A method and device for evaluating kidney stone burden based on CT plain scan images

Abstract

The application discloses a kidney stone load evaluation method and device based on CT plain scan images, and relates to the technical field of medical image processing. The method comprises the following steps: automatically analyzing and recording semantic coordinates of each lesion area, constructing a structured data set, and constructing a local feature space for the lesion. The semantic boundary of the suspected kidney stone in the CT plain scan image sequence is used to judge the lesion morphology by exploring the corresponding semantic features in the local feature space through a morphological method. A density space matrix is constructed by using the semantic features of the lesion, an optimal semantic density is calculated in the density space matrix by designing a search strategy corresponding to the lesion morphology, and the number of kidney stones is obtained by using an unsupervised clustering algorithm. According to the analysis results of the morphology and number of the kidney stones, the automatic volume measurement of the kidney stones is realized by using a graphics method. The application can realize the full-automatic and accurate measurement of the morphology, number and volume of the kidney stones in the CT plain scan image sequence.

Description

Technical Field

[0001] This application relates to the field of medical image processing technology, and in particular to a method and device for assessing kidney stone burden based on CT plain scan images. Background Technology

[0002] With the continuous development of medical imaging technology, kidney stone burden assessment based on CT plain scan images has become one of the important means of clinical diagnosis. However, existing assessment methods often rely on manual annotation and semi-automatic tool judgment, which is not only time-consuming and labor-intensive, but also easily affected by subjective factors, resulting in poor accuracy and consistency of assessment results. Therefore, it is particularly important to develop an efficient, accurate and automated method for kidney stone burden assessment. Summary of the Invention

[0003] The main objective of this application is to propose a method and dedicated device for assessing kidney stone burden based on CT plain scan images, thereby overcoming the shortcomings of existing technologies and achieving efficient and accurate assessment of kidney stone burden. This method combines medical clinical knowledge with unsupervised machine learning techniques, enabling efficient and accurate assessment of kidney stone burden in CT plain scan images without human intervention.

[0004] To achieve the above objectives, one aspect of this application proposes a method for assessing the burden of kidney stones based on plain CT images, the method comprising the following steps:

[0005] The semantic coordinates of each lesion region are automatically parsed and recorded to construct a structured dataset, where each lesion region is the semantic boundary of the suspected kidney stone in each CT plain scan image sequence.

[0006] A local feature space is constructed for the lesion, and the corresponding semantic features are explored in the local feature space using morphological methods to determine the lesion morphology;

[0007] A density space matrix is ​​constructed using the semantic features corresponding to the lesions, and a search strategy corresponding to the lesion morphology is designed to calculate the optimal semantic density in the density space matrix. Then, the number of kidney stones is obtained through an unsupervised clustering algorithm.

[0008] Based on the analysis of the morphology and number of kidney stones, an automated volume measurement of kidney stones was achieved using graphical methods.

[0009] In some embodiments, the automatic parsing and recording of semantic coordinates of each lesion region to construct a structured dataset includes the following steps:

[0010] The attributes of each lesion region are measured using morphological methods, and then the two-dimensional semantic coordinates of each lesion region are analyzed and recorded.

[0011] Based on the sequence number of the CT plain scan image sequence, the two-dimensional semantic coordinates of each lesion region are expanded into three-dimensional semantic coordinates, forming a structured dataset.

[0012] In some embodiments, constructing a local feature space for the lesion and exploring corresponding semantic features in the local feature space using morphological methods to determine the lesion morphology includes the following steps:

[0013] The number of lesion regions in the CT plain scan image sequence is counted and recorded using a dictionary. If the maximum number of lesion regions in the CT plain scan image is less than or equal to 1, the morphology of the lesion in the CT plain scan image is determined to be an isolated lesion. If the maximum number of lesion regions in the CT plain scan image is greater than 1, the morphology of the lesion in the CT plain scan image is tentatively defined as an interlaced lesion or a staghorn lesion.

[0014] For lesions tentatively classified as interlaced or staghorn, traverse all CT plain images of the lesion region with a number greater than 1, and use these as anchor points to associate them with the lesion regions contained in the preceding and following CT plain images to construct a local feature space.

[0015] The in-degree and out-degree of each anchor point in the local feature space are calculated using the degree analysis method. If the out-degree and in-degree of the lesion region are the same, the local feature space is marked as parallel; if the out-degree of the lesion region is greater than the in-degree, the local feature space is marked as bifurcation; if the in-degree of the lesion region is greater than the out-degree, the local feature space is marked as adhesion.

[0016] By traversing all the markers in the local feature space, if there are no bifurcated or adherent markers, the morphology of the lesion in the CT plain scan image is determined to be an interlaced lesion; if there are bifurcated or adherent markers, the morphology of the lesion in the CT plain scan image is determined to be a staghorn lesion.

[0017] In some embodiments, constructing a density space matrix using semantic features corresponding to the lesions includes the following steps:

[0018] A density space matrix is ​​obtained by calculating the semantic density between the lesion regions based on their three-dimensional semantic coordinates; wherein, the semantic density is used to describe the degree of spatial correlation between the lesion regions.

[0019] The search strategy corresponding to the designed lesion morphology calculates the optimal semantic density in the density space matrix, and then obtains the number of kidney stones through an unsupervised clustering algorithm, including the following steps:

[0020] Based on the specific morphology of the lesion, the domain-specific morphological structure and the corresponding anchor point image sequence number are obtained from the lesion area quantity dictionary and recorded in a dictionary manner; among them, the corresponding morphological structure of the interlaced lesion is parallel, and the corresponding morphological structure of the staghorn lesion is bifurcation and adhesion.

[0021] Based on different lesion morphologies and corresponding domain-specific morphological structures, the corresponding search objects are calculated.

[0022] The target search matrix is ​​obtained by adjusting the density space matrix according to the search object;

[0023] Based on the specific morphology of the lesion, the semantic density difference between the lesion regions is calculated by traversing the target search matrix. If the semantic density difference exceeds the density search upper limit of the target search matrix, the current density search upper limit is recorded.

