Costal cartilage evaluation method based on segmentation model and application of method

Abstract

The invention relates to a segmentation model-based costal cartilage assessment method and application thereof, and the method comprises the following steps: obtaining to-be-assessed costal cartilage 3D CT scanning image data, and carrying out the preprocessing, and obtaining the preprocessed image data; inputting the preprocessed image data into a pre-trained segmentation model for segmentation to obtain a segmentation result of each costal cartilage; and based on the segmentation result of each costal cartilage, evaluating by using a costal cartilage measurement model constructed based on a generalized additive model to obtain costal cartilage measurement indexes including the volume, length, cross sectional area, CT average value, calcification rate and calcification CT average value of each costal cartilage. Compared with the prior art, the method has the advantages of comprehensively and accurately evaluating various indexes of the costal cartilage and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a costal cartilage evaluation method based on a segmentation model and an application of the method. Background Art

[0002] Assessment of costal cartilage morphology and calcification is of great value in forensic age inference, orthopedic disease diagnosis, and thoracic surgery planning. Traditional assessment methods rely primarily on medical imaging technologies (such as X-rays, CT, and MRI) combined with manual interpretation. However, these methods are subject to high subjectivity, low efficiency, and poor consistency, making them difficult to meet the needs of precision medicine and large-scale screening. In recent years, with the rapid development of artificial intelligence technology, deep learning-based image segmentation models have demonstrated significant advantages in medical image analysis. However, automated assessment methods for this special anatomical structure, costal cartilage, still face many challenges.

[0003] There are many existing segmentation techniques in medical image segmentation methods, including:

[0004] Patent application CN114821052A discloses a method for segmenting 3D brain tumor MRI images based on a self-regulating strategy. This method uses a pixel-level segmentation network to extract shallow features to guide the training of a deeper network, enabling the deeper network to obtain more detailed information about the tumor's internal regions. Deep features obtained from an image-level classification network are then used to guide the training of the shallower network, preserving more semantic information about the tumor's core region. Patent application CN113344008A also discloses a deep learning-based method for high-throughput extraction of anatomical feature parameters from stem tissue. This method uses an encoder-decoder feature enhancer to extract high- and low-order feature maps and fuse them to obtain the strongest features.

[0005] With the development of precision medicine, the clinical demand for costal cartilage assessment has shifted from qualitative judgment to multidimensional quantitative analysis. However, there are currently few quantitative costal cartilage assessment indicators. Therefore, existing technologies cannot accurately quantitatively assess costal cartilage volume and calcification, making it difficult to provide effective data support for clinical diagnosis. Summary of the Invention

[0006] The purpose of the present invention is to provide a costal cartilage evaluation method based on a segmentation model and an application of the method for realizing digital measurement of costal cartilage indicators.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] A costal cartilage assessment method based on a segmentation model comprises the following steps:

[0009] Acquiring 3D CT scan image data of the costal cartilage to be evaluated, and performing preprocessing to obtain preprocessed image data;

[0010] Inputting the pre-processed image data into a pre-trained segmentation model for segmentation to obtain a segmentation result for each costal cartilage;

[0011] Based on the segmentation results of each costal cartilage, the costal cartilage measurement model constructed based on the generalized additive model is used for evaluation to obtain costal cartilage measurement indicators, including the volume, length, cross-sectional area, CT average value, calcification rate and CT average value of calcification of each costal cartilage.

[0012] Furthermore, the preprocessing operation includes cubic interpolation resampling to a preset resolution and data enhancement operations.

[0013] Furthermore, the segmentation result of each costal cartilage includes the segmented area, calcified area, non-calcified area, geometric shape and metadata of each costal cartilage, and the metadata includes the location and left and right side information of the costal cartilage.

