Mark point detection post-processing method and device, equipment and storage medium

The marker point detection post-processing method addresses inaccuracies in deep learning networks by identifying and correcting abnormal points, improving marker point detection accuracy and reducing costs through automated processes.

CN119090793APending Publication Date: 2024-12-06WUHAN UNITED IMAGING HEALTHCARE SURGICAL TECH CO LTD

Patent Information

Application Number
CN202310664990.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-06-05
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Existing deep learning networks for physiological marker detection in surgical planning produce inaccurate and unstable marker point outputs, often leading to false detections and increased costs due to the need for extensive training data and manual review.

Method used

A marker point detection post-processing method that includes identifying and correcting abnormal points using distance matrices, gradient fields, and signed distance fields to improve accuracy and stability of marker point detection.

Benefits of technology

The method effectively corrects marker point deviations and jumps, enhancing detection accuracy and reducing costs by automating the identification and correction of inaccuracies in marker point outputs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090793A_ABST
    Figure CN119090793A_ABST
Patent Text Reader

Abstract

The invention relates to a mark point detection post-processing method and device, equipment and a storage medium, and the method comprises the steps: obtaining a mark point detection result which comprises the position of a to-be-detected mark point, recognizing an abnormal point in the to-be-detected mark point according to the mark point detection result, and correcting the to-be-detected mark point according to the abnormal point. The problem that the mark points output by the deep learning network are inaccurate is solved, and correction of the mark points output by the deep learning network is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of physiological marker point detection, and in particular to a marker point detection post-processing method, device, equipment and storage medium. Background Art

[0002] In the application scenario of surgical planning, the calculation of physiological parameters is one of the important prerequisites for formulating personalized surgical planning plans, and physiological markers are important information to assist in calculating physiological parameters. In the current related technologies, deep learning technology is mainly used to detect physiological markers, and the trained neural network model is used to extract physiological marker information from the target tissue using the trained marker detection model. However, this method is limited by factors such as data volume, network structure, and data modality. The results that rely solely on the output of the deep learning network cannot stably output marker location information, and the results are sometimes mixed with false detection or poorly accurate positioning marker information.

[0003] In the related art, for the physiological markers output by the deep learning network, the method of improving the network structure is mainly used to improve the detection accuracy of the markers. However, this method requires sufficient training data samples. Or the method of manually reviewing the marker information output by the deep learning network is adopted, which will undoubtedly increase the cost of using the system and cause a waste of human resources.

[0004] At present, no effective solution has been proposed for the problem of inaccurate marking points output by deep learning networks in related technologies. Summary of the invention

[0005] The embodiments of the present application provide a marker point detection post-processing method, apparatus, device and storage medium to solve the problem of inaccurate marker points output by a deep learning network in the related art.

[0006] In a first aspect, an embodiment of the present application provides a marker point detection post-processing method, characterized by comprising:

[0007] Obtaining a marker point detection result, wherein the marker point detection result includes a position of a marker point to be detected;

[0008] According to the marking point detection result, identifying abnormal points in the marking points to be tested;

[0009] The marked point to be measured is corrected according to the abnormal point.

[0010] In some embodiments, according to the marker detection result, identifying abnormal points in the marker to be detected includes:

[0011] According to the detection result of the marking points, a first distance matrix and a second distance matrix are obtained, wherein the first distance matrix includes first distance data between each of the marking points to be measured and the rest of the marking points to be measured, and the second distance matrix includes the distance between each standard marking point and the rest of the standard marking points;

[0012] According to the first distance matrix and the second distance matrix, determining whether there are abnormal points in the marked points to be tested;

[0013] In the case that there are abnormal points in the mark points to be measured, the abnormal points in the mark points to be measured are identified.

