior mesenteric artery is located in the first layer slice of the original tensor matrix to obtain the target coordinates, including: binarizing the first layer slice and then using a circular mask for erosion to determine the inferior mesenteric artery region.

[0107] The size of the circular mask is matched with the size of the first layer slice.

[0108] Specifically, the segmentation device for the inferior mesenteric artery sets a preset threshold for the HU value based on the angiographic characteristics of the inferior mesenteric artery, and uses this preset threshold to perform binarization processing on the first slice. Then, it uses a circular mask with circular structuring elements to perform erosion operations on it, and outputs the foreground color region corresponding to the circular structuring elements as the inferior mesenteric artery region.

[0109] In this process, the size of the circular structural element in the circular mask is smaller than the size of the first layer slice, and is positively correlated with the size of the first layer slice.

[0110] The target coordinates are determined by weighted averaging of the pixel coordinates contained in the inferior mesenteric artery region.

[0111] Specifically, the segmentation device for the inferior mesenteric artery can perform a weighted average of the pixel coordinates contained in the region of the inferior mesenteric artery, and regard the calculated target coordinates as the center point of the inferior mesenteric artery.

[0112] For example, Figure 3 This is a flowchart illustrating the target coordinate positioning process provided by the present invention. For example... Figure 3 As shown in the figure, an embodiment of the present invention provides a specific implementation process for calculating target coordinates:

[0113] The first layer slice is binarized using a preset threshold T, and the binary image is labeled as B, as shown in the following formula:

[0114]

[0115] Where Ori(i,j) is the HU value of the pixel in the i-th row and j-th column of the first layer slice. Binary(i,j) is the binary value of the pixel in the i-th row and j-th column of the binary image B. This embodiment of the invention does not impose specific limitations on the value of the preset threshold T.

[0116] For example, based on a large amount of prior data, the preset threshold T can be set to 200.

[0117] Then, a circular mask C, obtained using a circular structuring element, is used to perform an erosion operation on the binary image B, acting as a filter to eliminate the background portion of the binary image B and retain only the foreground belonging to the inferior mesenteric artery. Finally, the average value of the pixel coordinates in the foreground portion of the circular region is used as the center coordinates of the inferior mesenteric artery, i.e., the target coordinates.

[0118] The size of the circular structural element in the circular mask C can be set manually or automatically according to the size of a single laparoscopic image.

[0119] For example, if the size of a single laparoscopic image is 512*512, the size of the circular structural element can be limited to between 10 and 35 pixels.

[0120] This invention employs a circular mask to perform erosion operations on the first binarized slice to determine the inferior mesenteric artery region. Then, the pixel coordinates within the inferior mesenteric artery region are used to calculate the target coordinates representing the center of the region. Subsequently, growth points are found by extending outwards from the center point to obtain the first tensor matrix. By performing morphological erosion operations on the binarized contrast image, the inferior mesenteric artery region in the image can be shrunk, eliminating errors caused by inaccurate region boundaries and maximizing the accuracy of coarse segmentation.

[0121] Based on any of the above embodiments, determining the second tensor matrix based on the first tensor matrix and the original tensor matrix includes: performing a morphological opening operation on the binarized original tensor matrix to obtain the third tensor matrix.

[0122] The third tensor matrix contains physiological tissue features from the original tensor matrix, excluding the inferior mesenteric artery.

[0123] Specifically, the segmentation device for the inferior mesenteric artery binarizes the original tensor matrix. Since the radius of the inferior mesenteric artery is relatively small compared to other tissues, when the binarized tensor matrix is ​​subjected to an opening operation of erosion followed by dilation, the features of the inferior mesenteric artery will be filtered out during the erosion process, while the adhesions of the aorta, vertebrae, and other tissues with obvious contours will be separated during the dilation process to obtain a third tensor matrix.

