Method and System for Identifying Odontogenic Jaw Cysts and Tumors Based on Deep Learning

Abstract

The present invention relates to the technical field of medical image processing, and particularly to a method and system for identifying odontogenic jaw cysts and tumors based on deep learning. The method includes: obtaining original jaw CBCT image data and performing preprocessing to generate a joint annotation dataset including lesion type, root resorption morphology annotation, and CT value annotation; constructing a multi-branch neural network model, training the multi-branch neural network model using the joint annotation dataset to obtain an odontogenic jaw lesion recognition model; obtaining the jaw CBCT image data of a target patient, inputting it into the odontogenic jaw lesion recognition model, and outputting the lesion type classification result, root resorption morphology, and bone destruction volume. The present invention can improve the diagnostic ability for micro-lesions, reduce missed diagnoses, realize the full-process auxiliary diagnosis from qualitative diagnosis to quantitative analysis, and improve the diagnostic speed.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly to a method and system for identifying odontogenic jaw cysts and tumors based on deep learning. Background Art

[0002] Odontogenic jaw tumors (such as ameloblastoma) and odontogenic jaw cysts (such as odontogenic keratocyst, dentigerous cyst, radicular cyst) are common diseases in the oral and maxillofacial region. Early diagnosis and precise treatment are crucial for the prognosis of patients.

[0003] However, traditional diagnostic methods mainly rely on doctors' clinical experience and subjective judgment of imaging examinations (such as X-rays, CT, etc.), and have the following limitations:

[0004] Strong subjectivity: Different doctors may have differences in the interpretation of the same image, resulting in inconsistent diagnostic results. And when reading CBCT films, the gray value is device-dependent, which may have an impact when doctors adjust the gray value to observe the lesion density.

[0005] Low efficiency: Doctors need to spend a lot of time reading films, which is difficult to meet the growing medical needs.

[0006] Difficult early diagnosis: Some odontogenic jaw cysts and tumors have no obvious symptoms and atypical imaging manifestations in the early stage, and are prone to missed diagnosis or misdiagnosis. Especially some junior dentists with insufficient experience are prone to overlook or even have difficulty differentiating odontogenic jaw cysts and tumors, resulting in missed diagnosis, misdiagnosis, and mistreatment, and patients cannot receive corresponding treatment in a timely manner at an early stage.

[0007] Although there are currently methods for assisting in lesion diagnosis through models, there is currently no auxiliary diagnosis scheme for odontogenic lesions, and existing models also have the following limitations: insufficient classification performance, difficult to distinguish lesions with blurred boundaries (such as ameloblastoma and keratocyst); poor adaptability to complex anatomical structures (such as multilocular tumors and oval cysts); lack of multi-modal feature fusion, existing models only focus on the overall features of images, do not jointly analyze detailed information such as root resorption patterns and dynamic CT values, and still require manual further detailed diagnosis, wasting patients' consultation time.

[0008] Therefore, there is an urgent need for a method and system for identifying odontogenic jaw cysts and tumors based on deep learning, which can improve the diagnostic ability for micro-lesions, reduce missed diagnosis, misdiagnosis, and mistreatment, realize the full-process auxiliary diagnosis from qualitative diagnosis to quantitative analysis, and improve the diagnostic speed. Summary of the Invention

[0009] To solve the above technical problems, the present invention provides a method and system for identifying odontogenic jaw cysts and tumors based on deep learning, which can improve the diagnostic ability for micro-lesions, reduce missed diagnoses, misdiagnoses, and mistreatments, realize the full-process assisted diagnosis from qualitative diagnosis to quantitative analysis, and improve the diagnostic speed.

[0010] The present invention provides a method for identifying odontogenic jaw cysts and tumors based on deep learning, comprising the following steps:

[0011] S1. Obtain the original jaw CBCT image data and perform preprocessing to generate a combined annotation dataset including lesion type, root resorption morphology annotation, and CT value annotation;

[0012] S2. Construct a multi-branch neural network model, and use the combined annotation dataset to train the multi-branch neural network model to obtain an odontogenic jaw lesion recognition model; wherein, the multi-branch neural network model includes: a main branch and two auxiliary branches;

[0013] The first auxiliary branch is a root resorption interpretation module for outputting the root resorption morphology according to the input jaw CBCT image data;

[0014] The second auxiliary branch is a dynamic CT value analysis module for outputting the bone destruction volume according to the input jaw CBCT image data;

[0015] The main branch is an odontogenic jaw lesion recognition module for outputting the predicted result of the lesion type according to the input jaw CBCT image data, root resorption morphology, and bone destruction volume;

[0016] S3. Obtain the jaw CBCT image data of the target patient, input it into the odontogenic jaw lesion recognition model, and output the lesion type classification result, root resorption morphology, and bone destruction volume.

