Vertebral body segmentation method based on frequency-enhanced multi-granularity context

Abstract

The invention discloses a frequency-enhanced multi-granularity context-based vertebral body segmentation method, and belongs to the fields of deep learning, computer vision and image segmentation. Comprising the following steps: firstly, collecting and dividing data, obtaining image data of CT and MRI of a vertebral body in a three-dimensional imaging mode, and dividing the image data into a training set and a test set; secondly, preprocessing, training and post-processing the image data; thirdly, training a network model, and training the multi-granularity context segmentation model based on frequency enhancement on the training set to obtain a segmentation model; and finally, testing the segmentation model effect. According to the method, the information loss is avoided by introducing the sampling operation under the wavelet transform, the low-frequency and high-frequency features are respectively strengthened, the accuracy and detail representation of the segmentation result can be improved, and important technical support is provided for precise medical treatment. According to the method, multi-granularity context information can be fused more effectively, image details are reserved, the segmentation precision is improved, and the method is suitable for various clinical application scenes.

Description

Technical Field

[0001] The present invention belongs to the fields of deep learning, computer vision, and image segmentation, and relates to a spinal vertebrae segmentation method based on frequency-enhanced multi-granularity context. Background Art

[0002] Medical image segmentation plays a vital role in modern healthcare. Leveraging advances in imaging technology and deep learning, it enables efficient and accurate analysis of complex medical images. Accurately identifying and extracting lesions not only improves diagnostic accuracy, supports personalized treatment plans, but also promotes medical research and education. High-quality segmentation results facilitate real-time monitoring of disease progression and adjustment of treatment strategies, thereby improving patient prognosis and quality of life. Therefore, medical image segmentation is not only a critical bridge connecting imaging data with clinical applications but also a key driver of the development of precision medicine.

[0003] In orthopedics, the spine serves as the body's second lifeline, supporting the normal functioning of various organs and closely connecting to the entire body through its central nervous system. The development of modern digital orthopedics relies on accurate segmentation of 3D medical imaging data, which is crucial in clinical applications such as disease diagnosis, preoperative planning, and surgical navigation. Existing vertebral segmentation methods typically employ an encoder-decoder architecture and have achieved remarkable results. However, these approaches often overlook the critical impact of downsampling techniques. Max pooling and bilinear interpolation are commonly used downsampling methods, but they often result in a significant loss of fine-grained details during the downsampling process. This loss of detail is particularly detrimental in blurry images, as preserving fine-grained details is crucial for accurate segmentation. Furthermore, vertebral segmentation requires sufficient contextual information to distinguish between different vertebrae. Existing convolutional neural networks (CNNs) rely on local convolution kernels to extract features, which limits their ability to model global and geometric features. While Transformer-based methods excel at capturing long-range dependencies, their self-attention mechanisms carry high computational complexity. Recently, state-space models (SSMs) have been proposed to capture long-range dependencies with linear complexity, providing a promising solution to the limitations of CNNs and Transformers. However, SSMs lack consideration of spatially varying context, which reduces their sensitivity to multi-scale contextual information of similar vertebrae and thus weakens their ability to distinguish vertebrae.

[0004] Therefore, developing an efficient and accurate vertebral segmentation method that can comprehensively address these challenges remains an urgent problem and is of great significance for improving medical quality and efficiency. Summary of the Invention

[0005] In response to the shortcomings of the above-mentioned existing methods, the present invention proposes a spinal vertebrae segmentation method based on frequency-enhanced multi-granularity context, introduces wavelet transform downsampling to avoid information loss, and enhances low-frequency and high-frequency features respectively, thereby improving the performance of spinal vertebrae segmentation.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for segmenting spinal vertebrae based on frequency-enhanced multi-granularity context, the method comprising the following steps:

[0008] Step 1: Collect and divide data, obtain vertebral CT and MRI image data in 3D imaging mode, and divide the image data into training and test sets;

[0009] Step 2: Standardized process construction for image data preprocessing, training and post-processing:

[0010] The image data preprocessing, training, and post-processing processes are fully integrated into the nnU-Net framework, establishing an efficient and standardized training process. This ensures that every step from image data preparation to final model evaluation meets best practice standards. Specifically:

[0011] In step 2.1, during the data preprocessing phase, the image data obtained in step 1 is first standardized, including grayscale normalization and resampling. Subsequently, the image is resized by cropping or padding to fit the network input size requirements.

