A glioma segmentation method based on XLSTM-HVED network

Abstract

The application discloses a brain neuroglioma missing mode segmentation method based on an XLSTM-HVED neural network, and comprises the following steps: step 1, acquiring a neuroglioma data set, and dividing the data set into a training set and a test set; step 2, constructing an XLSTM-HVED neural network model for neuroglioma segmentation, wherein the XLSTM-HVED neural network model comprises a SAVE encoder, a Vision XLSTM attention module, a segmentation decoder, a reconstruction decoder and a DuSFE module; step 3, training the XLSTM-HVED neural network model by using the training set, and optimizing parameters of the XLSTM-HVED neural network model in combination with characteristics of neuroglioma images; and step 4, testing and evaluating the obtained XLSTM-HVED network model by using the test set, and finally realizing brain neuroglioma missing mode segmentation.

Description

Technical Field

[0001] The present invention relates to the technical fields of computer vision deep learning and medical image processing, and specifically to a glioma segmentation method based on an XLSTM-HVED network. Background Art

[0002] Magnetic resonance imaging (MRI) is the gold standard for the diagnosis and management of diffuse gliomas, providing crucial information on tumor size, location, and temporal changes in morphology. Different MRI modalities, such as T1, T1-Gd, T2, and FLAIR, provide complementary information about brain tumors. T1 and T2 modalities provide information on vasogenic edema in subacute stroke; T1-Gd, enhanced by contrast agents, reveals information related to vasculature and the blood-brain barrier; and FLAIR provides general information about stroke lesions. Therefore, in clinical practice, using multimodal MRI for segmentation can help reduce uncertainty and improve segmentation performance compared to using single-modality MRI. However, in practice, missing modalities often occur. Due to limited patient tolerance, scan time constraints, and image corruption, some modalities may be lost. In such cases, the segmentation performance of traditional algorithms can be significantly reduced. Consequently, accurate MRI segmentation in the absence of modalities has attracted considerable attention in recent years, and its practical significance is self-evident. For example, Sharma et al. designed a multimodal generative adversarial network (MM-GAN) to synthesize missing MRI pulse sequences. Dorent et al. designed a heterogeneous modality variational encoder-decoder (HVED) for tumor segmentation and modality completion in the absence of multimodality. Chen et al. proposed a multimodal brain tumor segmentation framework based on feature decoupling and gated fusion, aiming to enhance robustness in the absence of multimodality. Liu et al. designed a multimodal masked autoencoder (M3AE) for brain tumor segmentation in the absence of multimodality. Zhang et al. proposed a multimodal medical Transformer model (mmFormer) for brain tumor segmentation in the absence of multimodality. However, these methods still have problems such as limited robustness to missing modalities, low efficiency of information fusion between modalities, and low segmentation accuracy. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the present invention proposes a glioma segmentation method based on the XLSTM-HVED network.

[0004] In order to solve the above technical problems, the technical solution of the present invention is:

[0005] A glioma segmentation method based on an XLSTM-HVED network comprises the following steps:

[0006] Step 1: Acquire a glioma dataset and divide the dataset into a training set and a test set; the dataset includes multimodal data, and the multimodal data includes at least one of four magnetic resonance imaging modalities: T1, T1c, T2, and FLAIR;

[0007] Step 2: Construct an XLSTM-HVED neural network model for glioma segmentation, which includes a SAVE encoder, a Vision XLSTM attention module, a segmentation decoder, a reconstruction decoder, and a DuSFE module.

[0008] Step 3: Use the training set to train the XLSTM-HVED neural network model and optimize the parameters of the XLSTM-HVED neural network model based on the characteristics of glioma images.

[0009] Step 4: Use the test set to test and evaluate the obtained XLSTM-HVED network model, and finally achieve the missing mode segmentation of brain glioma.

[0010] Preferably, the SAVE encoder adopts an MVAE encoder.

[0011] Preferably, the MVAE encoder comprises a DRB dimensionality reduction block.

[0012] Preferably, the Vision XLSTM attention module adopts the Vision-lstm module.

[0013] Preferably, the DuSFE module includes a cSFE module and an sSFE module.