[0124] The third tensor matrix is ​​a binary matrix with the same size as the original tensor matrix. Pixels with a value of 1 correspond to tissues with more obvious anatomical features than the inferior mesenteric artery, while pixels with a value of 0 correspond to tissues with less obvious anatomical features than the inferior mesenteric artery.

[0125] The second tensor matrix is ​​determined based on the first and third tensor matrices.

[0126] Specifically, the segmentation device for the inferior mesenteric artery can directly perform matrix subtraction on the first tensor matrix and the third tensor matrix, using the third tensor matrix as a mask for other tissues besides the inferior mesenteric artery, and peeling it off from the first tensor matrix. The pixel values ​​of the parts that are not peeled off are set to 1, and the rest are set to 0, so as to form a second tensor matrix for characterizing the inferior mesenteric artery.

[0127] This invention, based on morphological opening operations on the binarized original tensor matrix, yields a third tensor matrix containing aortic and vertebral features but excluding inferior mesenteric artery features. Then, matrix operations are used to extract features that are only present in the first tensor matrix but not in the third tensor matrix, which are then used as foreground features to generate a second tensor matrix. This method can obtain all tissue features except for the inferior mesenteric artery by smoothing the object's contour, breaking narrow necks, and eliminating fine protrusions. The inferior mesenteric artery features are then indirectly extracted from the complete tissue features, overcoming the difficulty of feature extraction due to the small size of the inferior mesenteric artery.

[0128] Based on any of the above embodiments, determining the second tensor matrix based on the first tensor matrix and the third tensor matrix includes: performing a matrix inner product operation on the first tensor matrix and the third tensor matrix to determine the fourth tensor matrix.

[0129] The fourth tensor matrix contains the physiological tissue characteristics shared by the first and third tensor matrices.

[0130] Specifically, the segmentation device for the inferior mesenteric artery performs a matrix inner product operation on the first tensor matrix and the third tensor matrix, extracts the tissue features contained in the third tensor matrix from the first segmented tensor matrix, and generates a fourth tensor matrix.

[0131] The fourth tensor matrix is ​​a binary matrix with the same size as the first and third tensor matrices. Pixels with a value of 1 correspond to tissues shared by both matrices in the initial segmentation of the inferior mesenteric artery (i.e., other tissues in the coarse region besides the actual inferior mesenteric artery, usually manifested as other small defective areas), while pixels with a value of 0 are tissues that exist only in the first or third tensor matrix.

[0132] Among them, matrix inner product operations include, but are not limited to, matrix dot product and Hadamard product.

[0133] Perform matrix subtraction on the first and fourth tensor matrices to determine the second tensor matrix.

[0134] Specifically, the segmentation device for the inferior mesenteric artery performs matrix subtraction on the first tensor matrix and the fourth tensor matrix, removing small defective regions belonging to the fourth tensor matrix from the initially segmented first tensor matrix to form a second tensor matrix that ultimately contains only the inferior mesenteric artery.

[0135] The second tensor matrix is ​​a binary matrix with the same size as the first and fourth tensor matrices. Pixels with a value of 1 correspond to tissues that actually belong to the inferior mesenteric artery in the initially segmented coarse region, while pixels with a value of 0 represent other tissues besides the inferior mesenteric artery.

[0136] In this embodiment of the invention, the minor imperfection features within the region obtained from the first segmentation layer are first preserved in a fourth tensor matrix by the inner product of the first and third tensor matrices. Then, by calculating the difference between the first and fourth tensor matrices, features that are only retained in the first tensor matrix but not included in the fourth tensor matrix are extracted as foreground features to generate a second tensor matrix. This method can eliminate noise introduced by other tissues surrounding the inferior mesenteric artery during feature extraction.

[0137] Based on any of the above embodiments, performing region growing on the second tensor matrix to obtain an image of the inferior mesenteric artery includes: processing the second tensor matrix using a three-dimensional region growing algorithm based on the second original seed to obtain an image of the inferior mesenteric artery.

[0138] The second original seed is determined based on the second tensor matrix.