[0012] In step 2.2, during training, we leverage the nnU-Net framework's automatic configuration capabilities to dynamically adjust hyperparameters, such as the learning rate, batch size, and loss function type, to adapt to the characteristics of different datasets. Furthermore, we introduce data augmentation operations during training, such as random rotation, scaling, elastic deformation, and intensity perturbations, to increase data diversity and further improve the model's robustness and generalization capabilities.

[0013] Step 2.3: In the post-processing stage, the prediction results are optimized to reduce false positive areas and boundary burrs. Specific operations include removing isolated points based on connected domain analysis and smoothing the boundaries using morphological closing operations.

[0014] Step 3: Train the network model. Train the frequency-enhanced multi-granularity context module (FMCM) on the training set, observe the changes in the loss function value during training, and terminate training when the loss value converges to obtain the trained segmentation model. This segmentation model uses a wavelet transform downsampling module (WTD) and a wavelet transform upsampling module (WTU) to build an encoder-decoder structure. It integrates a high-frequency feature refinement module (HFR) and a multi-granularity state space model (MG-SSM) to enhance the ability to capture features of different granularities. Specifically:

[0015] Step 3.1: In the encoding stage, the features extracted by the encoder are first decomposed into high-frequency components and low-frequency components through the wavelet transform-based downsampling module WTD. The high-frequency feature enhancement module HFR and the multi-granularity state space model MG-SSM are then applied to generate enhanced feature representations. These enhanced features are then fused to generate the final features, effectively capturing semantic and detail information. Specifically:

[0016] In step 3.1.1, the wavelet-based downsampling module WTD uses discrete wavelet transform (DWT) to process the features extracted by the encoder. This operation can effectively alleviate the irreversible distortion problem that occurs in the traditional downsampling process and reduce information loss.

[0017] Define the encoder The features extracted in each stage are , which is expressed as:

[0018] (1)

[0019] in, is the low frequency component; is the high frequency component. In the discrete wavelet transform (DWT), by Apply a low-pass filter to each axis ( ) and high-pass filter ( ), these 8 sub-band components are obtained, including a low-frequency component ( and 7 high-frequency components ( , the resolution of these components is half of the input feature resolution in the current encoding stage, achieving lossless downsampling.

[0020] In step 3.1.2, low-frequency components contain most of the basic information of the image, while high-frequency components can capture complex details, textures, and edge information. Therefore, feature enhancement processing is performed on both low-frequency components and high-frequency components at the same time:

[0021] (2)

[0022] (3)

[0023] in, and Represent the enhanced low-frequency features and high-frequency features respectively, is the multi-granularity state space model MG-SSM, It is the high-frequency feature enhancement module HFR;

[0024] In step 3.1.2.1, during the enhancement of low-frequency components, the multi-granularity state space model MG-SSM first uses the residual block to process the features extracted by the encoder to capture the local detail information contained therein. Then, three parallel 3D depth-separable convolutions with different expansion rates are used to capture multi-scale depth features with different receptive fields. These multi-scale deep features are input into three visual state space modules (VSSMs) of different scales to learn multi-granularity context information. The whole process is as follows:

[0025] (4)

[0026] (5)

[0027] in, Represent multi-scale deep features; Represents the expansion rate, including three different expansion rates; represents depth-wise separable convolution; represents the residual block; Represents the low-frequency component of the input; Represents the features of the first branch of the visual state space module VSSM; represents the spatial state model, Represents SILU activation function; 1 Represents a convolution layer with a convolution kernel of 1; represents a linear layer; Representation layer normalization operation. This process addresses the limitations of a single visual state space module (VSSM) in capturing spatial variation features and provides multi-granularity contextual information to distinguish similar vertebrae.

[0028] Then, the extracted multi-scale depth features Fusion is performed and input into the second branch of VSSM to obtain the features of the second branch :

[0029] (6)

[0030] in, represents the features of the second branch, which are shared among the three VSSMs; Represents a join operation.

[0031] The features of the two branches of VSSM are combined using the Hadamard product to obtain the final output:

[0032] (7)

[0033] in, The expansion rate is The output of VSSM; represents the Hadamard product.

[0034] The output of each VSSM is obtained by formula (7): , and finally fused to obtain enhanced low-frequency features .

