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

Abstract

The embodiment of the present application is applicable to the field of computer-assisted medical technology, and provides a method for identifying joint landmark points, a computer device, and a computer program product, the method comprising: acquiring medical image data with a first resolution; determining the initial position of each joint landmark point in the medical image data; aligning the initial landmark point set with the statistical shape model according to the initial position to obtain a target alignment matrix; in the case where there are missing landmark points in the medical image data, using the target alignment matrix to complete the missing landmark points; projecting each of the completed joint landmark points onto the joint model surface corresponding to the medical image data, and outputting joint landmark point identification result data with a second resolution. By adopting the above method, joint landmark points can be identified efficiently and accurately, and the accuracy and reliability of medical image data analysis and processing can be improved.

Description

Technical Field

[0001] The embodiments of the present application belong to the field of computer-assisted medical technology, and in particular, relate to a method for identifying joint landmark points, a computer device, and a computer program product. Background Art

[0002] Automatic recognition of human joint landmarks holds important applications in medical image analysis, orthopedic surgical planning, and assisted diagnosis. However, existing recognition methods suffer from issues such as inaccuracy, stability, and insufficient recognition rates. These issues collectively result in a low success rate for landmark recognition, severely impacting the accuracy and efficiency of subsequent medical analysis and diagnosis. The limitations of existing landmark recognition technology are particularly pronounced in complex cases, making it unable to meet the demands of precision medicine. Summary of the Invention

[0003] In view of this, embodiments of the present application provide a method for identifying joint landmark points, a computer device, and a computer program product to improve the efficiency and accuracy of identifying human joint landmark points.

[0004] A first aspect of an embodiment of the present application provides a method for identifying joint landmark points, comprising:

[0005] Acquiring medical image data having a first resolution, the first resolution being smaller than an original resolution of the medical image data;

[0006] determining an initial position of each joint landmark point in the medical image data;

[0007] According to the initial position, registering an initial landmark point set consisting of the joint landmark points with a pre-constructed statistical shape model to obtain a target registration matrix, wherein the statistical shape model is constructed based on a standard data set, the standard data set being a data set pre-labeled with joint landmark points;

[0008] In the case where there are missing landmark points in the medical image data, using the target registration matrix to complete the missing landmark points;

[0009] The completed joint landmark points are projected onto the joint model surface corresponding to the medical image data, and joint landmark point recognition result data with a second resolution is output, where the second resolution is greater than the first resolution.

[0010] A second aspect of an embodiment of the present application provides a device for identifying joint landmark points, comprising:

[0011] an acquisition module, configured to acquire medical image data having a first resolution, where the first resolution is smaller than an original resolution of the medical image data;

[0012] a determination module, configured to determine an initial position of each joint landmark point in the medical image data;

[0013] a registration module for registering, based on the initial positions, an initial landmark point set consisting of the joint landmark points with a pre-constructed statistical shape model to obtain a target registration matrix, wherein the statistical shape model is constructed based on a standard dataset, the standard dataset being a data set pre-labeled with joint landmark points;

[0014] a gap filling module, configured to fill in the missing landmark points using the target registration matrix when there are missing landmark points in the medical image data;

[0015] A projection module, configured to project the completed joint landmark points onto a joint model surface corresponding to the medical image data;

[0016] An output module is used to output joint landmark point recognition result data with a second resolution, where the second resolution is greater than the first resolution.

[0017] A third aspect of 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. When the processor executes the computer program, the computer device implements the method described in the first aspect above.

[0018] A fourth aspect of the embodiments of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a computer, the method described in the first aspect above is implemented.

[0019] A fifth aspect of the embodiments of the present application provides a computer program product, including a computer program, which, when executed, enables the method described in the first aspect to be executed.

[0020] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0021] In an embodiment of the present application, by processing medical imaging data into data with a lower first resolution, a computer device can use a deep neural network to determine the initial position of each joint point. On this basis, by aligning the initial landmark point set composed of each joint landmark point with a pre-built statistical shape model, a target alignment matrix can be obtained, so that the missing landmark points can be completed using the target alignment matrix. After projecting the completed joint landmark points onto the surface of the joint model, the computer device can output joint landmark point recognition result data with a higher second resolution to complete the recognition of all landmark points. Applying the recognition method provided in an embodiment of the present application, joint landmark points in medical imaging data can be efficiently and accurately identified, and improving the accuracy and reliability of medical imaging data processing helps provide a more solid foundation for the formulation of personalized treatment plans and the optimization of surgical planning.

[0022] The joint landmark point recognition method provided by the embodiments of this application has broad application prospects in the medical field, and can play a key role in orthopedic clinical practice in particular. For example, in terms of surgical planning, the precise landmark point recognition technology provided by the embodiments of this application can provide strong support for complex surgeries such as joint replacements. By accurately locating key anatomical structures, surgeons can develop more personalized and precise surgical plans, significantly improving the accuracy and success rate of intraoperative registration.

[0023] In addition, the joint landmark point identification method provided in the embodiment of the present application can also play an important role in the field of auxiliary diagnosis, especially in the early diagnosis and disease assessment of bone and joint diseases such as arthritis and osteoporosis. The application of this method can provide key reference points for the automated analysis of images such as CT and MRI, greatly improving the efficiency and accuracy of image interpretation. This is because accurate landmark point positioning helps the diagnostic system provide more standard data posture and observation surface, allowing doctors to more accurately analyze the patient's condition and thus implement timely intervention treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] To more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0025] Figure 1 is a schematic diagram of a method for identifying joint landmark points provided in an embodiment of the present application;

[0026] Figure 2 This is a schematic diagram of a deep neural network processing flow provided by an embodiment of the present application;

[0027] Figure 3 This is a schematic diagram of the process of segmenting medical image data using a deep neural network provided in an embodiment of the present application;

[0028] Figure 4 This is a schematic diagram of a process for constructing a shape statistical model provided by an embodiment of the present application;

[0029] Figure 5 This is a flowchart of landmark point registration and gap filling provided by an embodiment of the present application;

[0030] Figure 6 This is a schematic diagram of a joint landmark point projection process provided by an embodiment of the present application;

[0031] Figure 7 is a schematic diagram of a joint landmark point optimization process provided in an embodiment of the present application;

[0032] Figure 8 This is a schematic diagram of a joint landmark point recognition process provided by an embodiment of the present application;

[0033] Figure 9 is a schematic diagram of a joint landmark point recognition device provided by an embodiment of the present application;

[0034] Figure 10 This is a schematic diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present application with unnecessary details.