[0024] The minimum semantic density is calculated based on the density search upper limit of the current batch, and then compared with the optimal semantic density of the previous batch to obtain the optimal semantic density.

[0025] Based on the optimal semantic density and lesion characteristics, the lesion regions with achievable density are grouped together using an unsupervised clustering algorithm to obtain one or more clusters, and the number of clusters is the number of lesions; wherein, the lesion characteristics are that the lesions exist in at least two consecutive CT plain scan images.

[0026] In some embodiments, the method further includes the following steps:

[0027] Verify whether the cluster belongs to a kidney stone lesion;

[0028] The number of clusters that are statistically verified to be kidney stone lesions is used as the number of kidney stones.

[0029] In some embodiments, the automated volume measurement of kidney stones using graphical methods based on the analysis results of the morphology and number of kidney stones includes the following steps:

[0030] The pixels in the lesion area belonging to kidney stones are classified into two categories: foreground and background. Then, for the foreground area, the coverage between each pixel and its surrounding neighboring pixels is calculated using a graphics method. The foreground is kidney stones, and the background is non-kidney stones.

[0031] The coverage of each lesion area is obtained by traversing all pixels belonging to the kidney stone lesion area. Then, the volume of the kidney stone is calculated by combining the slice thickness and pixel spacing of the CT plain scan image.

[0032] To achieve the above objectives, another aspect of this application proposes a kidney stone burden assessment device based on CT plain scan images, the device comprising:

[0033] The dataset construction unit is used to automatically parse and record the semantic coordinates of each lesion region to construct a structured dataset, wherein each lesion region is the semantic boundary of the suspected kidney stone in each CT plain scan image in the CT plain scan image sequence.

[0034] The lesion morphology judgment unit is used to construct a local feature space for lesions and to explore the corresponding semantic features in the local feature space through morphological methods to judge the lesion morphology.

[0035] The kidney stone number determination unit is used to construct a density space matrix using the semantic features corresponding to the lesions, and to design a search strategy corresponding to the lesion morphology to calculate the optimal semantic density in the density space matrix, and then obtain the number of kidney stones through an unsupervised clustering algorithm.

[0036] The kidney stone volume determination unit is used to automatically measure the volume of kidney stones based on the analysis results of their morphology and number, using graphical methods.

[0037] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0038] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0039] The embodiments of this application include at least the following beneficial effects:

[0040] This application achieves automated and high-precision assessment of kidney stone burden in CT plain scan images by introducing unsupervised machine learning technology and medical clinical knowledge. Specifically, this method can automatically and accurately determine the morphology of kidney stones by analyzing the distribution characteristics and morphological features of the lesion area. Furthermore, this method utilizes unsupervised machine learning technology and graphics methods to automatically assess the number and volume of kidney stones without human intervention, significantly improving assessment efficiency and reducing human error. In summary, the kidney stone burden assessment method and its dedicated device based on CT plain scan images proposed in this application significantly improve the accuracy and efficiency of kidney stone burden assessment through technological innovation, making the assessment process more objective and standardized, and providing important technical support for clinical diagnosis and treatment. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A flowchart illustrating a method for assessing kidney stone burden based on CT plain scan images, provided in an embodiment of this application;

[0043] Figure 2 A schematic diagram of a kidney stone burden assessment device based on CT plain scan images provided in this application embodiment;

[0044] Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0046] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0047] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0049] Before providing a detailed description of the embodiments of this application, the various morphologies of kidney stones involved in this application are described as follows:

[0050] Solitary kidney stones: limited to a single kidney stone or multiple kidney stones that do not intersect in spatial dimensions, appearing as a single lesion area in each CT plain scan image.

[0051] Interleaved kidney stones: These consist of multiple kidney stones, but each stone does not branch out or attach to other stones. In plain CT scans, interleaved kidney stones appear as multiple lesion areas on each scan, each belonging to a different kidney stone.

[0052] Staghorn kidney stones: Staghorn kidney stones are similar in shape to interlaced kidney stones, but differ in that the stones themselves may branch or adhere to each other. Staghorn kidney stones appear as multiple lesion areas on each CT scan, which may belong to the same or different kidney stones.

[0053] This application provides a method for assessing the burden of kidney stones based on CT plain scan images, relating to the field of medical image processing technology. The method provided in this application can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing knowledge extraction methods, but is not limited to the above forms.

[0054] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0055] Reference Figure 1 This application provides a method for assessing kidney stone burden based on CT plain scan images. This method may include, but is not limited to, steps S110 to S140, as follows:

[0056] Before performing S110, this embodiment may further include a step of acquiring lesion regions. For example, this embodiment can identify multiple lesion regions from a CT plain scan image sequence, specifically:

[0057] The lesion boundary constraint model is used to identify and delineate the lesion area from the CT plain scan image sequence. The lesion boundary constraint model can be used to roughly divide the lesion area, and the detailed lesion area can be further identified by combining morphological analysis.

[0058] The steps to determine the lesion area may include:

[0059] 1. Constructing a lesion boundary constraint model based on a convolutional neural network (CNN): A pre-trained deep convolutional neural network is used to fine-tune specific lesion features in CT scan sequences to identify and accurately delineate lesion regions. This lesion boundary constraint model not only effectively identifies lesion regions but also maintains high-precision segmentation in complex backgrounds, exhibiting higher robustness and adaptability compared to traditional thresholding or edge detection methods.

[0060] 2. Determining attributes by combining morphological operations and lesion features: Morphological operations (such as opening and closing operations) are applied to the lesion regions identified by the lesion boundary constraint model to further refine the lesion region outline. The attributes of each lesion region (stone region or non-stone region) are then determined by combining features such as texture and size. This step not only enhances the clarity of the lesion region boundaries but also provides detailed lesion feature parameters for subsequent burden assessment.

[0061] S110: Automatically parse and record the semantic coordinates of each lesion region to construct a structured dataset, where each lesion region is the semantic boundary of the suspected kidney stone in each CT plain scan image sequence.

