Tibia shaft fracture bone healing state evaluation model training method, evaluation method and system

Abstract

The invention provides a tibia shaft fracture bone healing state evaluation model training method, evaluation method and system, and the method comprises the steps: constructing a training sample set, each sample comprises a tibia shaft X-ray image, and adding a label for each sample; an evaluation model is constructed, the evaluation model comprises a detection module and a scoring module, and the scoring module does not adopt a discrete classification mode any more, but directly outputs continuous scoring values through a full-connection layer; the evaluation model takes the tibia shaft X-ray image as input, the detection module detects to obtain a predicted fracture area, the scoring module scores the bone healing state of the predicted fracture area to obtain a predicted score value, and the predicted score value is used as output of the evaluation model; and training the detection module by adopting the training sample set, fixing parameters of the detection module after training is completed, and then training the scoring module until a preset performance is achieved, thereby obtaining the tibia shaft fracture bone healing state evaluation model. The evaluation model provided by the invention is high in efficiency and high in accuracy, and supports automatic continuous scoring.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a tibial shaft fracture bone healing status assessment model training method, assessment method and system. Background Art

[0002] Tibia shaft fracture is one of the most common long bone fractures. Accurate clinical assessment of its healing degree is crucial for developing a reasonable rehabilitation plan. For example, whether a patient can perform weight-bearing training and when to remove internal fixators such as plates all depend on the assessment of fracture healing status. At present, doctors mainly use the mRUST scoring system (modified Radiographic Union Score for Tibia Fractures) to determine fracture healing through X-ray images. This scoring system requires doctors to independently score multiple cortical bone parts of the frontal and lateral X-rays, which is time-consuming, labor-intensive and highly subjective. Especially when multiple doctors are required to collaborate in scoring, it is inefficient and prone to subjective errors.

[0003] There is no method that can automatically score mRUST, but there are some similar studies, such as the location and classification of fractures. Aertong et al. used the YOLO model to detect forearm fractures in children, and its average accuracy was close to that of professional radiologists and significantly better than that of ordinary doctors. A study pointed out that YOLO's average accuracy in fracture detection is better than other advanced algorithms. However, YOLO is mainly designed for fast target detection and is not effective for complex classification tasks, so it needs to be used in combination with other networks. Therefore, if fracture location and mRUST scoring are performed directly based on a one-stage model, these subtle and gradually changing features may not be accurately identified and evaluated, resulting in inaccurate scoring results. In contrast, SwinTransformer performs better in crack feature extraction. Its self-attention mechanism and hierarchical design can capture features of different ranges at different levels, while convolutional networks expand the receptive field by increasing depth and continuous convolution operations, which may introduce feature loss. Liu et al. combined YOLO and SwinTransformer to classify femoral fractures by AO / OTA, and the results showed that this method is more accurate than using YOLO alone. However, bone healing requires observation of changes such as callus formation, gradual connection of bone bridges, and blurring of fracture lines. Whether this algorithm is applicable to bone healing research still needs to be verified.

[0004] Bone healing is a dynamic physiological process that goes through multiple stages from fracture to complete healing, including inflammation, repair, and remodeling, involving a series of continuous physiological processes such as cell proliferation, matrix formation, and mineralization. However, traditional classification methods usually use a cross-entropy loss function, which ignores the information of the label order at different stages of bone healing, resulting in the mRUST score being unable to accurately reflect the healing status in clinical practice. Summary of the invention

[0005] In view of this, the embodiments of the present invention provide a tibial shaft fracture bone healing status assessment model training method, assessment method and system to eliminate or improve one or more defects existing in the prior art.

[0006] In one aspect, the present invention provides a method for training a tibial shaft fracture bone healing status assessment model based on mRUST, the method comprising the following steps: Constructing a training sample set, wherein the training sample set includes a plurality of samples, each sample includes a tibial shaft X-ray image; annotating a true fracture area and a true score value for each tibial shaft X-ray image; Constructing an evaluation model, the evaluation model includes a detection module and a scoring module, the scoring module removes the Softmax function, and the fully connected layer directly outputs the result; the evaluation model takes the tibial shaft X-ray image as input, the detection module detects the tibial shaft X-ray image to obtain a predicted fracture area, the scoring module scores the bone healing state of the predicted fracture area to obtain a predicted score value, and the predicted score value is used as the output of the evaluation model; The detection module is trained using the training sample set. After the training is completed, the parameters of the detection module are fixed, and then the scoring module is trained until the preset performance is achieved to obtain a tibial shaft fracture bone healing state assessment model.