[0036] As mentioned above, there are many problems with joint landmark point recognition in related technologies. The main problems include:

[0037] 1. Recognition difficulties caused by anatomical variations: Due to anatomical abnormalities caused by pathological features such as pelvic dysplasia and joint fusion, traditional recognition methods based on fixed templates and heat map regression methods based on simple deep learning are difficult to adapt to clinical data, resulting in a significant decrease in recognition accuracy.

[0038] 2. Errors caused by inconsistent data collection posture: Patients are unable to maintain the standard scanning posture for various reasons, resulting in deviations between the spatial position relationship of the image data and the training data, which in turn affects the performance of the recognition algorithm.

[0039] 3. Other surgeries lead to complex and degraded images: Bone defects and artifacts caused by implants in other surgeries not only change the original bone anatomical structure, but also introduce a large number of metal artifacts, which greatly increases the difficulty of landmark point identification.

[0040] 4. Insufficient ability to handle missing landmark points: Existing methods are insufficient to deal with missing landmark points due to various reasons and cannot effectively fill in the missing key anatomical landmark points.

[0041] 5. Trade-off between computational efficiency and accuracy: Improving recognition accuracy usually requires more complex algorithms and higher computational costs, which conflicts with the real-time requirements in clinical practice.

[0042] To address these issues, the present invention provides an intelligent joint landmark point identification method that overcomes data collection bias, completes missing landmark points, and maintains computational efficiency. This method not only improves the accuracy and reliability of medical image analysis but also provides a more robust foundation for developing personalized treatment plans and optimizing surgical planning.

[0043] The technical solution of this application is described below through specific embodiments.

[0044] Reference Figure 1 , which shows a schematic diagram of a method for identifying joint landmark points provided by an embodiment of the present application, which may specifically include the following steps:

[0045] S101: Acquire medical image data with a first resolution.

[0046] This method can be applied to a computer device, that is, the executor of this method can be a computer device. By executing the various steps of the method provided in the embodiment of the present application, the computer device can efficiently and accurately identify joint landmark points in medical imaging data. The above-mentioned computer device can be a computer-assisted medical device, such as a medical device with medical imaging data processing function or other types of desktop computers, cloud servers and other devices. The embodiment of the present application does not limit the type of computer device. Unless otherwise specified, the joint landmark points or landmark points mentioned in the embodiment of the present application refer to the same features, namely joint landmark points.

[0047] Medical imaging data can be three-dimensional imaging data obtained by scanning a patient using imaging equipment. For example, computed tomography (CT) data can be obtained by scanning a patient's joints or the entire body using CT equipment. Therefore, the medical imaging data can be imaging data of a portion of a patient's joints or imaging data of the entire patient.

[0048] In a possible implementation of an embodiment of the present application, a computer device can be communicatively connected to an imaging device. After acquiring the patient's imaging data, an imaging device such as a CT device can transmit the data to a computer device for joint landmark point recognition. After receiving the original medical imaging data, the computer device can downsample the data to downsample the original resolution medical imaging data to data of a first resolution. The above-mentioned first resolution is smaller than the original resolution. For example, assuming that the original resolution of the medical imaging data acquired by the CT device is 1500×1500 pixels, it can be processed into data with a lower first resolution through downsampling, for example, a first resolution of 750×750 pixels. Through downsampling, the number of pixels to be recognized by subsequent computer devices and deep neural networks configured in the computer devices can be reduced, so that the resolution of the downsampled medical imaging data is as close as possible to the resolution of the data that can actually be processed by the computer device or deep neural network, which facilitates the computer device or deep neural network to understand the structural relationship of the medical imaging data, thereby improving processing efficiency and recognition accuracy.

[0049] S102: Determine the initial position of each joint landmark point in the medical image data.

[0050] The initial position may refer to the approximate position of each joint landmark point preliminarily determined in the medical image data.

[0051] In one possible implementation of an embodiment of the present application, a computer device may use an image processing algorithm to process medical image data, preliminarily determine the areas where various joint landmark points in the image may exist, and quickly locate the approximate position of each landmark point.

[0052] In another possible implementation of the embodiment of the present application, a computer device can use a deep neural network configured therein to process the medical imaging data of the first resolution, output the approximate position of each joint landmark point, and lay the foundation for subsequent fine positioning.

[0053] Specifically, the computer device may be configured with a deep neural network comprising multiple channels. After acquiring medical image data, the computer device may use the deep neural network to segment the image data to obtain one or more connected regions corresponding to each joint landmark point. The computer device may calculate the initial position of each joint landmark point based on the one or more connected regions corresponding to each joint landmark point.

[0054] In an embodiment of the present application, the number of channels used by the deep neural network to process medical image data can be equal to the number of joint landmark points to be identified. In this way, each channel can be used to segment the data in the area where a joint landmark point is located. Since the number of joint landmark points that need to be identified in each recognition process is not necessarily exactly the same, the number of channels included in the deep neural network can be more. In each recognition process, the channels actually involved in the processing can be less than or equal to the number of channels actually included in the deep neural network.

[0055] like Figure 2 , which is a schematic diagram of a deep neural network processing flow provided by an embodiment of the present application. Figure 2 The processing flow shown in can be an example of using a deep neural network to process medical image data and output the initial positions of each joint landmark point.

[0056] See also Figure 2 , the deep neural network used in the embodiment of the present application can be constructed based on the U-Net structure. Figure 2 The raw data in this context may refer to medical image data at its original resolution, and the compressed data obtained through downsampling or other scaling processing may be medical image data at a lower first resolution. After the first-resolution medical image data is input into a deep neural network, the deep neural network can segment the data and output corresponding segmentation results. These segmentation results can indicate specific areas where landmark points may be located.

[0057] Combine Figure 3 The figure is a schematic diagram of the process of segmenting medical image data using a deep neural network provided by an embodiment of the present application. Figure 3 Part (a) shows that the medical imaging data is the CT image data X of the current case, and the deep neural network is used to The process of performing segmentation to obtain the CT segmentation result L representing the area where the landmark point may be located is also called the label region segmentation process. Where w, h, and d represent the size of the input CT image data X.

[0058] On this basis, computer equipment can then use deep neural networks The segmentation result L is further semantically segmented to locate the landmark area of ​​each landmark point. The landmark area is also the approximate position of each landmark point.