[0017] Further, in S1, obtaining the original jaw CBCT image data and performing preprocessing to generate a combined annotation dataset including lesion type, root resorption morphology annotation, and CT value annotation includes:

[0018] S11. Perform anatomically-guided multi-region cropping on the original jaw CBCT image data to generate a number of image patches covering the complete root and the bone structure within the preset range of the root;

[0019] S12. Label the lesion type, category label of the root resorption morphology, and CT value based on the image patches; wherein, the lesion type includes cysts and tumors; the root resorption morphology includes no resorption, bevel-shaped resorption, serrated resorption, and truncated resorption;

[0020] S13. Construct a combined annotation dataset according to the labeled image patches.

[0021] Further, in S2, the loss function of the multi-branch neural network model adopts a joint loss function, and its calculation formula is as follows:

[0022] L = αL main + βL aux1 + γL aux2 ;

[0023] where L represents the joint loss function, α, β, and γ are the influence weights of each branch respectively, L main represents the loss function of the main branch, and L aux1 represents the loss function of the first auxiliary branch, and L aux2 represents the loss function of the second auxiliary branch.

[0024] Further, the odontogenic jaw lesion recognition module adopts a ResUnet neural network, and a dynamic attention gating module is embedded before each upsampling operation in the decoder of the ResUnet neural network, which is used to receive the features output by the encoder and the features output by the two auxiliary branches, and generate spatial attention weights, and obtain the features after multi-branch fusion according to the spatial attention weights and the features output by the encoder.

[0025] Further, the root resorption interpretation module adopts a CNN neural network.

[0026] Further, the dynamic CT value analysis module uses wavelet transform to calculate the volume of bone destruction.

[0027] Further, using wavelet transform to calculate the volume of bone destruction includes:

[0028] Sa. Obtain the CT values of the CBCT image data;

[0029] Sb. Use wavelet transform to decompose the CT values of the CBCT image data to obtain wavelet coefficients at different scales;

[0030] Sc. Extract gradient features according to the wavelet coefficients and perform normalization processing;

[0031] Sd. Determine the voxel points belonging to the bone destruction area according to the normalized gradient features and a preset threshold;

[0032] Se. Determine the total volume of bone destruction according to all the voxel points belonging to the bone destruction area.

[0033] The present invention also provides a system for identifying odontogenic jaw cysts and tumors based on deep learning, which is used to implement the method for identifying odontogenic jaw cysts and tumors based on deep learning as described above. The system includes the following modules:

[0034] A data acquisition module, configured to acquire original jaw CBCT image data and perform preprocessing to generate a combined annotation dataset including lesion types, root resorption morphology annotations, and CT value gradient annotations;

[0035] A model construction module, connected to the data acquisition module, configured to construct a multi-branch neural network model, and use the combined annotation dataset to train the multi-branch neural network model to obtain an odontogenic jaw lesion recognition model;

[0036] An output module, connected to the model construction module, configured to acquire the jaw CBCT image data of a target patient, input it into the odontogenic jaw lesion recognition model, and output the lesion type classification result, root resorption morphology, and bone destruction volume.

[0037] The embodiments of the present invention have the following technical effects:

[0038] In this solution, through anatomically guided multi-region cropping of the original jaw CBCT image data, image blocks covering the complete root and the bone structure within the preset range of the root are generated, making subsequent annotation and analysis more targeted and capable of effectively extracting key information related to odontogenic jaw lesions; by constructing a multi-branch neural network model and training it using a combined annotation dataset including lesion types, root resorption morphology annotations, and CT value annotations, comprehensive judgment is made based on multi-faceted information, and the lesion type, root resorption morphology, and bone destruction volume can be identified more comprehensively and accurately, thereby improving the diagnostic accuracy of odontogenic jaw cysts and tumors, enhancing the diagnostic ability for minor lesions, reducing missed diagnoses, achieving a full-process assisted diagnosis from qualitative diagnosis to quantitative analysis, and improving the diagnostic speed; the root resorption interpretation module and the dynamic CT value analysis module respectively focus on the analysis of root resorption morphology and bone destruction volume, providing supplementary information for the main branch odontogenic jaw lesion recognition module, while the odontogenic jaw lesion recognition module embeds a dynamic attention gating module before each upsampling operation of the decoder, which can fuse the features output by the encoder and the features output by the two auxiliary branches, and by generating spatial attention weights, highlight the features that are important for lesion recognition, further improving the accuracy of lesion recognition. Description of the Drawings