[0007] In some embodiments of the present invention, before constructing the training sample set, the tibial shaft X-ray image is preprocessed, including: Normalizing the tibial shaft X-ray image to scale pixel values ​​to a range of 0 to 1; increasing the contrast of the tibial shaft X-ray image using adaptive histogram equalization; The data enhancement technology is used to randomly transform the tibial shaft X-ray image to generate multiple different versions of the image.

[0008] In some embodiments of the present invention, an evaluation model is constructed, wherein the evaluation model includes a detection module and a scoring module, and further includes: The detection module adopts the YOLOv9 model and CSPDarkNet as the backbone network. Through programmable gradient information technology, the gradient propagation path, direction and size are adaptively adjusted according to actual needs when calculating the gradient, so as to optimize the gradient transfer and improve the detection accuracy.

[0009] In some embodiments of the present invention, the detection module detects the tibial shaft X-ray image to obtain a predicted fracture area, including: Inputting the tibial shaft X-ray image into the detection module; The detection module extracts the features of the tibial shaft X-ray image through a convolutional network to generate a feature map; generates an anchor frame on the feature map, and predicts boundary parameters for each anchor frame to obtain the predicted fracture area.

[0010] In some embodiments of the present invention, before the scoring module scores the bone healing state of the predicted fracture area, the method further includes: The pixels of the predicted fracture area are adjusted, and when the size becomes smaller after adjustment, black edges are filled to keep the size consistent.

[0011] In some embodiments of the present invention, when the detection module is trained using the training sample set, the following steps are included: The intersection-and-union ratio of the predicted fracture area and the actual fracture area is calculated, the predicted fracture area is screened according to a preset intersection-and-union ratio threshold, and the detection module parameters are adjusted.

[0012] In some embodiments of the present invention, training the scoring module includes: A mean square error loss between the predicted score value and the true score value is constructed, and the scoring module is optimized with the goal of minimizing the mean square error loss.

[0013] On the other hand, the present invention also provides a method for evaluating bone healing status of tibial shaft fractures based on mRUST, the method comprising: Obtain an X-ray image of the tibial shaft to be evaluated; The tibial shaft X-ray image is input into the tibial shaft fracture bone healing state assessment model trained by any of the mRUST-based tibial shaft fracture bone healing state assessment model training methods mentioned above to generate a tibial shaft fracture bone healing state score value.

[0014] On the other hand, the present invention also provides a tibial shaft fracture bone healing status assessment system based on mRUST, and when the system is executed, it can implement the tibial shaft fracture bone healing status assessment method based on mRUST as described above, and the system includes: A data processing module, used for acquiring an X-ray image of the tibia shaft to be evaluated, and preprocessing the X-ray image of the tibia shaft; A status scoring module, comprising a tibial shaft fracture bone healing status assessment model trained based on any of the mRUST-based tibial shaft fracture bone healing status assessment model training methods mentioned above, and used for automatically assessing the tibial shaft fracture bone healing status based on the tibial shaft X-ray image; The evaluation output module is used to output the bone healing status score value of the tibial shaft fracture.

[0015] On the other hand, the present invention also provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of any of the methods mentioned above.

[0016] The present invention provides a training method, an evaluation method and a system for a tibial shaft fracture bone healing state evaluation model, comprising: constructing a training sample set, each sample comprising a tibial shaft X-ray image, and adding a label to each sample; constructing an evaluation model, comprising a detection module and a scoring module, wherein the scoring module no longer adopts a discrete classification method, but directly outputs a continuous scoring value through a fully connected layer; the evaluation model uses the tibial shaft X-ray image as input, and the detection module detects to obtain a predicted fracture area, and the scoring module scores the bone healing state of the predicted fracture area to obtain a predicted scoring value, and uses it as the output of the evaluation model; the training sample set is used to train the detection module, and after the training is completed, the parameters of the detection module are fixed, and then the scoring module is trained until the preset performance is reached, so as to obtain a tibial shaft fracture bone healing state evaluation model. The tibial shaft fracture bone healing state assessment model obtained by training based on the training method provided by the present invention can accurately detect the fracture area and assess the bone healing state, reduce subjective errors, and improve scoring accuracy; further, automatic scoring is performed based on the tibial shaft fracture bone healing state assessment model, thereby reducing the time and labor costs required for manual scoring and improving scoring efficiency; further, the Softmax function of the scoring module is removed, and the results are directly output by the fully connected layer, so that continuous scoring values ​​can be obtained, thereby more carefully reflecting the dynamic process of bone healing and overcoming the discreteness limitation of the existing mRUST scoring.