[0059] Specifically, in accordance with Figure 3 After obtaining the segmentation result L through the segmentation process shown in part (a), the computer device can extract the maximum connectivity channel by channel, and then calculate the maximum connectivity according to the segmentation result L. Figure 3The segmentation process shown in part (b) performs semantic segmentation on the maximum connectivity area to obtain the landmark area H. Therefore, Figure 3 The segmentation process shown in part (b) is also the landmark region segmentation process, where c represents the number of landmark points.

[0060] It should be noted that the deep neural network used to further semantic segmentation the segmentation result L obtained in the above steps is It can be the deep neural network used to segment CT image data X Same neural network, or different neural networks.

[0061] For example, a deep neural network based on the U-Net structure can be trained to have Figure 3 The segmentation function shown in parts (a) and (b) of and deep neural networks It can be the same neural network. Figure 3 The training data of the function shown in part (a) is the same as the training Figure 3 The training data for the functions shown in part (a) are different.

[0062] Alternatively, two deep neural networks can be constructed based on the U-Net structure, namely the deep neural network and deep neural networks , and then use the corresponding training data for training, so that the deep neural network have Figure 3 The function shown in part (a) of the deep neural network have Figure 3 The function shown in part (b) of and deep neural networks Configured in a computer device, it can process medical image data during the process of identifying joint landmark points and output the approximate position of each landmark point.

[0063] In an embodiment of the present application, the computer device can further accurately locate the approximate position of each landmark point based on the output of the deep neural network to obtain the initial position of each landmark point.

[0064] Specifically, the computer device can determine the maximum connected region among the one or more connected regions obtained by segmenting each channel of the deep neural network. This maximum connected region is also the region where the corresponding joint landmark point is located. Based on this, the computer device can calculate the center point of the maximum connected region corresponding to each channel as the initial position of the corresponding joint landmark point.

[0065] It should be noted that the joint landmark points obtained through deep neural network segmentation may include all landmark points to be identified, or only some of them. That is, after deep neural network processing of medical image data, some landmark points may not be identified. Therefore, the initial positions of the output joint landmark points do not include the positions of these unidentified landmark points.

[0066] S103 , registering an initial landmark point set consisting of the joint landmark points with a pre-built statistical shape model according to the initial position to obtain a target registration matrix.

[0067] Based on the initial position of each identified joint landmark point calculated in the above steps, the computer device can align the initial landmark point set with the pre-built statistical shape model to obtain a target registration matrix. The above initial landmark point set is composed of the identified joint landmark points.

[0068] In an embodiment of the present application, the statistical shape model can be constructed based on a standard dataset, which can be a data set pre-labeled with joint landmark points. For example, the standard dataset can be obtained by a doctor marking joint landmark points in a three-dimensional model. Based on the relevant data of each manually labeled joint landmark point, a computer device can construct the statistical shape model.

[0069] like Figure 4 , which is a flow chart of constructing a shape statistics model provided by an embodiment of the present application. Figure 4 The landmark dataset T in the data set may be a set of data points consisting of multiple landmark points extracted from the standard dataset. For example, the landmark dataset T may include N landmark points. The computer device may process the landmark dataset T and calculate the mean of the N landmark points. , eigenvalue and the eigenvector Statistical shape models (SSMs) are constructed by registering the landmarks within dataset T and combining them with principal components analysis (PCA). The SSMs capture the spatial relationships and variation patterns between landmark points, providing powerful prior knowledge for subsequent landmark point filling and optimization.

[0070] Based on the pre-constructed statistical shape model SSM, the computer device can align the identified joint landmark points to the model SSM to obtain a target registration matrix.

[0071] like Figure 5 FIG. 1 is a flow chart of landmark point registration and gap filling provided by an embodiment of the present application. Figure 5 The whole process of aligning the landmark points identified by the deep neural network with the statistical shape model SSM to obtain the target alignment matrix and completing the missing landmark points based on the target alignment matrix is ​​shown.

[0072] Specifically, see Figure 5 , Figure 5 The initial point set shown in Ps is the set of all points to be identified, the number of which is c, so the point set P s ={ P s1 ……P sc In the process of landmark point registration, the computer device can register each joint landmark point in the medical image data, i.e., the initial point set, based on the initial position of the identified landmark point. P s Perform null value detection to obtain a non-null landmark index number set I 0, the non-empty landmark index number set I 0 is obtained by sorting the joint landmark points identified from the medical imaging data.

[0073] The computer device can index the non-empty landmark sequence number set I 0 is registered with the statistical shape model SSM to obtain the initial registration matrix M 0, and calculate the initial registration matrix M The Euclidean distance between point clouds corresponding to 0 D 0, the distance D 0 can be used to update the initial registration matrix later M 0 is used to compare the distances corresponding to other registration matrices. Figure 5 In , Regi(., .) represents point cloud registration.

[0074] In order to update the above initial registration matrix M 0, the computer device can be based on a non-empty landmark index number set I 0 determines multiple optional point number sets. Figure 5 As shown, the optional point sequence number set may include N types, and these N optional point sequence number sets may be selected from the non-empty landmark index sequence number set I 0. That is:

[0075]

[0076] The above expression is a permutation and combination formula, which means Select K elements from the elements to combine. Represents a non-empty landmark index sequence number set I 0. For example, if =14, K=4, then the number of N optional point sequence sets obtained based on the above expression is: =1001.

[0077] In a possible implementation of the embodiment of the present application, any optional point number set may include at least 4 identified joint landmark points, that is, the value of K may be a natural number greater than or equal to 4. The number of joint landmark points included in each optional point number set may be equal.

[0078] For example, taking K=4 as an example, the computer device can select from the non-empty landmark index sequence number set I 0 to form an optional point number set. For example, the initial point set P s There are c landmark points in P s1 ，……，P sc , assuming that the identified landmark points are P s1 ，P s2 ，P s3 ，P s4 ，P s5 ，P s7 ，……，P sc , that is, the landmark point P s6 If not recognized by the deep neural network, the multiple optional point sequence number sets created can include:

[0079] Optional point sequence number set 1:P s1 ，P s2 ，P s3 ，P s4 ;

[0080] Optional point number set 2: P s1 ，P s2 ，P s3 ，P s5 ;

[0081] Optional point sequence number set 3: P s1 ，P s2 ，P s3 ，P s7 ;

[0082] Optional point number set 4: P s1 ，P s2 ，P s3 ，P s8 ;

[0083] …

[0084] Optional point number set N: P sc-3 ，P sc-2 ，P sc-1 ，P sc ;

[0085] The computer device can align each optional point number set with the statistical shape model. In this process, Figure 5 As shown, the computer device can obtain the initial sequence number set corresponding to the minimum Euclidean distance between point clouds by cyclically registering N times. I 0 and the sequence number set to be updated. The above sequence number set to be updated can be a non-empty landmark index sequence number set I 0 and the initial sequence number set I 1. For example, the sequence number set to be updated can be expressed as I 2= I 0- I 1.

