Eye socket lymphoma automatic typing model construction method, device, equipment and medium

Abstract

The first aspect of the invention is to provide a construction method of an orbital lymphoma automatic typing model. The second aspect of the invention is to provide an orbital lymphoma automatic typing device, which is characterized by comprising: an MRI data acquisition and preprocessing module; a texture feature extraction module; a spatial feature extraction module; the invention relates to an orbital lymphoma typing model. The third aspect of the invention is to provide an electronic device. The fourth aspect of the invention is to provide a computer readable storage medium, an executable instruction is stored on the computer readable storage medium, and when the instruction is executed by the processor, the processor executes the device. According to the method, the MALT lymphoma and the non-MALT lymphoma of the orbit are typed by using the MRI data, so that a reference basis is provided for the selection of a subsequent treatment scheme.

Description

Technical Field

[0001] The present invention relates to a method, device, equipment and medium for constructing an automatic classification model of orbital lymphoma based on MRI data, belonging to the technical fields of medical imaging and computer image automatic recognition. Background Art

[0002] Orbital lymphoma is the most common malignant tumor in the adult orbital region, presenting with various symptoms such as exophthalmos, globe displacement, swelling, edema, pain, ptosis, vision changes, limited eye movement and diplopia, etc. The most common histological subtype of orbital lymphoma is extranodal marginal zone B-cell lymphoma of mucosa-associated lymphoid tissue (i.e., MALT lymphoma), followed by non-MALT lymphoma (including diffuse large B-cell lymphoma, mantle cell lymphoma, follicular lymphoma, etc.). Generally, non-MALT lymphoma results in a worse prognosis and requires more aggressive treatment methods. Precise subtype identification and staging are key considerations in the treatment of orbital lymphoma. The differentiation between MALT and non-MALT lymphoma is difficult due to similar symptoms and the lack of unique diagnostic features.

[0003] The histological result of biopsy is the gold standard for differentiating different subtypes of orbital lymphoma. However, biopsy is an invasive procedure. If the tumor is close to the orbital apex or around the optic nerve, biopsy will be very difficult and there are surgical risks, such as eye movement disorders, vision decline, etc. Therefore, there is an urgent need for an alternative non-invasive diagnostic confirmation method to assist pathologists and ophthalmologists in making diagnoses. Magnetic resonance imaging (MRI) is a non-invasive examination method that can be used to evaluate the location, morphology, internal structure and characteristic manifestations of orbital lymphoma.

[0004] Currently, there are mainly two methods for diagnosing orbital lesions using MRI: First, radiologists summarize the MRI image characteristics of different types of lesions based on experience. For example, "spherical indentation" in the image is more likely to appear in invasive lymphoma subtypes, etc. Second, radiomics is used to extract texture features from orbital MRI images to build a model for differentiating benign tumors and malignant tumors. The above methods have been applied in the classification of other orbital diseases, but not in the classification of orbital MALT lymphoma and orbital non-MALT lymphoma. The main reason is that the MRI images of each subtype of orbital lymphoma are relatively close in terms of gray scale, texture, morphology and position, and it is difficult to classify them with a single feature. Summary of the Invention

[0005] The technical problem to be solved by the present invention is as follows: General radiomics technology models by extracting various texture features from orbital MRI images to distinguish different lesions. However, the MRI images of various subtypes of orbital lymphoma are relatively close in terms of gray scale and texture, making it difficult to perform high-precision classification solely using texture features.

[0006] To solve the above technical problem, the first aspect of the present invention provides a method for constructing an automatic classification model for orbital lymphoma, which is characterized by including the following steps:

[0007] Step 1: Obtain the orbital MRI data of the patient and perform annotation and preprocessing on the obtained data;

[0008] Step 2: Extraction of spatial features:

[0009] Use a schematic diagram of the orbital spatial relationship to represent the spatial position relationship of each structure within the orbit. In the figure, for the spatial features of orbital lymphoma, the centroid is used to represent each structure, and the midpoint connecting the two eyeballs is introduced as a reference. Consider: the relative position relationship between normal structures, the relative positions between lymphoma lesions and normal structures in different samples, the relationship from the reference point to the eyeballs and lymphoma lesions, and quantify the position features of lymphoma by measuring the distances between lymphoma lesions and other structures, thereby obtaining spatial feature information;

[0010] Step 3: Extract the texture features of the data obtained in Step 1;

[0011] Step 4: Training of the classification model:

[0012] Establish a classification model, randomly divide the data set processed in Step 2 and Step 3 into a training set and a test set, and before classification, average and normalize the features extracted from different sequences;

[0013] Perform classification training on the spatial features and texture features of lymphoma respectively, then fuse the two features for classification training, establish a hybrid model, and finally output the prediction probabilities of positive and negative samples. The class with a larger prediction probability is judged as the final classification.