[0017] Additional advantages, purposes, and features of the present invention will be described in part in the following description, and will become apparent to those skilled in the art after studying the following, or may be learned from the practice of the present invention. The purposes and other advantages of the present invention may be achieved and obtained by the structures specifically indicated in the specification and the accompanying drawings.

[0018] Those skilled in the art will appreciate that the objectives and advantages that can be achieved with the present invention are not limited to the above specific description, and the above and other objectives that can be achieved by the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of the present application, and do not constitute a limitation of the present invention. In the drawings: Figure 1 Schematic diagram of the steps of a tibial shaft fracture bone healing status assessment model training method based on mRUST in one embodiment of the present invention.

[0020] Figure 2 Schematic diagram of the flow of a tibial shaft fracture bone healing status assessment model training method based on mRUST in one embodiment of the present invention.

[0021] Figure 3 Schematic diagram of the structure of an evaluation model in one embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0023] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions according to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0024] It should be emphasized that the term “include / comprises” when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

[0025] It should also be noted that, unless otherwise specified, the term “connection” herein may refer not only to a direct connection but also to an indirect connection involving an intermediate.

[0026] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the accompanying drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

[0027] In order to solve the problems of low efficiency, strong subjectivity and low automation of the existing mRUST scoring technology for fracture bone healing status, and the problem that the existing mRUST scoring technology based on deep learning ignores the information of label order at different stages of bone healing status, resulting in the problem that the mRUST scoring cannot accurately reflect the healing status in clinical practice, the present invention provides a training method for a tibial shaft fracture bone healing status assessment model based on mRUST, such as Figure 1 As shown, the method includes the following steps S101-S103: Step S101: construct a training sample set, the training sample set includes multiple samples, each sample includes a tibial shaft X-ray image; annotate each tibial shaft X-ray image with a true fracture area and a true score value.

[0028] Step S102: construct an evaluation model, which includes a detection module and a scoring module. The scoring module removes the Softmax function and directly outputs the result by the fully connected layer. The evaluation model takes the tibial shaft X-ray image as input, and the detection module detects the tibial shaft X-ray image to obtain a predicted fracture area. The scoring module scores the bone healing state of the predicted fracture area to obtain a predicted score value, and the predicted score value is used as the output of the evaluation model.

[0029] Step S103: The detection module is trained using the training sample set. After the training is completed, the parameters of the detection module are fixed, and then the scoring module is trained until the preset performance is achieved to obtain a tibial shaft fracture bone healing status assessment model.

[0030] like Figure 2 Shown is a flowchart of the training method of the tibial shaft fracture bone healing status assessment model based on mRUST.

[0031] In step S101, a training sample set is constructed.

[0032] In some embodiments, a large number of tibial shaft X-ray images at different healing stages are collected from hospitals, and the available tibial shaft X-ray images are screened according to preset conditions to construct a data set. A plurality of experienced orthopedic surgeons annotate the tibial shaft X-ray images with the fracture area and the healing status score. When the opinions of multiple orthopedic surgeons are not consistent, the more senior orthopedic surgeon makes the final judgment.

[0033] In some embodiments, a preprocessing operation is performed on each tibial shaft X-ray image in the data set, including: The tibial shaft X-ray images are normalized and the pixel values ​​are scaled to the range of 0 to 1, so as to avoid the problem of unstable training or difficult convergence due to different pixel value ranges during model training.

[0034] Increase the contrast of tibial shaft X-ray images using adaptive histogram equalization.

[0035] Data augmentation technology is used to perform random transformations on tibial shaft X-ray images, such as rotation, flipping, scaling, etc., to generate multiple different versions of images to improve the robustness and generalization ability of the model and reduce the risk of overfitting.

[0036] In some embodiments, the constructed data set is divided into a training set (training sample set) and a test set according to a preset ratio, such as 80% of the data set as a training set and 20% of the data set as a test set. The training set is used to train the model, and the test set is used to test the model performance.

[0037] In step S102, an evaluation model is constructed, which includes a detection module and a scoring module. The detection module is used to detect and locate the fracture area of ​​the preprocessed tibial shaft X-ray image, and the scoring module is used to score the bone healing state of the fracture area.

[0038] In some embodiments, Figure 3 As shown in the figure, the detection module uses the YOLOv9 model, and the scoring module uses the SwinTransformer V2 model. The scoring module needs to remove the Softmax function, and the fully connected layer directly outputs the result, changing the model from a classification task to a regression task, thereby outputting continuous scoring values, overcoming the limitation of the discreteness of mRUST scoring in the prior art.

