Method and apparatus for predicting bone maturity

Abstract

The embodiment of the present application provides a bone maturity prediction method and device, the method comprises the following steps: obtaining target finger bone series data and target carpal bone series data corresponding to target hand bone image data respectively; inputting the target hand bone image data and the target finger bone series data into a preset finger bone maturity prediction model, so that the finger bone maturity prediction model outputs a finger bone maturity prediction value for bone age evaluation; and inputting the target hand bone image data and the target carpal bone series data into a preset carpal bone maturity prediction model, so that the carpal bone maturity prediction model outputs a carpal bone maturity prediction value for bone age evaluation. The present application can effectively improve the efficiency, automation and intelligence of the bone maturity prediction process, reduce the labor and time cost, and effectively improve the comprehensive coverage and accuracy of the bone maturity prediction result.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and more particularly to the field of artificial intelligence technology, specifically to a method and apparatus for predicting bone maturity. Background Technology

[0002] Bone age is a key indicator of biological maturity. In normal human skeletal development, the rate of ossification, the timing of epiphyseal and metaphyseal closure, and their morphological changes all exhibit certain regularities. This regularity, expressed over time, is known as bone age. However, due to significant differences in development among children of different races and regions, there are inconsistencies in skeletal maturity among children of the same age, making bone age a less reliable direct indicator of growth and development. Bone maturity, as an intermediate variable in bone age, reflects bone development and is independent of race and region. Therefore, bone maturity prediction can be used as an indicator to address the racial and regional differences in bone age.

[0003] Currently, existing methods for predicting bone maturity typically involve manually reading bone X-ray images to assess bone maturity. However, in actual clinical applications, this method is extremely inefficient, requires a significant amount of time and effort from the reader, and is highly dependent on the reader's expertise. Consequently, existing methods for predicting bone maturity suffer from problems such as low prediction efficiency, high labor and time costs, and poor prediction accuracy. Summary of the Invention

[0004] To address the problems in the prior art, this application provides a method and apparatus for predicting bone maturity, which can effectively improve the efficiency, automation and intelligence of the bone maturity prediction process, reduce labor and time costs, and effectively improve the comprehensiveness and accuracy of the bone maturity prediction results.

[0005] To solve the above-mentioned technical problems, this application provides the following technical solution:

[0006] In a first aspect, this application provides a method for predicting bone maturity, including:

[0007] Acquire the target phalangeal bone series data and target carpal bone series data corresponding to the target hand bone image data;

[0008] The target hand bone image data and target finger bone series data are input into a preset finger bone maturity prediction model so that the finger bone maturity prediction model outputs finger bone maturity prediction values ​​for bone age assessment.

[0009] Furthermore, the target hand bone image data and the target wrist bone series data are input into a preset wrist bone maturity prediction model so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment.

[0010] Furthermore, it also includes:

[0011] Using the predicted maturity values ​​of the finger bones and wrist bones, the bone age evaluation result corresponding to the target hand bone image data is determined from a preset bone maturity lookup table; wherein, the bone maturity lookup table is used to store the correspondence between each predicted maturity value of the finger bones, the predicted maturity value of the wrist bones, and the bone age evaluation result;

[0012] Output the bone age assessment results.

[0013] Furthermore, the acquisition of the target phalanx series data and target carpal series data corresponding to the target hand bone image data includes:

[0014] Acquire X-ray image data of hand bones;

[0015] The background is removed from the hand bone X-ray image data based on a preset background segmentation model to obtain the corresponding target hand bone image data.

[0016] Target phalangeal bone series data are extracted from the target hand bone image data, and target wrist bone series data are extracted from the target hand bone image data.

[0017] Further, the extraction of target phalanx series data from the target hand bone image data includes:

[0018] Based on a preset key point detection network, each key point is identified in the target hand bone image data;

[0019] Based on the RUS standard data specified in the TW3 scoring method, finger bone series key points were selected from each of the key points;

[0020] Extract the corresponding finger bone series image data of each finger bone series key point from the target hand bone image data to form the corresponding target finger bone series data.

[0021] Further, the extraction of target wrist bone series data from the target hand bone image data includes:

[0022] Based on the preset target detection network and the C standard data specified by the TW3 scoring method, the image data of each wrist bone series is extracted from the target hand bone image data to form the corresponding target wrist bone series data.

[0023] Furthermore, it also includes:

[0024] Acquire multiple historical hand bone X-ray image data;

[0025] Based on a preset background segmentation model, background removal is performed on each of the historical hand bone X-ray image data to obtain the historical hand bone image data corresponding to each of the historical hand bone X-ray image data.

[0026] Extract the corresponding historical finger bone series data from each of the historical hand bone image data, and extract the corresponding historical wrist bone series data from each of the historical hand bone image data.

[0027] Generate finger bone training sets and wrist bone training sets corresponding to each of the historical finger bone series data;

[0028] A finger bone maturity prediction model for outputting finger bone maturity prediction values ​​is trained using the historical hand bone image data and the finger bone training set. Similarly, a wrist bone maturity prediction model for outputting wrist bone maturity prediction values ​​is trained using the historical hand bone image data and the wrist bone training set.

[0029] Furthermore, both the finger bone maturity prediction model and the carpal bone maturity prediction model include a preset target deep neural network;

[0030] The target deep neural network includes interconnected feature extraction units and feature concatenation units;

[0031] The feature extraction unit includes: a first deep neural network layer for inputting hand bone image data, and a second deep neural network layer for inputting finger bone image data or wrist bone image data, and both the first deep neural network layer and the second deep neural network layer are connected to the feature splicing unit.

[0032] The feature splicing unit includes at least one fully connected layer and a linear layer connected in sequence.

[0033] Secondly, this application provides a bone maturity prediction device, comprising:

[0034] The data acquisition module is used to acquire the target phalangeal bone series data and target carpal bone series data corresponding to the target hand bone image data, respectively;

[0035] The finger bone maturity prediction module is used to input the target hand bone image data and the target finger bone series data into a preset finger bone maturity prediction model so that the finger bone maturity prediction model outputs a finger bone maturity prediction value for bone age assessment.

[0036] The wrist bone maturity prediction module is used to input the target hand bone image data and the target wrist bone series data into a preset wrist bone maturity prediction model, so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment.