[0139] It should be noted that the second original seed refers to the original seed set in the first person segmentation layer regarding the region growth strategy.

[0140] Specifically, the segmentation device for the inferior mesenteric artery can automatically search for the pixel coordinates corresponding to the aorta in the second tensor matrix based on the characteristic that the inferior mesenteric artery originates from the anterior wall of the aorta, using them as the second original seed. It then triggers the 3D-RGA algorithm to calculate the HU gradient value between the neighboring pixels of the second original seed and the second original seed. Pixels whose HU gradient values ​​meet the requirements are used as new growth points. The process of discovering new growth points is iterated until no new growth points are found. Pixels corresponding to the discovered growth points belong to the inferior mesenteric artery and their values ​​are set to 1. Pixels that do not correspond to growth points belong to the inferior mesenteric artery and their values ​​are set to 0, thus forming an image of the inferior mesenteric artery.

[0141] For example, Figure 4 This is the second schematic flowchart of the inferior mesenteric artery segmentation method provided by the present invention. For example... Figure 4 As shown in the figure, this embodiment of the invention provides a complete implementation process of a method for segmenting the inferior mesenteric artery:

[0142] (1) Extracting information from enhanced CT: First, the original enhanced CT sequence data is used as input, the size of a single CT image is scaled to 512*512, and the slope value and intercept value are read from the corresponding DICOM file of the imaging device. Each pixel is converted into the original HU value, and each enhanced CT image is converted into a horizontal slice in the original tensor matrix in turn.

[0143] (2) Automatic identification of the abdominal aorta center: In enhanced CT images, the abdominal aorta usually shows a high HU value. That is, the first horizontal slice from bottom to top in the original tensor matrix is ​​binarized, and a circular structuring element is used to perform erosion operation on it. The coordinates of several points left in the circular area of ​​the image are weighted and averaged, and the resulting target coordinates are taken as the center of the aorta.

[0144] (3) Coarse segmentation based on 3D-RGA: After obtaining the target coordinates, they are set as the original seed points for 3D-RGA. 26 points are selected from the adjacent cubic regions of the original seed points, and their gradient values ​​with the original seed points are calculated. If the gradient values ​​meet the requirements, these points become new seed points. Furthermore, the HU value of the new seed points is also restricted. This process is repeated until there are no new seed points, and the set of all seed points represents the first segmentation result. However, the first result is coarse, including the aorta, inferior mesenteric artery, and even parts of the vertebrae.

[0145] (4) Digital subtraction: To refine the initial segmentation results, it is preferable to find a model that includes the aortic vertebrae but not the inferior mesenteric artery, so that it can be calculated using the formula:

[0146]

[0147] Among them, R fine Let R be the second tensor matrix containing the entire inferior mesenteric artery. rough T is the first tensor matrix obtained from the coarse segmentation based on 3D-RGA. template This is a fourth tensor matrix that includes only the aorta, vertebrae, and other tissues whose features are distinct from those of the inferior mesenteric artery. "ο" represents the Hadamard product.

[0148] (5) Fine segmentation based on 3D-RGA: To further optimize the segmentation results of the inferior mesenteric artery, a second 3D-RGA was designed. Due to anatomical reasons, the inferior mesenteric artery originates from the anterior wall of the aorta, and its origin can be easily found by searching the aorta. Finally, like the first 3D-RGA, the growth origin is automatically set in the first tensor matrix and iteration begins. When the iteration ends, a clean inferior mesenteric artery model, i.e., the inferior mesenteric artery image, can be obtained.

[0149] This invention performs a second segmentation on a second tensor matrix containing only the inferior mesenteric artery and other minor imperfections. Pixels entirely belonging to the inferior mesenteric artery region are used as seed growth points for aggregation growth, extracting an image of the inferior mesenteric artery containing only its features. This allows for fine-grained segmentation after refining the tissue features obtained in the initial segmentation, enabling the separate extraction of the vascular morphology of the inferior mesenteric artery and improving the accuracy and efficiency of inferior mesenteric artery segmentation.

