Bone age prediction method, device and equipment based on deep learning model

Abstract

The invention discloses a bone age prediction method, device and equipment based on a deep learning model, and belongs to the technical field of medical image analysis, and the bone age prediction method comprises the steps: carrying out the preprocessing of an original hand bone image of a current user, and obtaining a target hand bone image; inputting the target hand bone image into a trained region cutting network to obtain a target segmentation region; inputting the target hand bone image marking the target segmentation region into a trained deep learning model, so as to enable the deep learning model to extract hand bone features of the current user from the target segmentation region; and predicting the bone features of the hand bones by using a regression model to obtain bone age prediction data of the current user. According to the method, accurate skeleton region segmentation is performed by using the segmentation network, and the bone age prediction is performed in combination with the deep learning model, so that the interference of background noise can be avoided while effective skeleton features are extracted, and the prediction accuracy and reliability are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of medical image analysis technology, and more specifically, relates to a method, device, and equipment for predicting bone age based on a deep learning model. Background Technology

[0002] Accurate prediction of bone age in adolescents is crucial for medical diagnosis, developmental assessment, and health management. Traditional bone age assessment methods primarily rely on radiologists' manual analysis and interpretation of hand bone X-ray images. This method is not only time-consuming but also susceptible to subjective influences, leading to some inaccuracies in diagnostic results. With the development of artificial intelligence technology, especially the rapid advancement of deep learning, automated bone age prediction methods have gradually become a research hotspot.

[0003] In existing technologies, image-based bone age prediction methods mostly employ deep learning models, especially convolutional neural networks (CNNs) and deep residual networks (ResNets), to perform regression or classification predictions of bone age by learning the features of hand bone X-ray images. These methods can automatically extract high-dimensional features from images and avoid manual intervention, thus improving prediction efficiency and accuracy. However, most existing deep learning methods focus on image classification, neglecting the accuracy of bone age prediction. Summary of the Invention

[0004] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides a bone age prediction method, device and equipment based on a deep learning model. The purpose is to solve the technical problem that most existing deep learning methods focus on image classification and ignore the accuracy of bone age prediction.

[0005] To achieve the above objectives, according to one aspect of the present invention, a bone age prediction method based on a deep learning model is provided, comprising:

[0006] S1: Preprocess the original hand bone image of the current user to obtain the target hand bone image;

[0007] S2: Input the target hand bone image into the trained region segmentation network to obtain the target segmentation region;

[0008] S3: Input the target hand bone image labeled with the target segmentation region into the trained deep learning model so that it can extract the hand bone skeletal features of the current user from the target segmentation region;

[0009] S4: Use a regression model to predict the skeletal features of the hand bones to obtain the bone age prediction data of the current user.

[0010] Further, the S1 comprises: performing grayscale, normalization and noise removal on the original hand bone image of the current user acquired from the X-ray device to obtain the target hand bone image.

[0011] Further, the region segmentation network is U-Net or Mask R-CNN. Further, the S2 comprises: inputting the target hand bone image into the trained region segmentation network to make it separate the target segmentation region from the background and other irrelevant regions, ensuring that only the features of the target segmentation region are used in subsequent prediction.

[0012] Further, the deep learning model is a deep residual network ResNet. Further, the S3 comprises: extracting key information from the target hand bone image by the deep learning model, including shape features, structure features, size features, bone density features and bone growth region features of the skeleton, to serve as the hand bone skeleton features of the current user.

[0013] Further, the regression model is a deep neural network or a support vector machine.

[0014] According to another aspect of the present application, a bone age prediction device based on a deep learning model is provided, comprising:

[0015] a preprocessing module configured to preprocess an original hand bone image of a current user to obtain a target hand bone image;

[0016] a region segmentation module configured to input the target hand bone image into a trained region segmentation network to obtain a target segmentation region;

[0017] a feature extraction module configured to input the target hand bone image marked with the target segmentation region into a trained deep learning model to make it extract hand bone skeleton features of the current user from the target segmentation region;

[0018] a regression prediction module configured to predict the hand bone skeleton features by a regression model to obtain bone age prediction data of the current user.

[0019] According to another aspect of the present application, a bone age prediction device is provided, comprising a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the above method when executing the computer program.

[0020] According to another aspect of the present application, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps of the above method.

[0021] Overall, the above technical solutions conceived by the present application can achieve the following beneficial effects compared with the prior art:

[0022] (1) To address the shortcomings of existing methods, the present application utilizes segmentation networks such as U-Net and Mask R-CNN for precise skeletal region segmentation, combined with deep learning models for bone age prediction. This approach can extract effective skeletal features while avoiding background noise interference, thereby significantly improving the accuracy and reliability of the prediction.