[0014] Preferably, Step 1 includes the following steps:

[0015] Step 101: Obtain the orbital MRI data of the patient in DICOM format;

[0016] Step 102: Perform annotation on the MRI data obtained in the previous step, annotating the bilateral eyeballs, bilateral extraocular rectus muscles (usually including the medial rectus muscle, lateral rectus muscle, superior rectus muscle, and inferior rectus muscle), bilateral optic nerves, and the lesion range;

[0017] Step 103: Crop the blank edges of the picture;

[0018] Step 104: Normalize the data obtained in the previous step according to the mean and standard deviation of the MRI signal intensity;

[0019] Step 105: Crop and resample the normalized MRI data to generate the total data set.

[0020] Preferably, in Step 2, the graph Laplacian method is applied to extract the spatial feature information of each structure. Then, extracting the spatial features based on the graph Laplacian operator includes the following steps:

[0021] Construct an undirected graph G(V, E, A, W), where:

[0022] V = {v1, v2, v3, …, v N} is N nodes, and each node v n represents the spatial position of each tissue center;

[0023] E = {e1, e2, e3, …, e M} is M edges, and each edge is an unordered pair of any two nodes in the undirected graph;

[0024] A is the adjacency matrix: If there is an edge from node v i to node v j , then the element A ij in the i-th row and j-th column of the adjacency matrix A is 1; If there is no edge connecting node v i and node v j , then A ij = 0;

[0025] W is the weight matrix, and the element W ij in the i-th row and j-th column of the weight matrix W is the Euclidean distance between node v i and node v j ;

[0026] Calculate the Laplacian matrix L = D - W, where D is the degree matrix, D = diag(d1, d2, …, d n , …, d N ), and d n is the degree of node v n ;

[0027] Extract the eigenvalues representing the spatial features from the Laplacian matrix L.

[0028] Preferably, the texture classification features include angular second moment, contrast, correlation, and entropy, where:

[0029] The angular second moment is equal to the sum of the squares of the elements in the co-occurrence matrix, measures the gray-level change of the MRI texture, and reflects the uniformity of the gray-level distribution and the roughness of the texture;

[0030] The contrast is the moment of inertia near the diagonal of the gray-level co-occurrence matrix, which reflects the sharpness of the MRI and the depth of the texture grooves;

[0031] The correlation reflects the similarity of the elements of the spatial gray-level co-occurrence matrix in the row or column direction, which reflects the local gray-level correlation of the MRI;

[0032] The entropy reflects the randomness of the MRI texture.

[0033] Preferably, in step 4, the XGBoost method is used to establish the classification model.

[0034] Preferably, in step 4, when establishing the training set and the test set, the 5-fold cross-validation method is used to reduce the risk of overfitting.

[0035] The second aspect of the present invention is to provide an automatic classification device for orbital lymphoma, which is characterized by including:

[0036] An MRI data acquisition and preprocessing module, configured to obtain the orbital MRI data of a patient and preprocess the obtained data;

[0037] A texture feature extraction module, configured to extract the lymphoma texture features of the data output by the MRI data acquisition and preprocessing module;

[0038] A spatial feature extraction module, configured to extract the lymphoma spatial features of the data output by the MRI data acquisition and preprocessing module;

[0039] An orbital lymphoma classification model, configured to automatically classify orbital MALT or non-MALT based on the lymphoma spatial features and the lymphoma texture features.

[0040] Preferably, the MRI data acquisition and preprocessing module further includes an MRI data acquisition unit and an MRI data preprocessing unit, wherein:

[0041] The MRI data acquisition unit is configured to acquire the orbital MRI data of a patient in the DICOM format;

[0042] The MRI data preprocessing unit is configured to preprocess the orbital MRI data of a patient obtained by the MRI data acquisition unit, including:

[0043] Cropping the blank edges of the picture;

[0044] Normalizing the MRI data according to the average value and the standard deviation of the MRI signal intensity;

[0045] Cropping and resampling the normalized MRI data.

