Bone age assessment method and system based on AI model and bone age instrument

Abstract

The invention relates to the technical field of medical image processing, artificial intelligence and the like, and provides a bone age assessment method and system based on an AI model and a bone age instrument. A detected left wrist X-ray image is preprocessed through gray scale standardization, noise filtering and a self-supervised learning model, the ratio of the width of the distal radius epiphysis to the width of the metaphysis, the length of the metaphysis of the third palm bone and graded skeleton features of the maturity of auxiliary ossification points are extracted through a Vision Transform, and multi-modal data fusion is conducted on the extracted skeleton features and demographic information. And a multi-layer perceptron model is adopted to predict the fused multi-dimensional bone age related data, and a bone age evaluation report is output, so that errors in bone age prediction of children and teenagers are effectively reduced, and the detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of medical image processing and artificial intelligence technology, and in particular to a method and system for automatically evaluating bone age of wrist X-ray images using a Transformer model. Background Art

[0002] Bone age is an important indicator for measuring the developmental maturity of children and adolescents in medicine. Accurate bone age assessment is of great significance for growth and development monitoring, clinical treatment plan formulation, and athlete selection. Traditional bone age assessment methods mainly rely on experienced radiologists or pediatricians to manually interpret X-ray images and make bone age judgments based on the Greulich-Pyle (GP) method or the Tanner-Whitehouse (TW) method. These methods rely on the personal experience of doctors and are significantly affected by the experience and fatigue of the readers. In addition, manual interpretation is not efficient enough for large-scale population screening and real-time monitoring, and there is an error accumulation problem in manual feature extraction. Although methods using convolutional neural networks (CNNs) have been explored for bone age assessment, CNNs have limitations in capturing long-range dependencies and global feature modeling, and are still insufficient for complex ossification points, age-group adaptability, and multimodal fusion integration.

[0003] In summary, existing bone age assessment methods have problems such as low degree of automation and insufficient accuracy. Summary of the invention

[0004] In view of the deficiencies in the above-mentioned prior art, the present invention provides a bone age assessment method and a bone age instrument based on an AI model, so that the bone age instrument can automatically perform accurate feature extraction and bone age prediction, thereby improving the accuracy and efficiency of bone age prediction.

[0005] In a first aspect, the present invention provides a bone age assessment method based on an AI model, comprising:

[0006] Obtaining an X-ray image of the left wrist of the examinee detected in a bone age meter and recording the demographic information of the examinee;

[0007] Preprocessing the X-ray image to obtain a preprocessed image;

[0008] Performing feature extraction on the preprocessed image through the trained Vision Transformer model to obtain the bone features of the subject;

[0009] Performing multimodal data fusion on the subject's skeletal features and the subject's demographic information to obtain multi-dimensional bone age-related data;

[0010] Processing the multi-dimensional bone age-related data through a trained deep learning regression model to evaluate bone age;

[0011] Based on the bone age assessment results of the deep learning regression model, a bone age assessment report is output.

[0012] In a second aspect, the present invention provides a bone age assessment system based on an AI model, and the bone age assessment system based on an AI model uses the above-mentioned bone age assessment method based on an AI model.

[0013] In a third aspect, the present invention provides a bone age meter, comprising a main control module, which runs the above-mentioned bone age assessment method based on the AI ​​model.

[0014] Compared with the prior art, the present invention has the following beneficial effects:

[0015] The present invention provides a bone age assessment method, system and bone age instrument based on an AI model. The method comprises the following steps: obtaining an X-ray image of the left wrist of a subject detected in the bone age instrument and recording the demographic information of the subject, preprocessing the X-ray image to obtain a preprocessed image, extracting features of the preprocessed image through a trained Vision Transformer model to obtain the bone features of the subject, performing multimodal data fusion on the bone features of the subject and the demographic information of the subject to obtain multi-dimensional bone age-related data, processing the multi-dimensional bone age-related data through a trained deep learning regression model to assess bone age, and outputting a bone age assessment report based on the bone age assessment result of the deep learning regression model, so that the bone age instrument can automatically perform accurate feature extraction and bone age prediction, thereby improving the accuracy and efficiency of bone age prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention and do not constitute an improper limitation of the present invention. Some specific embodiments of the present invention will be described in detail in an exemplary and non-restrictive manner with reference to the drawings. The same reference numerals in the drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. In the drawings:

[0017] Figure 1 It is a flow chart of a bone age assessment method of a bone age meter according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only an embodiment of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention.