[0039] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0040] Figure 1It is a flowchart of a method for identifying odontogenic jaw cysts and tumors based on deep learning provided by an embodiment of the present invention;

[0041] Figure 2 It is a schematic diagram of a panoramic view of the anatomical structure of a normal mandible and teeth and a cross-sectional view of the dental arch provided by an embodiment of the present invention;

[0042] Figure 3 It is a schematic diagram of a panoramic view of the anatomical structure of an odontogenic jaw tumor with serrated root resorption and a cross-sectional view of the dental arch provided by an embodiment of the present invention;

[0043] Figure 4 It is a schematic diagram of a panoramic view of the anatomical structure of an odontogenic jaw tumor with truncated root resorption and a cross-sectional view of the dental arch provided by an embodiment of the present invention;

[0044] Figure 5 It is a schematic diagram of a panoramic view of the anatomical structure of an odontogenic jaw cyst without root resorption and a cross-sectional view of the dental arch provided by an embodiment of the present invention;

[0045] Figure 6 It is a schematic diagram of a panoramic view of the anatomical structure of an odontogenic jaw cyst with beveled root resorption and a cross-sectional view of the dental arch provided by an embodiment of the present invention;

[0046] Figure 7 It is a schematic structural diagram of a system for identifying odontogenic jaw cysts and tumors based on deep learning provided by an embodiment of the present invention. Detailed implementation manners

[0047] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be described clearly and completely below. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention fall within the scope protected by the present invention.

[0048] The embodiment of the present invention provides a method for identifying odontogenic jaw cysts and tumors based on deep learning, Figure 1 which is a flowchart of a method for identifying odontogenic jaw cysts and tumors based on deep learning provided by an embodiment of the present invention, Figure 2 which is a schematic diagram of a panoramic view of the anatomical structure of a normal mandible and teeth and a cross-sectional view of the dental arch provided by an embodiment of the present invention. Refer to Figure 1 and Figure 2 , and specifically includes:

[0049] S1. Obtain the original jaw CBCT (Cone Beam Computed Tomography) image data, and perform preprocessing to generate a combined annotation dataset including lesion types, root resorption morphology annotations, and CT value (Computed Tomography Value) annotations.

[0050] In some embodiments, S1 may include the following sub-steps:

[0051] S11. Perform anatomically-guided multi-region cropping on the original jaw CBCT image data to generate a number of image patches covering the complete roots and the bone structures within a preset range of the roots.

[0052] Crop the original jaw CBCT image according to different anatomical regions (such as the maxilla and mandible) and the specific positions of each tooth to ensure that each cropped image patch contains the complete root and the surrounding bone structure; exemplarily, the preset range may be the bone structure from the alveolar crest top to the lower border of the jaw and within the distance from the root surface to the outer bone plate of the jaw in the lateral direction.

[0053] S12. Label the lesion type, class labels of root resorption morphology, and CT values based on the image patches.

[0054] Figure 3 It is a schematic diagram of a panoramic view of the odontogenic jaw tumor with serrated root resorption and the dental anatomical structure and a cross-sectional view of the dental arch provided by an embodiment of the present invention. Figure 4 It is a schematic diagram of a panoramic view of the odontogenic jaw tumor with truncated root resorption and the dental anatomical structure and a cross-sectional view of the dental arch provided by an embodiment of the present invention. Figure 5 It is a schematic diagram of a panoramic view of the odontogenic jaw cyst without root resorption and the dental anatomical structure and a cross-sectional view of the dental arch provided by an embodiment of the present invention. Figure 6 It is a schematic diagram of a panoramic view of the odontogenic jaw cyst with inclined root resorption and the dental anatomical structure and a cross-sectional view of the dental arch provided by an embodiment of the present invention. Refer to Figures 3 - 6 , the lesion types of the odontogenic jaw include cysts and tumors; the root resorption morphologies include no resorption, inclined resorption, serrated resorption, and truncated resorption.