[0014] Preferably, the SAVE encoder encodes the multimodal input through the MVAE encoder, then processes the features obtained from the encoding through a convolutional layer and spatial attention, and uses the input multimodal features to perform attention processing on the obtained features to obtain MVAE multimodal features; at the same time, the encoded MVAE multimodal features are decoded;

[0015] The MVAE encoder fuses the mean μ and covariance ∑ of each modality into the latent variable z using Gaussian product. The sampled latent variable z is decoded into the image space through a reparameterization method. The features of the obtained image space are optimized by the DRB dimensionality reduction block to obtain DRB multimodal features.

[0016] The multimodal data input is input of at least one of the four magnetic resonance imaging modalities (T1, T1c, T2 and FLAIR).

[0017] The latent variable z is obtained by fusion sampling of the mean μ and covariance ∑ of each mode through Gaussian product.

[0018] Finally, the MVAE and DRB multimodal features are added to the unimodal latent variable z through the downsampling layer to obtain the shape features (B, C, H, W, D) and input into the Vision XLSTM encoder.

[0019] Where B refers to the number of input data in a batch, C refers to the number of channels of the input data, H refers to the height of the data, W refers to the width of the data, and D refers to the depth of the data.

[0020] Preferably, in the Vision XLSTM attention module, the shape features (B, C, H, W, D) are first flattened to obtain features (B, H*W*D, C), and then the features (B, H*W*D, C) are input into the Vision-lstm module; the Vision-lstm module first divides the input feature map of shape (B, H*W*D, C) into non-overlapping H*W*D tags through shared linear projection; then, a learnable position embedding is added to each tag to obtain the ViL feature.

[0021] Preferably, the ViL feature and the DRB multimodal feature are connected to a segmentation decoder and a reconstruction decoder, which output modality F1 and modality F2 respectively, and then the modality F1 and modality F2 are input to the DuSFE module.

[0022] Preferably, the cSFE module first compresses the modal F1 and modal F2 into vectors V1 and V2 using global average pooling, and then connects them in series and passes through a fully connected layer to obtain a fusion vector V fuse ; Then use V fuse Vectors R1 and R2 are obtained through two fully connected layers respectively, and finally channel multiplication is used to apply vectors R1 and R2 back to the two input modes to generate F 1_cSFE and F 2_cSFE , the expression is as follows:

[0023]

[0024] V fuse =w[V1, V2]+b

[0025] F 1_cSFE =σ(W1V fuse +b1)⊙F1,F 2_sSFE =σ(W2V fuse +b2)⊙F1

[0026] where c refers to the cth channel of F1 and F2, σ(·) refers to the sigmoid function, ⊙ refers to the channel-wise multiplication, [·] refers to the channel-wise concatenation operator, b, b1, and b2 refer to the bias terms in the neural network, and w1 and w2 refer to the weight terms in the neural network.

[0027] Preferably, in the sSFE module, the modalities F1 and F2 are first processed by a convolutional layer, then concatenated, and then subjected to another convolution layer to generate M fuse ; Then, M fuse fed into two separate convolutional layers, which are applied to the input pattern output F via spatial multiplication 1_sSFE and F 2_sSFE , the expression is as follows:

[0028] M fuse =K fuse *[K in1 *F1, K in2 *F2]

[0029]

[0030] Where * refers to the convolution operator, refers to the spatial multiplication operator, K in1 , K in2 Refers to two convolution kernels used to process the convolution layers of modality F1 and modality F2 respectively, K out1 , K out2 Refers to two for processing M fuse The convolution kernels of two independent convolutional layers.

[0031] Preferably, the training method of the XLSTM-HVED neural network model is: the XLSTM-HVED neural network model is trained using a training set, combined with the characteristics of glioma images, implemented through the PyTorch framework, and trained using NVIDIA Tesla V100; during the training process, the batch size is set to 2, the learning rate is set to 0.0001, and the training duration is set to 72h.

[0032] Preferably, in step 4, two methods are used for evaluation: Dice score to measure the degree of overlap of regions and HD95 score to measure the degree of overlap of boundaries.

