Automatic identification method, device and equipment for wisdom teeth of lower jaw, medium and product

Abstract

The invention discloses an automatic identification method, device and equipment for mandibular wisdom teeth, a medium and a product, and relates to the field of image processing.The method comprises the steps that an oral maxillofacial cone beam computed tomography image is obtained; segmenting and positioning the mandibular wisdom tooth region-of-interest in the oral and maxillofacial cone beam computed tomography image by adopting a pre-trained first deep learning model to obtain a mandibular wisdom tooth segmentation mask image, and cutting a mandibular wisdom tooth region-of-interest image from the oral and maxillofacial cone beam computed tomography image; adopting a pre-trained second deep learning model to perform three-dimensional segmentation on the mandibular wisdom tooth region-of-interest image to obtain a three-dimensional segmentation mask image; and according to the mandibular wisdom tooth region-of-interest image and the three-dimensional segmentation mask image, determining a mandibular wisdom tooth recognition result by adopting a pre-trained third deep learning model. The identification accuracy and the identification efficiency of the wisdom teeth of the lower jaw are improved.

Description

Technical Field

[0001] The present application relates to the field of image processing, and in particular to a method, device, equipment, medium and product for automatic recognition of mandibular wisdom teeth based on deep learning. Background Art

[0002] Wisdom teeth are also called third molars. Due to insufficient jaw bone space, wisdom teeth often cannot grow completely, resulting in impaction. There are many types of complications in mandibular wisdom tooth extraction surgery, and the incidence rate is high. Among them, nerve damage is one of the more serious complications, which may cause incomplete numbness of the lower lip and chin skin or abnormal sensations such as burning, tingling, and tingling, causing pain to the patient.

[0003] The most obvious risk factor for inferior alveolar nerve injury caused by mandibular wisdom tooth extraction is the proximity of the wisdom tooth root to the mandibular canal. When the wisdom tooth root apex is observed to be close to the mandibular canal on preoperative X-ray images, the possibility of exposure of the inferior alveolar nerve and vascular bundle during surgery increases, thereby increasing the risk of nerve injury. Other complications are closely related to the position of the mandibular wisdom tooth, the direction of impaction, the relationship with the adjacent teeth, and the morphology of the tooth root in the mandible. In order to avoid and reduce the risk of adverse postoperative complications, it is of great significance to accurately evaluate the position, morphology, and relationship between the mandibular wisdom tooth and the surrounding tissues before surgery.

[0004] Routine imaging examinations before wisdom tooth extraction include plain films and oral and maxillofacial cone beam computed tomography (CBCT). Plain film examinations are apical films or curved body films, which only provide two-dimensional planar information. For some cases, plain X-ray films can meet the needs of preoperative diagnosis. However, the crown and root morphology of mandibular impacted wisdom teeth varies greatly, the impaction position and direction are different, the relationship between the root and the mandibular canal is complex, and the two-dimensional examination method has image overlap and deformation, which often makes it impossible to accurately judge the spatial position of wisdom teeth and their relationship with surrounding anatomical structures.

[0005] CBCT can effectively avoid image overlap, provide comprehensive and accurate information about teeth, jaws, and mandibular canals, and help physicians clarify the morphology, position, eruption direction, and relationship with surrounding tissues of impacted wisdom teeth, so as to carry out comprehensive preoperative planning and risk assessment. However, CBCT also has a clinical burden caused by excessive data volume. The traditional diagnostic method is for doctors to browse CT images layer by layer, perform some image post-processing, and make an imaging diagnosis based on clinical information. A single CBCT scan has hundreds of images, and it is necessary to browse images at multiple levels when reading the films. Browsing and diagnosis are time-consuming. Most dentists have not received professional craniomaxillofacial imaging diagnosis training, and are not familiar with the anatomy and imaging diagnosis outside of the direct area of ​​interest - teeth, which poses a risk of misdiagnosis and missed diagnosis.

[0006] In the current medical environment, only the imaging departments of large dental hospitals have sufficient CBCT reading experience, but completing more than 100 CBCT imaging reports every day also requires a lot of time and energy from multiple doctors, and medical resources are seriously lacking. Therefore, the development of new intelligent medical technologies and the use of more advanced technical methods to assist dentists in efficient and accurate CBCT imaging diagnosis have become important medical needs. Summary of the invention

[0007] The purpose of this application is to provide a method, device, equipment, medium and product for automatic recognition of mandibular wisdom teeth, which can improve the recognition accuracy and efficiency of mandibular wisdom teeth.

