Tooth trauma image diagnosis system and method based on deep learning

Abstract

The invention provides a tooth trauma image diagnosis system and method based on deep learning, and belongs to the field of tooth diagnosis and treatment. The method comprises the steps of performing data enhancement on a plurality of collected first CBCT image data to obtain training data with a higher style diversification degree, and training a segmentation model and a classification model based on Swin Transform by using the training data; performing tooth image segmentation processing on the second CBCT image data of the target patient by using the trained segmentation model to obtain a plurality of single tooth images, and extracting single features and common features of each tooth from each single tooth image; classifying the single features by using a classification model to obtain a plurality of classification results, and deleting part of the classification results based on the common features; and generating a corresponding treatment scheme based on each remaining classification result after deletion. The tooth trauma diagnosis system achieves accurate diagnosis of tooth trauma and can assist doctors in improving diagnosis and treatment efficiency.

Description

Technical Field

[0001] The present invention relates to the field of dental diagnosis and treatment, and in particular to a dental trauma imaging diagnosis system and method based on deep learning. Background Art

[0002] Dental trauma is a type of acute injury that is common in adolescents, but due to its complex classification and inconsistent treatment standards, it can easily lead to poor prognosis and may cause permanent physical and mental damage to children. Therefore, designing an intelligent diagnosis solution for dental trauma based on a deep learning model will have broad application prospects.

[0003] However, the existing intelligent diagnosis schemes for dental trauma based on deep learning models still have the following major problems: (1) Deep learning models heavily rely on large-scale, high-quality annotated datasets. However, due to privacy issues and annotation costs, large-scale standard datasets with diverse styles and detailed annotations are generally insufficient, which limits the accuracy and robustness of deep learning models. (2) Existing dental image segmentation methods still have the problem of unclear segmentation areas, making it difficult to refine the fuzzy boundaries of adjacent teeth. (3) The accuracy of dental trauma classification needs to be improved.

[0004] Therefore, designing an improved intelligent diagnosis solution for dental trauma based on deep learning models to solve the above technical problems is a technical problem that needs to be solved urgently. Summary of the invention

[0005] In response to the above technical problems, the present invention provides a dental trauma imaging diagnosis method, system, electronic device, computer storage medium and computer program product based on deep learning.

[0006] The present invention discloses a method for dental trauma imaging diagnosis based on deep learning, the method comprising the following steps: Performing data enhancement on the collected first CBCT image data to obtain training data with a higher degree of style diversity, and using the training data to train a segmentation model and a classification model based on Swin Transformer; Using the trained segmentation model to perform tooth image segmentation processing on the second CBCT image data of the target patient to obtain a plurality of single tooth images, and extracting single features and common features of each tooth from each of the single tooth images; Using a classification model to classify the individual features to obtain a number of classification results, and deleting some of the classification results based on the common features; Based on the classification results remaining after the deletion, a corresponding treatment plan is generated.

[0007] Furthermore, data enhancement is performed on the collected first CBCT image data to obtain training data with a higher degree of style diversity, including: Performing Fourier transform processing on the original first CBCT image data to obtain frequency domain image data; Selecting noise data from other original first CBCT image data and converting it into frequency domain to obtain frequency domain noise data, and superimposing the frequency domain noise data with the frequency domain image data to obtain new frequency domain image data; Perform inverse Fourier transform on each new frequency domain image data to obtain a plurality of third CBCT image data, and construct a plurality of training data based on each of the first CBCT image data, the third CBCT image data and the corresponding category labels.

[0008] Furthermore, the frequency domain noise data is superimposed with the frequency domain image data to obtain new frequency domain image data, including: Determine a low-frequency part and a high-frequency part of the frequency-domain image data, and for the low-frequency part and the high-frequency part, respectively use a first adding ratio and a second adding ratio to weaken the frequency-domain noise data; wherein the first adding ratio is smaller than the second adding ratio; The weakened frequency domain noise data is superimposed on the frequency domain image data to obtain new frequency domain image data.

[0009] Furthermore, before superimposing the frequency domain noise data with the frequency domain image data, the method further includes: Determine the overall similarity between the first CBCT image data and other first CBCT image data, and obtain the enhancement quantity according to the overall similarity matching; wherein the enhancement quantity is positively correlated with the overall similarity; A plurality of groups of additive ratios corresponding to the enhancement amounts are generated for the first CBCT image data, and each group of the additive ratios includes a first additive ratio and a second additive ratio.