[0019] Embodiment 1

[0020] See also Figure 1 The present embodiment provides a bone age assessment method based on an AI model, which can be run in a main control module of a bone age meter. By obtaining an X-ray image of the left wrist of a subject detected in the bone age meter and recording the demographic information of the subject, the X-ray image is preprocessed to obtain a preprocessed image, and a feature is extracted from the preprocessed image through a trained Vision Transformer model to obtain the bone features of the subject. The bone features of the subject are subjected to multimodal data fusion with the demographic information of the subject to obtain multi-dimensional bone age-related data, and the multi-dimensional bone age-related data is processed through a trained deep learning regression model to assess bone age. According to the bone age assessment result of the deep learning regression model, a bone age assessment report is output, so that the bone age meter can automatically perform accurate feature extraction and bone age prediction, thereby improving the accuracy and efficiency of bone age prediction.

[0021] See also Figure 1 The X-ray monitoring method of the bone age meter provided in this embodiment includes the following steps:

[0022] S101, obtaining an X-ray image of the left wrist of the examinee detected in a bone age meter and recording the demographic information of the examinee;

[0023] S102, preprocessing the X-ray image to obtain a preprocessed image;

[0024] S103, extracting features from the preprocessed image using the trained Vision Transformer model to obtain the bone features of the subject;

[0025] S104, performing multimodal data fusion on the subject's skeletal features and the subject's demographic information to obtain multi-dimensional bone age-related data;

[0026] S105, processing the multi-dimensional bone age-related data through the trained deep learning regression model to evaluate the bone age;

[0027] S106. Output a bone age assessment report based on the bone age assessment results of the deep learning regression model.

[0028] It should be noted that in step S101, an X-ray image of the left wrist of the subject detected in the bone age meter is obtained. The bone structure of the left wrist is a common part for clinically judging bone age. At the same time, the demographic information of the subject (such as gender, height, weight, race, date of birth, etc.) is recorded. This information is closely related to bone age assessment, which can assist subsequent multimodal data fusion and regression analysis, and provide more reference dimensions for bone age assessment, thereby improving the accuracy and reliability of the assessment.

[0029] In step S102, considering that the original X-ray image usually contains noise, low contrast or artifacts, directly using it in a deep learning model will affect the accuracy of feature extraction. In this embodiment, by preprocessing the original image of the left wrist X-ray image (such as denoising, enhancement, distortion correction, cropping, etc.), the image quality can be significantly improved, the interference factors in the model training and reasoning process can be reduced, and the subsequent Vision Transformer model can be more easily focused on bone features, thereby improving the accuracy and stability of bone age judgment.

[0030] In step S103, the preprocessed image is subjected to feature extraction through the trained Vision Transformer model to obtain bone features. Vision Transformer is a visual feature extraction network based on the Transformer architecture. Compared with the traditional convolutional neural network (CNN), it has better performance in capturing long-range dependencies, global information, and detail features. This step extracts higher-dimensional and finer-grained feature information of the bone structure, provides a more representative feature description for subsequent bone age calculations, and provides high-quality feature input for the bone age assessment model, thereby effectively improving the assessment accuracy.

[0031] In step S104, in medical applications, relying solely on image features may not fully reflect individual differences, and introducing demographic information (such as age, gender, height and weight, etc.) can further improve the prediction accuracy of bone age. Multimodal data fusion can integrate features from different sources, allowing the model to more comprehensively understand the growth and development of the subject.

[0032] In step S105, the final prediction of bone age is a regression problem, and the multimodal data obtained by the previous fusion is input into a deep learning regression model (such as based on a fully connected network or other advanced network structures) for training and reasoning. The model can be trained in advance and annotated and corrected according to the actual clinical evaluation results to ensure the accuracy of the prediction, so that it can make detailed predictions for individual differences and reduce the shortcomings of relying solely on traditional statistics or empirical formulas. In this step, through the deep learning regression model, automated, high-precision, personalized bone age assessment can be achieved, and it is easy to update and iterate (continue training based on more data and new algorithms).