[0035] In step 3.1.2.2, during the high-frequency component enhancement process, the high-frequency feature enhancement module (HFR) consists of two paths: a key feature amplification branch and a noise filtering branch. Both paths utilize a spatial attention mechanism to dynamically modulate different high-frequency components during training.

[0036] In the key feature amplification branch, the spatial attention mechanism Defined as:

[0037] (8)

[0038] (9)

[0039] in, Represents the sigmoid operation, is the characteristic of the high-frequency component after amplification; The operation calculates the contribution of each high-frequency component to the spatial attention map; Represents the grouped maximum pooling operation, which is used to extract key features from each component; The operation aggregates and amplifies the key features of each high-frequency component; yes The convolutional layer is used to convert The feature dimension after the operation; Represents an element-wise multiplication operation.

[0040] The structure of the denoising path of the noise filtering branch is the same as the key feature amplification path, except that average pooling is used instead of maximum pooling:

[0041] (10)

[0042] (11)

[0043] in, Represents the output of the spatial attention mechanism; Represented as the characteristics of high-frequency components after denoising; Represents a group average pooling operation.

[0044] Step 3.1.2.3, spatial attention map Multiply each high frequency component to highlight key features, while A similar operation is performed to filter out noise from each high-frequency component. Finally, the enhanced and denoised high-frequency components are concatenated to obtain the enhanced high-frequency features. .

[0045] (12)

[0046] Step 3.1.3, the enhanced version of step 3.1.2 and Fusion is performed to obtain features rich in semantic information and detail information :

[0047] (13)

[0048] In step 3.2, during the decoding stage, the wavelet transform-based upsampling module WTU is used to perform feature fusion and resolution improvement, making full use of the information extracted by the encoder to achieve high-quality vertebrae segmentation.

[0049] Specifically, the wavelet transform-based upsampling module WTU performs upsampling in the decoding stage while effectively utilizing features of different scales extracted by the encoder:

[0050] (14)

[0051] (15)

[0052] in, Indicates that it comes from The characteristics of the encoder in the stage, It represents the first The characteristics of the decoder at each stage; Indicates Inverse Discrete Wavelet Transform (IDWT); , Represents the features after discrete wavelet transform DWT decomposition; Indicates the The characteristics of the decoder at this stage.

[0053] During the entire process of step 3, the encoder stage first applies the wavelet transform-based downsampling module (WTD) to achieve lossless downsampling. Next, feature enhancement modules, including the high-frequency feature enhancement module (HFR) and the multi-granularity state space model (MG-SSM), are used to further improve feature representation. Finally, in the decoder stage, the wavelet transform-based upsampling module (WTU) is used to restore the image resolution, resulting in the final segmentation model.

[0054] Step 4: Test the segmentation model effect by inputting the CT or MR image data in the test set into the trained segmentation model to generate segmented images and calculate relevant engineering indicators;

[0055] The engineering metrics include: using the Dice Similarity Coefficient (DSC) to evaluate the region-level similarity between the segmentation results obtained by the segmentation model and the true label, and using the 95% Hausdorff Distance (HD95) to measure boundary accuracy.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] (1) Through the high-frequency feature enhancement module (HFR) and the multi-granularity state-space model (MG-SSM), this method effectively utilizes contextual information at different granularities while preserving details, resulting in detailed and accurate segmentation results. This significantly improves the accuracy and reliability of vertebral segmentation. This method not only helps improve diagnostic accuracy and supports the development of personalized plans, but also optimizes surgical planning and postoperative evaluation processes, significantly improving patient treatment outcomes and recovery progress.

[0058] (2) The present invention introduces wavelet transform downsampling to avoid information loss and enhance low-frequency and high-frequency features respectively, thereby effectively solving the problems caused by image blur and high similarity of vertebral structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of the overall structure of the present invention.

[0060] Figure 2 It is a schematic diagram of the multi-granularity state space model of the present invention.

[0061] Figure 3 Schematic diagram of the high-frequency feature enhancement module of the present invention.

[0062] Figure 4 It is a schematic diagram of an upsampling module based on wavelet transform of the present invention.

[0063] Figure 5 This is a visualization graph on the Verse 2019 dataset.