[0010] Furthermore, the using of the classification model to classify the individual features to obtain a number of classification results, and deleting some of the classification results based on the common features, includes: Using a classification model to classify the monomer features, obtaining a number of first classification results; Extracting the position and posture information of the common features of each tooth based on the second CBCT image data, performing consistency evaluation on each of the position and posture information, and if the consistency evaluation result is higher than a preset judgment condition, determining that the common features are non-tooth trauma; The first classification results that are not dental trauma are determined as second classification results, and the second classification results in each of the first classification results are deleted.

[0011] The present invention also discloses a dental trauma imaging diagnosis system based on deep learning, the system comprising a training module, a segmentation module, a classification module, and an output module; The training module is used to perform data enhancement on the collected first CBCT image data to obtain training data with a higher degree of style diversity, and use the training data to train a segmentation model and a classification model based on Swin Transformer; The segmentation module is used to perform tooth image segmentation processing on the second CBCT image data of the target patient using the trained segmentation model to obtain multiple single tooth images, and extract single features and common features of each tooth from each single tooth image; The classification module is used to classify the monomer features using a classification model to obtain a number of classification results, and delete some of the classification results based on the common features; The output module is used to generate a corresponding treatment plan based on each of the classification results remaining after deletion.

[0012] Furthermore, the training module is used to: Performing Fourier transform processing on the original first CBCT image data to obtain frequency domain image data; Selecting noise data from other original first CBCT image data and converting it into frequency domain to obtain frequency domain noise data, and superimposing the frequency domain noise data with the frequency domain image data to obtain new frequency domain image data; Perform inverse Fourier transform on each new frequency domain image data to obtain a plurality of third CBCT image data, and construct a plurality of training data based on each of the first CBCT image data, the third CBCT image data and the corresponding category labels.

[0013] The present invention also discloses an electronic device, comprising one or more processors, a memory, and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, and when the instructions are executed by the processor, the deep learning-based dental trauma imaging diagnosis method as described in any one of the above items is implemented.

[0014] The present invention also discloses a computer storage medium, comprising computer instructions that can be executed on an electronic device, wherein the computer instructions are executed to implement the deep learning-based dental trauma imaging diagnosis method as described in any one of the above items.

[0015] The present invention also discloses a computer program product, which can be run on an electronic device to implement the deep learning-based dental trauma imaging diagnosis method as described in any one of the above items.

[0016] The beneficial effects of the present invention are at least: The solution of the present invention solves the problems of insufficient training data and low diversity of the segmentation model through data enhancement technology, so that the trained segmentation model can adapt to actual situations such as blurred boundaries between adjacent teeth, and has higher segmentation accuracy and robustness.

[0017] The solution of the present invention can also realize accurate classification of dental trauma types based on the individual characteristics and common characteristics of individual teeth, thereby improving the accuracy of generating subsequent treatment plans, and can assist in improving the decision-making efficiency of clinicians in the diagnosis, treatment and prognosis evaluation of dental trauma. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a schematic diagram of the main flow of a method for dental trauma imaging diagnosis based on deep learning disclosed in an embodiment of the present invention; Figure 2 It is a schematic diagram of a process of performing data enhancement on CBCT image data disclosed in an embodiment of the present invention; Figure 3 It is a scene schematic diagram of a dental trauma imaging diagnosis system based on deep learning disclosed in an embodiment of the present invention; Figure 4 It is a structural schematic diagram of a dental trauma imaging diagnosis system based on deep learning disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0021] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] like Figure 1 As shown, in view of the above technical problems, an embodiment of the present invention discloses a method for dental trauma imaging diagnosis based on deep learning, and the method comprises the following steps: S10, performing data enhancement on the collected first CBCT image data to obtain training data with a higher degree of style diversity, and using the training data to train a segmentation model and a classification model based on Swin Transformer.