[0046] A third aspect of the present invention provides an electronic device, comprising:

[0047] One or more processors;

[0048] A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to execute the above-mentioned device.

[0049] A fourth aspect of the present invention provides a computer-readable storage medium having executable instructions stored thereon, which when executed by a processor cause the processor to execute the above-mentioned device.

[0050] The present invention uses MRI data to classify orbital MALT lymphoma and orbital non-MALT lymphoma, providing a reference basis for the selection of subsequent treatment plans. Compared with the prior art solutions, it has the following beneficial effects:

[0051] (1) The position of orbital lymphoma in the orbit is variable and difficult to accurately describe. The present invention uses the mass points of each structure to represent the positions of the normal orbital structure and the lesion structure, and quantitatively extracts the spatial position features of the orbital lymphoma lesion using the graph Laplacian algorithm, providing important parameters for the training of the classification model. In addition, the present invention is expected to be used for the extraction of the spatial position features of other space-occupying lesions in the orbit.

[0052] (2) There is a lot of irrelevant information in orbital MRI images. The present invention first preprocesses the MRI images to remove the surrounding blank parts, saving storage space and reducing training time.

[0053] (3) The present invention combines traditional radiomics texture features with the newly proposed orbital spatial position features, achieving an improvement in the performance of the orbital lymphoma classification model. Moreover, the present invention is expected to be used for the diagnosis and prognosis models of other diseases in the orbit. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Schematically shows the preprocessed orbital MRI data, wherein A and B show the original images displayed at different MRI levels, C shows the annotated layer including the medial rectus muscle, lateral rectus muscle, optic nerve, eyeball and lymphoma lesion, and D shows the annotated layer including the superior rectus muscle, eyeball and lymphoma lesion;

[0055] Figure 2 Is a schematic diagram of the orbital spatial relationship. In the figure, the centroid is used to represent the position of each structure: the white sphere represents the eyeball, the black sphere represents the lymphoma lesion, and the orange sphere represents the optic nerve and extraocular muscles. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of this application.

[0057] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "including", "comprising", etc. used herein indicate the presence of the described features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0058] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0059] In the case of using expressions such as "at least one of A, B, and C, etc.", generally, it should be interpreted according to the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include, but is not limited to, a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0060] The detailed background technology may include other technical problems in addition to the technical problems solved by the independent claim.

[0061] The main anatomical structures in the orbit include the eyeball, four rectus muscles, and the optic nerve, and their relative positions in the orbit are relatively fixed. The progression of tumors in the orbit may cause changes in the relative positions between the tumors and adjacent orbital structures, which represents the spatial characteristics of this type. The texture characteristics of the tumor area are considered to be local changes caused by tumor progression. By integrating global (spatial) characteristics and local (texture) characteristics, the performance of the orbital lymphoma classification model can be enhanced.

[0062] Based on the above understanding, one aspect of the embodiments of the present invention is to disclose a method for constructing an automatic classification model of orbital lymphoma based on MRI data, including the following steps:

[0063] Step 1. Acquisition, annotation, and preprocessing of MRI data, which further includes the following steps:

[0064] Step 101. Acquire the orbital MRI data of the patient (including three sequences: axial T1-weighted, axial T2-weighted, and axial T1-enhanced), in DICOM format.

[0065] Step 102: Use the open-source software 3D Slicer (https: / / download.slicer.org / , version 4.11) to annotate the MRI data obtained in the previous step, annotating the bilateral eyeballs, bilateral lateral rectus muscles, bilateral optic nerves, and the lesion range.

[0066] Step 103: Crop the blank edges of the image.

[0067] Step 104: Normalize the data obtained in the previous step according to the mean and standard deviation of the MRI signal intensity.

[0068] Step 105: Crop and resample the normalized MRI data to generate the total dataset, as Figure 1 shown.

[0069] Step 2: Extraction of spatial features, which further includes the following steps:

[0070] Use the schematic diagram of orbital spatial relationships as Figure 2 shown to represent the spatial position relationships of various structures within the orbit.