[0062] It is understandable that each lesion region can correspond to a two-dimensional semantic coordinate in a planar coordinate system. This semantic coordinate, combined with the CT plain scan image sequence number, can correspond to a three-dimensional semantic coordinate in a spatial coordinate system. The combination of the three-dimensional semantic coordinates of all lesion regions serves as a structured dataset.

[0063] Furthermore, S110 may include S111 to S112:

[0064] S111: The attributes of each lesion region are measured by morphological methods, and then the two-dimensional semantic coordinates of each lesion region are analyzed and recorded;

[0065] The expression for the two-dimensional semantic coordinates is:

[0066] C ij (x c,ij ,y c,ij )=f r (R ij );

[0067] Among them, C ij (x c,ij ,y c,ij ) represents the two-dimensional semantic coordinates, x c,ij and y c,ij f represents the x and y coordinates of the two-dimensional semantic coordinates of the i-th lesion region in the j-th CT plain image, respectively; r Represents the regionprops function in morphological methods; R ij Let i be the lesion region in the j-th CT plain scan image, where j = 1, 2, ..., E, and E is the total number of CT plain scan images;

[0068] Specifically, in this embodiment, the regionprops function in morphological methods can be used to determine the semantic coordinates of each lesion region contained in each CT plain scan image sequence, with each lesion region corresponding to a two-dimensional semantic coordinate.

[0069] S112: Based on the sequence number of the CT plain scan image sequence, expand the two-dimensional semantic coordinates of each lesion region into three-dimensional semantic coordinates, and form a structured dataset;

[0070] The expression for the three-dimensional semantic coordinates is:

[0071] Cij (x c,ij ,y c,ij ,z j );

[0072] Among them, z j =j represents the sequence number of the CT plain scan image; the dataset is represented by D, defined as D = {C ij |i=1,2,...,n j ;j=1,2,...,E}, C in D ij n represents the three-dimensional semantic coordinates; j This indicates the number of lesion regions identified in the j-th CT plain scan image.

[0073] Specifically, by determining the three-dimensional semantic coordinates of each lesion area, three-dimensional modeling of each lesion area can be performed, thereby providing a data basis for subsequent steps such as analyzing the morphology, number, and volume of kidney stones.

[0074] S120: Construct a local feature space for the lesion, and use morphological methods to explore the corresponding semantic features in the local feature space to determine the lesion morphology.

[0075] Understandably, kidney stones can exist in various forms. For ease of classification, this embodiment can classify kidney stones into solitary, interlocking, and staghorn types.

[0076] Different types of kidney stones may correspond to different lesion areas. Therefore, this embodiment can determine the corresponding kidney stone morphology based on the characteristics of each lesion area, such as the number distribution and morphological structure.

[0077] Furthermore, S120 may include S121 to S123:

[0078] S121: The number of lesion regions in the CT plain scan image sequence is counted and recorded using a dictionary; if the maximum number of lesion regions in the CT plain scan image is less than or equal to 1, the morphology of the lesion in the CT plain scan image is determined to be an isolated lesion; if the maximum number of lesion regions in the CT plain scan image is greater than 1, the morphology of the lesion in the CT plain scan image is temporarily defined as an interlaced lesion or a staghorn lesion.

[0079] The expression for counting the number of lesion areas is:

[0080]

[0081] Among them, Dict m A dictionary for recording the number of lesion regions in the CT plain scan image sequence. D is a statistical function indexed by the z-axis.z The dataset indexed by the z-axis;

[0082] The formula for initially distinguishing the morphology of kidney stones is:

[0083]

[0084] in, The maximum search function is defined as follows: is represents isolated kidney stones, ss represents interlaced kidney stones, cs represents staghorn kidney stones, and M represents the morphology of the kidney stone.

[0085] S122: For lesions tentatively classified as interlaced or staghorn, traverse all CT plain images of the lesion regions with a number greater than 1, and use these as anchor points to associate them with the lesion regions contained in the preceding and following CT plain images to construct a local feature space.

[0086] The expression for the local feature space is:

[0087] L z =f E (f pd (D z-1 D z D z+1 ));

[0088] Among them, L z Let f represent the local feature space. E It is the Euclidean distance function, f pd It is a Pandas data constructor;

[0089] S123: The in-degree and out-degree of each anchor point in the local feature space are calculated using the degree analysis method. If the out-degree and in-degree of the lesion region are the same, the local feature space is marked as parallel; if the out-degree of the lesion region is greater than the in-degree, the local feature space is marked as bifurcation; if the in-degree of the lesion region is greater than the out-degree, the local feature space is marked as adhesion.

[0090] The expression for calculating the in-degree is:

[0091]

[0092] Where ε is L z The average of the sum of the mean and the standard deviation; It is a statistical function that counts zero numbers;

[0093] S124: Traverse all the markers in the local feature space. If there are no bifurcation or adhesion markers, the morphology of the lesion in the CT plain scan image is determined to be an interlaced lesion; if there are bifurcation or adhesion markers, the morphology of the lesion in the CT plain scan image is determined to be a staghorn lesion.

[0094] The expression for further distinguishing the morphology of kidney stones is:

[0095]

[0096] in, Represents the row space of a matrix. Represents the column space of a matrix. This represents the value corresponding to the a-th row and b-th column of the matrix;

[0097] This embodiment focuses on the number of lesion regions contained in each CT scan sequence. By analyzing the distribution of the number of lesion regions and the contextual relationship of image matching, the morphology of kidney stones can be determined. This embodiment can improve the efficiency of identifying the morphology of kidney stones and can identify the morphology of kidney stones without manual intervention.

[0098] S130: Construct a density space matrix using the semantic features corresponding to the lesions, and design a search strategy corresponding to the lesion morphology to calculate the optimal semantic density in the density space matrix. Then, obtain the number of kidney stones through an unsupervised clustering algorithm.

[0099] Understandably, isolated, interlaced, and staghorn lesions can all present with one or more stones. However, the assessment method for the number of kidney stones will differ based on their morphology. For isolated lesions, a simple assessment suffices. For interlaced lesions, it is necessary to identify the CT scan image with the most lesion areas and begin the analysis from that image. For staghorn lesions, it is necessary to identify multiple consecutive CT scan images showing variations in the number of lesion areas and begin the analysis from these multiple consecutive CT scan images.