[0064] Figure 6 It is a visualization diagram on the LUMBAR dataset. DETAILED DESCRIPTION

[0065] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0066] The present invention is specifically implemented as follows:

[0067] Step 1: Collect and divide data, obtain vertebral CT and MRI image data in 3D imaging mode, and divide the image data into training and test sets;

[0068] In this embodiment, it is first necessary to obtain medical imaging data for spinal vertebral segmentation. In order to ensure the diversity and breadth of the data, the present invention mainly uses public authoritative data sets, including the CT data set Verse 2019 and the MRI data set. These data sets provide rich multimodal data support for model training and verification. The Verse 2019 data set covers the vertebral segmentation of the cervical, thoracic and lumbar vertebrae, providing comprehensive coverage of the spinal region; while the LUMBAR data set focuses on the detailed segmentation of the lumbar spine, enhancing the professionalism and meticulousness of the data set in specific areas. The combination of the two not only increases the integrity and diversity of the data, but also ensures that the model can be effectively learned and verified under different anatomical structures and imaging conditions. The collected data needs to be divided into training sets and test sets for effective model training and evaluation.

[0069] Step 2: Standardized process construction for image data preprocessing, training and post-processing:

[0070] The image data preprocessing, training, and post-processing processes are fully integrated into the nnU-Net framework, establishing an efficient and standardized training process. This ensures that every step from image data preparation to final model evaluation meets best practice standards. Specifically:

[0071] In step 2.1, during the data preprocessing phase, the image data obtained in step 1 is first normalized, including grayscale normalization and resampling. Subsequently, the image is resized by cropping to fit the network input size requirements.

[0072] In step 2.2, during training, we leverage the nnU-Net framework's automatic configuration capabilities to dynamically adjust hyperparameters, including the learning rate, batch size, and loss function type, to adapt to the characteristics of different datasets. Furthermore, we introduce data augmentation operations during training, including random rotations, scaling, elastic deformations, and intensity perturbations, to increase data diversity and further improve the model's robustness and generalization capabilities.

[0073] Step 2.3: In the post-processing stage, the prediction results are optimized to reduce false positive areas and boundary burrs. Specific operations include removing isolated points based on connected domain analysis and smoothing boundaries using morphological closing operations.

[0074] Step 3: Train the network model.

[0075] After completing step 2, the training set is used to train the frequency-enhanced multi-granularity context-efficient vertebral segmentation model FMCM. This model constructs an encoder-decoder structure using a wavelet transform-based downsampling module WTD and a wavelet transform-based upsampling module WTU. It integrates a high-frequency feature enhancement module HFR and a multi-granularity state space model MG-SSM to enhance the ability to capture features of different granularities. Specifically:

[0076] Step 3.1: In the encoding stage, the features extracted by the encoder are first decomposed into high-frequency components and low-frequency components through the wavelet transform-based downsampling module WTD. The high-frequency feature enhancement module HFR and the multi-granularity state space model MG-SSM are then applied to generate enhanced feature representations. These enhanced features are then fused to generate the final features, effectively capturing semantic and detail information. Specifically:

[0077] In step 3.1.1, the wavelet-based downsampling module WTD uses discrete wavelet transform (DWT) to process the features extracted by the encoder. This operation can effectively alleviate the irreversible distortion problem that occurs in the traditional downsampling process and reduce information loss.

[0078] Define the encoder The features extracted in each stage are , which is expressed as:

[0079] (1)

[0080] in, is the low frequency component; is the high frequency component. In the discrete wavelet transform (DWT), by Apply a low-pass filter to each axis ( ) and high-pass filter ( ), these 8 sub-band components are obtained, including a low-frequency component ( and 7 high-frequency components ( , the resolution of these components is half of the input feature resolution in the current encoding stage, achieving lossless downsampling.

[0081] In step 3.1.2, low-frequency components contain most of the basic information of the image, while high-frequency components can capture complex details, textures, and edge information. Therefore, feature enhancement processing is performed on both low-frequency components and high-frequency components at the same time:

[0082] (2)

[0083] (3)

[0084] in, and Represent the enhanced low-frequency features and high-frequency features respectively, is the multi-granularity state space model MG-SSM, It is the high-frequency feature enhancement module HFR;

[0085] In step 3.1.2.1, during the enhancement of low-frequency components, the multi-granularity state space model MG-SSM first uses the residual block to process the features extracted by the encoder to capture the local detail information contained therein. Then, three parallel 3D depth-separable convolutions with different expansion rates are used to capture multi-scale depth features with different receptive fields. These multi-scale deep features are input into three visual state space modules (VSSMs) of different scales to learn multi-granularity context information. The whole process is as follows:

[0086] (4)

[0087] (5)

[0088] in, Represent multi-scale deep features; Represents the expansion rate, including three different expansion rates; represents depth-wise separable convolution; represents the residual block; Represents the low-frequency component of the input; Represents the features of the first branch of the visual state space module VSSM; represents the spatial state model, Represents SILU activation function; 1 Represents a convolution layer with a convolution kernel of 1; represents a linear layer; Representation layer normalization operation. This process addresses the limitations of a single visual state space module (VSSM) in capturing spatial variation features and provides multi-granularity contextual information to distinguish similar vertebrae.