[0071] For the spatial features of orbital lymphoma, use the centroid to represent each structure, simplifying the complex spatial interrelationships into extractable features. And introduce the midpoint of the line connecting the two eyeballs as a reference. Here, consider three types of relative position relationships: First, the relative position relationships between normal structures (eyeballs, extraocular muscles, optic nerves); Second, the relative positions between lymphoma lesions and normal structures are different in different samples; Third, the relationships from the midpoint to the eyeballs and lymphoma lesions.

[0072] Quantify the position features of lymphoma by measuring the distances between the lymphoma lesion and other structures. To synthesize these spatial relationships, we apply the graph Laplacian method to extract the spatial feature information of each structure. The graph Laplacian algorithm generates eight eigenvalues, representing the spatial interrelationships between the centroids of seven structures (eyeballs, superior rectus muscle, inferior rectus muscle, medial rectus muscle, lateral rectus muscle, optic nerve, lymphoma lesion) and the midpoint. The eigenvalues related to the lymphoma lesion mainly capture the spatial features of the lymphoma lesion as a hub connecting other structures.

[0073] In the above steps, the basic process of extracting eigenvalues based on the graph Laplacian operator is as follows:

[0074] Let G(V, E, A, W) represent a graph G with nodes V = {v1, v2, v3, …, v N}, edges E = {e1, e2, e3, …, e M}, an adjacent matrix A, and a weight matrix W.

[0075] Here, each node v n represents the spatial position of each organizational center, where n = 1, 2, …, N. Each edge e k =(v i , v j ) ∈ V×V is an unordered pair of an undirected graph, where i = 1, 2, …, N and j = 1, 2, …, N. If there is an edge from node v i to node v j , then the element A ij in the i-th row and j-th column of the adjacency matrix A is 1; if there is no edge connecting node v i and node v j , then A ij = 0. If A ij = 1, it means that node v i has an adjacency relationship with node v j . The element W ij in the i-th row and j-th column of the weight matrix W is the Euclidean distance between node v i and node v j , which serves as the weight of the edge connecting node v i and node v j . Obviously, both the adjacency matrix A and the weight matrix W are symmetric matrices because the graph G is an undirected graph. The Laplacian matrix can be defined as L = D - W, where D is the degree matrix, expressed as D = diag(d1, d2, …, d n , …, d N ), and d n is the degree of node v n . Finally, eigenvalues characterizing spatial features are extracted from the Laplacian matrix L.

[0076] Step 3: Extraction of texture features

[0077] The texture classification features used in the present invention include angular second moment (ASM), contrast, correlation, and entropy. ASM is equal to the sum of the squares of the elements in the co-occurrence matrix, which measures the gray-scale variation of the MRI texture and reflects the uniformity of the gray-scale distribution and the roughness of the texture. Contrast is the moment of inertia near the diagonal of the gray-level co-occurrence matrix, which reflects the sharpness of the MRI and the depth of the texture grooves. Correlation reflects the similarity of the elements in the spatial gray-level co-occurrence matrix in the row or column direction and reflects the local gray-scale correlation of the MRI. Entropy reflects the randomness of the MRI texture. These four indicators jointly describe the texture features of the MRI image.

[0078] Step 4: Training of the classification model

[0079] Based on the extraction of the texture features and spatial features of lymphoma, a classification model is established using the XGBoost method. The dataset is randomly divided into a training set and a test set at the same splitting ratio (4:1), and a 5-fold cross-validation method is adopted to reduce the risk of overfitting. Before classification, the features extracted from different sequences are averaged and normalized to ensure consistency.

[0080] The spatial features and texture features of lymphoma are classified and trained respectively. Then, the two features are fused for classification training to establish a hybrid model. Finally, the prediction probabilities of positive and negative samples are output, and the class with the larger prediction probability is judged as the final classification. In this way, the automatic typing of the sample as orbital MALT or non-MALT lymphoma is achieved.

[0081] The second aspect of the embodiments of the present invention discloses an automatic typing device for orbital lymphoma based on MRI data, including:

[0082] An MRI data acquisition and preprocessing module, which further includes an MRI data acquisition unit and an MRI data preprocessing unit, where:

[0083] The MRI data acquisition unit is used to acquire the patient's orbital MRI data (including three sequences of axial T1-weighted, axial T2-weighted, and axial T1-enhanced), and the format is DICOM.