[0033] The present invention has the following characteristics and beneficial effects:

[0034] Using the above technical solution, the present invention integrates the multimodal features and multimodal coding features obtained by variational learning through Vision XLSTM, deeply fuses multimodal information and enhances the capture of temporal and spatial features. The present invention interactively upsamples information through a cross-perception module on the encoder of the multi-task missing modality reconstruction and segmentation task, and freezes part of the pre-trained reconstruction task encoder to enhance cross-modal feature fusion while improving the performance of multi-task learning. The present invention proposes a single-multimodal variational encoder that uses multimodal feature attention to perceive important single-modal features and ultimately fuses single-modal features through a variational encoder, thereby improving the downsampling capability of missing data and capturing the intrinsic structure and distribution characteristics of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 This is the overall structure diagram of the XLSTM-HVED neural network model of the present invention;

[0037] Figure 2 Schematic diagram of the Vision XLSTM attention module of the present invention;

[0038] Figure 3 Schematic diagram of the channel squeeze fusion excitation (cSFE) and spatial squeeze fusion excitation (sSFE) modules in the squeeze-fusion-excitation cross-perception module of the present invention. DETAILED DESCRIPTION

[0039] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0040] In the description of the present application, it needs to be understood that the orientation or positional relationship indicated by the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second" and the like are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined with "first", "second" and the like can be explicitly or implicitly included one or more. In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0041] In the description of the present application, it needs to be understood that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood through specific circumstances.

[0042] The present application provides a glioma segmentation method based on XLSTM-HVED network, as shown in Figure 1 The method comprises the following steps:

[0043] Step 1, obtaining a neuroglioma data set, and dividing the data set into a training set and a test set.

[0044] Specifically, in the present embodiment, a neuroglioma Brats2024 data set is established. The neuroglioma Brats2024 data set is a multi-modal brain tumor segmentation challenge (BraTS) 2024 data set. The data set contains about 4500 cases of T1, T1ce, T2 and FLAIR MRI mode imaging data of patients with various gliomas. The image data is resampled to isotropic 1mm 3 resolution, and is registered with an anatomical template, and then skull stripping preprocessing is performed. The label is manually marked by an expert, and the segmentation target is three regions of whole tumor, enhance tumor and tumor core. The data set is randomly divided into 80% training set and 20% test set.

[0045] Step 2: Construct an XLSTM-HVED neural network model for glioma segmentation. The XLSTM-HVED neural network model includes a SAVE encoder, a Vision XLSTM attention module, a segmentation decoder, a reconstruction decoder, and a squeeze-fusion-excitation cross-perception module (DuSFE module). The SAVE encoder includes an MVAE encoder. The Vision XLSTM attention module uses the Vision-LSTM module. The DuSFE module includes channel squeeze-fusion-excitation (cSFE) and spatial squeeze-fusion-excitation (sSFE).

[0046] Step 3: Use the training set to train the XLSTM-HVED neural network model, and optimize the parameters of the XLSTM-HVED neural network model based on the characteristics of glioma images.

[0047] Specifically, in this example, the XLSTM-HVED neural network model was trained using the training set, taking into account the characteristics of glioma images. This model was implemented using the PyTorch framework and trained on an NVIDIA Tesla V100. During training, the batch size was set to 2, the learning rate was set to 0.0001, and the training duration was set to 72 hours.

[0048] Step 4: Use the test set to test and evaluate the obtained XLSTM-HVED network model, and finally achieve the missing mode segmentation of brain glioma.

[0049] Specifically, to evaluate the model's image classification performance, we compared XLSTM-HVED with other classification models on the Brats2024 glioma dataset. We used two standard metrics to evaluate classification performance: Dice scores, which measure the degree of region overlap, and HD95 scores, which measure the degree of boundary overlap. Their formulas are as follows:

[0050]

[0051] Where |X∩Y| is the intersection between X and Y, |X| and |Y| represent the number of elements in X and Y, x and y are the elements in X and Y, d represents the distance, and HD95 is calculated based on the 95th percentile of the distance between the boundary points in X and Y.

[0052] Finally, the trained and evaluated XLSTM-HVED model was used to perform missing modality segmentation of brain glioma.

[0053] The specific method is as follows: first, the SAVE encoder encodes the multimodal data input through the MVAE encoder, processes the features obtained from the encoding through convolutional layers and spatial attention, and uses the input multimodal features to perform attention processing on the obtained features to obtain MVAE multimodal features; at the same time, the encoded MVAE multimodal features are decoded.