[0089] Then, the extracted multi-scale depth features Fusion is performed and input into the second branch of VSSM to obtain the features of the second branch :

[0090] (6)

[0091] in, represents the features of the second branch, which are shared among the three VSSMs; Represents a join operation.

[0092] The features of the two branches of VSSM are combined using the Hadamard product to obtain the final output:

[0093] (7)

[0094] in, The expansion rate is The output of VSSM; represents the Hadamard product.

[0095] The output of each VSSM is obtained by formula (7): , and finally fused to obtain enhanced low-frequency features .

[0096] In step 3.1.2.2, during the high-frequency component enhancement process, the high-frequency feature enhancement module (HFR) consists of two paths: a key feature amplification branch and a noise filtering branch. Both paths utilize a spatial attention mechanism to dynamically modulate different high-frequency components during training.

[0097] In the key feature amplification branch, the spatial attention mechanism Defined as:

[0098] (8)

[0099] (9)

[0100] in, Represents the sigmoid operation, is the characteristic of the high-frequency component after amplification; The operation calculates the contribution of each high-frequency component to the spatial attention map; Represents the grouped maximum pooling operation, which is used to extract key features from each component; The operation aggregates and amplifies the key features of each high-frequency component; yes The convolutional layer is used to convert The feature dimension after the operation; Represents an element-wise multiplication operation.

[0101] The structure of the denoising path of the noise filtering branch is the same as the key feature amplification path, except that average pooling is used instead of maximum pooling:

[0102] (10)

[0103] (11)

[0104] in, Represents the output of the spatial attention mechanism; Represented as the characteristics of high-frequency components after denoising; Represents a group average pooling operation.

[0105] Step 3.1.2.3, spatial attention map Multiply each high frequency component to highlight key features, while A similar operation is performed to filter out noise from each high-frequency component. Finally, the enhanced and denoised high-frequency components are concatenated to obtain the enhanced high-frequency features. .

[0106] (12)

[0107] Step 3.1.3, the enhanced version of step 3.1.2 and Fusion is performed to obtain features rich in semantic information and detail information :

[0108] (13)

[0109] In step 3.2, during the decoding stage, the wavelet transform-based upsampling module WTU is used to perform feature fusion and resolution improvement, making full use of the information extracted by the encoder to achieve high-quality vertebrae segmentation.

[0110] Specifically, the wavelet transform-based upsampling module WTU performs upsampling in the decoding stage while effectively utilizing features of different scales extracted by the encoder:

[0111] (14)

[0112] (15)

[0113] in, Indicates that it comes from The characteristics of the encoder in the stage, It represents the first The characteristics of the decoder at each stage; Indicates Inverse Discrete Wavelet Transform (IDWT); , Represents the features after discrete wavelet transform DWT decomposition; Indicates the The characteristics of the decoder at this stage.

[0114] During the entire process of step 3, the encoder stage first applies the wavelet transform-based downsampling module (WTD) to achieve lossless downsampling. Next, feature enhancement modules, including the high-frequency feature enhancement module (HFR) and the multi-granularity state space model (MG-SSM), are used to further improve feature representation. Finally, in the decoder stage, the wavelet transform-based upsampling module (WTU) is used to restore the image resolution, resulting in the final segmentation model.

[0115] During this process, the network model continuously optimizes to minimize the loss function, thereby improving the quality of the segmentation results. As training progresses, it is crucial to monitor the changes in the loss function in real time. The loss value should gradually decrease, indicating that the model is learning and optimizing its parameters. By observing the convergence of the loss function, we can determine whether the training has achieved the desired results. When the loss value stabilizes after multiple iterations, the model has converged. At this point, the training process is considered to have reached an optimal state and can be terminated. After training, the optimal model weights are saved to facilitate inference and testing in subsequent steps.

[0116] Step 4: Test the model effect.