[0024] The present invention uses a segmentation model based on Swin Transformer to segment and obtain the images of each tooth. Before this, it is necessary to construct training data to fully train the segmentation model. As described in the background technology, due to privacy issues and annotation costs, there is a general shortage of large-scale standard data sets with diverse styles and detailed annotations, which limits the accuracy and robustness of deep learning models. In this regard, the present invention uses tooth image data enhancement technology to obtain sufficient and more diverse training data. The details are as follows: Collect some first CBCT image data. CBCT images are cone beam CT images, which can clearly show the three-dimensional structure of teeth and surrounding tissues. For example, collect CBCT images of 100 adolescent patients with dental trauma, which cover different types of dental trauma, such as tooth fracture, dislocation, tooth surface scratches, etc. Then, data enhancement is performed. Data enhancement is to make the data richer and more diverse by performing some transformation operations on the original data. Conventional enhancement methods include, for example, rotating the collected CBCT images clockwise or counterclockwise by a certain angle, such as 15 degrees, 30 degrees, etc.; scaling operations can also be performed to enlarge or reduce the image by a certain ratio, such as reducing the image to 80% of the original or enlarging it to 120%; and brightness adjustment, brightening or darkening the image, etc. Through these operations, the original 100 cases of image data can be turned into 300 or even more, which is conducive to constructing a sufficient amount of training data with a higher degree of style diversity. Finally, these data-enhanced CBCT image data are used to construct a sufficient first set of training data (the training data contains CBCT image data and the boundary markers between the teeth annotated, and the boundary markers are used as labels), and the Swin Transformer-based segmentation model is trained using it. Swin Transformer is a model that performs well in the field of image segmentation. Various data-enhanced dental image data are input into the segmentation model, allowing it to continuously learn how to distinguish and identify the shape and boundaries of teeth, so that it can accurately segment the dental images later. Among them, the classification model also needs to be pre-trained, and the training data used is also constructed based on the CBCT image data obtained after data enhancement, which will be explained in detail later.

[0025] It should be noted that the label can be annotated in the CBCT image data (or the segmented CBCT sub-image data) in the form of a pixel-level pseudo mask.

[0026] S20, using the trained segmentation model to perform tooth image segmentation processing on the second CBCT image data of the target patient to obtain a plurality of single tooth images, and extracting single features and common features of each tooth from each of the single tooth images.

[0027] The second CBCT image data of the target patient is received, and the tooth image segmentation processing is performed on the second CBCT image data using the segmentation model trained in step S20, so that each tooth in the CBCT image of the entire oral cavity is segmented separately to obtain multiple single tooth images. For example, if there are 32 teeth (including wisdom teeth) in the oral cavity, 32 separate tooth images will be obtained after segmentation, and each image contains only one tooth. The individual features and common features of each tooth are extracted from each individual tooth image. Individual features are all the extractable features of each tooth, for example, a certain front tooth may have an obvious fracture mark. Common features refer to some common features that all (or most) teeth have, such as structural features such as enamel and dentin, surface features, etc. These features are extracted from these individual tooth images by using convolutional networks and other methods for subsequent analysis.

[0028] S30, using a classification model to classify the individual features to obtain a number of classification results, and deleting some of the classification results based on the common features.

[0029] The present invention also pre-constructs and trains a classification model (preferably based on a neural network algorithm), and the classification model can be used to classify individual features, that is, to determine the possible trauma type of each tooth. For example, based on the individual features of the fracture marks of the incisor, such as shape and size, it is possible to determine whether the incisor belongs to different dental trauma classifications such as crown fracture, root fracture, crown-root combined fracture, and conventional surface scratches. The identification features corresponding to each dental trauma classification are pre-set, and the classification results obtained by the classification model may include multiple classification results, all of which are classification results that are higher than the confidence threshold, that is, the classification model may not be able to uniquely determine the type of trauma to which a certain tooth belongs, but instead outputs all of the multiple classification results that are higher than the confidence threshold.

[0030] Since there may be errors such as misidentification in the above classification results, for example, the classification result of a tooth shows that its enamel part presents an abnormal shape, which is very different from the common features of enamel of other normal teeth and injured teeth. If the classification result is determined to be wrong based on the common features, it will be deleted from all classification results to ensure that the remaining classification results are more reliable.

[0031] It should be noted that the classification model also needs to be pre-trained, and the second set of training data used for training is also constructed based on the CBCT image data obtained in step S10. The difference is that the second set of training data contains single images of multiple teeth and their labels, and the labels are the types of dental trauma.

[0032] S40, generating a corresponding treatment plan based on the classification results remaining after the deletion.

[0033] Based on the remaining classification results after deletion, the corresponding treatment plan is automatically generated. For example, if the classification result is that the target patient has one tooth with a crown fracture and another tooth with a dislocated tooth, then for the tooth with a crown fracture, if the fracture is not serious, a filling restoration treatment can be performed; for the dislocated tooth, a repositioning and fixation treatment can be performed.

[0034] This process can also be achieved through an automatic treatment plan generation model, which can learn the standardized diagnosis and treatment methods of different types of dental trauma, and combine the results of evidence-based medicine to perform prognosis assessment, and finally achieve intelligent generation of treatment plans. The automatic treatment plan generation model can be constructed using conventional classification algorithms (such as SVM, random forest, etc.), or based on a deep learning algorithm, which is not specifically limited in the present invention.