[0100] Furthermore, S130 may include S131 to S137:

[0101] S131: Calculate the semantic density between the lesion regions based on their three-dimensional semantic coordinates to obtain a density space matrix; wherein, the semantic density is used to describe the degree of spatial correlation between the lesion regions;

[0102] The expression for the density space matrix is:

[0103]

[0104] Among them, P i,j Represents the semantic density, (x i,y i ,z i ) and (x j ,y j ,z j The three-dimensional semantic coordinates are any two different coordinates; the density space matrix is ​​represented by P, where P includes the values ​​of each of the three-dimensional semantic coordinates. i,j ;

[0105] S132: Based on the specific morphology of the lesion, obtain the domain-specific morphological structure and the corresponding anchor point image sequence number from the lesion area quantity dictionary and record them in a dictionary manner; wherein, the corresponding morphological structure of the interlaced lesion is parallel, and the corresponding morphological structure of the staghorn lesion is bifurcation and adhesion.

[0106] The expression for recording unique morphological structures and corresponding anchor point image numbers in a dictionary format is as follows:

[0107]

[0108] in, It is a morphological search function used to search for the location of specific semantic features of a given morphology in a CT plain scan image sequence;

[0109] S133: Based on different lesion morphologies and the corresponding domain-specific morphological structures, the corresponding search objects are calculated.

[0110] The expression for the search object is:

[0111]

[0112] Among them, f p It is a localization function used to locate the position of specific semantic features in a CT plain scan image sequence;

[0113] S134: Adjust the density space matrix according to the search object to obtain the target search matrix;

[0114] The expression for the target search matrix is:

[0115]

[0116] Where P′ represents the target search matrix constructed by columns, and P″ represents the target search matrix constructed by rows; m and n represent the rows and columns of the density space matrix P, P m,k It is a matrix focused on k columns, P k ′ ,n It is a matrix focused on k rows, and sort represents the sorting function;

[0117] S135: Based on the specific morphology of the lesion, traverse the target search matrix to calculate the semantic density difference between the lesion regions. If the semantic density difference exceeds the density search upper limit of the target search matrix, record the current density search upper limit.

[0118] The expression for the upper limit of the density search is:

[0119] cutoff = f J (DDN(P u " +1,k ,P u ′ , ′ k ),φ)

[0120] Among them, f J Let φ represent the contrast function, where φ is the average of the mean and variance of the k-th column in P″; DDN represents the neighborhood discrepancy function, expressed as:

[0121]

[0122] Among them, P u " +1,k and P u ′ , ′ k These are two values ​​that need to be compared, where μ is the value of P. u " +1,k and P u ′ , ′ k The average value;

[0123] S136: Calculate the minimum semantic density based on the density search upper limit of the current batch, and compare it with the optimal semantic density of the previous batch to obtain the optimal semantic density;

[0124] The expression for the optimal semantic density is:

[0125]

[0126] Where EpS represents the semantic density of the current batch; EpS' represents the optimal semantic density of the previous batch. It is a minimum search function used to search for the minimum value. It is a minimum comparison function used to determine the minimum value between two input values;

[0127] S137: Based on the optimal semantic density and lesion characteristics, the lesion regions with achievable density are grouped together using an unsupervised clustering algorithm to obtain one or more clusters, and the number of clusters is the number of lesions; wherein, the lesion characteristics are that the lesions exist in at least two consecutive CT plain scan images;

[0128] The expression for calculating the number of kidney stones is:

[0129] Q = F cluster (D,EpS)

[0130] Among them, F cluster This is an unsupervised clustering algorithm, where Q represents the number of clusters containing lesions.

[0131] Specifically, in this embodiment, the three-dimensional semantic coordinates are clustered by calculating the semantic density between them. When the semantic density is less than the calculated optimal semantic density, the three-dimensional semantic coordinates can be considered to belong to the same cluster.

[0132] To improve the accuracy of the number of kidney stones, embodiments of this application may further include the following steps S138-S139:

[0133] S138: Verify whether the cluster is a kidney stone lesion;

[0134] The expression for verification is:

[0135]

[0136] Where f represents the verification result, υ i It is a valid value; when υ i A value of 1 indicates that the object attribute of the lesion area is kidney stones. i A value of 0 indicates that the lesion area is not a kidney stone;

[0137] S139: The number of clusters belonging to kidney stone lesions is statistically verified as the number of kidney stones.

[0138] It is understandable that this embodiment verifies whether each cluster is a kidney stone, removes clusters that are not kidney stones, and finally counts the number of clusters that are verified to be kidney stones to obtain the total number of kidney stones.

[0139] S140: Based on the analysis results of the morphology and number of kidney stones, use graphical methods to realize the automated volume measurement of kidney stones.

[0140] Furthermore, S140 may specifically include S141 to S142:

[0141] S141: Pixels belonging to the lesion area of ​​kidney stones are classified into two categories: foreground and background. Then, for the foreground area, the coverage between each pixel and its surrounding neighboring pixels is calculated using a graphics method. The foreground is kidney stones, and the background is non-kidney stones.

[0142] The expression for calculating the coverage rate is as follows:

[0143]

[0144] Among them, X i,j The pixel value corresponding to the i-th row and j-th column in the target region;

[0145] S142: Traverse all pixels belonging to the lesion area of ​​the kidney stone to obtain the coverage of each lesion area, and then combine the slice thickness and pixel spacing of the CT plain scan image to calculate the volume of the kidney stone.

[0146] The expression for the total volume of kidney stones is:

[0147]

[0148] Where V represents the total volume of the kidney stone. and ST represents the pixel spacing of the k-th CT plain scan image. k This represents the slice thickness of the k-th CT plain scan image.