[0008] To achieve the above objectives, this application provides the following solutions:

[0009] In a first aspect, the present application provides a method for automatically identifying mandibular wisdom teeth, comprising:

[0010] Oral and maxillofacial cone-beam computed tomography images were obtained;

[0011] Using a pre-trained first deep learning model, segment and locate the mandibular wisdom tooth interest region in the oral and maxillofacial cone beam computed tomography image to obtain a mandibular wisdom tooth segmentation mask image; the mandibular wisdom tooth segmentation mask image includes wisdom teeth and second molars;

[0012] According to the mandibular wisdom tooth segmentation mask image, cutting out a mandibular wisdom tooth region of interest image from the oral and maxillofacial cone beam computed tomography image;

[0013] Using a pre-trained second deep learning model, three-dimensionally segment the mandibular wisdom tooth region of interest image to obtain a three-dimensional segmentation mask image; the three-dimensional segmentation mask image includes wisdom teeth, second molars, mandibles, and mandibular canals;

[0014] According to the mandibular wisdom tooth area of ​​interest image and the three-dimensional segmentation mask image, a pre-trained third deep learning model is used to obtain the mandibular wisdom tooth recognition result; the mandibular wisdom tooth recognition result includes the impaction position, impaction direction, eruption depth, relationship with the second molar, relationship with the mandibular ascending ramus, root morphology, number of roots, relationship with the mandibular canal and buccal and lingual bone plate thickness.

[0015] In a second aspect, the present application provides a mandibular wisdom tooth automatic identification device, comprising:

[0016] An image acquisition module, used for acquiring oral and maxillofacial cone beam computed tomography images;

[0017] A first segmentation module is used to segment and locate the mandibular wisdom tooth interest region in the oral and maxillofacial cone beam computed tomography image using a pre-trained first deep learning model to obtain a mandibular wisdom tooth segmentation mask image; the mandibular wisdom tooth segmentation mask image includes wisdom teeth and second molars;

[0018] A cutting module, used for cutting out an image of a mandibular wisdom tooth region of interest from the oral and maxillofacial cone-beam computed tomography image according to the mandibular wisdom tooth segmentation mask image;

[0019] A second segmentation module is used to perform three-dimensional segmentation on the mandibular wisdom tooth region of interest image using a pre-trained second deep learning model to obtain a three-dimensional segmentation mask image; the three-dimensional segmentation mask image includes wisdom teeth, second molars, mandibles, and mandibular canals;

[0020] The image recognition module is used to obtain the mandibular wisdom tooth recognition result according to the mandibular wisdom tooth area of ​​interest image and the three-dimensional segmentation mask image by using a pre-trained third deep learning model; the mandibular wisdom tooth recognition result includes the impaction position, impaction direction, eruption depth, relationship with the second molar, relationship with the mandibular ascending ramus, root morphology, number of roots, relationship with the mandibular canal and buccal and lingual bone plate thickness.

[0021] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for automatic identification of mandibular wisdom teeth.

[0022] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for automatic identification of mandibular wisdom teeth.

[0023] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned method for automatic identification of mandibular wisdom teeth.

[0024] According to the specific embodiments provided in this application, this application has the following technical effects:

[0025] The present application provides a method, device, equipment, medium and product for automatic recognition of mandibular wisdom teeth. Through a pre-trained deep learning model, oral and maxillofacial cone beam computed tomography images are automatically segmented and located, and mandibular wisdom teeth are recognized. The method can quickly and accurately segment wisdom teeth and surrounding tissue structures, and accurately determine the position, shape and relationship of mandibular impacted wisdom teeth with surrounding tissues, thereby improving the recognition accuracy and efficiency of mandibular wisdom teeth. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 This is a diagram of an application environment of a method for automatically identifying mandibular wisdom teeth in one embodiment of the present application;

[0028] Figure 2 A schematic diagram of a flow chart of a method for automatically identifying mandibular wisdom teeth provided in one embodiment of the present application;