[0023] (2) The present application combines deep convolutional neural networks (CNN) and skeletal segmentation techniques such as U-Net and Mask R-CNN to accurately extract useful skeletal features from adolescent hand X-ray images. These models have strong image feature learning capabilities and can automatically identify and extract subtle differences in images, such as bone morphology, structure, and epiphyseal development. Combined with segmentation networks such as U-Net and Mask R-CNN, the image segmentation process can accurately distinguish between skeletal and background regions, ensuring that the prediction model only relies on skeletal features and is not affected by background noise. Through this fine segmentation method, the present application significantly improves the reliability of bone age prediction, especially in complex or low-quality X-ray images.

[0024] (3) The present application uses deep learning models to achieve automatic feature extraction and prediction during the training process, avoiding the complex process of relying on manual feature engineering in traditional methods. Traditional methods often require experts to design features based on experience, which not only requires a lot of time and effort, but also may introduce human error. Through deep learning models, especially convolutional neural networks (CNN) and deep residual networks (ResNet), the present application can automatically learn the most discriminative features from a large amount of image data, effectively eliminating the limitations of human intervention. This automated process not only improves prediction efficiency, but also enables the model to mine potential complex rules from massive data, improving the accuracy of bone age prediction.

[0025] (4) The deep learning-based bone age prediction method of the present application has strong adaptability and scalability, and can be applied to large-scale bone age evaluation tasks. As the data set increases, the model can gradually improve its prediction ability and provide more scientific and accurate basis for adolescent bone age screening, health management, and personalized medicine. In the clinical and public health fields, the present application can provide medical institutions with efficient and convenient bone age evaluation tools, especially in scenarios where a large number of patient X-ray images need to be processed, which can significantly improve the efficiency of diagnosis and reduce labor costs. At the same time, with the continuous enrichment of the training set, the model also has the potential for further optimization and self-improvement, and has broad application prospects and market value.

[0026] (5) The present application integrates multiple advanced techniques in deep learning, such as convolutional neural networks, deep residual networks, U-Net, etc., in a unified framework, and proposes an innovative bone age prediction model design through residual connection and multi-level feature extraction. Through this design, the present application can accurately assess the bone age of adolescents in different age groups, not only solving the limitations of traditional methods, but also further improving the processing capacity of complex data. Compared with traditional prediction methods based on artificial features and traditional machine learning models, the present application has significant advantages. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a flowchart of the bone age prediction method based on a deep learning model provided by Embodiment 1 of the present application. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0029] Embodiment 1

[0030] As shown in Figure 1 , the present embodiment provides a bone age prediction method based on a deep learning model, comprising: S1-S4.

[0031] S1: pre-processing the original hand bone image of the current user to obtain a target hand bone image.

[0032] S2: inputting the target hand bone image into a trained region segmentation network to obtain a target segmentation region. The goal of this process is to separate the bone region in the image from the background and other irrelevant regions, ensuring that only the features of the bone region are used in subsequent prediction, thereby improving the accuracy of prediction. Segmentation algorithms such as U-Net or Mask R-CNN are used, which can accurately identify and separate the bone region in the hand bone image, while avoiding the influence of background noise and other irrelevant regions on bone age prediction. Through segmentation, a clearer bone image can be extracted, making bone age prediction more accurate.

[0033] S3: input the target hand bone image marked with the target segmentation region into the trained deep learning model to extract the current user's hand bone skeletal features from the target segmentation region. The goal of this step is to extract key information from the hand bone X-ray image through the deep learning model, including the shape, structure, size, bone density, and bone growth area of the skeleton. The extracted features can be local features (such as the edges and contours of the skeleton) and global features (such as the overall shape and symmetry of the skeleton). Through the training of a convolutional neural network (CNN) or other deep learning model, these features can be automatically learned and extracted in different convolutional layers. This process avoids the complexity of manual feature extraction and can learn and process detailed features of the skeleton at multiple levels.

[0034] S4: use a regression model to predict the hand bone skeletal features to obtain the bone age prediction data of the current user. The goal of this step is to predict the bone age of adolescents based on the segmented skeletal region features. By using a regression model, especially a deep neural network (DNN), a support vector machine (SVM), or other regression models, the extracted skeletal features are mapped to the actual bone age value. This process usually trains the model through a training data set and optimizes the accuracy of the model by minimizing the error between the predicted value and the true bone age value. Deep neural networks can handle complex nonlinear relationships through multi-level feature learning, thereby improving the accuracy of bone age prediction.

[0035] Further, S1 includes: performing grayscale, normalization, and noise removal on the original hand bone image of the current user obtained from the X-ray device to obtain the target hand bone image. Specifically, the original hand bone X-ray image obtained from the X-ray device is preprocessed; the image quality is improved based on grayscale, normalization, and noise removal operations.

[0036] Further, the region segmentation network is U-Net or Mask R-CNN. Further, S2 includes: inputting the target hand bone image into the trained region segmentation network to separate the target segmentation region from the background and other irrelevant regions, ensuring that only the features of the target segmentation region are used in subsequent prediction.