[0037] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the bone maturity prediction method.

[0038] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the bone maturity prediction method described above.

[0039] As can be seen from the above technical solution, the present application provides a method and apparatus for predicting bone maturity. The method includes: acquiring target phalanx series data and target carpal series data corresponding to target hand bone image data; inputting the target hand bone image data and target phalanx series data into a preset phalanx maturity prediction model, so that the phalanx maturity prediction model outputs a predicted phalanx maturity value for bone age assessment; and inputting the target hand bone image data and the target carpal series data into a preset carpal maturity prediction model, so that the carpal maturity prediction model outputs a predicted carpal maturity value for bone age assessment. The system inputs the target finger bone series data and target wrist bone series data corresponding to the target hand bone image data into a preset model. It can obtain the corresponding finger bone maturity prediction value and wrist bone maturity prediction value based on the same target hand bone image data. This can effectively improve the efficiency, automation and intelligence of the bone maturity prediction process, effectively reduce labor and time costs, and effectively improve the comprehensiveness and accuracy of the bone maturity prediction results. In this way, it can provide a more reliable and comprehensive data foundation for subsequent determination of children's bone age based on bone maturity, so that the bone age evaluation results are not limited by region or race, and thus effectively improve the applicability of the bone maturity prediction method. Attached Figure Description

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

[0041] Figure 1 This is a schematic diagram illustrating the relationship between the bone maturity prediction device, the client device, and the X-ray image acquisition device in the embodiments of this application.

[0042] Figure 2 This is a schematic diagram of the anatomical structure corresponding to the TW3-RUS series.

[0043] Figure 3 This is a schematic diagram of the anatomical structure corresponding to the TW3-C series.

[0044] Figure 4 This is a schematic diagram of the first process of the bone maturity prediction method in the embodiments of this application.

[0045] Figure 5 This is a schematic diagram of the second process of the bone maturity prediction method in the embodiments of this application.

[0046] Figure 6 This is a schematic diagram of the first step 100 in the bone maturity prediction method in the embodiments of this application.

[0047] Figure 7 This is a schematic diagram of the second process of step 100 in the bone maturity prediction method in the embodiments of this application.

[0048] Figure 8 This is a schematic diagram of the third process of step 100 in the bone maturity prediction method in the embodiments of this application.

[0049] Figure 9 This is a schematic diagram of the third process of the bone maturity prediction method in the embodiments of this application.

[0050] Figure 10 This is a schematic diagram of the structure of the target deep neural network in the embodiments of this application.

[0051] Figure 11 This is a schematic diagram of the bone maturity prediction device in the embodiments of this application.

[0052] Figure 12 This is a schematic diagram of the training process of the finger bone maturity prediction model provided in the application example of this application.

[0053] Figure 13 This is a schematic diagram illustrating background removal provided in the application example of this application.

[0054] Figure 14 This is a schematic diagram illustrating the process of obtaining the TW3-RUS series region of interest, provided in the application example of this application.

[0055] Figure 15 This is a schematic diagram of the training process of the wrist bone maturity prediction model provided in the application example of this application.

[0056] Figure 16 This is a schematic diagram illustrating the process of obtaining the TW3-C series region of interest, provided in the application example of this application.

[0057] Figure 17 This is a diagram illustrating the structure of the finger bone maturity prediction model provided in the application example of this application.

[0058] Figure 18 This is a diagram illustrating the structure of the wrist bone maturity prediction model provided in the application example of this application.

[0059] Figure 19 This is a schematic diagram illustrating the process of predicting finger bone maturity using the finger bone maturity prediction model provided in the application example of this application.

[0060] Figure 20 This is a schematic diagram illustrating the process of predicting wrist bone maturity using the wrist bone maturity prediction model provided in the application examples of this application.

[0061] Figure 21 This is a schematic diagram of the structure of the electronic device in the embodiments of this application. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] It should be noted that the bone maturity prediction method and apparatus disclosed in this application can be used in the fields of image recognition and artificial intelligence, as well as in any field other than image recognition and artificial intelligence.

[0064] Bone age is a key indicator of biological maturity. In normal human skeletal development, the rate of ossification, the timing of epiphyseal and metaphyseal closure, and their morphological changes all exhibit certain regularities. This regularity, expressed over time, is called bone age. Bone age assessment, which evaluates the actual age of a person based on the evolutionary process of skeletal development, has wide applications in clinical medicine, forensic identification, and sports. In clinical medicine, comparing bone age with actual age, combined with physical examination, can help identify factors affecting children's growth and development and endocrine disorders, allowing for timely intervention and treatment. Bone age assessment is extremely important in many fields. For example, in sports, bone age can effectively determine the age of participants in youth competitions and serves as an indicator for selecting young athletes, determining their developmental level, and developing scientific training programs.

[0065] Thodberg et al. developed the BoneXpert automated bone age assessment system, which has been approved for clinical use in some regions. BoneXpert uses an Active Appearance Model (AAM), a computer vision algorithm that reconstructs the contours of the 13 bones in the hand. The automated system then determines the overall bone age based on bone shape, texture, and strength, using either GP or TW techniques. For example, Son et al. proposed a complete end-to-end bone age assessment system based on the TW3 scoring method.

[0066] However, due to significant differences in development among children of different races and regions, there are inconsistencies in skeletal maturity among children of the same age, making bone age a less reliable direct indicator of children's growth and development. For example, BoneXpert is only effective for children in a specific region; applying the system to children outside that region will result in significant errors. Furthermore, this system does not utilize the wrist bone, although the wrist bone is important for assessing skeletal maturity in infants and young children. Another example is the complete end-to-end bone age assessment system based on the TW3 scoring method proposed by Son et al., which uses bone age as an indicator of children's developmental stage, making it unsuitable for widespread clinical use due to racial and regional limitations.

[0067] Therefore, this application considers bone maturity, as specified in the TW3 scoring criteria, as an intermediate variable for bone age. It reflects bone development and is independent of race, region, etc. Thus, bone maturity assessment can be used as an indicator to address the differences in bone age caused by racial and regional variations. However, current methods of bone maturity assessment typically involve manually reading bone X-ray images. In practical clinical applications, this method is highly inefficient, requiring significant time and effort from the reader and heavily relying on their expertise. Consequently, existing bone maturity assessment methods suffer from low efficiency, high labor and time costs, and poor accuracy.