[0150] Figure 5 This is a schematic diagram of the structure of the inferior mesenteric artery segmentation device provided by the present invention. Based on any of the above embodiments, such as... Figure 5 As shown, the device includes a preprocessing module 510 and a segmentation module 520, wherein:

[0151] The preprocessing module 510 is used to convert the pixel values ​​of each abdominal imaging image in the abdominal imaging image sequence into CT values ​​to obtain the original tensor matrix; wherein the arrangement order of the abdominal imaging image sequence is the same as the arrangement order of the slices in the original tensor matrix.

[0152] The segmentation module 520 is used to input the original tensor matrix into the segmentation model to obtain the inferior mesenteric artery image output by the segmentation model.

[0153] The segmentation model is trained based on the original tensor matrix of the samples and the corresponding arterial region labels. The segmentation model includes a first segmentation layer, a feature extraction layer, and a second segmentation layer.

[0154] Accordingly, the segmentation module 520 includes a first segmentation submodule, a feature extraction submodule, and a second segmentation submodule, wherein:

[0155] The first segmentation submodule is used to perform region growing on the original tensor matrix to obtain the first tensor matrix.

[0156] The feature extraction submodule is used to determine the second tensor matrix based on the first tensor matrix and the original tensor matrix.

[0157] The second segmentation submodule is used to perform region growing on the second tensor matrix to obtain images of the inferior mesenteric artery.

[0158] Optionally, the first segmentation submodule includes a target coordinate acquisition unit and a first segmentation unit, wherein:

[0159] The target coordinate acquisition unit is used to locate the inferior mesenteric artery in the first slice of the original tensor matrix and obtain the target coordinates.

[0160] The first segmentation unit is used to process the original tensor matrix using the target coordinates as the first original seed and the three-dimensional region growing algorithm to obtain the first tensor matrix.

[0161] Optionally, the target coordinate acquisition unit includes a positioning subunit and a coordinate transformation subunit, wherein:

[0162] The positioning subunit is used to binarize the first layer slice and then etch it using a circular mask to determine the inferior mesenteric artery region.

[0163] The coordinate transformation subunit is used to perform a weighted average of the coordinates of the pixels contained in the inferior mesenteric artery region to determine the target coordinates.

[0164] The size of the circular mask is matched with the size of the first layer slice.

[0165] Optionally, the feature extraction submodule includes a raw tensor matrix processing unit and a second tensor matrix acquisition unit, wherein:

[0166] The original tensor matrix processing unit is used to perform morphological opening operations on the binarized original tensor matrix to obtain a third tensor matrix.

[0167] The second tensor matrix acquisition unit determines the second tensor matrix based on the first and third tensor matrices.

[0168] The third tensor matrix contains physiological tissue features from the original tensor matrix, excluding the inferior mesenteric artery.

[0169] Optionally, the second tensor matrix acquisition unit includes a first operation subunit and a second operation subunit, wherein:

[0170] The first operation subunit is used to perform a matrix inner product operation on the first tensor matrix and the third tensor matrix to determine the fourth tensor matrix.

[0171] The second operation subunit is used to perform matrix subtraction on the first tensor matrix and the fourth tensor matrix to determine the second tensor matrix.

[0172] The fourth tensor matrix contains the physiological tissue characteristics shared by the first and third tensor matrices.

[0173] Optionally, the second segmentation submodule is specifically used to process the second tensor matrix based on the second original seed using a three-dimensional region growing algorithm to obtain an image of the inferior mesenteric artery.

[0174] The second original seed is determined based on the second tensor matrix.

[0175] The inferior mesenteric artery segmentation device provided in this embodiment of the invention is used to execute the inferior mesenteric artery segmentation method of the present invention. Its implementation method is consistent with the implementation method of the inferior mesenteric artery segmentation method provided in this invention, and can achieve the same beneficial effects, which will not be repeated here.