[0149] Specifically, for the problem of volume estimation of irregular objects such as kidney stones, this embodiment proposes a volume analysis scheme based on graphic combination: the kidney stone is regarded as an irregular shape composed of multiple squares and triangles. A square represents four pixels within the target area, and a triangle represents three out of four pixels within the target area. Therefore, this embodiment calculates the volume of the irregularly shaped kidney stone by counting the number of squares and triangles within the target area.

[0150] In some optional implementations, the volume calculation step may include: 1. First, based on the quantity assessment results, extract the cluster corresponding to each kidney stone, and then extract the corresponding image sequence number and three-dimensional semantic coordinates from the cluster. 2. Based on the image sequence number and three-dimensional semantic coordinates, map back to the original lesion area and calculate the coverage of the lesion area to obtain the area of ​​the lesion area. 3. Based on all the sequence numbers and coordinates, as well as the pixel spacing and slice thickness corresponding to the image, obtain the volume of the kidney stone.

[0151] The following section will provide a detailed introduction and explanation of the solutions in the embodiments of this application, using specific application examples.

[0152] Optionally, the steps in this embodiment may include obtaining the boundary of the lesion area and assessing the kidney stone burden.

[0153] Step 1: Obtain the boundary of the lesion area.

[0154] The steps for determining the boundaries of the lesion area may include:

[0155] 1. Constructing a lesion boundary constraint model based on a convolutional neural network (CNN): A pre-trained deep convolutional neural network is used to fine-tune specific lesion features in CT scan sequences to identify and accurately delineate lesion regions. This lesion boundary constraint model not only effectively identifies lesion regions but also maintains high-precision segmentation in complex backgrounds, exhibiting higher robustness and adaptability compared to traditional thresholding or edge detection methods.

[0156] 2. Determining attributes by combining morphological operations and lesion features: Morphological operations (such as opening and closing operations) are applied to the lesion regions identified by the lesion boundary constraint model to further refine the lesion region outline. The attributes of each lesion region (stone region or non-stone region) are then determined by combining features such as texture and size. This step not only enhances the clarity of the lesion region boundaries but also provides detailed lesion feature parameters for subsequent burden assessment.

[0157] Step Two: Kidney Stone Burden Assessment. The detailed steps are as follows:

[0158] Step 1: 3D Modeling: Let the j-th CT scan image be Image. j , where j = 1, 2, ..., E, and E is the total number of CT plain scan images. For the i-th lesion region R in the j-th CT plain scan image. ij The two-dimensional semantic coordinates C are extracted using the regionprops function in morphological methods. ij (x c,ij ,y c,ij This process can be formalized as C. ij (x c,ij ,y c,ij )=f r (R ij ), where x c,ij and y c,ij Let f represent the x and y coordinates of the two-dimensional semantic coordinates of the i-th lesion in the j-th CT plain image, respectively. r This refers to the `regionprops` function. To use the corresponding CT plain image slice number as the z-axis coordinate, each two-dimensional semantic coordinate C... ij (x c,ij ,y c,ij Expanded to three-dimensional semantic coordinates C ij (x c,ij ,y c,ij ,z j ), where z j =j represents the sequence number of the CT plain scan image. Finally, the three-dimensional semantic coordinates of all lesion regions in all CT plain scan images are compiled into a structured dataset D, defined as D = {C ij|i=1,2,...,n j ;j=1,2,...,E}, where n j This represents the number of lesions identified in the j-th CT plain scan image. This dataset D represents the distribution of lesions in three-dimensional space, with each three-dimensional semantic coordinate C... ij (x c,ij ,y c,ij ,z j () indicates the specific location of the lesion in the CT plain scan image sequence.

[0159] Step 2: Morphological Analysis: Clinicians typically focus on CT scans showing a large number of kidney stones or changes in their distribution within those areas. They then determine the morphology of the kidney stones by matching the context of the regions. Utilizing this diagnostic knowledge, the morphological analysis algorithm proposed in this embodiment can identify the morphology of kidney stones without manual intervention.

[0160] Step 2.1: Local Feature Space Construction: First, the number of lesion regions in each CT plain scan image is counted. This contains crucial information essential for determining the lesion morphology. This process can be formulated as: Dict m ={z|f s z (D z )}, where f s z This is a statistical function indexed by the z-axis, used to calculate the number of lesion regions on each plain CT image. Typically, the morphology of kidney stones exhibits localized variations. For interlaced kidney stones, the morphology shows lesions that are spatially parallel. For staghorn kidney stones, the morphology shows lesions that bifurcate or adhere to each other. For f... s z (D z When the value of )>1, meaning the CT scan image contains multiple lesion areas, a local feature space L is constructed to reflect changes in the morphology of the kidney stone. This process can be formalized as: L z =f E (f pd (D z-1 D z D z+1 )), where f E It is the Euclidean distance function, f pd It is a Pandas data constructor.

[0161] Step 2.2: Domain-Specific Semantic Feature Exploration: For solitary kidney stones, each CT scan image contains only a single lesion region, while for interlocking and staghorn kidney stones, multiple stone regions may coexist in a single CT scan image. This feature can be effectively used to identify solitary kidney stones. The process can be formulated as follows: in The maximum search function is used, where is represents isolated kidney stones, ss represents interlaced kidney stones, and cs represents staghorn kidney stones. Regarding interlaced and staghorn kidney stones, the former is characterized by parallelism, exhibiting a one-to-one regional correlation in the local feature space, while the latter is characterized by bifurcation or adhesion, potentially showing a one-to-many or many-to-one relationship in the local feature space. Therefore, exploring the in-degree and out-degree relationships between lesion regions in the local feature space can effectively distinguish them. This process can be formulated as follows: Where I is the index function, and the formula is: Where ε is L z The average of the sum of the mean and standard deviation, N is the value of s. i The total number, It is a statistical function that counts zero numbers.

[0162] Step 3: Quantitative Assessment: For different types of kidney stones, clinicians typically choose appropriate assessment methods based on the specific circumstances. For solitary kidney stones, a comprehensive assessment is usually performed from start to finish. For interlocking and staghorn kidney stones, the focus is usually concentrated on areas with specific semantic features. Utilizing this diagnostic knowledge, the proposed quantitative analysis algorithm adaptively employs appropriate strategies to automatically estimate the number of kidney stones.