[0068] Clinically, bone age assessment commonly employs the TW3 scoring method or methods based on it. The TW3 scoring method scores 20 relevant bones individually to obtain a bone maturity score for each bone. Therefore, it is possible to consider using a deep neural network to predict bone maturity based on phalangeal series data (e.g., 13 key bones), and then using a second deep neural network to predict bone maturity based on the corresponding hand bone image data. Finally, the optimal prediction result is selected from the two neural networks. However, this approach, because the prediction object of the second deep neural network is the acquired X-ray image itself, does not inherently possess an equal evaluation standard with the method of predicting bone maturity based on phalangeal series data. Furthermore, the final bone age assessment requires selecting the best result from the two predictions, which fails to meet the reliability requirements of the optimal bone maturity result and prolongs the time required to determine bone age.

[0069] Based on this, after extensive experimental research, this application found that it is first necessary to select relevant data that are relatively equal to and comparable to the phalangeal series data (RUS series) for model prediction. Therefore, the C series data specified by TW3 is considered to be used. Furthermore, there is no need to choose the best between the prediction results generated by the two models; instead, the bone age assessment result can be directly determined based on the correspondence between the two prediction results, individually or simultaneously, and bone age. This effectively improves the reliability and comprehensiveness of bone maturity results and enhances the efficiency of bone age assessment. In addition, this application also provides machine learning models specifically for the TW3-RUS and TW3-C systems, as well as various data processing methods, which are explained in detail through the following embodiments and application examples.

[0070] Based on this, and addressing the problems of low efficiency, long processing time, and poor accuracy in existing bone maturity prediction methods, this application provides a bone maturity prediction method. This method involves inputting target phalangeal bone series data corresponding to target hand bone image data into a preset phalangeal bone maturity prediction model, which outputs predicted phalangeal bone maturity values ​​for bone age assessment. Similarly, it involves inputting target wrist bone series data corresponding to the target hand bone image data into a preset wrist bone maturity prediction model, which outputs predicted wrist bone maturity values ​​for bone age assessment. This method can obtain predicted phalangeal and wrist bone maturity values ​​from the same target hand bone image data, effectively improving the efficiency, automation, and intelligence of the bone maturity prediction process. It also effectively reduces labor and time costs and improves the comprehensiveness and accuracy of the bone maturity prediction results. This provides a more reliable and comprehensive data foundation for subsequent bone age determination based on bone maturity, making bone age assessment results unrestricted by region or ethnicity, thus significantly increasing the applicability of the bone maturity prediction method.

[0071] Based on the foregoing, this application also provides a bone maturity prediction device for implementing the bone maturity prediction method provided in one or more embodiments of this application. This bone maturity prediction device can be a server. (See also...) Figure 1 The bone maturity prediction device can communicate and connect sequentially with various client devices and X-ray image acquisition devices, either independently or through a third-party server. The bone maturity prediction device can receive bone maturity prediction instructions sent by client devices and receive X-ray image data of the child's left hand bones sent by the image acquisition device according to the bone maturity prediction instructions. Then, it converts the hand bone X-ray image data into target hand bone image data, obtains the target finger bone series data and target wrist bone series data corresponding to the target hand bone image data, and inputs the target hand bone image data and target finger bone series data into a preset finger bone maturity prediction model so that the finger bone maturity prediction model outputs finger bone maturity prediction values ​​for bone age assessment.

[0072] Furthermore, the target hand bone image data and the target wrist bone series data are input into a preset wrist bone maturity prediction model, so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment. Then, the finger bone maturity prediction value and the wrist bone maturity prediction value can be output to the client device, so that the user can determine the bone age assessment result of the child whose hand bone X-ray image data was taken based on the finger bone maturity prediction value and the wrist bone maturity prediction value. Alternatively, after obtaining the finger bone maturity prediction value and the wrist bone maturity prediction value, the bone age assessment result corresponding to the target hand bone image data can be determined from a locally stored bone maturity lookup table, and then the bone age assessment result can be directly sent to the client device, so that the user can quickly and reliably obtain the bone age assessment result from the client device.

[0073] The bone maturity prediction portion of the aforementioned bone maturity prediction device can be executed on a server as described above. Alternatively, in another practical application, all operations can be completed on the client device. The choice can be made based on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed on the client device, the client device may further include a processor for the specific processing of bone maturity prediction.

[0074] It is understood that the client device can include any mobile device capable of running applications, such as smartphones, tablet computers, network set-top boxes, portable computers, personal digital assistants (PDAs), in-vehicle devices, and smart wearable devices. Among these, the smart wearable devices can include smart glasses, smartwatches, smart bracelets, etc.

[0075] For example, a user only needs to transmit hand bone X-ray image data to a certain application on a smartphone. The application can then call the bone maturity prediction function of a local or remote server to execute the bone maturity prediction method provided in this application embodiment, and generate or receive corresponding finger bone maturity prediction values, wrist bone maturity prediction values, and bone age evaluation results, etc. This allows the user to obtain the corresponding prediction or evaluation results simply by transmitting hand bone X-ray image data, which can effectively improve the user experience.

[0076] Based on the above examples, in order to further improve the user experience and the convenience of bone maturity prediction, a module with bone maturity prediction function can be integrated into an X-ray acquisition device for collecting hand bone X-ray image data. This allows the user to collect hand bone X-ray image data without having to input the data. The X-ray acquisition device can then control its own module with bone maturity prediction function to execute the bone maturity prediction method provided in this application embodiment, and directly display the finger bone maturity prediction value, wrist bone maturity prediction value, and bone age evaluation results on the corresponding display device.

[0077] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0078] The server and the client device can communicate using any suitable network protocol, including those not yet developed as of the date of this application. Such network protocols may include, for example, TCP / IP, UDP / IP, HTTP, HTTPS, etc. Furthermore, such network protocols may also include RPC (Remote Procedure Call Protocol) and REST (Representational State Transfer Protocol) used on top of the aforementioned protocols.

[0079] In one or more embodiments of this application, in order to obtain the region of interest related to the TW3 scoring method, it is divided into two parts: one part is to obtain the region of interest for the TW3-RUS phalanx series, and the other part is to obtain the region of interest for the TW3-C (Carpal) carpal series.