[0176] This invention utilizes laparoscopic image sequences from different torsions, converting them into raw tensor matrices using medical image processing techniques. These raw tensor matrices are then used as input to a segmentation model. A first segmentation layer aggregates and grows pixels with inferior mesenteric artery features within the raw tensor matrix. A feature extraction layer then performs digital subtraction on the initially segmented tensor matrix. Finally, a second segmentation layer refines the subtracted tensor matrix, resulting in a secondary segmentation that extracts only the inferior mesenteric artery features. This approach enables coarse-grained segmentation of complex tissue features, followed by fine-grained segmentation after refining the initially segmented tissue features. This allows for the separate extraction of the inferior mesenteric artery's vascular morphology, improving the accuracy and efficiency of inferior mesenteric artery segmentation.

[0177] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call logic instructions in the memory 630 to execute a segmentation method for the inferior mesenteric artery. This method includes: converting the pixel values ​​of each peritoneal imaging image in the peritoneal imaging sequence into CT values ​​to obtain an original tensor matrix; wherein the arrangement order of the peritoneal imaging image sequence is the same as the arrangement order of the slices in the original tensor matrix; inputting the original tensor matrix into a segmentation model to obtain an image of the inferior mesenteric artery output by the segmentation model; wherein the segmentation model is trained based on the sample original tensor matrix and the corresponding arterial region labels of the sample original tensor matrix; the segmentation model includes: a first segmentation layer for performing region growing on the original tensor matrix to obtain a first tensor matrix; a feature extraction layer for determining a second tensor matrix based on the first tensor matrix and the original tensor matrix; and a second segmentation layer for performing region growing on the second tensor matrix to obtain the image of the inferior mesenteric artery.

[0178] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0179] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the inferior mesenteric artery segmentation method provided by the above methods. The method includes: converting the pixel values ​​of each peritoneal imaging image in the peritoneal imaging image sequence into CT values ​​to obtain an original tensor matrix; wherein the arrangement order of the peritoneal imaging image sequence is the same as the arrangement order of the slices in the original tensor matrix; inputting the original tensor matrix into a segmentation model to obtain an inferior mesenteric artery image output by the segmentation model; wherein the segmentation model is trained based on the sample original tensor matrix and the arterial region labels corresponding to the sample original tensor matrix; the segmentation model includes: a first segmentation layer for performing region growing on the original tensor matrix to obtain a first tensor matrix; a feature extraction layer for determining a second tensor matrix based on the first tensor matrix and the original tensor matrix; and a second segmentation layer for performing region growing on the second tensor matrix to obtain an inferior mesenteric artery image.

[0180] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for segmenting the inferior mesenteric artery provided by the methods described above. This method includes: converting the pixel values ​​of each peritoneal imaging image in a peritoneal imaging sequence into CT values ​​to obtain an original tensor matrix; wherein the arrangement order of the peritoneal imaging image sequence is the same as the arrangement order of the slices in the original tensor matrix; inputting the original tensor matrix into a segmentation model to obtain an image of the inferior mesenteric artery output by the segmentation model; wherein the segmentation model is trained based on a sample original tensor matrix and the corresponding arterial region labels labeled in the sample original tensor matrix; the segmentation model includes: a first segmentation layer for performing region growing on the original tensor matrix to obtain a first tensor matrix; a feature extraction layer for determining a second tensor matrix based on the first tensor matrix and the original tensor matrix; and a second segmentation layer for performing region growing on the second tensor matrix to obtain an image of the inferior mesenteric artery.