[0039] After preprocessing such as cropping and image enhancement, the tibial X-ray image is input into the evaluation model. The convolutional network of the YOLOv9 model first extracts the features of the tibial X-ray image, generates a feature map, generates an anchor box on the feature map, and predicts the boundary parameters for each anchor box to obtain the predicted fracture area. The boundary parameters include the horizontal coordinate of the center point , the vertical coordinate of the center point ,width ,high and confidence score , the confidence score indicates the probability that the anchor box contains physical. Calculate the intersection over union (IoU) of the predicted fracture area and the true fracture area, select the predicted fracture areas that meet the threshold, and resize the predicted fracture areas that meet the threshold. Figure 3 The predicted fracture area is represented by ROI (Region of Interest).

[0040] The predicted fracture area detected by the YOLOv9 model is then input into the SwinTransformer V2 model for mRUST scoring. Since the Softmax function is removed, continuous scoring values ​​are directly output through the fully connected layer to obtain the predicted scoring value, which is used as the final output of the evaluation model.

[0041] In some embodiments, the size of the predicted fracture area is adjusted. For example, the predicted fracture area is adjusted from 256x256 pixels to 224x224 pixels, and the size is kept consistent by filling with black borders.

[0042] In some embodiments, the detection module adopts the YOLOv9 model and uses CSPDarkNet as the backbone network. Through programmable gradient information technology, the gradient propagation path, direction and size are adaptively adjusted according to actual needs when calculating the gradient to optimize the gradient transfer and improve the detection accuracy.

[0043] In step S103, the detection module is firstly trained using the tibial shaft X-ray image.

[0044] In some embodiments, the intersection-and-union ratio of the predicted fracture area obtained by the detection module and the actual fracture area is calculated, the predicted fracture area is screened according to a preset intersection-and-union ratio threshold, and the detection module parameters are adjusted.

[0045] In some embodiments, when training the detection module, a stochastic gradient descent optimizer is used, the batch size is 16, the initial learning rate is 0.001, the weight decay is 0.0005, the momentum is 0.937, the image size is 640x640 pixels, 100 epochs are trained, and an early stopping strategy is adopted to ensure model stability.

[0046] After the training is completed, the parameters of the detection module are fixed, and then the scoring module is trained.

[0047] In some embodiments, the mean square error loss between the predicted score value and the true score value obtained by constructing the scoring module is used to optimize the scoring module with the goal of minimizing the mean square error loss.

[0048] In some embodiments, when training the scoring module, an Adam optimizer is used, a batch size of 32, an initial learning rate of 0.001, 60 epochs are trained, and an early stopping strategy is used to prevent overfitting.

[0049] After the training is completed, the tibial shaft fracture bone healing status assessment model is finally obtained.

[0050] In some embodiments, the test set mentioned in step S101 is used to test the trained tibial shaft fracture bone healing state assessment model to evaluate the performance of the model. The YOLOv9 model is mainly evaluated from the aspects of macro-average precision, macro-average recall, macro-average F1 score, and average precision under different IoU thresholds; the SwinTransformer V2 model is mainly evaluated from the aspects of mean absolute error, mean square error, root mean square error, and determination coefficient.

[0051] The present invention also provides a method for evaluating bone healing status of tibial shaft fractures based on mRUST, comprising the following steps S201-S202: Step S201: Acquire an X-ray image of the tibial shaft to be evaluated.

[0052] Step S202: inputting the tibial shaft X-ray image into the tibial shaft fracture bone healing state assessment model trained based on the mRUST-based tibial shaft fracture bone healing state assessment model training method to generate a tibial shaft fracture bone healing state score value.

[0053] The present invention also provides a mRUST-based tibial shaft fracture bone healing state assessment system, which can implement the mRUST-based tibial shaft fracture bone healing state assessment method when executed. The system includes: The data processing module is used to obtain the tibial shaft X-ray image to be evaluated and perform preprocessing on the tibial shaft X-ray image, such as cropping, image enhancement and other operations.

[0054] The status scoring module includes a tibial shaft fracture bone healing status assessment model trained based on the mRUST-based tibial shaft fracture bone healing status assessment model training method, and is used to automatically assess the tibial shaft fracture bone healing status based on tibial shaft X-ray images.

[0055] The evaluation output module is used to output the bone healing status score value of the tibial shaft fracture.

[0056] Corresponding to the above method, the present invention also provides an electronic device, which includes a computer device, the computer device includes a processor and a memory, the memory stores computer instructions, the processor is used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the electronic device implements the steps of the method described above.