[0014] Furthermore, the steps of constructing the costal cartilage measurement model include:

[0015] Each costal cartilage metric is assumed to obey the Box-Cox power index distribution, and a generalized additive model is constructed for each costal cartilage metric, wherein the expression of the generalized additive model is:

[0016] y~D(μ,σ,θ,τ)

[0017]

[0018] Where y is the result vector, that is, the index value, D is the density function, μ is the location, σ is the scale, θ is the skewness, τ is the kurtosis, g() is the link function, X1, X2, X3, and X4 are covariates, which are the fixed effect design matrices affecting the location parameter, scale parameter, skewness parameter, and kurtosis parameter of the costal cartilage volume, respectively; β1, β2, β3, and β4 are vectors of fixed effect parameters, J is the total number of smoothing points, and s is the jth nonparametric or nonlinear function. It is a penalized b-spline function to fit nonlinear influencing factors, including age and gender;

[0019] Iteratively fitting the generalized additive model using iterative weighted least squares method to obtain an initial costal cartilage measurement model for each costal cartilage measurement index;

[0020] For the initial costal cartilage measurement model of each costal cartilage measurement index, the GAIC value is calculated, and the model corresponding to the minimum GAIC value is selected as the final costal cartilage measurement model, wherein the calculation expression of the GAIC value is:

[0021] GAIC=GD+(k×df)

[0022] where GAIC is the GAIC value, GD is the global deviation, k is the penalty associated with each degree of freedom, and df is the total effective degrees of freedom.

[0023] Furthermore, the steps of performing iterative fitting using iterative weighted least squares method include:

[0024] Initialization parameters: Preliminary estimation of parameters μ, σ, θ and τ;

[0025] Update the linear predictor: For each parameter (μ, σ, θ, τ), calculate the current residual and use penalized B-spline to estimate the nonlinear term, and repeat this step until the parameter change is less than the set threshold, completing the iterative fitting process.

[0026] Furthermore, the step of obtaining the costal cartilage measurement index includes:

[0027] Determine a covariate X for each costal cartilage according to the segmentation result of each costal cartilage, wherein the covariate includes age, gender, and costal cartilage position and left / right side information corresponding to the costal cartilage;

[0028] The covariate X is input into the constructed costal cartilage measurement model, the μ, σ, θ and τ parameters of the Box-Cox power index distribution are calculated, and various costal cartilage measurement indices are further calculated.

[0029] Furthermore, it also includes judging whether the goodness of fit of the costal cartilage measurement model meets the preset conditions, and the specific steps include:

[0030] The q-normal plot and worm plot are used to determine whether the skewness and kurtosis of the residual of the costal cartilage measurement model are within the ideal range. If not, the costal cartilage measurement model is reconstructed. If so, the Filliben correlation coefficient of the costal cartilage measurement model is further determined to be greater than a preset threshold.

[0031] If yes, it is considered that the costal cartilage measurement model meets the preset conditions.

[0032] If not, it is considered that the costal cartilage measurement model meets the preset conditions and the costal cartilage measurement model needs to be reconstructed.

[0033] Furthermore, the method further comprises the following steps:

[0034] The growth rate of the digitized costal cartilage metric with age is calculated, and the calculation expression of the growth rate is:

[0035]

[0036] Where, Growth rate k is the growth rate of the indicator at age k, y(Age k+1 )、y(Age k ) is the predicted value of the GAMLSS fitting indicator at age k+1 and k.

[0037] Furthermore, the segmentation model is trained using a 3D TransUNet17 network model, and the segmentation model includes a CNN-based encoder, a Transformer-based decoder, and a mask classifier. The segmentation steps of the segmentation model include:

[0038] Inputting the preprocessed image data into the CNN-based encoder to extract multi-scale local features;

[0039] Inputting the multi-scale features into the Transformer-based decoder and utilizing an attention mechanism to capture global features, wherein each query in the attention mechanism represents a segment of rib cartilage and the query interacts with the features over multiple optimization cycles;

[0040] Based on the global features, a mask classifier is used to perform instance-level segmentation to obtain a segmentation result.

[0041] The present invention also provides an application of the costal cartilage evaluation method based on the segmentation model described above, wherein the method is used to generate a costal cartilage detection report, wherein the costal cartilage detection report includes the name, gender, date, and various costal cartilage measurement indicators of the evaluation subject.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] (1) The present invention utilizes the segmentation performance of the segmentation model in artificial intelligence technology to segment each costal cartilage in the 3D CT scan image of the costal cartilage, and on this basis uses the costal cartilage measurement model constructed based on the generalized additive model to evaluate each indicator, obtain the measurement results of each indicator, and realize the digital measurement of costal cartilage indicators.