[0029] Figure 3 This is a schematic diagram of the processing process of the first deep learning model in one embodiment of the present application;

[0030] Figure 4 The image of the mandibular wisdom tooth interest area cut out at the lower left position in one embodiment of the present application;

[0031] Figure 5 The image of the mandibular wisdom tooth interest area cut out at the lower right position in one embodiment of the present application;

[0032] Figure 6 This is a schematic diagram of the processing process of the second deep learning model in one embodiment of the present application;

[0033] Figure 7 Schematic diagram of the processing process of the third deep learning model in one embodiment of the present application

[0034] Figure 8 A schematic diagram of functional modules of a mandibular wisdom tooth automatic identification device provided in one embodiment of the present application;

[0035] Fig. 9 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0037] First, some technical terms involved in the embodiments of the present application are introduced.

[0038] Deep learning: Deep learning is a subfield of machine learning based on learning algorithms of artificial neural networks, especially with multiple layers of nonlinear transformations (i.e., deep). These networks are able to learn multi-level representations of data, thus excelling in areas such as image and speech recognition, natural language processing, etc.

[0039] Convolutional neural networks (CNN): CNN is a deep learning architecture that is particularly suitable for processing data with grid-like topological structures, such as images. CNN can automatically and layer-by-layer extract features through structures such as convolutional layers, pooling layers, and fully connected layers, without the need for manual feature design, so it is very popular in image recognition and classification tasks.

[0040] Artificial intelligence (AI): AI is a branch of computer science that aims to create machines or software that can perform tasks that are usually performed by humans. These tasks include visual recognition, language understanding, decision-making, and translation. The goal of AI is to enable computers to simulate human cognitive functions.

[0041] CBCT: CBCT is a medical imaging technique that uses a cone-shaped X-ray beam that rotates around the patient to acquire three-dimensional volumetric data. This technique is commonly used in dental and maxillofacial imaging, and it provides high-quality three-dimensional images to aid doctors in diagnosis and treatment planning.

[0042] Impacted mandibular third molar (IMTM): IMTM refers to a condition in which the mandibular third molar (wisdom tooth) fails to erupt normally and is partially or completely covered by the gums or jawbone. This may cause pain, infection, or damage to adjacent teeth, and sometimes requires surgical removal.

[0043] Mandibular second molar: The mandibular second molar is the second molar located on the mandible, between the first molar and the third molar (wisdom tooth), and it is one of the main teeth in the chewing process.

[0044] Mandibular Canal: The mandibular canal is a canal located within the mandible that contains the inferior alveolar nerve and blood vessels. This canal runs from the back to the front of the mandible and provides sensation and nutrition to the mandibular teeth and jaw.

[0045] The inferior alveolar nerve (IAN) is a branch of the trigeminal nerve and is responsible for the sensation of the mandibular teeth, gums, and lower lip. It transmits sensory signals through the mandibular canal and is a nerve that requires special attention during dental surgery to avoid damage that may cause loss of sensation or pain.

[0046] The practical scenarios of wisdom tooth extraction surgery need to consider the following aspects:

[0047] (1) Preoperative evaluation: Before performing wisdom tooth extraction surgery, the doctor will use X-ray imaging and other means to evaluate the position, shape, and relationship of the wisdom tooth with surrounding tissues.

[0048] (2) Surgical risks: Mandibular wisdom tooth extraction surgery may be accompanied by a variety of complications, including nerve damage, adjacent tooth damage, alveolar bone fracture, and root displacement.

[0049] (3) Nerve damage: Damage to the inferior alveolar nerve may cause abnormal sensations such as numbness, burning, tingling, or tingling sensation in the skin of the lower lip and chin.

[0050] (4) Risk factors: The proximity of the wisdom tooth root to the mandibular canal is an obvious risk factor for nerve injury. Preoperative X-ray images show that when the wisdom tooth root apex is close to the mandibular canal, the risk of nerve injury during surgery increases.

[0051] The main purpose of this application is to use artificial intelligence technology to automate and optimize the recognition process of wisdom teeth in CBCT images, improve the accuracy and efficiency of wisdom tooth recognition results, and provide more precise guidance for subsequent clinical treatment.