[0080] Specifically, TW3-RUS refers to the phalangeal bone series in the TW3 scoring method, where RUS is an abbreviation for radius, ulna, and short bones. The anatomical structures (13 bone fragments) corresponding to this TW3-RUS series are as follows: Figure 2 As shown.

[0081] exist Figure 2 The 13 bone fragments specifically include: radius 3, ulna 28, first metacarpal 19, third metacarpal 9, fifth metacarpal 14, proximal thumb femur 20, third proximal phalanx 4, fifth proximal phalanx 11, third intermediate phalanx 12, fifth intermediate phalanx 23, distal thumb bone 8, third distal phalanx 17, and fifth distal phalanx 21. The RUS standard data specified in the TW3 scoring method mentioned in one or more embodiments of this application refers to the data such as the specified positions of these 13 bone fragments. The phalanx series data refers to these 13 bone fragments, and the phalanx series key points refer to the locations of these 13 bone fragments.

[0082] Specifically, TW3-C refers to the carpal bone series in the TW3 scoring method, where C is an abbreviation for carpal. The anatomical structures (7 bone pieces) corresponding to this TW3-C series are as follows: Figure 3 As shown.

[0083] exist Figure 3 The seven bone fragments specifically include: triquetrum C1, hamate C2, lunate C3, capitate C4, scaphoid C5, trapezium C6, and trapezium C7. The C-standard data specified in the TW3 scoring method mentioned in one or more embodiments of this application refers to the specified positions and other data of these seven bone fragments. The carpal bone series data refers to these seven bone fragments, and the carpal bone series key points refer to the locations of these seven bone fragments.

[0084] The following embodiments and application examples will be described in detail.

[0085] To address the problems of low efficiency, long processing time, and poor accuracy in existing bone maturity prediction methods, this application provides an embodiment of a bone maturity prediction method, see [link to embodiment]. Figure 4 The bone maturity prediction method based on the bone maturity prediction device specifically includes the following:

[0086] Step 100: Obtain the target phalangeal bone series data and target wrist bone series data corresponding to the target hand bone image data.

[0087] Understandably, target hand bone image data is usually generated from X-ray images of the left hand of children to further ensure the reliability of bone maturity prediction. However, in certain special scenarios, it can also be generated from image data of adults, the right hand, or other acquisition types, depending on the actual needs.

[0088] In one or more embodiments of this application, the phalangeal series data refers to: the radius, ulna, first metacarpal, third metacarpal, fifth metacarpal, proximal thumb femur, third proximal phalanx, fifth proximal phalanx, third intermediate phalanx, fifth intermediate phalanx, distal thumb bone, third distal phalanx, and fifth distal phalanx as specified in the TW3 evaluation method.

[0089] In one or more embodiments of this application, the wrist bone series data refers to the triquetrum, hamate, lunate, capitate, scaphoid, trapezium, and trapezium as specified in the TW3 evaluation method.

[0090] However, the finger bone series data and carpal bone series data mentioned in this application are not limited to the data specified in the TW3 evaluation method above. The bone pieces can be added or removed as the specified bone pieces in the subsequent finger bone series and carpal bone series change. The specific settings can be made according to actual needs.

[0091] Step 200: Input the target hand bone image data and target finger bone series data into a preset finger bone maturity prediction model so that the finger bone maturity prediction model outputs finger bone maturity prediction values ​​for bone age assessment.

[0092] In step 200, the finger bone maturity prediction model can employ a deep neural network or other available prediction models, and the target hand bone image data and target finger bone series data can be simultaneously input into the finger bone maturity prediction model as input data, and the finger bone maturity prediction value output by the finger bone maturity prediction model can be any score within a predefined range of finger bone maturity scores.

[0093] Step 300: Input the target hand bone image data and the target wrist bone series data into a preset wrist bone maturity prediction model so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment.

[0094] In step 300, the wrist bone maturity prediction model can employ a deep neural network or other available prediction models, and the target hand bone image data and target wrist bone series data can be simultaneously input into the wrist bone maturity prediction model as input data, and the wrist bone maturity prediction value output by the wrist bone maturity prediction model can be any score within a predefined range of wrist bone maturity scores.

[0095] In one or more embodiments of this application, "target" data refers to the data being processed during model application, and "historical" data refers to the data used during model training. The data structures and types of the two are the same or similar. For example, target hand bone image data and historical hand bone image data both refer to hand bone image data. However, target hand bone image data is the image data for which bone maturity prediction is currently to be performed, while historical hand bone image data is hand bone image data for which bone maturity values ​​have been obtained. The model is trained using these historical data with known labels (bone maturity values).

[0096] To further improve the accuracy of bone maturity prediction, different machine learning models can be used as their respective prediction models in steps 200 and 300.

[0097] It is understood that steps 300 and 200 are steps executed after step 100. Steps 300 and 200 can be executed simultaneously or at different times, depending on the actual application needs. If device operator combination is considered, they can be executed in time-sharing. If further improvement of bone maturity prediction efficiency is considered, they can be executed simultaneously. This application does not limit this.

[0098] As described above, the bone maturity prediction method provided in this application provides a preset finger bone maturity prediction model and a wrist bone maturity prediction model. By inputting the target finger bone series data and target wrist bone series data corresponding to the target hand bone image data into the preset models, the method can obtain the corresponding finger bone maturity prediction value and wrist bone maturity prediction value based on the same target hand bone image data. This can effectively improve the efficiency, automation, and intelligence of the bone maturity prediction process, effectively reduce labor and time costs, and effectively improve the comprehensiveness and accuracy of the bone maturity prediction results. As a result, it can provide a more reliable and comprehensive data foundation for subsequent determination of children's bone age based on bone maturity, making the bone age evaluation results not limited by region or race, and thus effectively improving the applicability of the bone maturity prediction method.

[0099] To provide a more reliable and comprehensive data foundation for subsequent determination of children's bone age based on bone maturity, an embodiment of the bone maturity prediction method provided in this application is described below. Figure 5 The bone maturity prediction method further includes the following content after steps 200 and 300:

[0100] Step 400: Apply the predicted maturity values ​​of the finger bones and the predicted maturity values ​​of the wrist bones to determine the bone age evaluation result corresponding to the target hand bone image data from a preset bone maturity lookup table; wherein, the bone maturity lookup table is used to store the correspondence between each predicted maturity value of the finger bones, the predicted maturity value of the wrist bones, and the bone age evaluation result.