[0117] After model training is complete, the vertebral segmentation model is evaluated using the CT and MRI image data in the test set. The specific steps are as follows: (1) Loading saved weights: Load the best model weights saved during training. (2) Inputting test data: Input the CT and MRI images in the test set into the trained model to generate vertebral segmentation results. (3) Evaluating the segmentation effect: By calculating indicators such as the Dice similarity coefficient (DSC) and the 95% Hausdorff distance (HD95), the difference between the segmentation results output by the model and the true labels is compared to evaluate the model performance and compare with other methods for verification.

[0118] The current model achieved a DSC of 79.92% and an HD95 performance index of 3.04 on the Verse dataset, and a DSC of 79.13% and an HD95 of 17.13 on the LUMBAR dataset. Figure 5 and Figure 6 The visual segmentation results of the proposed model FMCM on the Verse2019 dataset and the LUMBAR dataset are shown respectively, and compared with other methods (such as nnU-Net, UMamba, MambaClnix, Segmamba and FMCM). Figure 5 The following figure shows the segmentation results of the thoracic, cervical, and lumbar regions in the Verse2019 dataset, from top to bottom. It can be seen that in areas with similar vertebrae, FMCM can more accurately distinguish adjacent vertebrae, demonstrating superior segmentation capabilities compared to other methods. Figure 6 The results show that FMCM has advantages in segmentation of the lumbar spine on the LUMBAR dataset. The results show that the model can achieve more accurate vertebral segmentation than existing methods.

[0119] The above-described embodiments merely express the implementation methods of the present invention, but should not be understood as limiting the scope of the present invention. It should be pointed out that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, which all fall within the scope of protection of the present invention.

Claims

1. A method for vertebral segmentation based on frequency-enhanced multi-granularity context, characterized in that: The spinal vertebra segmentation method comprises the following steps: Step 1: Collect and divide data, obtain vertebral CT and MRI image data in 3D imaging mode, and divide the image data into training and test sets; Step 2: Preprocess, train and post-process the image data; Step 3: Train the network model. Train the frequency-enhanced multi-granularity contextual segmentation model (FMCM) on the training set, observe the changes in the loss function value during training, and terminate the training when the loss value converges to obtain a trained segmentation model. This segmentation model constructs an encoder-decoder structure by using a wavelet transform-based downsampling module (WTD) and a wavelet transform-based upsampling module (WTU). It integrates a high-frequency feature enhancement module (HFR) and a multi-granularity state space model (MG-SSM) to enhance the ability to capture features of different granularities. Step 4: Test the effectiveness of the segmentation model. Input the CT and MR image data in the test set into the trained segmentation model, generate segmented images, and calculate relevant engineering indicators.

2. The method for segmenting spinal vertebrae based on frequency-enhanced multi-granularity context according to claim 1, characterized in that: In step 2, the image data preprocessing, training, and post-processing processes are integrated into the nnU-Net framework. Specifically: Step 2.1: In the preprocessing stage, the image data obtained in step 1 is first standardized, including grayscale normalization and resampling operations; then, the image size is adjusted by cropping or padding to make it fit the network input size requirements; Step 2.2: During the training process, the automatic configuration function of the nnU-Net framework is used to dynamically adjust the hyperparameters to adapt to the characteristics of different datasets. At the same time, data augmentation operations are introduced during training to increase data diversity. Step 2.3: In the post-processing stage, the prediction results are optimized to reduce false positive areas and boundary burrs.

3. The method for segmenting spinal vertebrae based on frequency-enhanced multi-granularity context according to claim 1, characterized in that: The step three is specifically as follows: Step 3.1: In the encoding stage, the features extracted by the encoder are first decomposed into high-frequency components and low-frequency components through the wavelet transform-based downsampling module WTD. The high-frequency feature enhancement module HFR and the multi-granularity state space model MG-SSM are applied respectively to generate enhanced feature representations. These enhanced features are then fused to generate the final features, effectively capturing semantic and detail information. In step 3.2, during the decoding phase, the wavelet transform-based upsampling module (WTU) is used to perform feature fusion and resolution enhancement, making full use of the information extracted by the encoder to achieve high-quality vertebrae segmentation and obtain the final segmentation model.