[0084] The MRI data preprocessing unit is used to preprocess the patient's orbital MRI data obtained by the MRI data acquisition unit, including:

[0085] Cropping the blank edges of the slices;

[0086] Normalizing the MRI data according to the average value and standard deviation of the MRI signal intensity;

[0087] Cropping and resampling the normalized MRI data.

[0088] The spatial feature extraction module is used to extract the spatial features of lymphoma from the data output by the MRI data acquisition and preprocessing module.

[0089] The texture feature extraction module is used to extract the texture features of lymphoma from the data output by the MRI data acquisition and preprocessing module.

[0090] The orbital lymphoma typing model is used to automatically type orbital MALT or non-MALT based on the spatial features of lymphoma and the texture features of lymphoma.

[0091] The third aspect of the embodiments of the present invention discloses an electronic device including a processor, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) or programs loaded from a storage section into a random access memory (RAM). The processor may include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application-specific integrated circuit (ASIC)), etc. The processor may also include on-board memory for caching purposes. The processor may include a single processing unit or multiple processing units for performing different actions of the orbital lymphoma automatic typing device according to the embodiments of the present disclosure.

[0092] In the RAM, various programs and data required for the operation of the electronic device are stored. The processor, ROM, and RAM are connected to each other via a bus. The processor implements the above-mentioned orbital lymphoma automatic typing device by executing programs in the ROM and / or RAM. It should be noted that the programs may also be stored in one or more memories other than the ROM and RAM. The processor may also perform various operations of the method flow according to the embodiments of the present disclosure by executing programs stored in the one or more memories.

[0093] According to the embodiments of the present disclosure, the electronic device may further include an input / output (I / O) interface, and the input / output (I / O) interface is also connected to the bus. The electronic device may further include one or more of the following components connected to the I / O interface: an input section including a keyboard, a mouse, etc.; an output section including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section including a hard disk, etc.; and a communication section including a network interface card such as a LAN card, a modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface as needed. Removable media, such as magnetic disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on the drive as needed so that computer programs read from them can be installed into the storage section as needed.

[0094] The fourth aspect of the embodiments of the present invention further provides a computer-readable storage medium, which may be included in the device / device described in the above embodiments; or may exist separately without being assembled into the device / device. The above computer-readable storage medium carries one or more programs, and when the one or more programs are executed, the above-mentioned orbital lymphoma automatic typing device is implemented.

[0095] According to embodiments of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In embodiments of the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device. For example, according to embodiments of the present disclosure, a computer-readable storage medium may include the ROM 402 and / or the RAM 403 described above and / or one or more memories other than the ROM 402 and the RAM 403.

[0096] Those skilled in the art can understand that the features recited in the various embodiments and / or claims of the present disclosure can be combined or combined in various ways, even if such combinations or combinations are not explicitly recited in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features recited in the various embodiments and / or claims of the present disclosure can be combined and combined in various ways. All such combinations and / or combinations fall within the scope of the present disclosure.

[0097] The above describes the embodiments of the present disclosure. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described separately above, this does not mean that the measures in each embodiment cannot be used advantageously in combination. The scope of the present disclosure is defined by the appended claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art can make various substitutions and modifications, and all such substitutions and modifications should fall within the scope of the present disclosure.

Claims

1. A method for constructing an automatic classification model of orbital lymphoma, characterized in that, It includes the following steps: Step 1: Obtain the orbital MRI data of the patient, annotate and preprocess the obtained data, and obtain the total data set; Step 2: Extraction of spatial features: Use the schematic diagram of orbital spatial relationship to represent the spatial position relationship of each structure in the orbit. In the figure, for the spatial features of orbital lymphoma, the centroid is used to represent each structure, and the midpoint connecting the lines of the two eyeballs is introduced as a reference. And consider: the relative position relationship between normal structures, the relative positions between lymphoma lesions and normal structures in different samples are different, the relationship from the reference point to the eyeballs and lymphoma lesions, and the position characteristics of lymphoma are quantified by measuring the distances between lymphoma lesions and other structures, so as to obtain spatial feature information; Step 3: Extract the texture features of the data obtained in Step 1; Step 4: Training of the classification model: Establish a classification model, randomly divide the data set processed in Step 2 and Step 3 into a training set and a test set, and before classification, average and normalize the features extracted from different sequences; Classify and train the spatial features and texture features of lymphoma respectively. Then fuse the two features for classification training, establish a hybrid model, and finally output the prediction probabilities of positive and negative samples. The class with the larger prediction probability is judged as the final classification.