[0101] Step 500: Output the bone age assessment results.

[0102] It is understood that outputting the bone age assessment result can specifically refer to sending the bone age assessment result to the client device, or it can refer to sending it to a preset display device for display. In addition, in order to increase readability and further improve the efficiency and convenience for doctors and other users to obtain bone age assessment results, after obtaining the bone age assessment result, the bone age assessment result, the predicted value of finger bone maturity, and the predicted value of wrist bone maturity can be used to generate a corresponding relationship diagram based on preset graphic display rules, and the relationship diagram can be displayed on the display device.

[0103] As can be seen from the above description, the bone maturity prediction method provided in this application determines the bone age evaluation result corresponding to the target hand bone image data from a preset bone maturity comparison table based on the predicted maturity values ​​of the finger bones and wrist bones. This can effectively improve the comprehensiveness and accuracy of the bone maturity prediction results, thereby providing a more reliable and comprehensive data foundation for subsequent determination of children's bone age based on bone maturity. Compared with the method of selecting the best of the two prediction results before conducting bone age assessment, this method can not only improve the bone age evaluation but also effectively improve the efficiency of bone age evaluation.

[0104] To improve the efficiency, reliability, and intelligence of background removal, an embodiment of the bone maturity prediction method provided in this application is described below. Figure 6 Step 100 of the bone maturity prediction method specifically includes the following:

[0105] Step 110: Acquire X-ray image data of hand bones.

[0106] Step 120: Based on a preset background segmentation model, perform background removal on the hand bone X-ray image data to obtain the corresponding target hand bone image data.

[0107] In step 120, the background segmentation model can be selected from networks commonly used in medical image segmentation, such as U-Net. U-Net consists of a downsampling stage and an upsampling stage. Its network structure contains only convolutional and pooling layers, without fully connected layers. Shallower, high-resolution layers are used to solve pixel localization, while deeper layers are used to solve pixel classification, thus achieving semantic-level image segmentation. The U-Net structure includes a contraction path that captures contextual information and a symmetrical expansion path that allows for precise localization. This method can complete end-to-end training with very little data; that is, the input is one image, and the output is also one image.

[0108] Step 130: Extract target phalangeal bone series data from the target hand bone image data, and extract target wrist bone series data from the target hand bone image data.

[0109] As can be seen from the above description, the bone maturity prediction method provided in this application can effectively improve the efficiency, reliability and intelligence of background removal by using a background segmentation model to remove background from the hand bone X-ray image data. This can further improve the accuracy and effectiveness of extracting target finger bone series data and target wrist bone series data from the target hand bone image data, thereby further improving the reliability and accuracy of the bone maturity prediction results.

[0110] To improve the intelligence, efficiency, reliability, and specificity of extracting phalanx series image data corresponding to each key point of each phalanx series from the target hand bone image data, an embodiment of the bone maturity prediction method provided in this application is presented, see [link to embodiment]. Figure 7 Step 130 of the bone maturity prediction method specifically includes the following:

[0111] Step 131: Identify each key point in the target hand bone image data based on a preset key point detection network.

[0112] In step 131, the keypoint detection network can be a heatmap regression-based keypoint detection network: the SCN network. The SCN network (Spatial Configuration-Net) can achieve high keypoint recognition accuracy on a small dataset thanks to its unique network structure. Its network structure consists of two parts: one part is used for predicting blurred local keypoints in the image, and the other part is responsible for improving the robustness and accuracy of the prediction in the first part.

[0113] Step 132: Select finger bone series key points from each of the key points according to the RUS standard data specified in the TW3 scoring method.

[0114] Step 133: Extract the corresponding finger bone series image data of each finger bone series key point from the target hand bone image data to form the corresponding target finger bone series data.

[0115] Step 134: Extract target wrist bone series data from the target hand bone image data.

[0116] As can be seen from the above description, the bone maturity prediction method provided in this application, by adopting the RUS standard data and key point detection network specified by the TW3 scoring method, can effectively improve the intelligence, efficiency, reliability and specificity of extracting the corresponding finger bone series image data of each finger bone series key point from the target hand bone image data, thereby effectively improving the efficiency, automation and intelligence of the bone maturity prediction process, and effectively reducing labor and time costs.

[0117] To improve the efficiency of bone maturity prediction, in one embodiment of the bone maturity prediction method provided in this application, see [link to embodiment]. Figure 8 Step 134 of the bone maturity prediction method specifically includes the following:

[0118] Step 135: Based on the preset target detection network and the C standard data specified by the TW3 scoring method, extract the image data of each wrist bone series from the target hand bone image data to form the corresponding target wrist bone series data.

[0119] Understandably, the YOLO v4 algorithm can be chosen for object detection networks. The YOLO v4 algorithm is based on the original YOLO object detection architecture and adopts optimization strategies from the CNN field. It has been optimized to varying degrees in various aspects such as data processing, backbone network, network training, activation function, and loss function, making it an efficient and powerful object detection model.

[0120] As can be seen from the above description, the bone maturity prediction method provided in this application, by adopting the C standard data and target detection network specified by the TW3 scoring method, can effectively improve the intelligence, efficiency, reliability and specificity of extracting each series of wrist bone image data from the target hand bone image data, thereby effectively improving the efficiency, automation and intelligence of the bone maturity prediction process, and effectively reducing labor and time costs.

[0121] To improve the effectiveness and reliability of automated batch approval, an embodiment of the bone maturity prediction method provided in this application is described below. Figure 9 The bone maturity prediction method may further include the following specific components before steps 200 and 300:

[0122] Step 010: Acquire multiple historical hand bone X-ray image data.

[0123] Step 020: Based on the preset background segmentation model, perform background removal on each of the historical hand bone X-ray image data to obtain the historical hand bone image data corresponding to each of the historical hand bone X-ray image data.

[0124] Step 030: Extract the corresponding historical finger bone series data from each of the historical hand bone image data, and extract the corresponding historical wrist bone series data from each of the historical hand bone image data.