[0054] It should be noted that the MAVE encoder is in the condition-independent mode X=x1...x n Developed in

[15] , this module is based on variational autoencoders (VAEs) and handles missing data by expanding multimodal inputs.

[0055] The MVAE encoder includes a DRB dimension reduction block.

[0056] The MVAE encoder obtains the latent variable z by fusing the mean μ and covariance ∑ of each modality using Gaussian product sampling. The sampled latent variable z is decoded into the image space by a reparameterization method. The features of the obtained image space are optimized by the DRB dimensionality reduction block to obtain DRB multimodal features.

[0057] The multimodal input is input of at least one of the four magnetic resonance imaging modalities (T1, T1c, T2, and FLAIR).

[0058] Finally, the MVAE and DRB multimodal features are added to the unimodal latent variable z through the downsampling layer to obtain the shape features (B, C, H, W, D) and input into the Vision XLSTM encoder.

[0059] Where B refers to the number of input data in a batch, C refers to the number of channels of the input data, H refers to the height of the data, W refers to the width of the data, and D refers to the depth of the data.

[0060] Specifically, in this embodiment, the SAVE encoder is provided with four layers, each layer is connected in sequence, the output of the previous layer of SAVE encoder is used as the input of the next layer of SAVE encoder, and the output of the last layer of SAVE encoder is input into the Vision XLSTM attention module.

[0061] Further, such as Figure 2As shown in the figure, in the Vision XLSTM attention module, the shape features (B, C, H, W, D) are first flattened to obtain the features (B, H*W*D, C), and then the features (B, H*W*D, C) are input into the Vision-lstm module; the Vision-lstm module first divides the input feature map of shape (B, H*W*D, C) into non-overlapping H*W*D tags through shared linear projection; then, a learnable position embedding is added to each tag to obtain the ViL feature.

[0062] It can be understood that the core of the Vision-lstm module is the mLSTM block with matrix storage and covariance update rule.

[0063] Further configuration of this embodiment is as follows: Figure 3 As shown, the ViL feature and the DRB multimodal feature are connected to a segmentation decoder and a reconstruction decoder, and the segmentation decoder and the reconstruction decoder output modality F1 and modality F2 respectively, and then the modality F1 and modality F2 are input to the DuSFE module.

[0064] Specifically, the cSFE module first uses global average pooling to compress the modal F1 and modal F2 into vectors V1 and V2, and then concatenates them through the fully connected layer to obtain the fusion vector V fuse ; Then use V fuse Vectors R1 and R2 are obtained through two fully connected layers respectively, and finally channel multiplication is used to apply vectors R1 and R2 back to the two input modes to generate F 1_cSFE and F 2_cSFE , the expression is as follows:

[0065]

[0066] V fuse =w[V1, V2]+b

[0067] F 1_cSFE =σ(W1V fuse +b1)⊙F1,F 2_sSFE =σ(W2V fuse +b2)⊙F1

[0068] Here, c refers to the c-th channel of F1 and F2, σ(·) refers to the sigmoid function, ⊙ refers to the channel-wise multiplication, [·] refers to the channel-level concatenation operator, b, b1, and b2 refer to the bias terms in the neural network, and w1 and w2 refer to the weight terms in the neural network.

[0069] Specifically, in the sSFE module, the modalities F1 and F2 are first processed by a convolutional layer, then concatenated, and then subjected to another convolution layer to generate Mfuse ; Then, M fuse fed into two separate convolutional layers, which are applied to the input pattern output F via spatial multiplication 1_sSFE and F 2_sSFE , the expression is as follows:

[0070] M fuse =K fuse *[K in1 *F1, K in2 *F2]

[0071]

[0072] Where * refers to the convolution operator, refers to the spatial multiplication operator, K in1 , K in2 Refers to two convolution kernels used to process the convolution layers of modality F1 and modality F2 respectively, K out1 , K out2 Refers to two for processing M fuse The convolution kernels of two independent convolutional layers.