[0044] (2) The present invention utilizes the fitting properties based on the generalized additive model to fit various indicators, and uses the iterative weighted least squares method to perform iterative fitting in the costal cartilage measurement model, determines the most appropriate distribution of each indicator based on the GAIC value, and judges the goodness of fit, etc., so as to construct an accurate costal cartilage measurement model and obtain accurate measurement results.

[0045] (3) The present invention can provide reference for doctors and patients through costal cartilage assessment before surgery and provide effective data support for clinical diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Schematic diagram of the method flow of the present invention;

[0047] Figure 2 Schematic diagrams of the costal cartilage of the present invention are shown in Figure 3D, wherein (a) is a 3D model of the costal cartilage of a 35-year-old female, (b) is a 3D model of the costal cartilage of a 34-year-old male, (c) is a 3D model of the costal cartilage of a 33-year-old male, and (d) is a 3D model of the costal cartilage of a 28-year-old male.

[0048] Figure 3 This is the costal cartilage detection report of the present invention. DETAILED DESCRIPTION

[0049] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0050] Example 1

[0051] This embodiment provides a costal cartilage assessment method based on a segmentation model. Figure 1 As shown, the method includes the following steps:

[0052] S1. Build a segmentation model based on the costal cartilage 3D CT scan image dataset.

[0053] 1. Image dataset:

[0054] This implementation collected a large number of healthy and unhealthy rib cartilage 3D CT scan images of different genders and ages as the training set and test set of the model, such as Figure 2 As shown, there are calcification areas of different locations and sizes on the unhealthy costal cartilage.

[0055] 2. Segmentation model structure:

[0056] The 3D TransUNet network structure has proven highly successful in medical image segmentation tasks. Therefore, this example trains the segmentation model based on the settings of 3D TransUNet17. This network structure reformulates the traditional pixel-by-pixel classification problem as a region-based mask classification task. This design helps maintain the anatomical continuity of the costal cartilage structure, a key consideration for subsequent surgical planning applications. The network structure includes the following:

[0057] CNN-based encoder: The CNN encoder extracts local features from the 3D CT image, focusing on fine details such as the edges and boundaries of the costal cartilage. The encoder gradually reduces the image resolution while learning increasingly abstract representations of the cartilage structure. This step captures fine anatomical details, which are critical for accurately predicting cartilage boundaries.

[0058] Transformer-based decoder: Leveraging the power of attention mechanisms, the decoder employs a Transformer mechanism to model global spatial relationships across the entire 3D CT volume. Unlike traditional approaches that process each pixel independently, the model in this embodiment starts with a set of learnable queries—each query potentially representing a segment of costal cartilage. These queries interact with image features through multiple optimization cycles, enabling the model to understand the complex relationship between cartilage and surrounding tissue. This approach allows the segmentation model to produce coherent, anatomically plausible segments that maintain structural integrity, which is challenging when segmenting costal cartilage but clinically important.

[0059] Mask classification: Instead of classifying individual pixels, our approach performs instance-level segmentation to generate cartilage segments and then associates them with specific costal cartilage levels. During training, the model learns to align the predicted segments with their correct anatomical roles using Hungarian matching optimization. The algorithm ensures the consistency of the anatomical mapping by finding the best assignment between the prediction and the ground truth. During inference, the model directly predicts the anatomical label for each generated mask, producing the final segmentation for further clinical use. The segmentation result includes the segmented area, calcified area, non-calcified area, geometry and metadata of each costal cartilage. The metadata includes the costal cartilage location and left-right side information, e.g., the left costal cartilage in case 6.