[0014] In some embodiments, judging whether there are abnormal points in the marker points to be tested according to the first distance matrix and the second distance matrix includes:

[0015] Calculate the product of a plurality of first distance data between each of the to-be-tested marking points and the remaining to-be-tested marking points in the first distance matrix and a plurality of second distance data between each of the corresponding standard marking points and the remaining standard marking points in the second distance matrix;

[0016] Add the products of a plurality of the first distance data and the corresponding second distance data under the same mark point to be measured, to obtain the similarity between each mark point to be measured and the corresponding standard mark point;

[0017] According to the similarity, it is determined whether there are abnormal points in the marked points to be tested.

[0018] In some embodiments, when there are abnormal points in the marker points to be tested, identifying the abnormal points in the marker points to be tested includes:

[0019] Calculate the absolute difference between a plurality of first distance data between each of the to-be-tested marking points and the remaining to-be-tested marking points in the first distance matrix and a plurality of second distance data between the corresponding each of the standard marking points and the remaining standard marking points in the second distance matrix;

[0020] Select the maximum absolute difference value under each of the mark points to be measured, and record the position index of the maximum absolute difference value;

[0021] Perform statistics on the position indexes, select the position index with the largest number of occurrences, and select the abnormal point according to the position index with the largest number of occurrences;

[0022] The abnormal points are eliminated, and the abnormal point identification and elimination are continued for the remaining marked points to be tested until all the abnormal points in the marked points to be tested are identified.

[0023] In some embodiments, calibrating the marker point to be measured includes:

[0024] Eliminate the first distance data of the abnormal point in the first distance matrix and the second distance data of the standard marking point corresponding to the abnormal point in the second distance matrix;

[0025] Calculate a mapping relationship between first distance data in the first distance matrix after elimination and second distance data in the second distance matrix after elimination;

[0026] According to the mapping relationship and the coordinates of the standard point corresponding to the abnormal point, the first corrected coordinates of the abnormal point are obtained.

[0027] In some embodiments, calibrating the marker point to be measured further includes:

[0028] Eliminate the first distance data of the abnormal point in the first distance matrix and the second distance data of the standard marking point corresponding to the abnormal point in the second distance matrix;

[0029] Calculate the proportionality coefficient of each first distance data in the first distance matrix after elimination relative to each corresponding second distance data in the second distance matrix after elimination, and obtain the average value of the proportionality coefficients under each of the marker points to be measured as a scaling factor;

[0030] According to the scaling factor and the second distance data in the second distance matrix, the theoretical distance data from the outlier point to the remaining marked points to be measured are calculated, and the second corrected coordinates of the outlier point are obtained by solving the theoretical distance data.

[0031] In some embodiments, the method further comprises:

[0032] Generate a gradient field according to the image where the marker point to be measured is located;

[0033] The coordinates of the marker point to be measured are corrected according to the gradient field so that the marker point to be measured converges to the local maximum point of the gradient field.

[0034] In some embodiments, the method further comprises:

[0035] Segmenting the image where the marker point to be measured is located, and generating a signed distance field according to the segmentation result;

[0036] The marker point to be measured is corrected according to the signed distance field, so that the marker point to be measured converges to a point where a local minimum value of the signed distance field is located.

[0037] In a second aspect, an embodiment of the present application provides a marker point detection post-processing device, comprising:

[0038] An acquisition module, used for acquiring a mark point detection result, wherein the mark point detection result includes a position of a mark point to be detected;

[0039] An identification module, used for identifying abnormal points among the marker points to be tested according to the marker point detection result;

[0040] A correction module is used to correct the marked point to be tested according to the abnormal point.

[0041] In a third aspect, an embodiment of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the marker point detection post-processing method as described in any one of the first aspects above when executing the computer program.

[0042] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, the steps of the marker point detection post-processing method as described in any one of the above-mentioned first aspects are implemented.

[0043] Compared with the related art, the marker point detection post-processing method, device, equipment and storage medium provided in the embodiments of the present application obtain marker point detection results, wherein the marker point detection results include the positions of the marker points to be tested, identify abnormal points in the marker points to be tested according to the marker point detection results, and correct the marker points to be tested according to the abnormal points, thereby solving the problem of inaccurate marker points output by the deep learning network and realizing the correction of the marker points output by the deep learning network.