[0033] In step S106, a bone age assessment report is output based on the bone age assessment results of the deep learning regression model. The bone age assessment ultimately needs to output the results in a clinically readable and easy-to-use form to provide doctors or patients with intuitive bone age analysis. In addition to the bone age value, the bone age can also be compared with the actual age or a reference interval can be given to obtain a bone age assessment report to help clinically evaluate the growth and development of the subject.

[0034] In some preferred embodiments, preprocessing the X-ray image to obtain a preprocessed image includes:

[0035] Performing grayscale standardization and noise filtering on the X-ray image;

[0036] The self-supervised learning model masked autoencoder is used to perform feature enhancement on the X-ray image after grayscale standardization and noise filtering to obtain a preprocessed image.

[0037] It should be noted that grayscale standardization refers to standardizing the grayscale values ​​of the original X-ray image so that it has a zero mean and a unit standard deviation. Grayscale standardization can reduce the interference caused by imaging differences.

[0038] The algorithm for grayscale standardization of the X-ray image is:

[0039]

[0040] Among them, I norm (x, y) is the gray value at the pixel coordinate (x, y) after grayscale standardization of the original X-ray image. (x,y) Gray value of the original X-ray image at pixel coordinate (x, y). μ and σ are the mean and standard deviation of the gray distribution of the original X-ray image, respectively.

[0041] It should be noted that after grayscale normalization of the X-ray image, in order to further reduce noise interference, it is preferred to use Gaussian filtering to smooth the grayscale normalized image again. A 7×7 Gaussian kernel can be used to perform convolution operation on the grayscale normalized image to smooth random noise while retaining the edge information of the key bone structure of the wrist as much as possible.

[0042] It should be noted that the self-supervised learning model masked autoencoder (abbreviated as: MAE encoder) randomly blocks some pixels (or image blocks) in the input image, forcing the model to "reconstruct" the blocked parts according to the contextual information of other regions, thereby learning more comprehensive and in-depth bone features. In bone age assessment, this higher quality feature representation enables the model to more accurately capture the development status of each part and enhance the robustness of the model. The MAE encoder performs feature enhancement on the X-ray image, including: dividing the X-ray image processed by grayscale standardization and Gaussian filtering into several image blocks, and randomly blocking a certain proportion (such as 30% to 50%). The MAE encoder only encodes the unmasked image blocks and learns a compact feature vector; the MAE decoder reconstructs the blocked part according to the feature vector. By continuously optimizing the reconstruction error, the model gradually improves the ability to restore the missing area. After the training is completed, the MAE encoder part is retained as a feature extractor to encode the new input image and obtain the "feature enhanced" image representation, and then these features are input into the subsequent bone age assessment model (Vision Transformer) to improve the overall assessment effect.

[0043] In some preferred embodiments, the bone characteristics include: the ratio of the width of the distal radial epiphysis to the metaphysis, the length of the metaphysis of the third metacarpal bone, and the maturity grade of the auxiliary ossification point.

[0044] It should be noted that the distal radial epiphysis and metaphysis are important indicators reflecting the degree of bone development. The distal radial epiphysis and metaphysis show regular changes with age and are easy to observe and measure. The ratio of the width of the distal radial epiphysis to the metaphysis can not only significantly reflect the regular changes in bone development, but also has the characteristics of being easy to observe and measure. The dimensionless ratio eliminates the influence of imaging conditions, providing efficient and reliable indicator support for bone age assessment. The length of the third metaphysis is a key indicator of metacarpal growth and development, which is highly correlated with bone age. The measurement position is fixed and has good repeatability. Auxiliary ossification points are located at specific positions of the carpal bones or phalanges. They are some secondary ossification centers (such as pisiform bones, sesamoid bones, etc.), which will show different morphologies in X-ray images during bone maturation. According to the morphology, density and fusion of ossification points with the surrounding bones, they can be divided into different levels: no ossification points (level 0), ossification points appear (level 1), ossification points develop (level 2) and ossification points close (level 3), to reflect the development stage of the auxiliary ossification points. Numerical indicators such as the ratio of the width of the distal radial epiphysis to the metaphysis and the length of the third metacarpal metaphysis can quantitatively reflect the morphological changes of bones during growth and development, and the maturity grading of auxiliary ossification points further supplements the judgment of the speed of bone development through the appearance and development of ossification centers. By combining these characteristics, bone age assessment is more accurate and comprehensive.