[0060] 3. Training and validation of segmentation model:

[0061] To accommodate CT scans from different sources, this embodiment provides a standardized preprocessing pipeline. The number or number of CT scan image examinations is resampled to a resolution of 2.22mm×1.45mm×1.45mm using cubic interpolation to standardize the input while preserving anatomical details. During training, this embodiment uses patch-based processing (96×160×128 pixels) with an overlap of 50% to handle memory constraints while maintaining spatial context. Data augmentation techniques, including random rotation (±15°), are used to enhance the generalizability of the model. The model is trained using a composite loss function that combines cross entropy and dice loss, and is balanced to address the class imbalance problem inherent in medical image segmentation. Training is performed using the Adam optimizer with an initial learning rate of 0.08, and model selection is performed using moving average optimization.

[0062] To evaluate the performance of the segmentation model described above in a controlled setting, this implementation conducted experiments on an internal test set. As described in the Dataset section, this test set consists of 1,203 samples randomly selected from leading centers. This implementation compared the segmentation model with several state-of-the-art segmentation networks, including nnUNet18, Swin UNETR19, and CoTr20. Evaluation metrics included DSC, Intersection over Union (IoU), F1 score, precision, and recall. Furthermore, this implementation assessed the robustness of the model by analyzing its performance on scans with varying levels of costal cartilage visibility. The internal test set evaluation aimed to establish a baseline performance benchmark and identify potential areas for model improvement. To validate the generalizability of the segmentation model, this implementation collected an external multi-center test set consisting of 438 patients. The data was annotated following the same guidelines as the internal dataset. This implementation evaluated the model's performance using the same metrics used in internal validation, focusing on its ability to maintain high accuracy across different imaging conditions and patient demographics. External validation aimed to demonstrate the model's robustness and clinical applicability in real-world settings, ensuring its reliability for wider deployment.

[0063] S2. Use the costal cartilage measurement model based on the generalized additive model to evaluate and obtain the costal cartilage measurement index.

[0064] 1. Constructing a costal cartilage measurement model:

[0065] This example uses the GAMLSS and GAMLSS2 software packages for canonical modeling using the generalized additive models of location, shape, and scale (GAMLSS) framework. Available in R v2024.04.2 (Build 764), this fitting method has been applied to modeling height, weight, brain volume, and related growth indices with high accuracy and reliability.

[0066] GAMLSS is used to simulate the volume and calcification of costal cartilage. This method has been used by the World Health Organization to develop child growth standards and consists of four parameters: location (μ), scale (σ), skewness (θ), and kurtosis (τ). Specifically, the definition of GAMLSS is as follows:

[0067] y~D(μ,σ,θ,τ)

[0068]

[0069] In the above formula, y is the outcome vector, that is, the indicator value, D is the general density function, g() is the link function of the model applied to each component, X is the fixed-effect design matrix, time delay is the fixed-effect parameter vector, parameters are the cartilage volume location parameter, scale parameter, skewness parameter, and kurtosis parameter, J is the total number of smoothing points, and s is the jth nonparametric or nonlinear function.

[0070] The GAMLSS framework in our study uses the RS algorithm (iterative weighted least squares) to fit the initial costal cartilage metric model for each costal cartilage metric. This algorithm does not include the expected value of the cross derivative and has been shown to be more appropriate. The specific iterative fitting steps include:

[0071] Initialization parameters: Preliminary estimation of parameters μ, σ, θ and τ;

[0072] Update the linear predictor: For each parameter (μ, σ, θ, τ), calculate the current residual and use penalized B-spline to estimate the nonlinear term, and repeat this step until the parameter change is less than the set threshold, completing the iterative fitting process.

[0073] In addition, this embodiment uses the generalized Akaike Information Criterion (GAIC) to determine the most appropriate distribution for each indicator, and selects the model corresponding to the minimum GAIC value as the final costal cartilage measurement model. The definition of GAIC is as follows:

[0074] GAIC=GD+(k×df)

[0075] In the above formula, GD represents the global deviance. In addition, df represents the total effective degrees of freedom used in the model, and k represents the penalty associated with each degree of freedom.

[0076] In order to test whether the above-mentioned fitted costal cartilage measurement model meets the preset conditions, the goodness of fit is also used for judgment:

[0077] The q-normal plot and worm plot are used to determine whether the skewness and kurtosis of the residual of the costal cartilage measurement model are within the ideal range. If not, the costal cartilage measurement model is reconstructed. If so, the Filliben correlation coefficient of the costal cartilage measurement model is further determined to be greater than a preset threshold.