[0044] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0046] Figure 1 It is a hardware structure block diagram of a terminal of a marking point detection post-processing method according to an embodiment of the present application;

[0047] Figure 2 is a flow chart of a marker point detection post-processing method according to an embodiment of the present application;

[0048] Figure 3 This is an overall flow chart of a post-processing method for marker point detection according to an embodiment of the present application;

[0049] Figure 4 is a schematic diagram of a marking point deviation phenomenon in one embodiment of the present application;

[0050] Figure 5 It is a schematic diagram of a marking point jump phenomenon of an embodiment of the present application;

[0051] Figure 6 is a flow chart of a method for automatically identifying abnormal points according to an embodiment of the present application;

[0052] Figure 7 is a flow chart of an abnormal point error correction method according to an embodiment of the present application;

[0053] Figure 8 This is a flowchart of an abnormal point error correction method according to another embodiment of the present application.

[0054] Fig. 9 is a flow chart of a method for automatically adhering marking points according to an embodiment of the present application;

[0055] Fig.10 is a flow chart of a method for automatically adhering marking points according to another embodiment of the present application;

[0056] Fig.11 It is a hardware structure diagram of a marker point detection post-processing device according to an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0058] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0059] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0060] Unless otherwise defined, the technical terms or scientific terms involved in this application should be understood by people with ordinary skills in the technical field to which this application belongs. The words "one", "a", "a kind of", "the" and the like involved in this application do not indicate a quantitative limitation and may represent the singular or plural. The terms "include", "comprise", "have" and any of their variations involved in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units inherent to these processes, methods, products or devices. The "multiple" involved in this application refers to two or more.

[0061] The method embodiment provided in this embodiment can be executed in a terminal, a computer or a similar computing device. For example, running on a terminal, Figure 1 is a hardware structure block diagram of a terminal of the marker point detection post-processing method of this embodiment. Figure 1 As shown, the terminal may include one or more ( Figure 1 Only one is shown in the figure) processor 102 and memory 104 for storing data, wherein processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA. The above terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is for illustration only and does not limit the structure of the above terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0062] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the post-processing method of the marker detection in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0063] The transmission device 106 is used to receive or send data via a network. The above network includes a wireless network provided by the communication provider of the terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0064] At present, the mainstream solution for automatic detection of anatomical landmarks in medical images is deep neural network detection. However, when the training samples of the deep learning network are insufficient, the robustness of the detection results it outputs is poor. When faced with data collected by different hospitals, different equipment, and different technicians, it often lacks stability and often causes problems with misdetection of landmarks.

[0065] Therefore, in this embodiment, a post-processing method for marker detection is provided. Figure 2 is a flowchart of the post-processing method of the marker point detection in the embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0066] Step S201, obtaining a marker point detection result, wherein the marker point detection result includes a position of a marker point to be detected.

[0067] Among them, the marking points include the marking points to be tested output by the deep learning network according to the medical image, and the positions of the marking points to be tested include the position coordinates of the marking points to be tested output by the deep learning network in the medical image.

[0068] Step S202: identifying abnormal points among the marker points to be tested according to the marker point detection result.

[0069] Among them, according to the position coordinates of each marker point to be tested in the medical image, the distribution of the marker point to be tested on the medical image is obtained. The marker point corresponding to the marker point to be tested on the same medical image is selected, and the position coordinates of each standard marker point on the medical image are obtained, so as to obtain the distribution of the standard marker point on the medical image. Among them, the standard marker point can be obtained by manual marking, or after obtaining multiple accurate standard marker point distributions, the distribution template of the standard marker point is obtained with the help of a statistical shape model, so as to obtain the position distribution information of the standard marker point. After obtaining the position distribution of the marker point to be tested and the standard marker point, it is judged whether there are abnormal points in the marker point to be tested according to the position relationship between the marker point to be tested and the standard marker point.