[0035] The solution of the present invention solves the problems of insufficient training data and low diversity of segmentation models through data enhancement technology, so that the segmentation model obtained through training can adapt to actual situations such as blurred boundaries between adjacent teeth, and has higher segmentation accuracy and robustness. At the same time, the solution of the present invention can also accurately classify the types of dental trauma based on the individual characteristics and common characteristics of individual teeth, thereby improving the accuracy of generating subsequent treatment plans, and can assist in improving the decision-making efficiency of clinicians in the diagnosis, treatment and prognosis evaluation of dental trauma.

[0036] like Figure 3 As shown, after segmenting and classifying the tooth image data, the corresponding dental trauma classification results, treatment plans, etc. are output to the human-computer interaction platform for the doctor's reference. The output results may also include the positional relationship between the traumatized tooth and the alveolar socket, or the traumatized tooth and its classification results are directly marked in the CBCT image data. There is no specific limitation on this.

[0037] Optionally, in order to alleviate the problems of insufficient dental image data and single style, the present invention adopts Fourier transform-based data enhancement technology to add noise from other data samples to the training samples to generate new data with different styles, which helps the model learn high-level semantic information from diverse data distributions and improves the versatility and clinical applicability of the model. Figure 2 , as follows: Data enhancement is performed on the collected first CBCT image data to obtain training data with a higher degree of style diversity, including: S11, performing Fourier transform processing on the original first CBCT image data to obtain frequency domain image data.

[0038] First, the original first CBCT image data (such as images of teeth and surrounding tissues) is Fourier transformed, which converts the image from a spatial pixel representation to a frequency component representation. Through Fourier transform, the image can be decomposed into a combination of sine and cosine waves of different frequencies. Each frequency component carries different aspects of the image information. The low-frequency part reflects the general outline of the image, and the high-frequency part reflects the detailed characteristics of the image.

[0039] CBCT images are essentially two-dimensional images, and Fourier transform is usually performed using two-dimensional discrete Fourier transform (2D-DFT). Suppose the CBCT image data is ,in ; ; and are the number of rows and columns of CBCT image data respectively. Its two-dimensional discrete Fourier transform formula is: ; In the formula, is the transformed frequency domain image data, , , is an imaginary unit .

[0040] In actual calculation, starting from the upper left corner pixel of the image, for each pixel , its pixel value Multiplying complex exponential terms , and then accumulate all the pixels of the entire image. The change of the value will obtain the transformation results at different frequencies, thus constructing the complete frequency domain image data. hour, It represents the image The DC component of is the sum of all pixel values. and The increase in the absolute value corresponds to a higher frequency component, which reflects the detail information in the CBCT image.

[0041] S12, selecting noise data from other original first CBCT image data and converting it into frequency domain to obtain frequency domain noise data, and superimposing the frequency domain noise data with the frequency domain image data to obtain new frequency domain image data.

[0042] Noise data are selected from other original first CBCT image data, and these noise data refer to image segments that are different from the current first CBCT image data but contain similar elements. These image segments can come from multiple other first CBCT image data. The selected noise data is also converted to the frequency domain in the above manner to obtain frequency domain noise data.

[0043] Then, the frequency domain noise data is superimposed with the frequency domain image data obtained previously. By superimposing noise, new features and changes can be introduced in the frequency domain, making the final generated image more diverse in style. Different noise data will bring different changes to the frequency domain image, thereby generating new frequency domain image data of various styles. For example, if there are more high-frequency components in the frequency domain noise data, the high-frequency part of the new frequency domain image data after superposition will be enhanced, so that the final generated third CBCT image data will show more changes in details, such as more complex texture on the tooth surface.

[0044] S13, performing inverse Fourier transform on each new frequency domain image data to obtain a plurality of third CBCT image data, and constructing a plurality of training data based on each of the first CBCT image data, the third CBCT image data and the corresponding category labels.

[0045] The superimposed new frequency domain image data is subjected to inverse Fourier transform to convert the new frequency domain data back to the spatial domain to obtain the third CBCT image data that can be used for training. Inverse Fourier transform is the inverse process of Fourier transform, which recombines the signal in the frequency domain into the pixel form of the image. The formula for inverse transform is as follows: ; When calculating the inverse transform, the frequency domain image is also Each point is operated, and the frequency domain value Multiplying complex exponential terms , and then for all Add up and finally divide by Get the original image in The pixel value at Through this process, the frequency information contained in the frequency domain is recombined into an image in the spatial domain, and a CBCT image with practical significance is restored.