[0181] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0182] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for segmenting the inferior mesenteric artery, characterized in that, include: The pixel values ​​of each laparography image in the sequence are converted into CT values ​​to obtain the original tensor matrix; wherein the order of the laparography image sequence is the same as the order of the slices in the original tensor matrix. The original tensor matrix is ​​input into the segmentation model to obtain the inferior mesenteric artery image output by the segmentation model; The segmentation model is trained based on the original tensor matrix of the sample and the corresponding arterial region labels labeled on the original tensor matrix of the sample; the segmentation model includes: The first segmentation layer is used to perform region growing on the original tensor matrix to obtain the first tensor matrix; A feature extraction layer is used to determine a second tensor matrix based on the first tensor matrix and the original tensor matrix. The determination of the second tensor matrix based on the first tensor matrix and the original tensor matrix includes: performing a morphological opening operation on the binarized original tensor matrix to obtain a third tensor matrix; performing a matrix inner product operation on the first tensor matrix and the third tensor matrix to determine a fourth tensor matrix; and performing a matrix subtraction operation on the first tensor matrix and the fourth tensor matrix to determine the second tensor matrix. The third tensor matrix contains physiological tissue features from the original tensor matrix other than the inferior mesenteric artery, and the fourth tensor matrix contains physiological tissue features shared by the first tensor matrix and the third tensor matrix. The second segmentation layer is used to perform region growing on the second tensor matrix to obtain the image of the inferior mesenteric artery.

2. The method for segmenting the inferior mesenteric artery according to claim 1, characterized in that, The step of performing region growing on the original tensor matrix to obtain the first tensor matrix includes: The inferior mesenteric artery is located in the first slice of the original tensor matrix to obtain the target coordinates; Using the target coordinates as the first original seed, the original tensor matrix is ​​processed using a three-dimensional region growing algorithm to obtain the first tensor matrix.

3. The method for segmenting the inferior mesenteric artery according to claim 2, characterized in that, The inferior mesenteric artery is located in the first slice of the original tensor matrix to obtain the target coordinates, including: After binarizing the first layer slice, erosion was performed using a circular mask to determine the inferior mesenteric artery region. The target coordinates are determined by weighted averaging of the pixel coordinates contained in the inferior mesenteric artery region. The size of the circular mask matches the size of the first layer slice.

4. The method for segmenting the inferior mesenteric artery according to claim 1, characterized in that, The step of performing region growing on the second tensor matrix to obtain the image of the inferior mesenteric artery includes: Based on the second original seed, the second tensor matrix is ​​processed using a three-dimensional region growing algorithm to obtain the image of the inferior mesenteric artery; The second original seed is determined based on the second tensor matrix.

5. A device for segmenting the inferior mesenteric artery, characterized in that, include: The preprocessing module is used to convert the pixel values ​​of each laparography image in the laparography image sequence into CT values ​​to obtain the original tensor matrix; wherein the arrangement order of the laparography image sequence is the same as the arrangement order of the slices in the original tensor matrix. The segmentation module is used to input the original tensor matrix into the segmentation model to obtain the inferior mesenteric artery image output by the segmentation model; The segmentation model is trained based on the original tensor matrix of the sample and the corresponding arterial region labels labeled on the original tensor matrix of the sample; the segmentation model includes: The first segmentation layer is used to perform region growing on the original tensor matrix to obtain the first tensor matrix; A feature extraction layer is used to determine a second tensor matrix based on the first tensor matrix and the original tensor matrix. The determination of the second tensor matrix based on the first tensor matrix and the original tensor matrix includes: performing a morphological opening operation on the binarized original tensor matrix to obtain a third tensor matrix; performing a matrix inner product operation on the first tensor matrix and the third tensor matrix to determine a fourth tensor matrix; and performing a matrix subtraction operation on the first tensor matrix and the fourth tensor matrix to determine the second tensor matrix. The third tensor matrix contains physiological tissue features from the original tensor matrix other than the inferior mesenteric artery, and the fourth tensor matrix contains physiological tissue features shared by the first tensor matrix and the third tensor matrix. The second segmentation layer is used to perform region growing on the second tensor matrix to obtain the image of the inferior mesenteric artery.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method for segmenting the inferior mesenteric artery as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for segmenting the inferior mesenteric artery as described in any one of claims 1 to 4.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for segmenting the inferior mesenteric artery as described in any one of claims 1 to 4.

Images