[0070] Step S203: calibrate the mark point to be measured according to the abnormal point.

[0071] Among them, there are two main abnormal phenomena in the marking points obtained after deep learning network processing. One is the deviation of marking points, such as Figure 4 As shown in the figure, the marker point appears inside or outside the target tissue near the target position and cannot be attached to the surface of the target tissue. Another phenomenon is the marker point jump phenomenon, such as Figure 5 As shown, the markers appear in an area far from the target location.

[0072] In the above steps S201 to S203, by obtaining the marker point detection results, wherein the marker point detection results include the positions of the marker points to be tested, identifying the abnormal points in the marker points to be tested according to the marker point detection results, and correcting the marker points to be tested according to the abnormal points, the automatic detection and correction of the deviation and jump phenomena occurring in the marker points output by deep learning are realized, the efficiency of abnormal point processing is improved, and the abnormal points of different abnormal types can be adapted to be corrected by different methods, which effectively reduces the occurrence of different abnormal phenomena in the marker points to be tested, solves the problem of inaccurate marker points output by the deep learning network, reduces the cost of marker point detection, and improves the detection accuracy of marker points.

[0073] Figure 3 is an overall flow chart of the post-processing method for marker point detection in an embodiment of the present application. Figure 3As shown, the marker point detection post-processing method of the embodiment of the present application mainly includes three parts, namely, automatic identification of outliers, automatic error correction of outliers, and automatic bordering of marker points. Automatic identification of outliers is used to identify outliers in the marker points to be tested, outlier error correction is used to correct outliers that deviate from the target position, and automatic bordering of marker points is used to correct the marker points to be tested that are not bordering with the surface of the target tissue. First, the detection results output by the deep learning network are obtained, and the outliers of the marker point detection results are automatically identified. If the outliers exist, the outliers are automatically corrected, otherwise the outliers are skipped, and finally all marker points are automatically bordered. Through the marker point detection post-processing method of the embodiment of the present application, the marker points output by the deep neural network can be comprehensively corrected, the deviation and jump phenomena of the marker points are avoided, and the accuracy of the marker points output by the deep learning network is improved.

[0074] In some of the embodiments, according to the marker point detection results, identifying abnormal points among the marker points to be tested includes: according to the marker point detection results, obtaining a first distance matrix and a second distance matrix, wherein the first distance matrix includes first distance data between each marker point to be tested and the remaining marker points to be tested, and the second distance matrix includes the distance between each standard marker point and the remaining standard marker points; according to the first distance matrix and the second distance matrix, judging whether there are abnormal points among the marker points to be tested; if there are abnormal points among the marker points to be tested, identifying the abnormal points among the marker points to be tested.

[0075] Among them, the first distance data includes the Euclidean distance between each mark point to be tested and all other mark points to be tested, and the second distance data includes the Euclidean distance between each standard mark point corresponding to each mark point to be tested and all other standard mark points. The first distance data and the second distance data are recorded in the form of a matrix to generate an m×n first distance matrix and an m×n second distance matrix, wherein each column in the distance matrix represents a mark point, and each row represents the distance data from a mark point corresponding to the column to another mark point, and the first distance data of each position in the first distance matrix corresponds to the second distance data of each position in the second distance matrix. By recording the distance data in the form of a matrix, the position distribution of the mark points to be tested and the standard mark points is converted into an intuitive matrix form, which is convenient for subsequent data processing and calculation, and improves the efficiency of post-processing of mark point detection.

[0076] In some embodiments, judging whether there are abnormal points in the mark points to be tested based on the first distance matrix and the second distance matrix includes: calculating the product of multiple first distance data between each mark point to be tested and the remaining mark points to be tested in the first distance matrix and multiple second distance data between each corresponding standard mark point and the remaining standard mark points in the second distance matrix; adding the products of multiple first distance data and corresponding second distance data under the same mark point to be tested to obtain the similarity between each mark point to be tested and the corresponding standard mark point; judging whether there are abnormal points in the mark points to be tested based on the similarity.