[0057] The embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the aforementioned edge computing server deployment method are implemented. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0058] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0059] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0060] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0061] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A training method for evaluating the bone healing status of tibial shaft fractures based on mRUST, characterized in that: The method comprises the following steps: Constructing a training sample set, wherein the training sample set includes a plurality of samples, each sample includes a tibial shaft X-ray image; annotating a true fracture area and a true score value for each tibial shaft X-ray image; Constructing an evaluation model, the evaluation model includes a detection module and a scoring module, the scoring module removes the Softmax function, and the fully connected layer directly outputs the result; the evaluation model takes the tibial shaft X-ray image as input, the detection module detects the tibial shaft X-ray image to obtain a predicted fracture area, the scoring module scores the bone healing state of the predicted fracture area to obtain a predicted score value, and the predicted score value is used as the output of the evaluation model; The detection module is trained using the training sample set. After the training is completed, the parameters of the detection module are fixed, and then the scoring module is trained until the preset performance is achieved to obtain a tibial shaft fracture bone healing state assessment model.

2. The mRUST-based tibial shaft fracture bone healing status assessment model training method according to claim 1, characterized in that: Before constructing the training sample set, the tibial shaft X-ray image is preprocessed, including: Normalizing the tibial shaft X-ray image to scale pixel values ​​to a range of 0 to 1; increasing the contrast of the tibial shaft X-ray image using adaptive histogram equalization; The data enhancement technology is used to randomly transform the tibial shaft X-ray image to generate multiple different versions of the image.

3. The mRUST-based tibial shaft fracture bone healing status assessment model training method according to claim 1, characterized in that: Construct an evaluation model, which includes a detection module and a scoring module, and also includes: The detection module adopts the YOLOv9 model and CSPDarkNet as the backbone network. Through programmable gradient information technology, the gradient propagation path, direction and size are adaptively adjusted according to actual needs when calculating the gradient, so as to optimize the gradient transfer and improve the detection accuracy.

4. The mRUST-based tibial shaft fracture bone healing status assessment model training method according to claim 1, characterized in that: The detection module detects the tibial shaft X-ray image to obtain a predicted fracture area, including: Inputting the tibial shaft X-ray image into the detection module; The detection module extracts the features of the tibial shaft X-ray image through a convolutional network to generate a feature map; generates an anchor frame on the feature map, and predicts boundary parameters for each anchor frame to obtain the predicted fracture area.

5. The mRUST-based tibial shaft fracture bone healing status assessment model training method according to claim 1, characterized in that: Before the scoring module scores the bone healing state of the predicted fracture area, the method further includes: The pixels of the predicted fracture area are adjusted, and when the size becomes smaller after adjustment, black edges are filled to keep the size consistent.

6. The mRUST-based tibial shaft fracture bone healing status assessment model training method according to claim 1, characterized in that: When the detection module is trained using the training sample set, it includes: The intersection-and-union ratio of the predicted fracture area and the actual fracture area is calculated, the predicted fracture area is screened according to a preset intersection-and-union ratio threshold, and the detection module parameters are adjusted.

7. The mRUST-based tibial shaft fracture bone healing status assessment model training method according to claim 1, characterized in that: When training the scoring module, it includes: A mean square error loss between the predicted score value and the true score value is constructed, and the scoring module is optimized with the goal of minimizing the mean square error loss.

8. A method for evaluating bone healing status of tibial shaft fracture based on mRUST, characterized in that: The method comprises: Obtain an X-ray image of the tibial shaft to be evaluated; The tibial shaft X-ray image is input into the tibial shaft fracture bone healing state assessment model trained by the tibial shaft fracture bone healing state assessment model training method based on mRUST according to any one of claims 1 to 7 to generate a tibial shaft fracture bone healing state score value.

9. A tibial shaft fracture bone healing status assessment system based on mRUST, characterized in that: When the system is executed, it is possible to implement the method for evaluating bone healing status of tibial shaft fractures based on mRUST as claimed in claim 8, and the system comprises: A data processing module, used for acquiring an X-ray image of the tibia shaft to be evaluated, and preprocessing the X-ray image of the tibia shaft; A state scoring module, comprising a tibial shaft fracture bone healing state assessment model trained by the mRUST-based tibial shaft fracture bone healing state assessment model training method according to any one of claims 1 to 7, and used for automatically assessing the tibial shaft fracture bone healing state according to the tibial shaft X-ray image; The evaluation output module is used to output the bone healing status score value of the tibial shaft fracture.

10. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images