[0055] Among them, the lesion structure of the odontogenic jaw tumor (ameloblastoma): mostly shows a multilocular shape, with a relatively high CT value, and the roots show serrated or truncated resorption.

[0056] The lesion structure of the odontogenic jaw cyst (radicular cyst, dentigerous cyst, keratocyst): shows a round or oval transparent shadow with a neat edge, a relatively low CT value, and rarely shows root resorption, and the absorption method is mostly inclined.

[0057] CT value is an important parameter in image data. It can reflect the density information of tissues and also plays an important role in judging the nature of lesions.

[0058] In some embodiments, before CT value annotation, the original jaw CBCT image data is first subjected to dynamic CT value standardization. The gray-scale range is adjusted according to different density zones of the jaw to highlight the contrast between the lesion area and the surrounding tissues. The standardization formula is as follows:

[0059] ;

[0060] where, I norm (x, y, z) represents the CT value of the voxel point with the standardized coordinates (x, y, z), I(x, y, z) represents the original CT value of the voxel point with the coordinates (x, y, z), I max represents the maximum CT value in the original jaw CBCT image data, and I min represents the minimum CT value in the original jaw CBCT image data.

[0061] S13. Construct a joint annotation dataset based on the annotated image patches.

[0062] S2. Construct a multi-branch neural network model, and use the joint annotation dataset to train the multi-branch neural network model to obtain an odontogenic jaw lesion recognition model.

[0063] In some embodiments, the multi-branch neural network model includes: a main branch and two auxiliary branches;

[0064] The first auxiliary branch is a root resorption interpretation module for outputting the root resorption morphology according to the input jaw CBCT image data;

[0065] The second auxiliary branch is a dynamic CT value analysis module for outputting the bone destruction volume according to the input jaw CBCT image data;

[0066] The main branch is an odontogenic jaw lesion recognition module for outputting the predicted result of the lesion type according to the input jaw CBCT image data, the root resorption morphology, and the bone destruction volume.

[0067] In some embodiments, the odontogenic jaw lesion recognition module adopts a ResUnet neural network, and a dynamic attention gating module is embedded before each upsampling operation in the decoder of the ResUnet neural network. The dynamic attention gating module is used to receive the features output by the encoder and the features output by the two auxiliary branches, and generate spatial attention weights. According to the spatial attention weights and the features output by the encoder, the features after multi-branch fusion are obtained.

[0068] Specifically, the calculation process of the dynamic attention gating module to obtain the features after multi-branch fusion based on the features output by the encoder and the features output by the two auxiliary branches is as follows:

[0069] ;

[0070] ;

[0071] Among them, σ represents the sigmoid activation function, represents the feature concatenation operation, Conv represents the convolution operation, W att represents the spatial attention weight, F absorb represents the features output by the root resorption discrimination module, F grad represents the features output by the dynamic CT value analysis module, F main represents the features output by the encoder, F en represents the features after multi-branch fusion, represents element-wise multiplication, + represents residual connection.

[0072] In some embodiments, the root resorption interpretation module uses a CNN neural network.

[0073] In some embodiments, the dynamic CT value analysis module uses wavelet transform to calculate the bone destruction volume.

[0074] In some embodiments, the loss function of the multi-branch neural network model uses a joint loss function, and the calculation formula is as follows:

[0075] L = αL main + βL aux1 + γL aux2 ;

[0076] Among them, L represents the joint loss function, α, β, and γ are the influence weights of each branch respectively, L main represents the loss function of the main branch, L aux1 represents the loss function of the first auxiliary branch, L aux2 represents the loss function of the second auxiliary branch.

[0077] Exemplarily, the loss function of the main branch can use the Focal loss function to alleviate the problem of sample imbalance in lesion categories and balance the model's attention to the two types of samples;

[0078] The loss function of the first auxiliary branch can use a loss function that fuses the Dice loss function and weighted cross-entropy to optimize the low quota of the segmentation region and alleviate the problem of class imbalance when the absorption region accounts for a small proportion;

[0079] The loss function of the second auxiliary branch can use MSE (mean square error) as the loss function.

[0080] In some embodiments, calculating the volume of bone destruction using wavelet transform includes:

[0081] Sa. Obtain the CT values of the CBCT image data.

[0082] In some embodiments, wavelet transform calculation can be performed based on the standardized CT values.

[0083] Sb. Use wavelet transform to decompose the CT values of the CBCT image data to obtain wavelet coefficients of different scales.