[0078] If yes, it is considered that the costal cartilage measurement model meets the preset conditions.

[0079] If not, it is considered that the costal cartilage measurement model meets the preset conditions and the costal cartilage measurement model needs to be reconstructed.

[0080] 2. Costal cartilage measurement model evaluation:

[0081] Based on the costal cartilage measurement model constructed above, we begin to measure various costal cartilage indicators. The specific measurement steps include:

[0082] Determine a covariate X for each costal cartilage according to the segmentation result of each costal cartilage, wherein the covariate includes age, gender, and costal cartilage position and left / right side information corresponding to the costal cartilage;

[0083] The covariate X is input into the constructed costal cartilage measurement model, the μ, σ, θ and τ parameters of the Box-Cox power index (BCPE) distribution are calculated, and various costal cartilage measurement indices are further calculated.

[0084] Each evaluation indicator can be measured through the above measurement process to obtain the corresponding indicator value, including the volume, length, cross-sectional area, CT average value, calcification rate, and CT average value of calcification of each costal cartilage. In this example, the costal cartilage of a patient was evaluated and the indicator measurement results in Table 1 were obtained as follows:

[0085] Table 1 Index measurement results

[0086]

[0087] This embodiment also fits the growth rate of the digital measurement index of the costal cartilage with age to predict the future index value, which is also provided for reference by doctors and patients. The calculation expression of the growth rate is:

[0088]

[0089] Where, Growth rate k is the growth rate of the indicator at age k, y(Age k+1 )、y(Age k ) is the predicted value of the GAMLSS fitting indicator at age k+1 and k.

[0090] In addition, a comparison of costal cartilage indices was performed:

[0091] Continuous data were expressed as mean ± standard deviation (SD) or median (interquartile range, IQR), and Student's t-test or Mann-Whitney U test were used for consistency assessment when appropriate. Categorical data were expressed as numbers and percentages and analyzed using chi-square test or Fisher's exact test when appropriate. All analyses were performed using SPSS 25.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism 8 (GraphPad Software, San Diego, California, USA). A two-tailed P value < 0.05 was considered statistically significant.

[0092] Example 2

[0093] This embodiment provides an application of a costal cartilage assessment method based on a segmentation model, which is used to generate a costal cartilage detection report according to the method of embodiment 1. The costal cartilage detection report includes the name, gender, date, and costal cartilage measurement indicators of the assessment subject. The generated costal cartilage detection report is as follows: Figure 3 shown.

[0094] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0095] It will be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0096] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0097] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0099] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0100] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A costal cartilage assessment method based on a segmentation model, characterized in that: The following steps are involved: Acquiring 3D CT scan image data of the costal cartilage to be evaluated, and performing preprocessing to obtain preprocessed image data; Inputting the pre-processed image data into a pre-trained segmentation model for segmentation to obtain a segmentation result for each costal cartilage; Based on the segmentation results of each costal cartilage, the costal cartilage measurement model constructed based on the generalized additive model is used for evaluation to obtain costal cartilage measurement indicators, including the volume, length, cross-sectional area, CT average value, calcification rate and CT average value of calcification of each costal cartilage.

2. The costal cartilage assessment method based on a segmentation model according to claim 1, characterized in that: The preprocessing operation includes cubic interpolation resampling to a preset resolution and data enhancement operation.

3. The costal cartilage assessment method based on a segmentation model according to claim 1, characterized in that: The segmentation result of each costal cartilage includes the segmented area, calcified area, non-calcified area, geometric shape and metadata of each costal cartilage, and the metadata includes the location and left and right side information of the costal cartilage.