[0086] Specifically, the computer device can register each optional point sequence number set with the statistical shape model SSM and calculate the Euclidean distance between the point clouds under the current registration result. In this way, after completing the registration of all N optional point sequence number sets, the Euclidean distance results between N point clouds can be obtained. The computer device can compare these N distances and determine the optional point sequence number set corresponding to the minimum distance value as the initial sequence number set. I 1. The initial sequence number set I 1 is the best set of optional point numbers when K landmark points are selected.

[0087] Then, the computer device can update the serial number set I One or more joint boundary points in 2 are included in the initial sequence number set I 0, and repeat the operation of registering the sequence number set with other joint boundary points with the statistical shape model SSM to obtain the target index sequence number set I f , the target index number set I f The corresponding registration matrix M f That is, the optimal target registration matrix.

[0088] Specifically, the computer device can be updated according to the sequence number set to be updated I The index of each joint landmark point in 2, the joint landmark point with the smallest index value is included in the initial sequence number set I 0, and align the sequence number set after incorporating the joint landmark point with the smallest index value with the statistical shape model SSM to obtain the corresponding Euclidean distance between point clouds.

[0089] For example, assuming that the initial sequence number set I 0 is the optional point sequence set 1 in the above example, which contains the landmark points P s1 ，P s2 ，P s3 and P s4 , since the landmark points not recognized by the deep neural network are landmark points P s6 , so the sequence number set to be updated I The joint boundary point with the smallest index value in 2 is P s5 , before including the landmark point in the initial sequence number set I The sequence number set obtained after 0 contains a total of 5 landmark points, namely P s1 ，P s2 ，Ps3 ，P s4 and P s5 The computer device can register the sequence number set containing the above five landmark points with the statistical shape model SSM to obtain the corresponding Euclidean distance and distance difference between point clouds. For example, the Euclidean distance between point clouds can be Figure 5 The distance shown in D t Correspondingly, the distance difference can be the Euclidean distance between the point clouds corresponding to the current sequence number set. D t With the initial registration matrix M Euclidean distance corresponding to 0 D The difference between 0, that is D t - D 0.

[0090] The computer device can judge the above distance difference and determine its relationship with the distance threshold T. Figure 5 As shown, if the distance difference D t - D 0 is greater than or equal to the distance threshold T, which means that compared with the initial sequence number set I 0, include the joint landmark point with the smallest current index value (i.e. the landmark point in the above example P s5 ) and no better sequence number set is obtained. At this time, the computer device can discard the joint landmark point with the smallest current index value from the sequence number set to be updated. D t - D 0 is less than the distance threshold T, which means that compared with the initial sequence number set I 0, a better sequence number set can be obtained by incorporating the joint landmark point with the smallest current index value. At this time, the computer device can P s5 ) is retained in the initial sequence number set I 0. Taking the above example as an example, the sequence number set is updated to include P s1 ，P s2 ，P s3 ，P s4 and P s5 A total of 5 landmark points.

[0091] The computer device can repeatedly perform the above steps according to the index of each joint landmark point in the set to be updated, add the joint landmark point with the smallest index value into the initial set and perform registration, and finally obtain the target index set. I f and target registration matrix M f .

[0092] For example, in the above example, P s1 ，P s2 ，P s3 ，P s4 and P s5 After the collection of 5 landmark points, due to the landmark points P s6 Not recognized, so the sequence number set is pending update I The landmark point with the smallest index value in 2 is the landmark point P s7 After the landmark point is included in the sequence number set, it can be obtained P s1 ，P s2 ，P s3 ，P s4 ，P s5 and P s7 There are a total of 6 landmark point number sets. The computer device can repeatedly perform the registration process to align the number set with the statistical shape model SSM, and determine whether to retain the landmark point in the number set by comparing the Euclidean distance between the point clouds. P s7 Or abandon.

[0093] The above registration operation can be repeated multiple times until the Euclidean distance between point clouds is found. D t With the initial registration matrix M Euclidean distance corresponding to 0 D 0 is less than the distance threshold T and contains the largest number of landmark points, the sequence number set at this time can be used as the final index sequence I f , which is the target index sequence number set. It can be considered that the final index sequence I f The landmark points included in are all accurately identified landmark points and are not included in the final index sequence above. I fThe landmark points in need to be completed in the subsequent steps. I f Then, the target registration matrix is ​​obtained accordingly. M f .

[0094] S104 : When there are missing landmark points in the medical image data, use the target registration matrix to complete the missing landmark points.

[0095] In an embodiment of the present application, if not all landmark points in the medical image data are identified after segmentation processing by the deep neural network, that is, there are missing landmark points, the computer device can use the target alignment matrix obtained above to complete the missing landmark points and complete the landmark point filling process.

[0096] like Figure 5 As shown, using the target registration matrix M f When completing missing landmark points, you can first use the target index number set I f Determine the empty landmark index number set I e , the empty landmark index number set I e It represents the set of landmark points that need to be filled.

[0097] In the embodiment of the present application, the empty landmark index number set I e It represents the landmark points that have not been accurately identified before the landmark point filling operation compared to all the landmark points to be identified. Since the landmark points that have been accurately identified are the final index sequence I f The landmark points contained in , so the above empty landmark index number set I e Can be represented as the original sequence number set I With the final index sequence I f The difference, that is I e = I - I f .

[0098] The computer device can then use the target registration matrix M f , the statistical shape model SSM and the empty landmark index number set I e The corresponding standard landmark points are transformed into the medical image data to fill in the missing landmark points. Figure 5 As shown, the process can be expressed as:

[0099]

[0100] In the above expression, Indicates the index number set of the empty landmark I e The point set obtained after completion is, Represents the set of index numbers of empty landmarks in the statistical shape model SSM I e The above expression is obtained by using the target registration matrix M f The inverse matrix of is used to transform the standard landmark points to obtain the completed point set.