[0046] Based on the original first CBCT image data, the generated third CBCT image data and their corresponding category labels (such as the first category label marking the tooth boundary, or the second category label of normal teeth, tooth trauma type, etc.), multiple training data are constructed. These training data contain the original data and the new data after data enhancement, which enriches the diversity of the training set and helps to improve the training effect and generalization ability of the model. It should be noted that the generated training data is divided into two groups, one for the segmentation model and the other for the classification model, which will not be repeated here.

[0047] Among them, after the inverse Fourier transform, the generated new data sample, i.e., the third CBCT image data, needs to be subjected to boundary clipping and filtering processing to remove edge outliers and high-frequency noise that may be generated by the noise addition and transformation process, so as to improve the quality of the new data sample.

[0048] Optionally, superimposing the frequency domain noise data with the frequency domain image data to obtain new frequency domain image data includes: Determine a low-frequency part and a high-frequency part of the frequency-domain image data, and for the low-frequency part and the high-frequency part, respectively use a first adding ratio and a second adding ratio to weaken the frequency-domain noise data; wherein the first adding ratio is smaller than the second adding ratio; The weakened frequency domain noise data is superimposed on the frequency domain image data to obtain new frequency domain image data.

[0049] In this embodiment, the frequency domain image data contains information of different frequencies, wherein the low frequency part mainly reflects the overall features of the teeth in the image, such as the general outline and shape, such as the overall shape of the teeth, the general structure of the alveolar bone, etc.; the high frequency part focuses on the details of the teeth in the image, such as the texture of the tooth surface, tiny damage marks, etc. The low frequency and high frequency parts can be distinguished according to the frequency range or through a specific filtering algorithm. For example, a low-pass filter can be used to extract the low frequency part in the frequency domain image data, and a high-pass filter can be used to obtain the high frequency part.

[0050] Since the low-frequency part has a great influence on the overall structure and main features of the dental CBCT image, if too much low-frequency noise is added, the basic shape and contour of the image may be greatly changed, affecting the authenticity and availability of the data. For example, if too much noise is added to the low-frequency part, the overall shape of the tooth may appear distorted, which is not conducive to subsequent diagnosis and other operations based on the image. Therefore, the present invention adopts a smaller first addition ratio so that the impact of the added low-frequency noise on the main structure of the image is controlled within a smaller range, while increasing the diversity of the data and retaining the key features of the original image to the greatest extent.

[0051] At the same time, considering that the high-frequency part mainly reflects the image details. Properly increasing the amount of noise added in the high-frequency part can introduce more detail changes and enhance the diversity of the data. For example, details such as tiny cracks and wear marks on the surface of teeth are reflected in the high-frequency part. Properly increasing the high-frequency noise can simulate more different details and allow the segmentation model to learn richer features. However, this ratio cannot be too large, otherwise it may introduce too much messy high-frequency noise, covering up the real detail information.

[0052] The frequency domain noise data weakened by different proportions are superimposed on the low-frequency and high-frequency parts of the frequency domain image data. For the low-frequency part, the noise weakened by the first addition ratio is added to the low-frequency image data, and for the high-frequency part, the noise weakened by the second addition ratio is added to the high-frequency image data, and finally the new frequency domain image data is obtained. The new frequency domain image data after superposition not only maintains the similarity with the original image in the overall structure, but also increases the diversity in the details, providing rich frequency domain information for the subsequent generation of CBCT images with diverse styles through inverse Fourier transform, which helps to improve the training effect and generalization ability of the segmentation model.

[0053] It should be noted that, before and after the superposition, the marking position of the tooth boundary label is actually unchanged, so the newly generated third CBCT image data also contains the first category label of the corresponding first CBCT image data.

[0054] Optionally, before superimposing the frequency domain noise data with the frequency domain image data, the method further includes: Determine the overall similarity between the first CBCT image data and other first CBCT image data, and obtain the enhancement quantity according to the overall similarity matching; wherein the enhancement quantity is positively correlated with the overall similarity; A plurality of groups of additive ratios corresponding to the enhancement amounts are generated for the first CBCT image data, and each group of the additive ratios includes a first additive ratio and a second additive ratio.

[0055] In this embodiment, the style diversity of the first CBCT image data can be improved by data enhancement processing, thereby improving the style diversity of the training data. However, the significance of different first CBCT image data is also different. For example, some first CBCT image data are significantly different from other first CBCT image data, while other first CBCT image data are similar to some other first CBCT image data. If too much data enhancement processing is performed on the latter, the third CBCT image data obtained by the enhancement processing is actually very similar to the original first CBCT image data, which will lead to a decrease in the overall style diversity of the first CBCT image data collected in this batch.