[0163] Step 3.1: Semantic Density Spatialization: Euclidean distance is used to evaluate the semantic density between the three-dimensional semantic coordinates of each lesion region. The formula is as follows: Where (x) i ,y i ,z i ) and (x j ,y j ,z j These are two different three-dimensional semantic coordinates. After the above steps, a density space matrix P reflecting the relationships between various lesion regions can be obtained.

[0164] Step 3.2: Semantic Density Search Using Clinical Knowledge: In clinical settings, methods for assessing different morphologies of kidney stones exhibit significant differences. Fixed semantic density search strategies become inflexible when faced with various kidney stone morphologies. This work proposes a morphology-specific semantic density search strategy from a clinical perspective. This strategy integrates clinical judgment knowledge into the semantic density search process, calculating the optimal semantic density through interaction with the density space matrix and lesion morphology. Therefore, the search process can be summarized in the following steps: constructing the search object, designing the search behavior, and defining the search termination.

[0165] Step 3.2.1: Construction of Search Objects: The representation of search objects constitutes the initial and crucial stage of semantic density search. The search objects for each step are constructed from a medical perspective. The construction of search objects involves adjusting the density space matrix based on regions with domain-specific semantic features. This process is defined as follows: Where m and n represent the rows and columns of the density space matrix P, P m,k It is a matrix focused on k columns, P k ′ ,n It is a matrix focused on k rows, where sort represents the sorting function, k is the search object, and the formula is: Where f p Dict is a localization function used to locate specific semantic features within a CT plain image sequence. q It is a dictionary that records unique semantic features and their corresponding image locations, with the following formula: It is a morphological search function used to search for the location of specific semantic features of a given morphology in a CT plain image sequence.

[0166] Step 3.2.2: Search Behavior Design: If adjacent 3D semantic coordinates belong to the same cluster, their deviation should remain within a stable range. This knowledge helps in calculating the semantic density difference cutoff, which distinguishes the semantic density difference between adjacent semantic coordinates by calculating their discrete values. This process can be formulated as: cutoff = f J (DDN(P″ u+1,k ,P″ u,k ),φ), where f J Let φ represent the contrast function, where φ is the average of the mean and variance of the k-th column in P″, and DDN represents the neighborhood discrepancy function, which can be formulated as: Where P u " +1,k and P u ′ , ′ k The two values ​​to be compared are μ and P. u "+1,k and P u ′ , ′ k The average value is then used. Subsequently, the optimal semantic density Eps can be determined by comparing the optimal semantic density of the current batch with the optimal semantic density of previous batches. This process can be formulated as follows: in It is a minimum search function used to search for the minimum value in an input list or dictionary. It is a minimum comparison function used to determine the minimum value between two input values.

[0167] Step 3.2.3: Definition of search termination: The termination condition of the complete semantic density search process is that the search boundary exceeds the range of matrix P, which can be determined as: u≥m-1.

[0168] Step 3.3: Cluster Discovery: Density-based clustering algorithms can discover clusters with random shapes without predefining the number of clusters. The number of kidney stones can be easily determined by calculating the optimal semantic density using a semantic density search strategy, which can be formulated as: Q = F cluster (D,EpS), where F cluster It is an unsupervised clustering algorithm.

[0169] Step 3.4: Validity Verification: Since the search target includes all regions identified by the adaptive discriminative boundary algorithm, the clustering results include non-stone clusters and stone clusters. To distinguish between non-stone clusters and stone clusters, a validity check is designed to examine whether semantic coordinates belonging to non-stone areas exist in a certain cluster. Invalid semantic coordinates mean that the lesion region is judged to belong to non-stone areas in the lesion boundary constraint model. The designed validity verification method can be described as follows: Where f represents the verification result, υ i This is a valid value. Valid values ​​are only 1 and 0; the former indicates a semantic coordinate region of kidney stones, and the latter indicates a region without kidney stones. Therefore, for a given cluster, if the validation result is 1, the cluster is considered to contain kidney stones, except in other cases.

[0170] Step 4: Volume Assessment: For the volume estimation problem of irregular objects such as kidney stones, a volume analysis algorithm based on image combination is proposed. This algorithm treats a kidney stone as an irregular shape composed of multiple squares and triangles. A square represents four pixels within the target area, and a triangle represents three out of four pixels within the target area. Therefore, calculating the volume of this irregular shape mainly requires calculating the number of squares and triangles within the target area. In the proposed algorithm, the pixel attributes of the target area are first distinguished, and then the coverage ratio between each pixel and its neighboring pixels is calculated for pixels belonging to the kidney stone, expressed as: When traversing all pixels, the total volume of the kidney stone can be expressed as: in and ST represents the pixel spacing of the k-th CT plain scan image. k This represents the slice thickness of the k-th CT plain scan image.

[0171] The kidney stone burden assessment in this embodiment can be divided into three categories: kidney stone morphology assessment, kidney stone quantity assessment, and kidney stone volume assessment. In kidney stone morphology assessment, the morphology of the stones is classified into three types: isolated, interlaced, and staghorn. Morphology assessment utilizes prior clinical knowledge to construct a local feature space for areas with high variability or concentration of kidney stones. A morphological correlation algorithm is then designed to explore regional correlations within the local feature space, thereby achieving stone morphology analysis. In kidney stone quantity assessment, an adaptive analysis strategy is designed using prior clinical knowledge of analyzing the quantity of stones under different morphologies. This strategy combines domain-specific lesion semantic features to construct a density space, and then uses a discrete neighbor difference function to measure the semantic density difference between lesions within the density space. Through the interaction between semantic density difference information and lesion semantic features, this embodiment can adaptively and automatically perform quantity analysis on stones of different morphologies. In kidney stone volume assessment, since kidney stones are extremely irregular objects, regular volume calculation methods are not suitable for kidney stone volume assessment. Therefore, a graph combination segmentation algorithm is designed to accurately assess the kidney stone volume. In the designed algorithm, based on the pixel spacing and layer thickness in the CT plain scan image, the pixels of the kidney stone in the cross section of a single CT plain scan image are restored to their actual size. Then, through a graphic combination segmentation algorithm, the kidney stone is segmented into an irregular cube composed of multiple triangular prisms and cubes. The total volume of the triangular prisms and cubes is the volume of the kidney stone on the cross section of a single CT plain scan image. Finally, based on the connected component information obtained in the kidney stone quantity assessment, the volume of the kidney stone is calculated.