[0101] After completing the missing landmark point filling operation, we can get a set of all landmark points to be identified, that is, Figure 5 The patch landmark point set shown in P f .

[0102] S105 , projecting the completed joint landmark points onto the joint model surface corresponding to the medical image data.

[0103] In the embodiments of the present application, since the missing landmark points are filled using the target registration matrix to calculate the positions of the missing landmark points, it is possible that the calculated positions do not lie on the joint model surface corresponding to the medical image data. To address this situation, the computer device can project each of the filled joint landmark points onto the joint model surface.

[0104] like Figure 6 FIG. 1 is a schematic diagram of a joint landmark point projection process provided by an embodiment of the present application. Figure 6 As shown, before performing the landmark point projection, the computer device can determine the joint model surface corresponding to the medical image data by performing morphological corrosion on the medical image data. Figure 6 middle represents morphological erosion, which erodes the medical image data L. K represents the convolution kernel, and d(., .) represents the Euclidean distance between points in the calculated space. Morphological erosion shrinks the original medical image data L by one circle, preserving the surface area of ​​the model.

[0105] Specifically, if Figure 6 As shown, the medical image data L can be corroded using the morphological corrosion method, and its expression is: , the result after corrosion is the data after the original medical image data L is reduced by one circle (corroded the surface of the medical image data L). Then use the original medical image data L to subtract , the surface area of ​​the model can be obtained .

[0106] On this basis, the patched landmark point set obtained after completion can be P f Each landmark point in is projected onto the model surface.

[0107] Specifically, the computer device can target any completed joint landmark point p , calculate each point on the surface of the joint model s The distance between each joint landmark point except the current one , the distance The point corresponding to the minimum value in is taken as the projection point of the current joint landmark point on the surface of the joint model, and the projection point set is obtained, that is, Figure 6 Surface projected landmarks in The above points s are all points on the surface S of the joint model, that is, . Figure 6 The expression argmin represents the set of parameter values ​​that minimize the objective function, and its overall value is used to represent the points at each joint boundary. p The corresponding point on the joint model surface S ,point The set formed is the landmark point obtained after projection .

[0108] S106: Output joint landmark point recognition result data with a second resolution.

[0109] The aforementioned processing of medical image data is performed after downsampling the original image data and scaling it to a first resolution. The various joint landmark points located on the surface of the joint model obtained by the final processing are also presented in the medical image data of the first resolution, which has a lower resolution. Therefore, after projecting the completed landmark points onto the surface of the joint model, the computer device can also optimize the landmark points and output joint landmark point recognition result data with a second resolution. The above-mentioned second resolution is greater than the first resolution. For example, the second resolution can be the same as the original resolution of the medical image data.

[0110] like Figure 7 FIG. 1 is a schematic diagram of a joint landmark point optimization process provided by an embodiment of the present application. Figure 7 The optimization process shown, the computer device can use a deep neural network, such as a deep neural network Optimize and fine-tune each joint landmark point on the model surface to obtain the final set of landmark points The above deep neural network It can be the same as the aforementioned deep neural network and deep neural networks Neural networks with the same structure. For example, deep neural networks It can also be a neural network built based on the U-Net structure.

[0111] In an embodiment of the present application, by processing medical image data into data with a lower first resolution, a computer device can use a deep neural network to determine the initial position of each joint point. On this basis, by aligning the initial landmark point set composed of each joint landmark point with a pre-built statistical shape model, a target alignment matrix can be obtained, so that the missing landmark points can be supplemented using the target alignment matrix. After projecting the supplemented joint landmark points onto the surface of the joint model, the computer device can output joint landmark point recognition result data with a higher second resolution to complete the recognition of all landmark points. Using the recognition method provided in the embodiment of the present application, joint landmark points in medical image data can be efficiently and accurately identified, thereby improving the accuracy and reliability of medical image data processing. This method not only improves the recognition accuracy of human joint landmark points under complex pathology and non-standard postures, but also solves the problem of missing landmark points caused by lesions, surgery or acquisition postures, and also reduces the spatial dislocation caused by false detection by optimizing the landmark point positioning accuracy, thereby ensuring the accuracy of the recognition results.

[0112] It should be noted that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0113] For ease of understanding, the following introduces the joint landmark point recognition method provided in the embodiment of the present application with reference to a complete example.

[0114] In the embodiment of the present application, the recognition of joint landmark points using computer aids is mainly divided into a model training stage and an inference recognition stage. Among them, in the model training stage, the CT data set and the landmark point cloud data set can be used to train the deep neural network and the shape statistical model. The data volume of the training set can account for 80% of the data set, and the data augmentation method is used to augment the data set. The best model in the test set is taken as the final model. When constructing the shape statistical model, the point cloud in the data set can be aligned to the same coordinate system, and the PCA algorithm can be used to calculate the mean, eigenvalue and eigenvector of the data set. This process can be as follows Figure 4 shown.

[0115] In the inference recognition stage, the computer device can downsample the original resolution CT impact data of the joint area to 1 / 2 of the original size and then input it into the neural network. Segmentation is performed in , and the segmented joint labels are obtained By labeling the data Enter the neural network , we can get the landmark area The process can be as follows Figure 3 As shown. Using connectivity to extract landmark areas The initial position of each landmark point can be obtained. After registering each landmark point obtained by initial recognition and inputting it into the statistical shape model for completion, the repaired landmark point set can be obtained. The process can be as follows Figure 5 As shown. When using morphological methods to calculate labels The edge of the label and project the completed landmark point to the label On the edge of , all landmark points can be projected onto the joint model surface corresponding to the medical image data. This process can be as follows Figure 6 Finally, the original resolution CT of each edge point and the corresponding block is input into the neural network , we can get the fine-tuned landmarks The process can be as follows Figure 7 shown.

[0116] Based on the above introduction, the recognition of joint landmark points using computer aided methods mainly includes landmark center extraction based on deep neural network, point cloud registration based on greedy algorithm, missing point reconstruction based on shape statistical model, and landmark point surface refinement and high-resolution optimization. Figure 8 , is a schematic diagram of a joint landmark point recognition process provided by an embodiment of the present application. Figure 8 The detailed description of each step in the process is as follows:

[0117] (1) Landmark center extraction based on deep neural network

[0118] This step mainly includes low-resolution neural network segmentation and initial landmark point set extraction.