[0077] Among them, firstly, the first distance matrix A T With the second distance matrix A GT After normalization, we get A N_GT and A N_T , A N_GT and A N_T Perform point multiplication processing to obtain the point product matrix D. The practical significance of this step is to calculate the similarity between each corresponding first distance data and second distance data in the first distance matrix and the second distance matrix, so as to obtain the similarity between the mark point to be tested and the corresponding standard mark point. After obtaining the point product matrix D, calculate the sum of each column of the matrix D. If there is a column whose sum is less than the preset value T, it is considered that the similarity between the mark point to be tested and the standard mark point corresponding to the column does not meet the preset expectation, and the mark point to be tested corresponding to the column is determined to be an abnormal point, wherein the preset value T is an empirical value obtained through experiments.

[0078] In the above process, the similarity between the mark point to be tested and the corresponding standard mark point is calculated by matrix operation, so as to select the abnormal point according to the similarity, simplifying the selection process of the abnormal point to the operation process of the distance matrix, realizing the automatic identification of the abnormal point and improving the efficiency of the automatic identification of the abnormal point.

[0079] In some of the embodiments, when there are abnormal points in the mark points to be tested, identifying the abnormal points in the mark points to be tested includes: calculating the absolute difference between multiple first distance data between each mark point to be tested and the remaining mark points to be tested in the first distance matrix, and multiple second distance data between each corresponding standard mark point and the remaining standard mark points in the second distance matrix; selecting the largest absolute difference under each mark point to be tested, and recording the position index of the largest absolute difference; counting the position indexes, selecting the position index with the largest number of occurrences, and selecting the abnormal points according to the position index with the largest number of occurrences; eliminating the abnormal points, and continuing to identify and eliminate the abnormal points for the remaining mark points to be tested until all the abnormal points in the mark points to be tested are identified.

[0080] Among them, when there are outliers in the marked points to be tested, the outlier voting mechanism is entered. The specific voting process is: calculate the absolute difference matrix A between the first distance matrix and the second distance matrix absDiff = abs(A GT -A T ), select the absolute difference matrix A absDiff The maximum value of each column in the matrix is ​​obtained, and the position index Ind of the column corresponding to the maximum value is recorded, that is, the vote of the marker point to be tested corresponding to the column. After traversing the maximum values ​​of all columns of the absolute difference matrix, the number of position indexes that appear in each column is counted, that is, the number of votes for the position corresponding to each marker point to be tested. The position index appears the most times, that is, the more votes the marker point to be tested at this position gets. The marker point to be tested with the most votes is selected as the outlier. During the voting process, only one outlier is selected for each vote. For the case of multiple outliers, the selected outliers are eliminated from the marker points to be tested, and the remaining marker points to be tested are continuously identified and eliminated until all the outliers in the marker points to be tested are identified, and the identification and selection of all outliers can be completed.

[0081] Through multiple voting, each time the mark point to be tested with the most votes is selected as the outlier point. There is no need to sort all the mark points to be tested according to the number of votes. Only one mark point to be tested with the most votes is selected as the outlier point each time, which reduces the complexity of the outlier point automatic identification algorithm and improves the calculation speed and efficiency.

[0082] in, Figure 6 is a flow chart of the method for automatically identifying abnormal points in an embodiment of the present application. Figure 6 As shown, by generating distance matrices for the marker points to be tested and the standard marker points respectively, by operating the distance matrices, it is determined whether anomalies exist, and when anomalies exist, all anomalies are voted out, thereby realizing automatic identification of anomalies in the marker points to be tested and improving calculation efficiency.

[0083] In some embodiments, correcting the marker points to be measured includes: eliminating the first distance data of the abnormal points in the first distance matrix and the second distance data of the standard marker points corresponding to the abnormal points in the second distance matrix; calculating the mapping relationship between the first distance data in the first distance matrix after elimination and the second distance data in the second distance matrix after elimination; and obtaining the first corrected coordinates of the abnormal points according to the mapping relationship and the coordinates of the standard points corresponding to the abnormal points.