[0056] In view of the above technical problems, the present invention first calculates the similarity between each first CBCT image data and the remaining other first CBCT image data, and then aggregates all similarities (for example, calculates the average value or mode value or median value) to obtain the overall similarity. The higher the overall similarity, the greater the difference in style between the first CBCT image data and the other first CBCT image data, and vice versa, the closer the style of the first CBCT image data is to the other first CBCT image data.

[0057] Furthermore, based on the overall similarity obtained above, the number of reinforcements is matched and obtained. The matching process is based on a preset correlation relationship, that is, a positive correlation between the number of reinforcements and the overall similarity is established in advance, and then the overall similarity is matched and calculated with the positive correlation to obtain the corresponding number of reinforcements.

[0058] In this way, a larger enhancement number is set for the first CBCT image data with an independent style, that is, more groups of addition ratios are generated for it. In other words, more third CBCT image data are subsequently generated based on the first CBCT image data. Each group of addition ratios includes different first addition ratios and second addition ratios, but the first addition ratio and the second addition ratio should be within a preset reasonable range, and the details are not repeated here.

[0059] This embodiment of the present invention distinguishes the first CBCT image data with significant style from the first CBCT image data with insignificant style, and generates different numbers of third CBCT image data based on the first CBCT image data with significant style. In this way, the third CBCT image data can be generated mainly using the first CBCT image data with significant style. Since the first CBCT image data has a higher degree of style significance, it has more enhancement processing space, and the style differences between the third CBCT image data obtained will be relatively larger, which will not reduce the overall style diversity of the first CBCT image data collected in this batch.

[0060] Optionally, the using a classification model to classify the individual features to obtain a number of classification results, and deleting some of the classification results based on the common features, includes: Using a classification model to classify the monomer features, obtaining a number of first classification results; Extracting the position and posture information of the common features of each tooth based on the second CBCT image data, performing consistency evaluation on each of the position and posture information, and if the consistency evaluation result is higher than a preset judgment condition, determining that the common features are non-tooth trauma; The first classification results that are not dental trauma are determined as second classification results, and the second classification results in each of the first classification results are deleted.

[0061] In this embodiment, the classification model analyzes and judges the monomer features extracted from the monomer tooth images. The monomer features of each tooth contain information unique to the tooth, such as cracks on the tooth surface, the shape and position of the defect, etc. The classification model classifies these monomer features into different categories based on the pre-learned knowledge and patterns, thereby obtaining a series of first classification results. These results may include different types of tooth trauma, such as crown fracture, root fracture, etc., and may also include non-tooth trauma situations, that is, misidentified non-tooth trauma classifications.

[0062] Common features refer to some common features that all teeth have, such as the arrangement of teeth in the mouth, relative position and angle, etc. In this embodiment, common features refer to surface marks with specific shapes and distribution patterns on each tooth (or most teeth). These surface marks may be caused by trauma, that is, under the action of trauma, scars with similar features appear on the surface of multiple teeth, such as long strips of scratches formed by sharp metal on multiple teeth. The above situation belongs to dental trauma.

[0063] However, these surface marks may also be non-dental trauma. For example, teenagers may wear braces for a long time. The brackets and archwires of fixed braces are connected to the teeth. The enamel at the attachment of the brackets will be demineralized due to continuous external force and changes in the microenvironment in the oral cavity, which is manifested as local color changes and the formation of chalky plaques. During the wearing and removal of removable braces, they will frequently rub against the teeth, which may cause fine scratches on the tooth surface. In addition, wearing braces will also change the difficulty of cleaning in the mouth, causing food residues and bacteria to accumulate more easily in the contact area between the braces and the teeth, accelerating the demineralization of the tooth surface, and forming obvious marks over time. Such marks will be clearly presented in the CBCT image, causing the classification model to misclassify them as the corresponding type of dental trauma, such as surface enamel damage. Obviously, these surface marks belong to non-dental trauma, and the corresponding dental trauma classification results are wrong.