[0084] In some embodiments, the Daubechies wavelet function can be used; the calculation formula is as follows:

[0085] ;

[0086] where j represents the scale level of the wavelet function, W j (x, y, z) represents the wavelet coefficient of the j-th scale, I norm (m, n, p) represents the standardized CT value of the voxel point with coordinates (m, n, p), (m, n, p) is the coordinate of a reference voxel point, (x, y, z) represents the coordinate of the current voxel point, φ j (x - m, y - n, z - p) represents the wavelet basis function at the j-th scale, indicating the offset of the current voxel point relative to the reference voxel point in three-dimensional space.

[0087] Sc. Extract gradient features based on the wavelet coefficients and perform normalization processing.

[0088] Calculate the gradient of the wavelet coefficients of each layer to highlight the local changes of the influence. The calculation formula is as follows:

[0089] ;

[0090] where G j (x, y, z) represents the gradient of the wavelet coefficient of the j-th scale.

[0091] The calculation formula for normalizing the gradient is as follows:

[0092] ;

[0093] where G norm,j (x, y, z) represents the normalized gradient of the wavelet coefficient of the j-th scale, G max,j represents the maximum value of the gradient of the wavelet coefficient of the j-th scale, G min,j represents the minimum value of the gradient of the wavelet coefficient of the j-th scale.

[0094] Sd. Determine the voxel points belonging to the bone destruction area according to the normalized gradient features and a preset threshold.

[0095] In some embodiments, the destruction threshold T can be set according to expert experience or historical data.

[0096] For each voxel point, determine whether its normalized gradient value is greater than the destruction threshold T. If it is greater, then determine that the voxel point belongs to the bone destruction area:

[0097] ;

[0098] Where D(x, y, z) represents the classification of the voxel point. When D(x, y, z) = 1, it means that the voxel point belongs to the bone destruction area. When D(x, y, z) = 0, it means that the voxel point belongs to the normal area.

[0099] Se. Determine the total volume of bone destruction according to all the voxel points belonging to the bone destruction area.

[0100] Calculate the number of all voxel points with D(x, y, z) = 1 to obtain the total volume of bone destruction. The calculation formula is as follows:

[0101] ;

[0102] Where V destroy represents the total volume of bone destruction.

[0103] S3. Obtain the jaw CBCT image data of the target patient, input it into the odontogenic jaw lesion recognition model, and output the lesion type classification result, the root resorption morphology, and the bone destruction volume.

[0104] This solution generates image blocks covering the entire root and the bone structure within the preset range of the root by performing anatomically-guided multi-region cropping on the original jaw CBCT image data, making subsequent annotation and analysis more targeted and enabling the effective extraction of key information related to odontogenic jaw lesions. By constructing a multi-branch neural network model and training it using a joint annotation dataset containing lesion type, root resorption morphology annotation, and CT value annotation, and making judgments by integrating information from multiple aspects, it can more comprehensively and accurately identify the lesion type, root resorption morphology, and bone destruction volume, thereby improving the diagnostic accuracy of odontogenic jaw cysts and tumors, enhancing the diagnostic ability for minor lesions, reducing missed diagnoses, achieving a full-process assisted diagnosis from qualitative diagnosis to quantitative analysis, and improving the diagnostic speed. The root resorption interpretation module and the dynamic CT value analysis module focus on the analysis of root resorption morphology and bone destruction volume respectively, providing supplementary information for the main branch odontogenic jaw lesion recognition module. The odontogenic jaw lesion recognition module embeds a dynamic attention gating module before each upsampling operation in the decoder, which can fuse the features output by the encoder and the features output by the two auxiliary branches, and further improve the accuracy of lesion recognition by generating spatial attention weights to highlight the features important for lesion recognition.

[0105] An embodiment of the present invention provides a system for identifying odontogenic jaw cysts and tumors based on deep learning. Figure 7 FIG. is a schematic structural diagram of the system for identifying odontogenic jaw cysts and tumors based on deep learning provided by an embodiment of the present invention. Refer to Figure 7 This system is used to implement the above-mentioned method for identifying odontogenic jaw cysts and tumors based on deep learning. The system includes the following modules:

[0106] The data acquisition module is used to acquire the original jaw CBCT image data, perform preprocessing, and generate a joint annotation dataset containing lesion type, root resorption morphology annotation, and CT value gradient annotation.

[0107] The model construction module is connected to the data acquisition module and is used to construct a multi-branch neural network model, train the multi-branch neural network model using the joint annotation dataset, and obtain an odontogenic jaw lesion recognition model.