[0084] After the abnormal point is automatically identified, the identified abnormal point needs to be corrected to the vicinity of the area where the point normally appears. This embodiment provides an abnormal point correction method, firstly obtaining the information of the abnormal point, including the first distance data of the abnormal point, removing the first distance data of the abnormal point from the first distance matrix, and obtaining the first distance matrix D after the removal.T , and remove the second distance data of the standard marking points corresponding to the abnormal points from the second distance matrix to obtain the second distance matrix D after removal GT ; Calculate D T With D GT In the embodiment of the present application, the Pullman transform T is used to calculate the Pullman transform T from the second distance matrix after elimination to the first distance matrix after elimination, so that D T =T*D GT After transformation, obtain the point coordinates P of the standard marking point corresponding to the abnormal point that has been removed in the marking point to be tested GT , restore the abnormal point P in the marked point to be tested o , that is, P o =T*P GT .

[0085] in, Figure 7 is a flow chart of an abnormal point error correction method according to an embodiment of the present application. Figure 7 As shown, by obtaining the outlier information, the outliers in the marked points to be tested are eliminated, the standard marked points are transformed according to the mapping relationship, and the coordinates of the standard marked points corresponding to the outliers are added to the marked points to be tested, thereby completing the restoration of the outlier coordinates, realizing the correction of the outliers, and improving the accuracy of the marked points output by the deep learning network.

[0086] In some embodiments, the correction of the marker points to be measured also includes: eliminating the first distance data of the abnormal points in the first distance matrix and the second distance data of the standard marker points corresponding to the abnormal points in the second distance matrix; calculating the proportional coefficient of each first distance data in the first distance matrix after elimination relative to each corresponding second distance data in the second distance matrix after elimination, and obtaining the average value of the proportional coefficient under each marker point to be measured as the scaling factor; calculating the theoretical distance data from the abnormal point to the remaining marker points to be measured according to the scaling factor and the second distance data in the second distance matrix, and solving according to the theoretical distance data to obtain the second corrected coordinates of the abnormal point.

[0087] After the abnormal point is automatically identified, the identified abnormal point needs to be corrected to the vicinity of the area where the store normally appears. This embodiment provides another abnormal point correction method, firstly obtaining abnormal point information, including first distance data of the abnormal point, removing the first distance data of the abnormal point from the first distance matrix, and obtaining the first distance matrix A after removal GT_fine , and remove the second distance data of the standard marking points corresponding to the abnormal points from the second distance matrix to obtain the second distance matrix A after removal T_fine, divide the second distance matrix after removing the outliers by the first distance matrix after removing the outliers to obtain the scaling matrix S, take the average of each column of the scaling matrix to obtain the scaling factor a. Multiply the second distance matrix corresponding to the standard marking point by the scaling factor a, calculate the theoretical true distance from the outliers to the remaining normal marking points, and restore the true first distance matrix A based on the theoretical true distance True Based on the least squares method, the theoretical true coordinates of the outliers are solved according to the true first distance matrix.

[0088] in, Figure 8 is a flow chart of an abnormal point error correction method according to another embodiment of the present application. Figure 8 As shown, by obtaining the outlier information, the outliers in the marked points to be tested are eliminated, and the scaling factors of the first distance matrix and the second distance matrix after eliminating the outliers are calculated. The real distance from the outlier to each normal marked point is calculated according to the scaling factor, and the theoretical real coordinates of the outlier are solved, thereby completing the solution of the outlier coordinates, realizing the correction of the outliers, and improving the accuracy of the marked points output by the deep learning network.

[0089] After correcting the selected abnormal points, all the marker points to be tested, regardless of whether they have been corrected, need to be processed once to ensure that the marker points to be tested appear on the target tissue surface. The automatic edge-adaptation of marker points is mainly based on the position information of the marker points to be tested, the medical image containing the tissue, or the target tissue segmentation result, and the coordinate information of the target tissue surface closest to the marker points to be tested is calculated through a search algorithm. The present application mainly provides the following two embodiments to realize the automatic edge-adaptation of marker points to be tested.