4. The costal cartilage assessment method based on a segmentation model according to claim 1, characterized in that: The steps of constructing the costal cartilage measurement model include: Each costal cartilage metric is assumed to obey the Box-Cox power index distribution, and a generalized additive model is constructed for each costal cartilage metric, wherein the expression of the generalized additive model is: y~D(μ,σ,θ,τ) Where y is the result vector, that is, the index value, D is the density function, μ is the location, σ is the scale, θ is the skewness, τ is the kurtosis, g() is the link function, X1, X2, X3, and X4 are covariates, which are the fixed effect design matrices affecting the location parameter, scale parameter, skewness parameter, and kurtosis parameter of the costal cartilage volume, respectively; β1, β2, β3, and β4 are vectors of fixed effect parameters, J is the total number of smoothing points, and s is the jth nonparametric or nonlinear function. It is a penalized b-spline function to fit nonlinear influencing factors, including age and gender; Iteratively fitting the generalized additive model using iterative weighted least squares method to obtain an initial costal cartilage measurement model for each costal cartilage measurement index; For the initial costal cartilage measurement model of each costal cartilage measurement index, the GAIC value is calculated, and the model corresponding to the minimum GAIC value is selected as the final costal cartilage measurement model, wherein the calculation expression of the GAIC value is: GAIC=GD+(k×df) where GAIC is the GAIC value, GD is the global deviation, k is the penalty associated with each degree of freedom, and df is the total effective degrees of freedom.

5. The costal cartilage assessment method based on a segmentation model according to claim 3, characterized in that: The steps of iterative fitting using iterative weighted least squares method include: Initialization parameters: Preliminary estimation of parameters μ, σ, θ and τ; Update the linear predictor: For each parameter (μ, σ, θ, τ), calculate the current residual and use penalized B-spline to estimate the nonlinear term, and repeat this step until the parameter change is less than the set threshold, completing the iterative fitting process.

6. The costal cartilage assessment method based on a segmentation model according to claim 3, characterized in that: The step of obtaining the costal cartilage measurement index comprises: Determine a covariate X for each costal cartilage according to the segmentation result of each costal cartilage, wherein the covariate includes age, gender, and costal cartilage position and left / right side information corresponding to the costal cartilage; The covariate X is input into the constructed costal cartilage measurement model, the μ, σ, θ and τ parameters of the Box-Cox power index distribution are calculated, and further various costal cartilage measurement indices are calculated.

7. The costal cartilage assessment method based on a segmentation model according to claim 1, characterized in that: It also includes determining whether the goodness of fit of the costal cartilage measurement model meets preset conditions, and the specific steps include: The q-normal plot and worm plot are used to determine whether the skewness and kurtosis of the residual of the costal cartilage measurement model are within the ideal range. If not, the costal cartilage measurement model is reconstructed. If so, the Filliben correlation coefficient of the costal cartilage measurement model is further determined to be greater than a preset threshold. If yes, it is considered that the costal cartilage measurement model meets the preset conditions. If not, it is considered that the costal cartilage measurement model meets the preset conditions and the costal cartilage measurement model needs to be reconstructed.

8. The costal cartilage assessment method based on a segmentation model according to claim 1, characterized in that: The following steps are also included: The growth rate of the digitized costal cartilage metric with age is calculated, and the calculation expression of the growth rate is: Where, Growth rate k is the growth rate of the indicator at age k, y(Age k+1 )、y(Age k ) is the predicted value of the GAMLSS fitting indicator at age k+1 and k.

9. The costal cartilage assessment method based on a segmentation model according to claim 1, characterized in that: The segmentation model is trained using a 3D TransUNet17 network model. The segmentation model includes a CNN-based encoder, a Transformer-based decoder, and a mask classifier. The segmentation steps of the segmentation model include: Inputting the preprocessed image data into the CNN-based encoder to extract multi-scale local features; Inputting the multi-scale features into the Transformer-based decoder and utilizing an attention mechanism to capture global features, wherein each query in the attention mechanism represents a segment of rib cartilage and the query interacts with the features over multiple optimization cycles; Based on the global features, a mask classifier is used to perform instance-level segmentation to obtain a segmentation result.

10. An application of the costal cartilage assessment method based on the segmentation model according to any one of claims 1 to 9, characterized in that: The method is used to generate a costal cartilage detection report, which includes the name, gender, date, and various costal cartilage measurement indicators of the evaluation subject.

Images