[0045] In some preferred embodiments, training the Vision Transformer model includes:

[0046] A training set of annotated wrist X-ray images is obtained, wherein the annotations include the ratio of the width of the distal radius epiphysis to the metaphysis, the length of the third metacarpal metaphysis, and the maturity grade of the auxiliary ossification points; the wrist X-ray images in the training set are divided into a number of image blocks; the image blocks are positionally encoded to obtain a sequence of image blocks with position information; the image block sequence is input into a Vision Transformer model, and feature learning is performed through a multi-head self-attention mechanism; for the numerical features of the ratio of the width of the distal radius epiphysis to the metaphysis and the length of the third metacarpal metaphysis, a mean square error loss function is used, and for the categorical features of the auxiliary ossification point maturity grade, a cross entropy loss function is used, and the error between the features output by the model and the manually annotated values ​​is calculated; based on the error, the parameters of the Vision Transformer model are optimized and updated until the model converges to obtain the trained Vision Transformer model.

[0047] It should be noted that the specific training process of the Vision Transformer model includes: obtaining a training set of annotated wrist X-ray images; annotating the following information for each image: the ratio of the width of the distal radial epiphysis to the metaphysis; the length of the third metacarpal metaphysis; the maturity level of the auxiliary ossification point. Each X-ray image is divided into blocks of 16×16 pixels; each image block is flattened into a vector and mapped to a feature vector through linear projection; position encoding is added to retain the spatial position information of the image to obtain an image block sequence, which is input into a multi-layer Transformer encoder; in each layer, it is calculated through a multi-head self-attention mechanism. After being processed by the multi-layer Transformer encoder, a feature vector is output, which contains a global understanding of the structures of the radius, metacarpal bones, and auxiliary ossification points. After the last layer of the Vision Transformer model, multiple fully connected layers are connected for regression and classification tasks respectively. In order to output the two continuous numerical indicators of the ratio of the width of the distal radial epiphysis to the metaphysis and the length of the third metacarpal metaphysis, two independent regression fully connected layers are connected for the ratio of the width of the distal radial epiphysis to the metaphysis and the length of the third metacarpal metaphysis, respectively. The number of output nodes in each layer is 1, which is used to predict the corresponding continuous bone features. For the classification task of the maturity level of the auxiliary ossification point, a fully connected layer is connected, and the number of output nodes is 4, corresponding to the four levels of ossification points: no ossification point (level 0), ossification point appearance (level 1), ossification point development (level 2), and ossification point closure (level 3). Based on the extracted ossification point feature vector, the model can select the category with the largest probability value as the final maturity classification of the auxiliary ossification point through the probability distribution of the classification output. For the numerical bone features of the ratio of the width of the distal radial epiphysis to the metaphysis and the length of the third metacarpal metaphysis, the mean square error loss function can be used. For the classification of bone features to assist the maturity level of ossification points, a cross entropy loss function can be used. The total loss is obtained by weighting the above mean square error loss and cross entropy loss. According to the total loss, the back propagation algorithm is combined with the optimizer Adam to continuously iteratively update the Vision Transformer model parameters, and the training is carried out until the loss function converges to obtain the trained Vision Transformer model. The Vision Transformer model is used to automatically extract bone features, achieve efficient feature extraction, and also bring more objective consistency evaluation, making up for the subjectivity of traditional manual judgment.

[0048] In some preferred embodiments, the skeletal features are subjected to multimodal data fusion with the demographic information of the subject, including: vectorizing the skeletal features to obtain a skeletal feature vector; standardizing and encoding the demographic information to obtain a demographic feature vector; and splicing the skeletal structure feature vector with the demographic feature vector of the subject to obtain the multi-dimensional bone age-related data.