[0037] Wherein, segmentation techniques such as U-Net or Mask R-CNN are used to accurately segment the skeletal region in the image. The goal of this process is to separate the skeletal region from the background and other irrelevant regions in the image to more accurately predict the bone age. Through accurate segmentation of the skeletal region, only the features of the skeletal region are input into the subsequent bone age prediction model, avoiding the influence of background noise.

[0038] Further, the deep learning model is a deep residual network ResNet. Further, S3 comprises: extracting key information from the target hand bone image by the deep learning model, including shape features, structure features, size features, bone density features, and bone growth region features of the bone, to serve as the hand bone skeletal features of the current user.

[0039] wherein the skeletal features are automatically extracted from the adolescent hand bone X-ray image using a deep learning model (such as a deep residual network ResNet, U-Net, etc.). The goal of this step is to extract key information from the image, such as the shape, structure, size, and bone density of the bone, through a deep learning model. By training these deep learning models, features can be automatically learned and extracted at multiple levels, thereby providing high-quality input for subsequent bone age prediction.

[0040] Further preferably, the loss function of the above-mentioned adolescent bone age prediction model is:

[0041]

[0042] wherein m is the number of hand bone X-ray images in the training set; Y i is the real bone age value corresponding to the ith hand bone X-ray image; is the predicted bone age value when the ith hand bone X-ray image is input; E i is the predicted residual error of the ith image.

[0043] Further, the regression model is a deep neural network or a support vector machine.

[0044] wherein based on the segmented bone region features, a regression model (such as a deep neural network, a support vector machine, etc.) is used for bone age prediction. S4: Using the regression model to predict the hand bone skeletal features to obtain the bone age prediction data of the current user. The goal of this step is to predict the bone age of adolescents according to the extracted skeletal features, usually through a regression model to output a specific age. This step uses a deep learning model for regression training, which can accurately predict the age according to the bone structure in the image.

[0045] Embodiment 2

[0046] The embodiment provides a bone age prediction device based on a deep learning model, comprising: a preprocessing module configured to preprocess an original hand bone image of a current user to obtain a target hand bone image; a region cutting module configured to input the target hand bone image into a trained region cutting network to obtain a target segmentation region; a feature extraction module configured to input the target hand bone image marked with the target segmentation region into a trained deep learning model, so that the deep learning model extracts hand bone skeleton features of the current user from the target segmentation region; and a regression prediction module configured to predict the hand bone skeleton features by using a regression model to obtain bone age prediction data of the current user.

[0047] Embodiment 3

[0048] The embodiment provides a bone age prediction device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0049] Embodiment 4

[0050] The embodiment provides a computer readable storage medium, which stores a computer program, and the computer program implements the steps of the above method when executed by a processor.

[0051] Those skilled in the art can easily understand that the above description is only the preferred embodiment of the present application, and is not intended to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A bone age prediction method based on a deep learning model, characterized in that, include: S1: Preprocess the original hand bone image of the current user to obtain the target hand bone image; S2: Input the target hand bone image into the trained region segmentation network to obtain the target segmentation region; S3: Input the target hand bone image labeled with the target segmentation region into the trained deep learning model so that it can extract the hand bone skeletal features of the current user from the target segmentation region; S4: Use a regression model to predict the skeletal features of the hand bones to obtain the bone age prediction data of the current user.

2. The bone age prediction method based on a deep learning model as described in claim 1, characterized in that, S1 includes: performing grayscale conversion, normalization, and noise removal on the original hand bone image of the current user obtained from the X-ray device to obtain the target hand bone image.

3. The bone age prediction method based on a deep learning model as described in claim 1, characterized in that, The region segmentation network is either U-Net or Mask R-CNN.

4. The bone age prediction method based on a deep learning model as described in claim 3, characterized in that, S2 includes: inputting the target hand bone image into a trained region segmentation network so that it separates the target segmented region from the background and other irrelevant regions, ensuring that only the features of the target segmented region are used in subsequent predictions.

5. The bone age prediction method based on a deep learning model as described in claim 1, characterized in that, The deep learning model is a deep residual network, ResNet.

6. The bone age prediction method based on a deep learning model as described in claim 5, characterized in that, The S3 includes: extracting key information from the target hand bone image using a deep learning model, including the shape features, structural features, size features, bone density features, and bone growth region features of the bones, as the hand bone skeletal features of the current user.

7. The bone age prediction method based on a deep learning model as described in claim 1, characterized in that, The regression model is a deep neural network or a support vector machine.

8. A bone age prediction device based on a deep learning model, characterized in that, include: The preprocessing module is used to preprocess the original hand bone image of the current user to obtain the target hand bone image; The region segmentation module is used to input the target hand bone image into a trained region segmentation network to obtain the target segmented region; The feature extraction module is used to input the target hand bone image labeled with the target segmentation region into the trained deep learning model so that it can extract the hand bone skeletal features of the current user from the target segmentation region. The regression prediction module is used to predict the skeletal features of the hand bones using a regression model to obtain the bone age prediction data of the current user.

9. A bone age prediction device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images