[0073] Specifically, in this embodiment, the DuSFE module is provided with two layers, and three groups of segmentation decoders and reconstruction decoders are provided. A group of segmentation decoders and reconstruction decoders are respectively provided at the front end of the first-layer DuSFE module, between the two-layer DuSFE modules, and at the output end of the second-layer DuSFE module. The output of the first group of segmentation decoders and reconstruction decoders serves as the input of the first-layer DuSFE module, the output of the first-layer DuSFE module serves as the input of the second group of segmentation decoders and reconstruction decoders, the output of the second group of segmentation decoders and reconstruction decoders serves as the output of the second-layer DuSFE module, and the output of the second-layer DuSFE module serves as the input of the third group of segmentation decoders and reconstruction decoders. The output of the third group of segmentation decoders and reconstruction decoders is discriminated by the discriminator, and finally the classification result is output by generating loss.

[0074] In addition, the DRB features input to the first group of segmentation decoders and reconstruction decoders come from the output of the DRB dimensionality reduction block in the third-layer SAVE encoder; the DRB features input to the second group of segmentation decoders and reconstruction decoders come from the output of the DRB dimensionality reduction block in the second-layer SAVE encoder; and the DRB features input to the third group of segmentation decoders and reconstruction decoders come from the output of the DRB dimensionality reduction block in the first-layer SAVE encoder.

[0075] It can be understood that in the above technical solution, the squeeze-fusion-excitation cross-perception module inputs the ViL features into the segmentation and reconstruction decoders, while the downsampled SAVE features are connected to the decoding process. During upsampling, the 8-dimensional and 16-dimensional features from the two decoders interact through the DuSFE module. The final feature tensors are obtained through the respective modules for segmentation and reconstruction. In pre-training, only the reconstruction decoder is trained, and the 2nd, 3rd and last modules are frozen. The DuSFE module integrates segmentation and reconstruction functions using channel squeeze fusion excitation (cSFE) and spatial squeeze fusion excitation (sSFE) for spatial recalibration.

[0076] Comparative Example:

[0077] The RA-HVED and RMBTs models were used as baselines for evaluation, and the model designed by the present invention was used to evaluate their performance. Both the baseline and the model designed by the present invention used the same backbone network for training and testing during the evaluation phase. The present invention selected U-HVED as the backbone network. Notably, the model of the present invention performed well across all evaluation metrics and was particularly effective in multimodal fusion scenarios.

[0078] In addition, the present invention compares the XLSTM-HVED model with several baseline models and state-of-the-art models. Compared with other advanced models such as mmFormer, the HD95 index of XLSTM-HVED performs well, especially in the case of multimodal fusion.

[0079] Ablation experiments analyzed the impact of the SAVE, ViLA, and SFECA modules on model performance. The results showed that removing ViLA significantly degraded segmentation performance in the T1 modality, with both the HD95 and Dice metrics worsening. Removing SAVE also reduced the overall performance of multimodal fusion, particularly when segmenting the entire tumor (WT). Removing the SFECA module significantly reduced segmentation performance in both T1c and T1 modalities, particularly with the HD95 metric worsening. This demonstrates that these three modules are crucial for improving model accuracy.

[0080] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. It will be apparent to those skilled in the art that various changes, modifications, substitutions, and variations of these embodiments, including components, without departing from the principles and spirit of the present invention are still within the scope of protection of the present invention.

Claims

1. A glioma segmentation method based on XLSTM-HVED network, characterized in that: The steps include: Step 1: Acquire a glioma dataset and divide the dataset into a training set and a test set; the dataset includes multimodal data, and the multimodal data includes at least one of four magnetic resonance imaging modalities: T1, T1c, T2, and FLAIR; Step 2: Construct an XLSTM-HVED neural network model for glioma segmentation, which includes a SAVE encoder, a Vision XLSTM attention module, a segmentation decoder, a reconstruction decoder, and a DuSFE module. The DuSFE module includes a cSFE module and an sSFE module; The SAVE encoder encodes the multimodal input through the MVAE encoder, then processes the features obtained from the encoding through convolutional layers and spatial attention, and uses the input multimodal data to perform attention processing on the obtained features to obtain MVAE multimodal features; at the same time, the encoded MVAE multimodal features are decoded; The MVAE encoder obtains a latent variable z by fusing the mean μ and covariance ∑ of each modality using Gaussian product sampling, and the sampled latent variable z is decoded into the image space by a reparameterization method to obtain image space features, and the image space features are optimized by the DRB dimensionality reduction block to obtain DRB multimodal features; Finally, the MVAE and DRB multimodal features are added to the unimodal latent variable z through the downsampling layer to obtain the shape features (B, C, H, W, D) and input into the Vision XLSTM encoder, where B refers to the number of input data in a batch, C refers to the number of channels of the input data, H refers to the height of the data, W refers to the width of the data, and D refers to the depth of the data; Step 3: Use the training set to train the XLSTM-HVED neural network model and optimize the parameters of the XLSTM-HVED neural network model based on the characteristics of glioma images. Step 4: Use the test set to test and evaluate the obtained XLSTM-HVED network model, and finally achieve the missing mode segmentation of brain glioma.