[0108] The output module is connected to the model construction module and is used to acquire the jaw CBCT image data of the target patient, input it into the odontogenic jaw lesion recognition model, and output the lesion type classification result, root resorption morphology, and bone destruction volume.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A system for identifying odontogenic jaw cysts and tumors based on deep learning, characterized in that, The system includes the following modules: A data acquisition module, which is used to acquire the original jaw CBCT image data, perform preprocessing, and generate a joint annotation dataset including lesion types, root resorption morphology annotations, and CT value gradient annotations; A model construction module, connected to the data acquisition module, which is used to construct a multi-branch neural network model, and use the joint annotation dataset to train the multi-branch neural network model to obtain an odontogenic jaw lesion recognition model; wherein, the multi-branch neural network model includes: a main branch and two auxiliary branches; The first auxiliary branch is a root resorption interpretation module, which is used to output the root resorption morphology according to the input jaw CBCT image data; The second auxiliary branch is a dynamic CT value analysis module, which is used to output the bone destruction volume according to the input jaw CBCT image data; The main branch is an odontogenic jaw lesion recognition module, which is used to output the predicted result of the lesion type according to the input jaw CBCT image data, root resorption morphology, and bone destruction volume; The odontogenic jaw lesion recognition module adopts a ResUnet neural network, and a dynamic attention gating module is embedded before each upsampling operation in the decoder of the ResUnet neural network, which is used to receive the features output by the encoder and the features output by the two auxiliary branches, and generate spatial attention weights, and obtain the features after multi-branch fusion according to the spatial attention weights and the features output by the encoder; Among them, the calculation process of the dynamic attention gating module to obtain the features after multi-branch fusion according to the features output by the encoder and the features output by the two auxiliary branches is as follows: ； ； Among them, σ represents the sigmoid activation function, represents the feature concatenation operation, Conv represents the convolution operation, W att represents the spatial attention weight, F absorb represents the feature output by the root resorption discrimination module, F grad represents the feature output by the dynamic CT value analysis module, F main represents the feature output by the encoder, F en represents the feature after multi-branch fusion, represents element-wise multiplication, + represents the residual connection; An output module, connected to the model construction module, which is used to acquire the jaw CBCT image data of the target patient, input it into the odontogenic jaw lesion recognition model, and output the lesion type classification result, root resorption morphology, and bone destruction volume.

2. The system for identifying odontogenic jaw cysts and tumors based on deep learning according to claim 1, characterized in that, Specifically, the data acquisition module is used for: Performing anatomically-guided multi-region cropping on the original jaw CBCT image data to generate a number of image patches covering the complete root and the bone structure within the preset range of the root; Based on the image patches, annotating the lesion type, category labels of the root resorption morphology, and CT values; wherein, the lesion types include cysts and tumors; the root resorption morphologies include no resorption, inclined plane resorption, serrated resorption, and truncated resorption; Constructing a joint annotation dataset according to the annotated image patches.

3. The system for identifying odontogenic jaw cysts and tumors based on deep learning according to claim 1, wherein In the model construction module, the loss function of the multi-branch neural network model adopts a joint loss function, and the calculation formula is as follows: L = αL main + βL aux1 + γL aux2 ; Among them, L represents the combined loss function, and α, β, and γ are the influence weights of each branch. L main represents the loss function of the main branch, and L aux1 represents the loss function of the first auxiliary branch, and L aux2 represents the loss function of the second auxiliary branch.

4. The system for identifying odontogenic jaw cysts and tumors based on deep learning according to claim 3, wherein The root resorption interpretation module adopts a CNN neural network.

5. The system for identifying odontogenic jaw cysts and tumors based on deep learning according to claim 3, characterized in that, The dynamic CT value analysis module uses wavelet transform to calculate the bone destruction volume.

6. The system for identifying odontogenic jaw cysts and tumors based on deep learning according to claim 5, wherein, The calculation of the bone destruction volume using wavelet transform includes: Sa. Obtaining the CT values of the CBCT image data; Sb. Using wavelet transform to decompose the CT values of the CBCT image data to obtain wavelet coefficients of different scales; Sc. Extracting gradient features according to the wavelet coefficients and performing normalization processing; Sd. Determining the voxel points belonging to the bone destruction region according to the normalized gradient features and a preset threshold. Se. Determine the total volume of bone destruction based on all voxel points belonging to the bone destruction region.

Images