2. The construction method of an automatic classification model for orbital lymphoma according to claim 1, characterized in that, The said Step 1 includes the following steps: Step 101: Obtain the orbital MRI data of the patient, in the DICOM format; Step 102: Annotate the MRI data obtained in the previous step, annotating the bilateral eyeballs, bilateral lateral rectus muscles, bilateral optic nerves and the lesion range; Step 103: Crop the blank edges of the picture; Step 104: Normalize the data obtained in the previous step according to the average value and standard deviation of the MRI signal intensity; Step 105: Crop and resample the normalized MRI data to generate the total data set.

3. The construction method of an automatic classification model for orbital lymphoma according to claim 1, characterized in that In the said Step 2, the Laplacian method is applied to extract the spatial feature information of each structure. Then the extraction of spatial features based on the Laplacian operator includes the following steps: Construct an undirected graph G(V, E, A, W), where: V = {v1, v2, v3, …, v N} is the N nodes, and each node v n represents the spatial location of each organizational center; E = {e1, e2, e3, …, e M} is M edges, and each edge is an unordered pair of any two nodes in the undirected graph; A is an adjacency matrix: If there is an edge from node v i to node v j , then the element A ij in the i-th row and j-th column of the adjacency matrix A is 1; If there is no edge connecting node v i and node v j , then A ij = 0; W is the weight matrix, and the element W at the i-th row and j-th column in the weight matrix W ij is the Euclidean distance between node v i and node v j ; Calculate the Laplacian matrix \(L = D - W\), where \(D\) is the degree matrix, \(D=\text{diag}(d_1,d_2,\ldots,d n ,\ldots,d N ), \(d n \) is the degree of node \(v n \), Extract the eigenvalues representing spatial features from the Laplacian matrix L.

4. The construction method of an automatic classification model for orbital lymphoma according to claim 1, wherein, The said texture classification features include angular second moment, contrast, correlation and entropy, where: The angular second moment is equal to the sum of the squares of the elements in the co-occurrence matrix, measuring the gray level change of the MRI texture, reflecting the uniformity of the gray level distribution and the roughness of the texture; The contrast is the moment of inertia near the diagonal of the gray level co-occurrence matrix, reflecting the sharpness of the MRI and the depth of the texture grooves; The correlation reflects the similarity of the elements in the spatial gray level co-occurrence matrix in the row or column direction, reflecting the local gray level correlation of the MRI; The entropy reflects the randomness of the MRI texture.

5. The construction method of an automatic classification model for orbital lymphoma according to claim 1, wherein, In Step 4, the XGBoost method is used to establish the said classification model.

6. The construction method of an automatic classification model for orbital lymphoma according to claim 1, characterized in that In Step 4, when establishing the training set and the test set, the 5-fold cross-validation method is adopted to reduce the risk of overfitting.

7. An automatic classification device for orbital lymphoma, characterized in that, It includes: The MRI data acquisition and preprocessing module is used to obtain the orbital MRI data of the patient and preprocess the obtained data; The texture feature extraction module is used to extract the lymphoma texture features of the data output by the MRI data acquisition and preprocessing module; A spatial feature extraction module for extracting the spatial features of lymphoma from the data output by the MRI data acquisition and preprocessing module; An orbital lymphoma typing model for automatically typing orbital MALT or non-MALT based on the spatial features and texture features of lymphoma.

8. The automatic classification device for orbital lymphoma according to claim 7, wherein The MRI data acquisition and preprocessing module further includes an MRI data acquisition unit and an MRI data preprocessing unit, where: The MRI data acquisition unit is used to acquire the orbital MRI data of the patient in DICOM format; The MRI data preprocessing unit is used to preprocess the orbital MRI data of the patient obtained by the MRI data acquisition unit, including: Cropping the blank edges of the image; Normalizing the MRI data according to the average value and standard deviation of the MRI signal intensity; Cropping and resampling the normalized MRI data.

9. An electronic device, comprising: One or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to execute the device according to claim 7 or 8.

10. A computer-readable storage medium having executable instructions stored thereon, which when executed by a processor cause the processor to execute the device according to claim 7 or 8.

Images