[0049] It should be noted that the bone structure characteristics include: the ratio of the width of the distal radial epiphysis to the metaphysis, the length of the metaphysis of the third metacarpal, and the maturity grade of the auxiliary ossification point. Demographic characteristics include: age, height, and gender. Among them, the numerical characteristics, the ratio of the width of the distal radial epiphysis to the metaphysis, the length of the metaphysis of the third metacarpal, age, and height are standardized by Z-score. The maturity grade of the auxiliary ossification point and gender characteristics (male, female) are represented by one-hot encoding. The one-hot encoding of the maturity grade of the auxiliary ossification point is expressed as maturity level 0 [1,0,0,0], maturity level 1 [0,1,0,0], maturity level 2 [0,0,1,0], and maturity level 3 [0,0,0,1]. The gender characteristics are encoded as male: [1,0] female: [0,1]. The skeletal structural feature vector is quantified as [ratio of distal radial epiphysis to metaphysis width, metaphysis of the third metacarpal bone, maturity grade of auxiliary ossification points], and the demographic feature vector is quantified as [age, height, male, female]. The skeletal structural feature vector and the demographic feature vector are directly merged into a new vector in the feature dimension by splicing operation to form a multimodal input feature vector. By multimodal data fusion of skeletal features and demographic information of the subjects, the model can utilize more comprehensive feature information and improve the accuracy and reliability of bone age assessment.

[0050] In some preferred embodiments, when training the deep learning regression model, it includes: using a multilayer perceptron model as the deep learning regression model; training the multilayer perceptron model to obtain the trained deep learning regression model.

[0051] It should be noted that the multilayer perceptron model is a simple but powerful feedforward neural network. Through the design of the input layer, multiple hidden layers and the output layer, it can effectively process multi-dimensional bone age-related features. The specific network structure of the multilayer perceptron model is as follows: Input layer: accepts the fused multi-dimensional bone age-related feature vector. Hidden layer: contains 3 layers, each with 128 neurons, and the activation function is ReLU. Output layer: outputs the predicted bone age value.

[0052] In some preferred embodiments, training the multilayer perceptron model includes: using the multi-dimensional bone age-related data as model input and the corresponding standard bone age value as training target; calculating the error based on the difference between the predicted bone age value and the standard bone age value, and updating the model parameters using the back propagation algorithm; repeating the steps of using the multi-dimensional bone age-related data as model input and the corresponding standard bone age value as training target, calculating the error based on the difference between the predicted bone age value and the standard bone age value, and updating the model parameters using the back propagation algorithm, until the prediction error drops to the expected range, so as to obtain the trained deep learning regression model.

[0053] It should be noted that the training process of the multi-layer perceptron model includes: inputting the multi-dimensional feature vector after multimodal fusion and the corresponding standard bone age value into the model, calculating the error between the predicted bone age value and the actual bone age value, and using the mean square error loss function; using the back propagation algorithm to update the network parameters by gradient; looping the above process, when the mean square error loss function loss is lower than the preset threshold, the training process ends and the trained multi-layer perceptron model is obtained. During the training process, the Adam optimizer is used, the learning rate is set to 0.001, the batch size is 32, and the number of training rounds is 100. To prevent overfitting, the training process is dynamically adjusted in combination with the early stopping strategy. By using the multi-layer perceptron model to regress and predict the fused feature vector, the nonlinear relationship between bone development indicators and demographic information can be effectively captured, and bone age can be accurately predicted.

[0054] In some preferred embodiments, a bone age assessment report is output based on the bone age assessment results of the deep learning regression model. The bone age assessment report may include: predicted bone age value, bone feature explanation, and demographic information, which are used to compare the prediction results and provide better diagnosis and consulting services for doctors and patients.

[0055] Embodiment 2

[0056] See also Figure 1The present embodiment provides a bone age assessment system based on an AI model. The bone age assessment system based on an AI model uses the above-mentioned bone age assessment method based on an AI model, obtains an X-ray image of the left wrist of the subject detected in the bone age meter and records the demographic information of the subject, pre-processes the X-ray image to obtain a pre-processed image, extracts features from the pre-processed image through a trained Vision Transformer model to obtain the bone features of the subject, performs multimodal data fusion on the bone features of the subject and the demographic information of the subject to obtain multi-dimensional bone age-related data, processes the multi-dimensional bone age-related data through a trained deep learning regression model, assesses the bone age, and outputs a bone age assessment report based on the bone age assessment result of the deep learning regression model, so that the bone age meter can automatically perform accurate feature extraction and bone age prediction, thereby improving the accuracy and efficiency of bone age prediction.