[0119] 1. Low-resolution neural network segmentation: First, the input low-resolution medical image data can be segmented using a neural network. After segmentation, the maximum connectivity is extracted channel by channel. Then the deep neural network is used Further semantic segmentation of the segmentation results can obtain relevant information about the area where each landmark point may be located. This step can quickly locate the approximate location of the landmark point, laying the foundation for subsequent fine-grained positioning.

[0120] like Figure 8 As shown, the current case CT image data can be processed low-resolution medical imaging data, using neural network After segmentation, a CT segmentation result can be obtained. This process is the label region segmentation process introduced in the above embodiment. Figure 8 The landmark region segmentation neural network shown in can be a deep neural network . Using deep neural networks After further processing of the CT segmentation results, the approximate region where each landmark point is located can be determined, completing the landmark region segmentation process described in the aforementioned embodiment.

[0121] The input and output data and processing flow involved in the above steps can be expressed as:

[0122] - Input: joint label data

[0123] - Processing: Use deep neural networks for semantic segmentation to obtain possible areas where landmark points may exist

[0124] - Output: Landmark area

[0125] 2. Extraction of initial landmark point set: Based on the segmentation result of the previous step, by extracting the maximum connected area from the probability map corresponding to each landmark point, its center can be calculated as the initial position of the landmark point.

[0126] Based on the calculated initial position of each landmark point, the initial landmark point set can be formed P s ={ P s1 ……P sc}.

[0127] The input and output data and processing flow involved in the above steps can be expressed as:

[0128] - Input: Segmented landmark area

[0129] - Processing: Extract the largest connected area for each segmentation channel and calculate its center as the initial landmark point

[0130] - Output: Initial landmark point set

[0131] (2) Point cloud registration based on greedy algorithm

[0132] This step mainly includes statistical shape model (SSM) construction and data registration optimization.

[0133] 3. Statistical Shape Model (SSM) Construction: This step uses landmark points from a standard dataset to construct a statistical shape model through in-dataset registration and principal component analysis. This model captures the spatial relationships and variation patterns between landmark points, providing powerful prior knowledge for subsequent landmark completion and optimization.

[0134] The input and output data and processing flow involved in the above steps can be expressed as:

[0135] - Input: Landmark point annotations in the standard dataset

[0136] - Processing: Align the data in the dataset to the same space and apply PCA analysis

[0137] - Output: SSM model

[0138] 4. Data registration optimization: This method uses the initial landmark point set P s When aligning with the SSM model, an innovative iterative alignment strategy is adopted. K (K>=4) points are selected from the optional point set for point-to-point alignment. After repeated multiple times, the optimal alignment matrix can be obtained, which is the target alignment matrix introduced in the above embodiment. M f This step effectively reduces the impact of falsely detected landmark points on the overall registration and improves the stability of the system.

[0139] The input and output data and processing flow involved in the above steps can be expressed as:

[0140] - Input: initial landmark point set, SSM model

[0141] - deal with:

[0142] a) Traverse the optional point set to complete the registration

[0143] b) Use greedy strategy to select the optimal registration matrix

[0144] c) Iteratively incorporate other points to distinguish between successful points and false positives

[0145] - Output: optimized registration matrix

[0146] (3) Missing point reconstruction based on shape statistical model

[0147] This step mainly includes SSM to complete the missing landmark points.

[0148] 5. SSM complements missing landmark points: the optimized registration matrix, i.e. the target registration matrix, can be used M f, transform the current landmark point set into the SSM mean space, and then use the SSM model to estimate and complete the missing landmark points. This step fills in the missing landmark points and ensures that the landmark points can be output completely.

[0149] The input and output data and processing flow involved in the above steps can be expressed as:

[0150] - Input: optimized registration matrix, SSM model

[0151] - Processing: Transform the standard points through the registration matrix and use SSM to fill in the missing points

[0152] - Output: Completed landmark point set

[0153] (IV) Landmark point surface refinement and high-resolution optimization

[0154] This step mainly includes surface projection correction and high-resolution refinement.

[0155] 6. Surface Projection Correction: By projecting the completed landmark point set onto the segmented 3D data surface, we ensure that all landmark points are located on the actual surface of the anatomical structure, such as the joint surface. This step corrects for any deviations introduced by the SSM completion and improves the anatomical accuracy of the landmark points.

[0156] The input and output data and processing flow involved in the above steps can be expressed as:

[0157] - Input: Completed landmark point set, joint label data

[0158] - Processing: Project the completed points to the nearest model surface

[0159] - Output: roughly corrected landmark point set

[0160] 7. High-resolution refinement: Finally, the original resolution medical imaging data and landmarks are input into the trained deep neural network. , each landmark point can be locally optimized. Deep Neural Network It can perform precise positioning only in a small range around the landmark point, ensuring the image quality of the output landmark recognition result and further improving the accuracy of landmark positioning.

[0161] The input and output data and processing flow involved in the above steps can be expressed as:

[0162] - Input: roughly rectified landmark point set, original high-resolution image

[0163] - Processing: Optimize each landmark point using a neural network trained on the original resolution dataset

[0164] - Output: Final refined landmark point set

[0165] The method for identifying joint landmark points provided in the embodiment of the present application fully considers the balance between efficiency and accuracy. This method quickly locates the approximate position of the landmark points from a low-resolution initial screening, uses a shape statistical model to supplement prior knowledge and process missing landmark points, and finally uses high-resolution refinement processing to ensure the high accuracy of the final result. Each step provides a basis for optimization for the next step, forming a closed-loop, self-improving recognition system. The application of this multi-stage, adaptive landmark point recognition method provided in the embodiment of the present application can effectively handle the landmark point recognition problem in complex pathological conditions and provide high-precision, high-reliability landmark point recognition results.

[0166] Reference Figure 9 , shows a schematic diagram of a joint landmark recognition device provided by an embodiment of the present application, which may specifically include an acquisition module 901, a determination module 902, a registration module 903, a gap filling module 904, a projection module 905, and an output module 906, wherein:

[0167] An acquisition module 901 is configured to acquire medical image data having a first resolution, where the first resolution is smaller than an original resolution of the medical image data;

[0168] a determination module 902 for determining an initial position of each joint landmark point in the medical image data;

[0169] a registration module 903 for registering, based on the initial positions, an initial landmark point set consisting of the joint landmark points with a pre-constructed statistical shape model to obtain a target registration matrix, wherein the statistical shape model is constructed based on a standard dataset, which is a data set pre-labeled with joint landmark points;

[0170] A gap filling module 904 is configured to fill in the missing landmark points using the target registration matrix when there are missing landmark points in the medical image data;

[0171] A projection module 905 is configured to project the completed joint landmark points onto a joint model surface corresponding to the medical image data;

[0172] The output module 906 is configured to output joint landmark point recognition result data having a second resolution, where the second resolution is greater than the first resolution.