[0064] In order to identify whether there are surface marks caused by wearing braces in the above classification results, for example, the present invention projects these common features into the second CBCT image data, and then the position and posture of these common features can be determined, where the position is the distribution area of ​​these surface marks on each tooth, and the posture is the orientation of these surface marks on each tooth. If the above common features of all teeth (i.e., the position and posture of the surface marks) have a high degree of consistency, it means that the distribution of these surface marks on the patient's teeth is highly consistent. At this time, it is determined to be a surface mark caused by wearing braces, that is, it is determined not to be a case of dental trauma, and the corresponding classification result is wrong and is deleted. On the contrary, if the above common features of all teeth, i.e., the position and posture of the surface marks, have a low consistency, it means that these surface marks are not consistent, and it is impossible that they are surface marks caused by wearing braces. They are determined to be a case of dental trauma, and the corresponding classification result is retained.

[0065] With such arrangement, the present invention realizes a more accurate classification of dental trauma, and a low proportion of erroneous classification results can significantly reduce the ineffective workload of doctors, which is more practical when classifying dental trauma for batches of patients.

[0066] It is understandable that the consistency evaluation result may be a consistency evaluation value, and a higher consistency evaluation value indicates a higher consistency of the common features, and vice versa. Correspondingly, the preset determination condition is, for example, that the consistency evaluation value is higher than the evaluation threshold.

[0067] The embodiment of the present invention also discloses a dental trauma imaging diagnosis system based on deep learning, such as Figure 4 As shown, the system includes a training module, a segmentation module, a classification module, and an output module; The training module is used to perform data enhancement on the collected first CBCT image data to obtain training data with a higher degree of style diversity, and use the training data to train a segmentation model and a classification model based on Swin Transformer; The segmentation module is used to perform tooth image segmentation processing on the second CBCT image data of the target patient using the trained segmentation model to obtain multiple single tooth images, and extract single features and common features of each tooth from each single tooth image; The classification module is used to classify the monomer features using a classification model to obtain a number of classification results, and delete some of the classification results based on the common features; The output module is used to generate a corresponding treatment plan based on each of the classification results remaining after deletion.

[0068] Optionally, the training module is used to: Performing Fourier transform processing on the original first CBCT image data to obtain frequency domain image data; Selecting noise data from other original first CBCT image data and converting it into frequency domain to obtain frequency domain noise data, and superimposing the frequency domain noise data with the frequency domain image data to obtain new frequency domain image data; Perform inverse Fourier transform on each new frequency domain image data to obtain a plurality of third CBCT image data, and construct a plurality of training data based on each of the first CBCT image data, the third CBCT image data and the corresponding category labels.

[0069] Optionally, the training module is used to: Determine a low-frequency part and a high-frequency part of the frequency-domain image data, and for the low-frequency part and the high-frequency part, respectively use a first adding ratio and a second adding ratio to weaken the frequency-domain noise data; wherein the first adding ratio is smaller than the second adding ratio; The weakened frequency domain noise data is superimposed on the frequency domain image data to obtain new frequency domain image data.

[0070] Optionally, the training module is further used to: before superimposing the frequency domain noise data with the frequency domain image data: Determine the overall similarity between the first CBCT image data and other first CBCT image data, and obtain the enhancement quantity according to the overall similarity matching; wherein the enhancement quantity is positively correlated with the overall similarity; A plurality of groups of additive ratios corresponding to the enhancement amounts are generated for the first CBCT image data, and each group of the additive ratios includes a first additive ratio and a second additive ratio.

[0071] Optionally, the classification module is used to: Using a classification model to classify the monomer features, obtaining a number of first classification results; Extracting the position and posture information of the common features of each tooth based on the second CBCT image data, performing consistency evaluation on each of the position and posture information, and if the consistency evaluation result is higher than a preset judgment condition, determining that the common features are non-tooth trauma; The first classification results that are not dental trauma are determined as second classification results, and the second classification results in each of the first classification results are deleted.

[0072] An embodiment of the present invention also discloses an electronic device, comprising one or more processors, a memory, and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, and the instructions are executed by the processor to implement the method described in any of the preceding items.

[0073] An embodiment of the present invention further discloses a computer storage medium, comprising computer instructions that can be executed on an electronic device, wherein the computer instructions implement any of the above methods when executed.

[0074] An embodiment of the present invention further discloses a computer program product, which can be run on an electronic device to implement any of the methods described above.

[0075] Each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0076] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0077] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A dental trauma imaging diagnosis method based on deep learning, characterized by: The method comprises the following steps: Performing data enhancement on the collected first CBCT image data to obtain training data with a higher degree of style diversity, and using the training data to train a segmentation model and a classification model based on Swin Transformer; Using the trained segmentation model to perform tooth image segmentation processing on the second CBCT image data of the target patient to obtain a plurality of single tooth images, and extracting single features and common features of each tooth from each of the single tooth images; Using a classification model to classify the individual features to obtain a number of classification results, and deleting some of the classification results based on the common features; Based on the classification results remaining after the deletion, a corresponding treatment plan is generated.