[0125] In step 030, key points can be identified in the historical hand bone image data based on a preset key point detection network; finger bone series key points are selected from the key points according to the RUS standard data specified by the TW3 scoring method; finger bone series image data corresponding to each finger bone series key point are extracted from the historical hand bone image data to form corresponding historical finger bone series data; and, based on the preset historical detection network and the C standard data specified by the TW3 scoring method, wrist bone series image data are extracted from the historical hand bone image data to form corresponding historical wrist bone series data.

[0126] Step 040: Generate the finger bone training set corresponding to each of the historical finger bone series data and the wrist bone training set corresponding to each of the historical wrist bone series data.

[0127] It is understood that each historical phalanx series in the phalanx training set contains its own corresponding label (phalanx maturity score), and each historical carpal series in the carpal training set contains its own corresponding label (carpal maturity score).

[0128] Step 050: Train a finger bone maturity prediction model for outputting finger bone maturity prediction values ​​using the historical hand bone image data and the finger bone training set; and train a wrist bone maturity prediction model for outputting wrist bone maturity prediction values ​​using the historical hand bone image data and the wrist bone training set.

[0129] As can be seen from the above description, the bone maturity prediction method provided in this application, by pre-training the finger bone maturity prediction model and the wrist bone maturity prediction model, can effectively improve the efficiency and automation of obtaining the predicted values ​​of finger bone maturity and wrist bone maturity for bone age assessment, and can effectively improve the prediction accuracy and reliability of the predicted values ​​of finger bone maturity and wrist bone maturity, so that the bone age assessment results are not limited by region or race, thereby effectively improving the applicability of the bone maturity prediction method.

[0130] To improve the reliability and accuracy of the finger bone maturity prediction model and the wrist bone maturity prediction model, an embodiment of the bone maturity prediction method provided in this application is described below. Figure 10 Both the finger bone maturity prediction model and the carpal bone maturity prediction model include a preset target deep neural network;

[0131] The target deep neural network includes interconnected feature extraction units and feature concatenation units;

[0132] The feature extraction unit includes: a first deep neural network layer for inputting hand bone image data, and a second deep neural network layer for inputting finger bone image data or wrist bone image data, and both the first deep neural network layer and the second deep neural network layer are connected to the feature splicing unit.

[0133] The feature splicing unit includes at least one fully connected layer and a linear layer connected in sequence.

[0134] In one or more embodiments of this application, the target deep neural network may be written as SMANet.

[0135] As can be seen from the above description, the bone maturity prediction method provided in this application embodiment can effectively improve the application reliability and prediction accuracy of the finger bone maturity prediction model and the wrist bone maturity prediction model by setting a target deep neural network with a special structure, thereby further improving the efficiency, automation and intelligence of the bone maturity prediction process.

[0136] From a software perspective, to address the problems of low efficiency, long processing time, and poor accuracy in existing bone maturity prediction methods, this application provides an embodiment of a bone maturity prediction device for executing all or part of the aforementioned bone maturity prediction method. See [link to embodiment]. Figure 11 The bone maturity prediction device specifically includes the following components:

[0137] The data acquisition module 10 is used to acquire the target phalangeal bone series data and the target wrist bone series data corresponding to the target hand bone image data, respectively.

[0138] The finger bone maturity prediction module 20 is used to input the target hand bone image data and the target finger bone series data into a preset finger bone maturity prediction model so that the finger bone maturity prediction model outputs a finger bone maturity prediction value for bone age assessment.

[0139] The wrist bone maturity prediction module 30 is used to input the target hand bone image data and the target wrist bone series data into a preset wrist bone maturity prediction model so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment.

[0140] The embodiments of the bone maturity prediction device provided in this application can be used to execute the processing flow of the bone maturity prediction method embodiments described above. Its functions will not be repeated here, but can be referred to the detailed description of the above method embodiments.

[0141] As described above, the bone maturity prediction device provided in this application provides a preset finger bone maturity prediction model and a wrist bone maturity prediction model. By inputting the target finger bone series data and target wrist bone series data corresponding to the target hand bone image data into the preset models, it can obtain the corresponding finger bone maturity prediction value and wrist bone maturity prediction value based on the same target hand bone image data. This can effectively improve the efficiency, automation, and intelligence of the bone maturity prediction process, effectively reduce labor and time costs, and effectively improve the comprehensiveness and accuracy of the bone maturity prediction results. In this way, it can provide a more reliable and comprehensive data foundation for subsequent determination of children's bone age based on bone maturity, so that the bone age evaluation results are not limited by region or race, thereby effectively improving the applicability of the bone maturity prediction method.

[0142] To further illustrate this approach, this application provides a bone maturity prediction method, achieving a fully automated end-to-end bone maturity prediction process. Existing automated bone age assessment systems directly predict bone age values ​​from left-hand X-ray images, leading to significant differences in bone age among children of different races and regions with the same level of bone development. Therefore, these automated assessment systems have strong limitations. This application uses bone maturity instead of bone age as an indicator of bone development, avoiding the aforementioned problems. Furthermore, this application's solution uses an artificial intelligence algorithm model based on the TW3 scoring method for bone maturity scoring. Therefore, this method has higher clinical acceptance. This application provides C-series scores and RUS-series scores from the TW3 scoring method, and then the scores of the phalanges and carpal bones can be accumulated to obtain the total RUS maturity score and the total C maturity score for comprehensive bone age determination.

[0143] First, see Figure 12 The training process for the finger bone maturity prediction model in the SMANet model (also known as the RUS series model, SMANet-RUS model, or TW3-RUS model) includes the following:

[0144] Because the original image obtained in this application contains a lot of background noise, which will affect the training effect of various subsequent models, the background of the original image is first removed using the U-Net network, a background segmentation model commonly used in medical images. See [link to relevant documentation]. Figure 13 .

[0145] Then, the regions of interest (ROIs) for the TW3-RUS phalanx series in the TW3 scoring method were obtained. To obtain the ROIs for the TW3-RUS series, this application first used a keypoint detection network based on heatmap regression (SCN network) to locate 37 keypoints on the image. Then, the corresponding ROIs for the TW3-RUS bone age assessment method were cropped centered on the keypoints related to TW3-RUS, and these ROIs were stitched together into an image. See [link to relevant documentation]. Figure 14 .

[0146] Finally, the whole hand image and the TW3-RUS region of interest are input into SMANet to train the TW3-RUS model.