[0090] In one of the embodiments, the method further includes: generating a gradient field according to the image where the marker point to be measured is located; and correcting the coordinates of the marker point to be measured according to the gradient field so that the marker point to be measured converges to a local maximum point of the gradient field.

[0091] in, Fig. 9 : is a flow chart of a method for automatically pasting marking points according to an embodiment of the present application. Fig. 9 As shown, the method includes: obtaining image data and coordinate information of the marker points to be measured, and generating a gradient field according to the image data. Based on the generated gradient field, the marker points to be measured are corrected one by one by the meanshift algorithm (mean shift algorithm), and the marker points to be measured are converged to the local maximum point of the gradient field, so that the marker points to be measured are corrected to the surface of the target tissue, the marker point jump phenomenon is prevented, and the accuracy of the marker points output by the deep learning network is improved.

[0092] In another embodiment, the method further includes: segmenting the image where the marker point to be tested is located, and generating a signed distance field according to the segmentation result; and correcting the marker point to be tested according to the signed distance field so that the marker point to be tested converges to a point where a local minimum value of the signed distance field is located.

[0093] in, Fig.10 is a flow chart of a method for automatically pasting marking points according to another embodiment of the present invention. Fig.10 As shown, the method includes: obtaining the segmentation result of the target tissue and the coordinate information of the marker points to be measured, and generating a signed distance field according to the segmentation result of the target tissue. Based on the signed distance field, the marker points to be measured are corrected one by one by the gradient descent method, and the marker points to be measured are converged to the point where the local minimum of the signed distance field is located, so as to realize the correction of the marker points to the surface of the target tissue, prevent the occurrence of marker point jump phenomenon, and improve the accuracy of the marker points output by the deep learning network.

[0094] The present application also provides a post-processing device for marking point detection. Fig.11 is a hardware structure diagram of the marking point detection post-processing device of an embodiment of the present application, such as Fig.11 As shown, the device comprises:

[0095] An acquisition module 111 is used to acquire a marker point detection result, wherein the marker point detection result includes a position of a marker point to be detected;

[0096] The identification module 112 is used to identify abnormal points in the marker points to be tested according to the marker point detection results;

[0097] The correction module 113 is used to correct the marked points to be tested according to the abnormal points.

[0098] It should be noted that the above modules can be functional modules or program modules, and can be implemented by software or hardware. For modules implemented by hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0099] An embodiment of the present application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the marker point detection post-processing method provided in the above embodiments are implemented.

[0100] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the marker point detection post-processing method provided in the above-mentioned embodiments are implemented.

[0101] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0102] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0103] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A post-processing method for marker point detection, characterized in that: include: Obtaining a marker point detection result, wherein the marker point detection result includes a position of a marker point to be detected; According to the marking point detection result, identifying abnormal points in the marking points to be tested; The marked point to be measured is corrected according to the abnormal point.

2. The marker point detection post-processing method according to claim 1, characterized in that: According to the marker point detection result, identifying abnormal points in the marker points to be detected includes: According to the detection result of the marking points, a first distance matrix and a second distance matrix are obtained, wherein the first distance matrix includes first distance data between each of the marking points to be measured and the rest of the marking points to be measured, and the second distance matrix includes the distance between each standard marking point and the rest of the standard marking points; According to the first distance matrix and the second distance matrix, determining whether there are abnormal points in the marked points to be tested; In the case that there are abnormal points in the mark points to be measured, the abnormal points in the mark points to be measured are identified.

3. The post-processing method for marker point detection according to claim 2, characterized in that: Judging whether there are abnormal points in the marker points to be tested according to the first distance matrix and the second distance matrix includes: Calculate the product of a plurality of first distance data between each of the to-be-tested marking points and the remaining to-be-tested marking points in the first distance matrix and a plurality of second distance data between each of the corresponding standard marking points and the remaining standard marking points in the second distance matrix; Add the products of a plurality of the first distance data and the corresponding second distance data under the same mark point to be measured, to obtain the similarity between each mark point to be measured and the corresponding standard mark point; According to the similarity, it is determined whether there are abnormal points in the marked points to be tested.