2. The method for dental trauma imaging diagnosis based on deep learning according to claim 1, characterized in that: Data enhancement is performed on the collected first CBCT image data to obtain training data with a higher degree of style diversity, including: Performing Fourier transform processing on the original first CBCT image data to obtain frequency domain image data; Selecting noise data from other original first CBCT image data and converting it into frequency domain to obtain frequency domain noise data, and superimposing the frequency domain noise data with the frequency domain image data to obtain new frequency domain image data; Perform inverse Fourier transform on each new frequency domain image data to obtain a plurality of third CBCT image data, and construct a plurality of training data based on each of the first CBCT image data, the third CBCT image data and the corresponding category labels.

3. The method for dental trauma imaging diagnosis based on deep learning according to claim 2, characterized in that: The frequency domain noise data is superimposed on the frequency domain image data to obtain new frequency domain image data, including: Determine a low-frequency part and a high-frequency part of the frequency-domain image data, and use a first adding ratio and a second adding ratio to weaken the frequency-domain noise data for the low-frequency part and the high-frequency part, respectively; wherein the first adding ratio is smaller than the second adding ratio; The weakened frequency domain noise data is superimposed on the frequency domain image data to obtain new frequency domain image data.

4. The method for dental trauma imaging diagnosis based on deep learning according to claim 3, characterized in that: Before superimposing the frequency domain noise data with the frequency domain image data, the method further includes: Determine the overall similarity between the first CBCT image data and other first CBCT image data, and obtain the enhancement quantity according to the overall similarity matching; wherein the enhancement quantity is positively correlated with the overall similarity; A plurality of groups of additive ratios corresponding to the enhancement amounts are generated for the first CBCT image data, and each group of the additive ratios includes a first additive ratio and a second additive ratio.

5. The method for dental trauma imaging diagnosis based on deep learning according to claim 1, characterized in that: Using a classification model to classify the individual features to obtain a number of classification results, and deleting some of the classification results based on the common features, including: Using a classification model to classify the monomer features, obtaining a number of first classification results; Extracting the position and posture information of the common features of each tooth based on the second CBCT image data, performing consistency evaluation on each of the position and posture information, and if the consistency evaluation result is higher than a preset judgment condition, determining that the common features are non-tooth trauma; The first classification results that are not dental trauma are determined as second classification results, and the second classification results in each of the first classification results are deleted.

6. A dental trauma imaging diagnosis system based on deep learning, characterized by: The system includes a training module, a segmentation module, a classification module, and an output module; The training module is used to perform data enhancement on the collected first CBCT image data to obtain training data with a higher degree of style diversity, and use the training data to train a segmentation model and a classification model based on Swin Transformer; The segmentation module is used to perform tooth image segmentation processing on the second CBCT image data of the target patient using the trained segmentation model to obtain multiple single tooth images, and extract single features and common features of each tooth from each single tooth image; The classification module is used to classify the monomer features using a classification model to obtain a number of classification results, and delete some of the classification results based on the common features; The output module is used to generate a corresponding treatment plan based on each of the classification results remaining after deletion.

7. The deep learning-based dental trauma imaging diagnostic system according to claim 6, characterized in that: The training module is used to: Performing Fourier transform processing on the original first CBCT image data to obtain frequency domain image data; Selecting noise data from other original first CBCT image data and converting it into frequency domain to obtain frequency domain noise data, and superimposing the frequency domain noise data with the frequency domain image data to obtain new frequency domain image data; Perform inverse Fourier transform on each new frequency domain image data to obtain a plurality of third CBCT image data, and construct a plurality of training data based on each of the first CBCT image data, the third CBCT image data and the corresponding category labels.

8. An electronic device, characterized in that: It includes one or more processors, a memory, and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions, and when the instructions are executed by the processor, the deep learning-based dental trauma imaging diagnosis method as described in any one of claims 1 to 5 is implemented.

9. A computer storage medium, characterized in that: It includes computer instructions that can be executed on an electronic device, and the computer instructions are executed to implement the deep learning-based dental trauma imaging diagnosis method as described in any one of claims 1 to 5.

10. A computer program product, characterized in that: The computer program product can be run on an electronic device to implement the deep learning-based dental trauma imaging diagnosis method as described in any one of claims 1 to 5.

Images