[0147] Secondly, see Figure 15 The training process for the wrist bone maturity prediction model in the SMANet model (also known as the C-series model, SMANet-C model, or TW3-C model) includes the following:

[0148] First, the background of the original image is removed using the U-Net network, a commonly used background segmentation model in medical imaging. (See [link]) Figure 13 .

[0149] Then, the region of interest (ROI) for the TW3-C series bone age assessment method was obtained. To obtain the ROI for the TW3-C series bone age assessment method, this application used the YOLO v4 object detection network to detect the entire wrist bone region. See [link to relevant documentation]. Figure 16 .

[0150] Finally, the whole hand image and the TW3-C region of interest are input into SMANet to train the TW3-C model.

[0151] In the above description, the training process of the finger bone maturity prediction model can be carried out simultaneously or at different times with the training process of the wrist bone maturity prediction model.

[0152] Among them, see Figure 17 The backbone network of the aforementioned finger bone maturity prediction model consists of two deep neural networks, Inceptionv3, used to extract features from the image. One Inceptionv3 network extracts features from the input integer image data, while the other extracts features from the input finger bone series data. Then, two fully connected layers (Dense) and a linear layer (Linear) are used to process the data to obtain a score. The purpose of the Dense layer is to extract the correlations between the previously extracted features through a non-linear transformation, and finally map them onto the output space. Each neuron in the Dense layer is connected to all neurons in the previous layer, achieving linear combination and linear transformation of the previous layer.

[0153] See Figure 18 The backbone network of the aforementioned wrist bone maturity prediction model consists of two Inceptionv3 networks used to extract image features. One Inceptionv3 network extracts features from the input integer image data, while the other extracts features from the input wrist bone series data. Then, two fully connected layers (Dense) and a linear layer (Linear) are used to process the data and obtain a score. The purpose of the Dense layer is to non-linearly transform the previously extracted features, extract the correlations between these features, and finally map them onto the output space. Each neuron is connected to all neurons in the previous layer, achieving linear combination and linear transformation of the previous layer.

[0154] Based on the finger bone maturity prediction model and carpal bone maturity prediction model obtained from the above training, bone maturity can be predicted from X-ray image data of a child's left hand. (See [link to relevant documentation]). Figure 19 and Figure 20 Specifically, it includes the following:

[0155] During prediction, an X-ray image of the child's left hand needs to be input into the bone maturity prediction device or system. After a series of image processing steps (refer to the training process), the system finally outputs two scores: the RUS series score and the C series score. Because this device is an end-to-end automated bone maturity prediction device, the entire processing is imperceptible during prediction; from the outside perspective, it simply appears as if an X-ray image is input, and then the RUS series and C series scores are obtained. Finally, the bone age is determined by comparing it with a bone age and bone maturity reference table for the child's region.

[0156] In one specific example, the model training process was as follows: The experiment used 5300 left-hand X-ray images of children from a hospital as the dataset. The experiment was conducted on a single NVIDIA RTX5000 GPU using PyTorch, optimized with the Adam algorithm. The batch size used during model training was 32, and the learning rate was 0.001. After 200 complete training epochs, the learning rate was reduced to 0.0002, and training ended after 1400 epochs. The bone maturity prediction process was as follows: The left-hand X-ray images of the children to be tested were input into the SMANet model, which output the child's TW3-RUS bone maturity score and TW3-C bone maturity score. The bone age could then be determined by consulting a bone age reference table for the child's region.

[0157] Based on this, the bone maturity prediction method and device provided in the application examples of this application can solve the problem that automatic bone age assessment systems cannot be widely promoted due to regional and racial issues, and solve the problem of low clinical acceptance of end-to-end automatic bone age assessment methods. By using bone maturity instead of bone age as an indicator to measure the degree of children's growth and development, the differences caused by racial and regional differences are avoided. At the same time, this application also adopts a bone maturity prediction method based on multi-region fusion deep network, which automates the entire prediction process.

[0158] From a hardware perspective, in order to address the problems of low efficiency, long processing time, and poor accuracy in existing bone maturity prediction methods, this application provides an embodiment of an electronic device for implementing all or part of the aforementioned bone maturity prediction method. The electronic device specifically includes the following components:

[0159] Figure 21 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 21 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 21 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0160] In one embodiment, the bone maturity prediction function can be integrated into a central processing unit (CPU). The CPU can be configured to perform the following controls:

[0161] Step 100: Obtain the target phalangeal bone series data and target wrist bone series data corresponding to the target hand bone image data.

[0162] Step 200: Input the target hand bone image data and target finger bone series data into a preset finger bone maturity prediction model so that the finger bone maturity prediction model outputs finger bone maturity prediction values ​​for bone age assessment.

[0163] Step 300: Input the target hand bone image data and the target wrist bone series data into a preset wrist bone maturity prediction model so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment.

[0164] As described above, the electronic device provided in this application embodiment, by presetting a finger bone maturity prediction model and a wrist bone maturity prediction model, and inputting the target finger bone series data and target wrist bone series data corresponding to the target hand bone image data into the presetting model, can obtain the corresponding finger bone maturity prediction value and wrist bone maturity prediction value respectively based on the same target hand bone image data. This can effectively improve the efficiency, automation, and intelligence of the bone maturity prediction process, effectively reduce labor and time costs, and effectively improve the comprehensiveness and accuracy of the bone maturity prediction results. In this way, it can provide a more reliable and comprehensive data foundation for subsequent determination of children's bone age based on bone maturity, so that the bone age evaluation results are not limited by region or race, thereby effectively improving the applicability of the bone maturity prediction method.

[0165] In another embodiment, the bone maturity prediction device can be configured separately from the central processing unit 9100. For example, the bone maturity prediction device can be configured as a chip connected to the central processing unit 9100, and the bone maturity prediction function can be realized through the control of the central processing unit.

[0166] like Figure 21 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 21 All components shown; in addition, the electronic device 9600 may also include Figure 21 For components not shown, please refer to existing technologies.

[0167] like Figure 21 As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0168] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0169] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0170] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0171] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device's communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0172] The communication module 9110 is a transmitter / receiver 9110 that transmits and receives signals via the antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processing unit 9100 to provide input signals and receive output signals, which can be the same as in a conventional mobile communication terminal.