4. The post-processing method for marker point detection according to claim 2, characterized in that: In the case where there are abnormal points in the mark points to be measured, identifying the abnormal points in the mark points to be measured includes: Calculate the absolute difference between a plurality of first distance data between each of the to-be-tested marking points and the remaining to-be-tested marking points in the first distance matrix and a plurality of second distance data between the corresponding each of the standard marking points and the remaining standard marking points in the second distance matrix; Select the maximum absolute difference value under each of the mark points to be measured, and record the position index of the maximum absolute difference value; Perform statistics on the position indexes, select the position index with the largest number of occurrences, and select the abnormal point according to the position index with the largest number of occurrences; The abnormal points are eliminated, and the abnormal point identification and elimination are continued for the remaining marked points to be tested until all the abnormal points in the marked points to be tested are identified.

5. The post-processing method for marker point detection according to claim 2, characterized in that: Correcting the mark point to be measured includes: Eliminate the first distance data of the abnormal point in the first distance matrix and the second distance data of the standard marking point corresponding to the abnormal point in the second distance matrix; Calculate a mapping relationship between first distance data in the first distance matrix after elimination and second distance data in the second distance matrix after elimination; According to the mapping relationship and the coordinates of the standard point corresponding to the abnormal point, the first corrected coordinates of the abnormal point are obtained.

6. The post-processing method for marker point detection according to claim 2, characterized in that: Correcting the mark point to be measured further includes: Eliminate the first distance data of the abnormal point in the first distance matrix and the second distance data of the standard marking point corresponding to the abnormal point in the second distance matrix; Calculate the proportionality coefficient of each first distance data in the first distance matrix after elimination relative to each corresponding second distance data in the second distance matrix after elimination, and obtain the average value of the proportionality coefficients under each of the marker points to be measured as a scaling factor; According to the scaling factor and the second distance data in the second distance matrix, the theoretical distance data from the abnormal point to the remaining marked points to be measured are calculated, and the second corrected coordinates of the abnormal point are obtained by solving the theoretical distance data.

7. The marker point detection post-processing method according to claim 1, characterized in that: The method further comprises: Generate a gradient field according to the image where the marker point to be measured is located; The coordinates of the marker point to be measured are corrected according to the gradient field so that the marker point to be measured converges to the local maximum point of the gradient field.

8. The marker point detection post-processing method according to claim 1, characterized in that: The method further comprises: Segmenting the image where the marker point to be measured is located, and generating a signed distance field according to the segmentation result; The marker point to be measured is corrected according to the signed distance field, so that the marker point to be measured converges to a point where a local minimum value of the signed distance field is located.

9. A post-processing device for marking point detection, characterized in that: include: An acquisition module, used for acquiring a mark point detection result, wherein the mark point detection result includes a position of a mark point to be detected; An identification module, used for identifying abnormal points among the marker points to be tested according to the marker point detection result; A correction module is used to correct the marked point to be tested according to the abnormal point.

10. A computer 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 computer program, the steps of the marker point detection post-processing method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the marker point detection post-processing method according to any one of claims 1 to 8 are implemented.

Citation Information

Patent Citations

  • Positioning Anatomical Landmarks In Volume Data Sets

    CN104112292A

  • Image correction method and device, electronic equipment and storage medium thereof

    CN116071272A

  • Brain MRI mark point detection method based on self-supervised learning and geometric constraint

    CN118447016A

  • Anomaly detection in dynamically evolving data and systems

    US9843596B1

Cited By

  • Joint landmark point identification method, computer equipment and computer program product

    CN120318218A

  • Method for identifying joint landmark points, computer equipment and computer program product

    CN120318218B