[0173] In a possible implementation of the embodiment of the present application, the determining module 902 may be specifically configured to:

[0174] Segmenting the medical image data using a deep neural network having multiple channels to obtain one or more connected regions corresponding to each joint landmark point, wherein the number of channels of the deep neural network is equal to the number of joint landmark points to be identified, and any one of the channels is used to segment data in the region where one joint landmark point is located;

[0175] Based on one or more connected areas corresponding to each of the joint landmark points, an initial position of each of the joint landmark points is calculated.

[0176] In the embodiment of the present application, the determining module 902 may also be used to:

[0177] For each channel segmentation of the deep neural network, one or more connected regions are obtained, and a maximum connected region among the one or more connected regions corresponding to each channel is determined, where the maximum connected region is the region where the corresponding joint landmark point is located;

[0178] The center point of the maximum connected area corresponding to each channel is calculated as the initial position of the corresponding joint landmark point.

[0179] In a possible implementation of the embodiment of the present application, the registration module 903 may be specifically configured to:

[0180] performing a null value detection on each of the joint landmark points in the medical image data according to the initial position to obtain a non-null landmark index sequence number set, wherein the non-null landmark index sequence number set is obtained by sorting each of the joint landmark points identified from the medical image data;

[0181] Determine a plurality of optional point sequence number sets based on the non-empty landmark index sequence number set;

[0182] By aligning each of the optional point sequence number sets with the statistical shape model, an initial sequence number set and a sequence number set to be updated are obtained, wherein the sequence number set to be updated is the difference between the non-empty landmark index sequence number set and the initial sequence number set;

[0183] After incorporating one or more joint landmark points in the sequence number set to be updated into the initial sequence number set, repeatedly aligning the sequence number set incorporating other joint landmark points with the statistical shape model to obtain a target index sequence number set, and the alignment matrix corresponding to the target index sequence number set is the target alignment matrix.

[0184] In the embodiment of the present application, the registration module 903 may also be used to:

[0185] According to the index of each joint landmark point in the to-be-updated sequence number set, the joint landmark point with the smallest index value is included in the initial sequence number set, and the sequence number set after the joint landmark point with the smallest index value is included is aligned with the statistical shape model to obtain the corresponding Euclidean distance and distance difference between point clouds;

[0186] If the distance difference is greater than or equal to the distance threshold, the joint landmark point with the smallest current index value is discarded from the set of sequence numbers to be updated;

[0187] If the distance difference is less than the distance threshold, retaining the joint landmark point with the smallest current index value in the initial sequence number set;

[0188] Repeat the steps of adding the joint landmark point with the smallest index value into the initial sequence number set and performing registration according to the index of each joint landmark point in the sequence number set to be updated, to obtain a target index sequence number set.

[0189] In a possible implementation manner of the embodiment of the present application, any of the optional point sequence number sets includes at least 4 identified joint landmark points, and the number of the joint landmark points included in each of the optional point sequence number sets is equal.

[0190] In a possible implementation of the embodiment of the present application, the gap filling module 904 may be specifically configured to:

[0191] Determine an empty landmark index sequence number set based on the target index sequence number set;

[0192] The target registration matrix is ​​used to transform each standard landmark point in the statistical shape model corresponding to the empty landmark index number set into the medical image data to complete the missing landmark points.

[0193] In a possible implementation of the embodiment of the present application, the projection module 905 may be specifically used to:

[0194] Determining a joint model surface corresponding to the medical image data by performing morphological corrosion on the medical image data;

[0195] For any of the completed joint landmark points, respectively calculating the distance between each point on the joint model surface and each of the other joint landmark points except the current joint landmark point;

[0196] The point corresponding to the minimum distance value is used as the projection point of the current joint landmark point on the surface of the joint model.

[0197] The present invention provides a device for identifying joint landmarks, which can be a computer device as described in the aforementioned method embodiments, or a processing unit or module within the computer device capable of performing the corresponding functions. This device can implement the various steps in the aforementioned method embodiments, efficiently and accurately identifying joint landmarks and improving the accuracy and reliability of medical imaging data analysis and processing.

[0198] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment part.

[0199] Reference Figure 10 , shows a schematic diagram of a computer device provided by an embodiment of the present application. Figure 10 As shown, the computer device 1000 in the embodiment of the present application includes: a processor 1010, a memory 1020, and a computer program 1021 stored in the memory 1020 and executable on the processor 1010. When the processor 1010 executes the computer program 1021, the steps of each embodiment of the above-mentioned joint landmark point recognition method are implemented, such as Figure 1 Alternatively, when the processor 1010 executes the computer program 1021, the functions of the modules / units in the above-mentioned device embodiments are realized, for example, Figure 9 Functions of modules 901 to 906 are shown.

[0200] Exemplarily, the computer program 1021 may be divided into one or more modules / units, which are stored in the memory 1020 and executed by the processor 1010 to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments may be used to describe the execution process of the computer program 1021 in the computer device 1000. For example, the computer program 1021 may be divided into an acquisition module, a determination module, a registration module, a gap filling module, a projection module, and an output module, with the specific functions of each module being as follows:

[0201] an acquisition module, configured to acquire medical image data having a first resolution, where the first resolution is smaller than an original resolution of the medical image data;

[0202] a determination module, configured to determine an initial position of each joint landmark point in the medical image data;

[0203] a registration module for registering, based on the initial positions, an initial landmark point set consisting of the joint landmark points with a pre-constructed statistical shape model to obtain a target registration matrix, wherein the statistical shape model is constructed based on a standard dataset, the standard dataset being a data set pre-labeled with joint landmark points;

[0204] a gap filling module, configured to fill in the missing landmark points using the target registration matrix when there are missing landmark points in the medical image data;

[0205] A projection module, configured to project the completed joint landmark points onto a joint model surface corresponding to the medical image data;

[0206] An output module is used to output joint landmark point recognition result data with a second resolution, where the second resolution is greater than the first resolution.