[0173] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored sound via the speaker 9131.

[0174] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the bone maturity prediction method in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the bone maturity prediction method in the above embodiments, where the execution subject is a server or client. For example, when the processor executes the computer program, it implements the following steps:

[0175] Step 100: Obtain the target phalangeal bone series data and target wrist bone series data corresponding to the target hand bone image data.

[0176] Step 200: Input the target hand bone image data and target finger bone series data into a preset finger bone maturity prediction model so that the finger bone maturity prediction model outputs finger bone maturity prediction values ​​for bone age assessment.

[0177] Step 300: Input the target hand bone image data and the target wrist bone series data into a preset wrist bone maturity prediction model so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment.

[0178] As described above, the computer-readable storage medium provided in this application embodiment, by pre-setting a finger bone maturity prediction model and a wrist bone maturity prediction model, and inputting the target finger bone series data and target wrist bone series data corresponding to the target hand bone image data into the pre-setting model, can obtain the corresponding finger bone maturity prediction value and wrist bone maturity prediction value respectively based on the same target hand bone image data. This can effectively improve the efficiency, automation, and intelligence of the bone maturity prediction process, effectively reduce labor and time costs, and effectively improve the comprehensiveness and accuracy of the bone maturity prediction results. In turn, it can provide a more reliable and comprehensive data foundation for subsequent determination of children's bone age based on bone maturity, so that the bone age evaluation results are not limited by region or race, thereby effectively improving the applicability of the bone maturity prediction method.

[0179] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0180] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0181] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0182] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0183] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for predicting bone maturity, characterized in that, include: Acquire X-ray image data of hand bones; The background is removed from the hand bone X-ray image data based on a preset background segmentation model to obtain the corresponding target hand bone image data. Based on a preset key point detection network, each key point is identified in the target hand bone image data; According to the RUS standard data specified in the TW3 scoring method, finger bone series key points are selected from each of the key points; finger bone series image data corresponding to each of the finger bone series key points are extracted from the target hand bone image data to form the corresponding target finger bone series data. Based on the preset target detection network and the C standard data specified by the TW3 scoring method, the image data of each wrist bone series is extracted from the target hand bone image data to form the corresponding target wrist bone series data. The target hand bone image data and target finger bone series data are input into a preset finger bone maturity prediction model so that the finger bone maturity prediction model outputs finger bone maturity prediction values ​​for bone age assessment; bone maturity is an intermediate variable of bone age, which reflects bone development and is also independent of race and region; bone maturity prediction is used as an indicator to solve the differences in bone age caused by race and region. In addition, the target hand bone image data and the target wrist bone series data are input into a preset wrist bone maturity prediction model so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment. Both the finger bone maturity prediction model and the carpal bone maturity prediction model include a preset target deep neural network; The target deep neural network includes interconnected feature extraction units and feature concatenation units; The feature extraction unit includes: a first deep neural network layer for inputting hand bone image data, and a second deep neural network layer for inputting finger bone image data or wrist bone image data, and both the first deep neural network layer and the second deep neural network layer are connected to the feature splicing unit. The feature splicing unit includes at least one fully connected layer and a linear layer connected in sequence.

2. The bone maturity prediction method according to claim 1, characterized in that, Also includes: Using the predicted maturity values ​​of the finger bones and wrist bones, the bone age evaluation result corresponding to the target hand bone image data is determined from a preset bone maturity lookup table; wherein, the bone maturity lookup table is used to store the correspondence between each predicted maturity value of the finger bones, the predicted maturity value of the wrist bones, and the bone age evaluation result; Output the bone age assessment results.

3. The bone maturity prediction method according to claim 1, characterized in that, Also includes: Acquire multiple historical hand bone X-ray image data; Based on a preset background segmentation model, background removal is performed on each of the historical hand bone X-ray image data to obtain the historical hand bone image data corresponding to each of the historical hand bone X-ray image data. Extract the corresponding historical finger bone series data from each of the historical hand bone image data, and extract the corresponding historical wrist bone series data from each of the historical hand bone image data. Generate finger bone training sets and wrist bone training sets corresponding to each of the historical finger bone series data; A finger bone maturity prediction model for outputting finger bone maturity prediction values ​​is trained using the historical hand bone image data and the finger bone training set. Similarly, a wrist bone maturity prediction model for outputting wrist bone maturity prediction values ​​is trained using the historical hand bone image data and the wrist bone training set.

4. A bone maturity prediction device, characterized in that, include: The data acquisition module is used to collect X-ray image data of the hand bones; The background is removed from the hand bone X-ray image data based on a preset background segmentation model to obtain the corresponding target hand bone image data. Based on a preset key point detection network, each key point is identified in the target hand bone image data; According to the RUS standard data specified in the TW3 scoring method, finger bone series key points are selected from each of the key points; finger bone series image data corresponding to each of the finger bone series key points are extracted from the target hand bone image data to form the corresponding target finger bone series data; based on the preset target detection network and the C standard data specified in the TW3 scoring method, each wrist bone series image data is extracted from the target hand bone image data to form the corresponding target wrist bone series data. The finger bone maturity prediction module is used to input the target hand bone image data and target finger bone series data into a preset finger bone maturity prediction model, so that the finger bone maturity prediction model outputs finger bone maturity prediction values ​​for bone age assessment; bone maturity is an intermediate variable of bone age, which reflects bone development and is also independent of race and region; bone maturity prediction is used as an indicator to solve the differences in bone age caused by race and region. The wrist bone maturity prediction module is used to input the target hand bone image data and the target wrist bone series data into a preset wrist bone maturity prediction model so that the wrist bone maturity prediction model outputs a wrist bone maturity prediction value for bone age assessment. Both the finger bone maturity prediction model and the carpal bone maturity prediction model include a preset target deep neural network; The target deep neural network includes interconnected feature extraction units and feature concatenation units; The feature extraction unit includes: a first deep neural network layer for inputting hand bone image data, and a second deep neural network layer for inputting finger bone image data or wrist bone image data, and both the first deep neural network layer and the second deep neural network layer are connected to the feature splicing unit. The feature splicing unit includes at least one fully connected layer and a linear layer connected in sequence.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the bone maturity prediction method according to any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the bone maturity prediction method according to any one of claims 1 to 3.

Images