4. The method for segmenting spinal vertebrae based on frequency-enhanced multi-granularity context according to claim 3, characterized in that: The step 3.1 is specifically as follows: Step 3.1.1, the wavelet-based downsampling module WTD uses discrete wavelet transform DWT to process the features extracted by the encoder; define the encoder first The features extracted in each stage are , which is expressed as: （1）， in, is the low frequency component; is the high frequency component; In discrete wavelet transform DWT, by Apply low-pass filter and high-pass filter to each axis of the image to obtain 8 sub-band components, including one low-frequency component and 7 high-frequency components, to achieve lossless downsampling; Step 3.1.2: Perform feature enhancement on both low-frequency and high-frequency components to obtain enhanced and : （2）， （3）， in, and Represent the enhanced low-frequency features and high-frequency features respectively, is the multi-granularity state space model MG-SSM, It is the high-frequency feature enhancement module HFR; Step 3.1.3, the enhanced version of step 3.1.2 and Fusion is performed to obtain features rich in semantic information and detail information : （13）。 5. The method for segmenting spinal vertebrae based on frequency-enhanced multi-granularity context according to claim 4, characterized in that: The step 3.1.2 is specifically as follows: Step 3.1.2.1, in the process of enhancing low-frequency components, the multi-granularity state space model MG-SSM first uses the residual block to process the features extracted by the encoder; then, three parallel 3D depth-separable convolutions with different expansion rates are used to capture multi-scale depth features with different receptive fields. ; The multi-scale deep features are input into three visual state space modules (VSSM) of different scales to learn multi-granularity context information; the whole process is as follows: （4）， （5）， in, Represent multi-scale deep features; Represents the expansion rate, including three different expansion rates; represents depth-wise separable convolution; represents the residual block; Represents the low-frequency component of the input; Represents the features of the first branch of the visual state space module VSSM; represents the spatial state model, Represents SILU activation function; 1 Represents a convolution layer with a convolution kernel of 1; represents a linear layer; Representation layer normalization operation; The extracted multi-scale deep features Fusion is performed and input into the second branch of VSSM to obtain the features of the second branch : （6）， in, represents the features of the second branch, which are shared among the three VSSMs; Indicates a connection operation; The features of the two branches of VSSM are combined using the Hadamard product to obtain the final output: （7）， in, The expansion rate is The output of VSSM; represents the Hadamard product; The output of each VSSM is obtained by formula (7): , and finally fused to obtain enhanced low-frequency features ; In step 3.1.2.2, during the enhancement of high-frequency components, the high-frequency feature enhancement module (HFR) consists of two paths: a key feature amplification branch and a noise filtering branch. Both paths use a spatial attention mechanism to dynamically modulate different high-frequency components during training. In the key feature amplification branch, the spatial attention mechanism Defined as: （8）， （9）， in, Represents the sigmoid operation, is the characteristic of the high-frequency component after amplification; The operation calculates the contribution of each high-frequency component to the spatial attention map; Represents the grouped maximum pooling operation, which is used to extract key features from each component; The operation aggregates and amplifies the key features of each high-frequency component; yes The convolutional layer is used to convert The feature dimension after the operation; Represents element-wise multiplication operation; The structure of the denoising path of the noise filtering branch is the same as the key feature amplification path, except that average pooling is used instead of maximum pooling: （10）， （11）， in, Represents the output of the spatial attention mechanism; Represented as the characteristics of high-frequency components after denoising; Represents the group average pooling operation; Step 3.1.2.3, spatial attention mechanism is multiplied by each high frequency component to enhance it, and Filter out the noise from each high-frequency component to perform denoising; finally, concatenate the enhanced and denoised high-frequency components to obtain the enhanced high-frequency features ; （12）。 6. The method for segmenting spinal vertebrae based on frequency-enhanced multi-granularity context according to claim 5, characterized in that: The step 3.2 is specifically as follows: The wavelet transform-based upsampling module WTU performs upsampling in the decoding stage and effectively utilizes features of different scales extracted by the encoder: （14）， （15）， in, Indicates that it comes from The characteristics of the encoder in the stage, It represents the first The characteristics of the decoder at each stage; represents the inverse discrete wavelet transform; , Represents the features after discrete wavelet transform DWT decomposition; Indicates the The characteristics of the decoder at this stage.

7. The method for segmenting spinal vertebrae based on frequency-enhanced multi-granularity context according to claim 1, characterized in that: The engineering metrics include: evaluating the region-level similarity between the segmentation results obtained by the segmentation model and the true label through the Dice similarity coefficient, and measuring the boundary accuracy using the 95% Hausdorff distance.

Images