[0057] The above embodiments are only preferred specific implementation modes of the present invention, and the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. The protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A bone age assessment method based on an AI model, characterized in that: The following steps are involved: Obtaining an X-ray image of the left wrist of the examinee detected in a bone age meter and recording the demographic information of the examinee; Preprocessing the X-ray image to obtain a preprocessed image; Performing feature extraction on the preprocessed image through the trained Vision Transformer model to obtain the bone features of the subject; Performing multimodal data fusion on the subject's skeletal features and the subject's demographic information to obtain multi-dimensional bone age-related data; Processing the multi-dimensional bone age-related data through a trained deep learning regression model to evaluate bone age; Based on the bone age assessment results of the deep learning regression model, a bone age assessment report is output.

2. The bone age assessment method based on the AI ​​model according to claim 1, characterized in that: Preprocessing the X-ray image to obtain a preprocessed image includes: Performing grayscale standardization and noise filtering on the X-ray image; The self-supervised learning model masked autoencoder is used to perform feature enhancement on the X-ray image after grayscale standardization and noise filtering to obtain a preprocessed image.

3. The bone age assessment method based on the AI ​​model according to claim 1, characterized in that: The bone characteristics include: the ratio of the width of the distal radial epiphysis to the metaphysis, the length of the metaphysis of the third metacarpal bone, and the maturity grade of the auxiliary ossification point.

4. The bone age assessment method based on the AI ​​model according to claim 3 is characterized in that: When training the VisionTransformer model, it includes: A training set of annotated wrist X-ray images is obtained, wherein the annotations include the ratio of the width of the distal radius epiphysis to the metaphysis, the length of the third metacarpal metaphysis, and the maturity grade of the auxiliary ossification points; the wrist X-ray images in the training set are divided into a number of image blocks; the image blocks are positionally encoded to obtain a sequence of image blocks with position information; the image block sequence is input into a Vision Transformer model, and feature learning is performed through a multi-head self-attention mechanism; for the numerical features of the ratio of the width of the distal radius epiphysis to the metaphysis and the length of the third metacarpal metaphysis, a mean square error loss function is used, and for the categorical features of the auxiliary ossification point maturity grade, a cross entropy loss function is used, and the error between the features output by the model and the manually annotated values ​​is calculated; based on the error, the parameters of the Vision Transformer model are optimized and updated until the model converges to obtain the trained Vision Transformer model.

5. The bone age assessment method based on the AI ​​model according to claim 1, characterized in that: The skeletal features are fused with the demographic information of the subject in multimodal data, including: vectorizing the skeletal features to obtain a skeletal feature vector; standardizing and encoding the demographic information to obtain a demographic feature vector; and concatenating the skeletal structure feature vector with the demographic feature vector of the subject to obtain the multi-dimensional bone age-related data.

6. The bone age assessment method based on the AI ​​model according to claim 1, characterized in that: When training the deep learning regression model, it includes: using a multilayer perceptron model as the deep learning regression model; training the multilayer perceptron model to obtain the trained deep learning regression model.

7. The bone age assessment method based on the AI ​​model according to claim 6, characterized in that: When training the multi-layer perceptron model, it includes: taking the multi-dimensional bone age-related data as model input and the corresponding standard bone age value as training target; calculating the error according to the difference between the predicted bone age value and the standard bone age value, and updating the model parameters using the back propagation algorithm; repeating the steps of taking the multi-dimensional bone age-related data as model input and the corresponding standard bone age value as training target, calculating the error according to the difference between the predicted bone age value and the standard bone age value, and updating the model parameters using the back propagation algorithm, until the prediction error drops to the expected range, so as to obtain the trained deep learning regression model.

8. The bone age assessment method based on the AI ​​model according to any one of claims 1 to 7, characterized in that: The bone age assessment report includes a predicted bone age value and an explanation of skeletal characteristics.

9. A bone age assessment system based on an AI model, characterized in that: The AI ​​model-based bone age assessment system uses the AI ​​model-based bone age assessment method as described in any one of claims 1-8.

10. A bone age meter, comprising a main control module, characterized in that: The main control module runs the bone age assessment method based on the AI ​​model as described in any one of claims 1-9.

Images