2. The glioma segmentation method based on the XLSTM-HVED network according to claim 1, characterized in that: The SAVE encoder adopts an MVAE encoder, and the MVAE encoder includes a DRB dimension reduction block.

3. The glioma segmentation method based on the XLSTM-HVED network according to claim 1, characterized in that: The Vision XLSTM attention module adopts the Vision-lstm module.

4. The glioma segmentation method based on the XLSTM-HVED network according to claim 3, characterized in that: In the Vision XLSTM attention module, the shape features (B, C, H, W, D) are first flattened to obtain the features (B, H*W*D, C), and then the features (B, H*W*D, C) are input into the Vision-lstm module; the Vision-lstm module first divides the input feature map of shape (B, H*W*D, C) into non-overlapping H*W*D tokens through shared linear projection; then, a learnable position embedding is added to each token to obtain the ViL feature.

5. The glioma segmentation method based on the XLSTM-HVED network according to claim 4, characterized in that: The ViL features and DRB multimodal features are connected to a segmentation decoder and a reconstruction decoder, which output modality F1 and modality F2 respectively, and then the modality F1 and modality F2 are input to the DuSFE module.

6. The glioma segmentation method based on the XLSTM-HVED network according to claim 5, characterized in that: The cSFE module first uses global average pooling to compress the modal F1 and modal F2 into vectors V1 and V2, and then connects them in series and passes through the fully connected layer to obtain the fusion vector V fuse ; Then use V fuse Vectors R1 and R2 are obtained through two fully connected layers respectively, and finally channel multiplication is used to apply vectors R1 and R2 back to the two input modes to generate F 1_cSFE and F 2_cSFE , the expression is as follows: V fuse =w[V1,V2]+b F 1_cSFE =σ(W1V fuse +b1)⊙F1,F 2_sSFE =σ(W2V fuse +b2)⊙F1 where c refers to the cth channel of F1 and F2, σ(·) refers to the sigmoid function, ⊙ refers to the channel-wise multiplication, [·] refers to the channel-wise concatenation operator, b, b1, and b2 refer to the bias terms in the neural network, and w1 and w2 refer to the weight terms in the neural network.

7. The glioma segmentation method based on the XLSTM-HVED network according to claim 5, characterized in that: In the sSFE module, modality F1 and modality F2 are first processed by a convolutional layer, then concatenated, and then subjected to another convolution layer to generate M fuse ; Then, M fuse fed into two separate convolutional layers, which are applied to the input pattern output F via spatial multiplication 1_sSFE and F 2_sSFE , the expression is as follows: M fuse =K fuse *[K in1 *F1,K in2 *F2] Where * refers to the convolution operator, refers to the spatial multiplication operator, K in1 , K in2 Refers to two convolution kernels used to process the convolution layers of modality F1 and modality F2 respectively, K out1 , K out2 Refers to two for processing M fuse The convolution kernels of two independent convolutional layers.

8. The glioma segmentation method based on the XLSTM-HVED network according to claim 1, characterized in that: The training method of the XLSTM-HVED neural network model is as follows: the XLSTM-HVED neural network model is trained using a training set, combined with the characteristics of glioma images, implemented using the PyTorch framework, and trained using an NVIDIA Tesla V100; during the training process, the batch size is set to 2, the learning rate is set to 0.0001, and the training duration is set to 72 hours.

9. The glioma segmentation method based on the XLSTM-HVED network according to claim 1, characterized in that: In step 4, two methods are used for evaluation: the Dice score is used to measure the degree of overlap of regions, and the HD95 score is used to measure the degree of overlap of boundaries.

Images