[0207] The computer device 1000 may be an electronic device capable of implementing each step or related functions in the aforementioned method embodiments. The computer device 1000 may be a desktop computer, a cloud server, or other devices. For example, the computer device 1000 may be a computer-assisted medical device. The computer device 1000 may include, but is not limited to, a processor 1010 and a memory 1020. Those skilled in the art will understand that Figure 10 This is only an example of the computer device 1000 and does not constitute a limitation of the computer device 1000. The computer device 1000 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 1000 may also include input and output devices, network access devices, buses, etc.

[0208] The processor 1010 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0209] The memory 1020 can be an internal storage unit of the computer device 1000, such as a hard drive or memory of the computer device 1000. The memory 1020 can also be an external storage device of the computer device 1000, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the computer device 1000. Furthermore, the memory 1020 can include both an internal storage unit of the computer device 1000 and an external storage device. The memory 1020 is used to store the computer program 1021 and other programs and data required by the computer device 1000. The memory 1020 can also be used to temporarily store data that has been output or is about to be output.

[0210] An embodiment of the present application further discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the methods described in the aforementioned embodiments are implemented.

[0211] An embodiment of the present application further discloses a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed by a computer, the methods described in the aforementioned embodiments are implemented.

[0212] An embodiment of the present application further discloses a computer program product, including a computer program. When the computer program is run on a computer, the computer is caused to execute the methods described in the aforementioned embodiments.

[0213] The above embodiments are intended only to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they may still modify the technical solutions described in the above embodiments or replace some of the technical features therein with equivalents; and such modifications or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present application and should be included within the scope of protection of the present application.

Claims

1. A method for identifying joint landmark points, characterized in that: include: Acquiring medical image data having a first resolution, the first resolution being smaller than an original resolution of the medical image data; determining an initial position of each joint landmark point in the medical image data; According to the initial position, registering an initial landmark point set consisting of the joint landmark points with a pre-constructed statistical shape model to obtain a target registration matrix, wherein the statistical shape model is constructed based on a standard data set, the standard data set being a data set pre-labeled with joint landmark points; In the case where there are missing landmark points in the medical image data, using the target registration matrix to complete the missing landmark points; Projecting each of the completed joint landmark points onto a joint model surface corresponding to the medical image data, and outputting joint landmark point recognition result data having a second resolution, wherein the second resolution is greater than the first resolution; wherein, according to the initial position, the initial landmark point set composed of each of the joint landmark points is registered with the pre-constructed statistical shape model to obtain a target registration matrix, including: according to the initial position, performing null value detection on each of the joint landmark points in the medical image data to obtain a non-empty landmark index sequence number set, wherein the non-empty landmark index sequence number set is obtained by sorting each of the joint landmark points identified from the medical image data; determining a plurality of optional point sequence number sets based on the non-empty landmark index sequence number set; obtaining an initial sequence number set and a sequence number set to be updated by aligning each of the optional point sequence number sets with the statistical shape model, wherein the sequence number set to be updated is the difference between the non-empty landmark index sequence number set and the initial sequence number set; after incorporating one or more joint landmark points in the sequence number set to be updated into the initial sequence number set, repeatedly aligning the sequence number set incorporating other joint landmark points with the statistical shape model to obtain a target index sequence number set, wherein the registration matrix corresponding to the target index sequence number set is the target registration matrix; The method of using the target registration matrix to complete the missing landmark points includes: determining an empty landmark index number set based on the target index number set; and using the target registration matrix to transform each standard landmark point in the statistical shape model corresponding to the empty landmark index number set into the medical image data to complete the missing landmark points.

2. The method according to claim 1, characterized in that Determining the initial position of each joint landmark point in the medical image data includes: Segmenting the medical image data using a deep neural network having multiple channels to obtain one or more connected regions corresponding to each joint landmark point, wherein the number of channels of the deep neural network is equal to the number of joint landmark points to be identified, and any one of the channels is used to segment data in the region where one joint landmark point is located; Based on one or more connected areas corresponding to each of the joint landmark points, an initial position of each of the joint landmark points is calculated.

3. The method according to claim 2, characterized in that The calculating the initial position of each joint landmark point based on one or more connected areas corresponding to each joint landmark point includes: For each channel segmentation of the deep neural network, one or more connected regions are obtained, and a maximum connected region among the one or more connected regions corresponding to each channel is determined, where the maximum connected region is the region where the corresponding joint landmark point is located; The center point of the maximum connected area corresponding to each channel is calculated as the initial position of the corresponding joint landmark point.

4. The method according to claim 1, wherein After incorporating one or more joint landmark points in the to-be-updated sequence number set into the initial sequence number set, repeatedly aligning the sequence number set incorporating other joint landmark points with the statistical shape model to obtain a target index sequence number set, including: According to the index of each joint landmark point in the to-be-updated sequence number set, the joint landmark point with the smallest index value is included in the initial sequence number set, and the sequence number set after the joint landmark point with the smallest index value is included is aligned with the statistical shape model to obtain the corresponding Euclidean distance and distance difference between point clouds; If the distance difference is greater than or equal to the distance threshold, the joint landmark point with the smallest current index value is discarded from the set of sequence numbers to be updated; If the distance difference is less than the distance threshold, retaining the joint landmark point with the smallest current index value in the initial sequence number set; Repeat the steps of adding the joint landmark point with the smallest index value into the initial sequence number set and performing registration according to the index of each joint landmark point in the sequence number set to be updated, to obtain a target index sequence number set.

5. The method according to claim 1, characterized in that Any of the optional point sequence number sets includes at least 4 identified joint landmark points, and the number of the joint landmark points included in each of the optional point sequence number sets is equal.

6. The method according to any one of claims 1 to 5, characterized in that The projecting of the completed joint landmark points onto the joint model surface corresponding to the medical image data includes: Determining a joint model surface corresponding to the medical image data by performing morphological corrosion on the medical image data; For any of the completed joint landmark points, respectively calculating the distance between each point on the joint model surface and each of the other joint landmark points except the current joint landmark point; The point corresponding to the minimum distance value is used as the projection point of the current joint landmark point on the surface of the joint model.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the computer device is caused to implement the method according to any one of claims 1 to 6.

8. A computer program product comprising a computer program, characterized in that When the computer program is executed, the method according to any one of claims 1 to 6 is performed.

Images