[0172] Existing methods for assessing kidney stones primarily rely on a simple determination of the association between the contours of kidney stones in sequential CT images and an approximate method to calculate volume. These methods are only applicable to patients with kidney stones and cannot adapt to real-world clinical situations. Furthermore, existing methods focus solely on volume assessment, lacking research on the morphology and number of kidney stones. Therefore, the kidney stone assessment method implemented in this embodiment comprehensively assesses the morphology, number, and volume of kidney stones. Utilizing professional diagnostic knowledge, it customizes a burden assessment method tailored to different stone morphologies from a clinical perspective, achieving automated and precise assessment of kidney stone burden.

[0173] This embodiment has the following technical features:

[0174] Lesion boundary constraint method: It can automatically identify and delineate the lesion boundary outline, and then use morphological operations (such as opening and closing operations) to further refine the lesion boundary, and then combine the texture, size and other features of the lesion to determine the attributes of the lesion area.

[0175] The method for assessing the burden of kidney stones involves two main approaches. First, it utilizes prior clinical knowledge to design a morphological correlation analysis algorithm for kidney stones, exploring the correlations between lesion regions to enable adaptive, automated morphological analysis. Second, it incorporates semantic features specific to the kidney stone domain to develop morphology-specific adaptive analysis strategies. These strategies use neighborhood discrepancy functions to measure semantic density differences between lesion regions, combined with validity validation methods to screen clusters that match stone characteristics. This allows for adaptive, automated quantitative analysis of kidney stones with different morphologies. Third, based on the cluster results from the quantitative analysis, the method uses a graphical combination method to calculate the volume of each lesion region within a cluster, and then calculates the total volume of the entire cluster to achieve kidney stone volume assessment.

[0176] Reference Figure 2 This application also provides a kidney stone burden assessment device based on CT plain scan images, which can realize the above-mentioned kidney stone burden assessment method based on CT plain scan images. The device includes:

[0177] The dataset construction unit is used to automatically parse and record the semantic coordinates of each lesion region to construct a structured dataset, wherein each lesion region is the semantic boundary of the suspected kidney stone in each CT plain scan image in the CT plain scan image sequence.

[0178] The lesion morphology judgment unit is used to construct a local feature space for lesions and to explore the corresponding semantic features in the local feature space through morphological methods to judge the lesion morphology.

[0179] The kidney stone number determination unit is used to construct a density space matrix using the semantic features corresponding to the lesions, and to design a search strategy corresponding to the lesion morphology to calculate the optimal semantic density in the density space matrix, and then obtain the number of kidney stones through an unsupervised clustering algorithm.

[0180] The kidney stone volume determination unit is used to automatically measure the volume of kidney stones based on the analysis results of their morphology and number, using graphical methods.

[0181] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0182] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned method for assessing kidney stone burden based on CT plain scan images. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0183] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0184] Please see Figure 3 , Figure 3 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0185] The processor 301 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0186] The memory 302 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 302 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 302 and called and executed by the processor 301 to execute the method for assessing kidney stone burden based on CT plain scan images according to the embodiments of this application.

[0187] Input / output interface 303 is used to implement information input and output;

[0188] The communication interface 304 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0189] Bus 305 transmits information between various components of the device (e.g., processor 301, memory 302, input / output interface 303, and communication interface 304);

[0190] The processor 301, memory 302, input / output interface 303, and communication interface 304 are connected to each other within the device via bus 305.

[0191] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for assessing kidney stone burden based on CT plain scan images.

[0192] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0193] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0194] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0195] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0196] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; 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.

[0197] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0198] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0199] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0200] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0201] The units described above 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0202] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0203] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or 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 multiple 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 of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0204] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for kidney stone burden assessment based on CT scan images, characterized in that, The method includes the following steps: The semantic coordinates of each lesion region are automatically parsed and recorded to construct a structured dataset, where each lesion region is the semantic boundary of the suspected kidney stone in each CT plain scan image sequence. A local feature space is constructed for the lesion, and the corresponding semantic features are explored in the local feature space using morphological methods to determine the lesion morphology; A density space matrix is ​​constructed using the semantic features corresponding to the lesions, and a search strategy corresponding to the lesion morphology is designed to calculate the optimal semantic density in the density space matrix. Then, the number of kidney stones is obtained through an unsupervised clustering algorithm. Based on the analysis of the morphology and number of kidney stones, an automated volume measurement of kidney stones was achieved using graphical methods. The construction of the density space matrix using the semantic features corresponding to the lesions includes the following steps: A density space matrix is ​​obtained by calculating the semantic density between the lesion regions based on their three-dimensional semantic coordinates; wherein, the semantic density is used to describe the degree of spatial correlation between the lesion regions. The search strategy corresponding to the designed lesion morphology calculates the optimal semantic density in the density space matrix, and then obtains the number of kidney stones through an unsupervised clustering algorithm, including the following steps: Based on the specific morphology of the lesion, the domain-specific morphological structure and corresponding anchor point image sequence number are obtained from the lesion area quantity dictionary and recorded in a dictionary manner; among them, the corresponding morphological structure of interlaced lesions is parallel, and the corresponding morphological structure of staghorn lesions is bifurcation and adhesion. Based on different lesion morphologies and corresponding domain-specific morphological structures, the corresponding search objects are calculated. The target search matrix is ​​obtained by adjusting the density space matrix according to the search object; Based on the specific morphology of the lesion, the semantic density difference between the lesion regions is calculated by traversing the target search matrix. If the semantic density difference exceeds the density search upper limit of the target search matrix, the current density search upper limit is recorded. The minimum semantic density is calculated based on the density search upper limit of the current batch, and then compared with the optimal semantic density of the previous batch to obtain the optimal semantic density. Based on the optimal semantic density and lesion characteristics, the lesion regions with achievable density are grouped together using an unsupervised clustering algorithm to obtain one or more clusters, and the number of clusters is the number of lesions; wherein, the lesion characteristics are that the lesions exist in at least two consecutive CT plain scan images.

2. The method of assessing kidney stone burden based on CT scan images according to claim 1, wherein, The automatic parsing and recording of semantic coordinates of each lesion region to construct a structured dataset includes the following steps: The attributes of each lesion region are measured using morphological methods, and then the two-dimensional semantic coordinates of each lesion region are analyzed and recorded. Based on the sequence number of the CT plain scan image sequence, the two-dimensional semantic coordinates of each lesion region are expanded into three-dimensional semantic coordinates, forming a structured dataset.

3. The method of assessing kidney stone burden based on CT scan images as claimed in claim 1, wherein, The process of constructing a local feature space for the lesion and exploring corresponding semantic features in the local feature space using morphological methods to determine the lesion morphology includes the following steps: The number of the lesion regions in the CT scan image sequence is counted and recorded by means of a dictionary; if the maximum number of the lesion regions in the CT scan image is less than or equal to 1, the shape of the lesion in the CT scan image is determined as an isolated lesion; if the maximum number of the lesion regions in the CT scan image is greater than 1, the shape of the lesion in the CT scan image is temporarily determined as an interlaced lesion or a stag-horn lesion; For the lesion temporarily determined as an interlaced lesion or a stag-horn lesion, all the CT scan images in which the number of the lesion regions is greater than 1 are traversed, and the lesion regions contained in the CT scan images before and after the anchor point are associated to construct a local feature space; The in-degree and out-degree of the anchor point in each local feature space are calculated by a degree analysis method; if the out-degree and the in-degree of the lesion region are the same, the local feature space is marked as parallel; if the out-degree of the lesion region is greater than the in-degree, the local feature space is marked as bifurcation; if the in-degree of the lesion region is greater than the out-degree, the local feature space is marked as adhesion; The marks of all the local feature spaces are traversed; if there is no bifurcation or adhesion mark, the shape of the lesion in the CT scan image is determined as an interlaced lesion; if there is a bifurcation or adhesion mark, the shape of the lesion in the CT scan image is determined as a stag-horn lesion.

4. The method of claim 1, wherein the method further comprises: The method further comprises the following steps: Verifying whether the cluster belongs to a kidney stone lesion; Counting the number of the clusters belonging to the kidney stone lesion as the number of kidney stones.

5. The method of assessing kidney stone burden based on CT scan images as claimed in claim 1, wherein, According to the shape and number analysis results of the kidney stones, the automatic volume measurement of the kidney stones is realized by using a graph method, comprising the following steps: The pixels of the lesion region belonging to the kidney stone are classified into foreground and background, and then the coverage rate between each pixel and the surrounding adjacent pixels of the region belonging to the foreground is calculated by using a graph method; wherein the foreground is the kidney stone, and the background is the non-kidney stone; The coverage rate of each lesion region is obtained by traversing all the pixels of the lesion region belonging to the kidney stone, and then the volume of the kidney stone is calculated by combining the slice thickness and the pixel spacing of the CT scan image.

6. A device for assessing kidney stone burden based on CT scan images, characterized by, The device comprises: A data set construction unit for automatically analyzing and recording the semantic coordinates of each lesion region, and constructing a structured data set, wherein each lesion region is the semantic boundary of a suspected kidney stone in each CT scan image in a CT scan image sequence; A lesion shape judgment unit for constructing a local feature space for the lesion, and exploring the corresponding semantic features in the local feature space to judge the shape of the lesion by using a morphological method; A kidney stone number determination unit for constructing a density space matrix by using the semantic features corresponding to the lesion, and designing a search strategy corresponding to the shape of the lesion to calculate the optimal semantic density in the density space matrix, and then obtaining the number of kidney stones by using an unsupervised clustering algorithm; A kidney stone volume determination unit for realizing the automatic volume measurement of the kidney stone by using a graph method according to the shape and number analysis results of the kidney stone; The density space matrix constructed by using the semantic features corresponding to the lesion comprises the following steps: The semantic density between the various lesion regions is calculated according to the three-dimensional semantic coordinates of the various lesion regions, and a density space matrix is obtained; wherein the semantic density is used to describe the correlation degree between the lesion regions in space; The search strategy corresponding to the designed lesion morphology is used to calculate the optimal semantic density in the density space matrix, and then an unsupervised clustering algorithm is used to obtain the number of kidney stones, including the following steps: According to the specific morphology of the lesion, the field-specific morphological structure and the corresponding anchor image sequence number are obtained from the number dictionary of the lesion regions, and are recorded in the form of a dictionary; wherein the corresponding morphological structure of the interlaced lesion is parallel, and the corresponding morphological structure of the antler-shaped lesion is bifurcation and adhesion; Based on different lesion morphologies and corresponding field-specific morphological structures, the corresponding search objects are calculated; According to the search objects, the density space matrix is adjusted to obtain a target search matrix; Based on the specific morphology of the lesion, the semantic density difference between the lesion regions is calculated by traversing the target search matrix, and if the semantic density difference exceeds the density search upper limit of the target search matrix, the current density search upper limit is recorded; The minimum semantic density is calculated according to the density search upper limit of the current batch, and the minimum semantic density is compared with the optimal semantic density of the previous batch to obtain the optimal semantic density; According to the optimal semantic density and the lesion characteristics, the lesion regions reachable by density are gathered together by an unsupervised clustering algorithm to obtain one or more clusters, and the number of clusters is the number of lesions; wherein the lesion characteristics are that the lesion exists in at least two consecutive CT plain scan images.

7. An electronic device, comprising: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method of any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 7. The computer program is executed by the processor to implement the method of any one of claims 1 to 5.

Images