[0052] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0053] The method for automatically identifying mandibular wisdom teeth provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, or it can be integrated on the server 104, or it can be placed on the cloud or other servers. The terminal 102 can send the oral and maxillofacial cone beam computerized tomography image to be processed to the server 104. After the server 104 receives the oral and maxillofacial cone beam computerized tomography image to be processed, it uses a deep learning model to automatically identify the mandibular wisdom teeth. The server 104 can feedback the mandibular wisdom tooth recognition result to the terminal 102. In addition, in some embodiments, the mandibular wisdom tooth automatic recognition method can also be implemented by the server 104 or the terminal 102 alone.

[0054] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0055] In an exemplary embodiment, Figure 2 As shown, a method for automatically identifying mandibular wisdom teeth is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, and the steps include the following steps 201 to 205.

[0056] Step 201, obtaining oral and maxillofacial cone beam computed tomography images.

[0057] Step 202, using a pre-trained first deep learning model, segment and locate the mandibular wisdom tooth interest region in the oral and maxillofacial cone-beam computed tomography image to obtain a mandibular wisdom tooth segmentation mask image.

[0058] The mandibular wisdom tooth segmentation mask image includes wisdom teeth and second molars.

[0059] In an exemplary embodiment, the first deep learning model is a 3D U-Net model or any other deep learning model with similar functions, such as a variant of the U-Net model including U-Net++, a DeepLab series, a Transformer model, and the like.

[0060] In order to ensure the accuracy and robustness of mandibular wisdom tooth recognition, this application uses a deep learning model to segment and locate the mandibular wisdom tooth interest area. Figure 3 As shown, the 3D U-Net model is used as the network structure of the deep learning model. After the model is trained, the CBCT image is input into the trained 3D U-Net model to obtain the mandibular wisdom tooth segmentation mask image. Since the diagnosis and classification of wisdom teeth involve the second molars, the second molars are also segmented together with the wisdom teeth.

[0061] Step 203 , based on the mandibular wisdom tooth segmentation mask image, a mandibular wisdom tooth region of interest image is cut out from the oral and maxillofacial cone-beam computed tomography image.

[0062] Specifically, a three-dimensional bounding box is generated based on the mandibular wisdom tooth segmentation mask image. The oral and maxillofacial cone beam computed tomography image is cropped based on the three-dimensional bounding box to obtain an image of the mandibular wisdom tooth region of interest. In a specific example, Figure 4 and Figure 5 As shown, two mandibular wisdom tooth region of interest images are obtained, namely, the mandibular wisdom tooth region of interest image at the lower left position and the mandibular wisdom tooth region of interest image at the lower right position.

[0063] Step 204: Use a pre-trained second deep learning model to perform three-dimensional segmentation on the mandibular wisdom tooth region of interest image to obtain a three-dimensional segmentation mask image.

[0064] The three-dimensional segmentation mask image includes wisdom teeth, second molars, mandibles and mandibular canals.

[0065] In an exemplary embodiment, the second deep learning model is a 3D U-Net model or any other deep learning model with similar functions, such as a variant of the U-Net model including U-Net++, a DeepLab series, a Transformer model, and the like.

[0066] Specifically, Figure 6 As shown in the figure, a 3D U-Net model is used as the network structure of the deep learning model. After the model is trained, the image of the mandibular wisdom tooth interest area is input into the trained 3D U-Net model to obtain a 3D segmentation mask image. Among them, the 3D U-Net model outputs multiple horizontal images, based on which the 3D images of the wisdom teeth, second molars, and mandible can be reconstructed, that is, the 3D segmentation mask image.

[0067] Step 205, according to the mandibular wisdom tooth area of ​​interest image and the three-dimensional segmentation mask image, a pre-trained third deep learning model is used to obtain a mandibular wisdom tooth recognition result.

[0068] The mandibular wisdom tooth identification results include the impaction position, impaction direction, eruption depth, relationship with the second molar, relationship with the mandibular ascending ramus, root morphology, number of roots, relationship with the mandibular canal, and buccal and lingual bone plate thickness.

[0069] Since there are many types of wisdom teeth and they involve other tissue structures, this application designs multiple classification networks based on deep learning to ensure the accuracy and robustness of the recognition results. The recognition of wisdom teeth is mainly based on the morphology of wisdom teeth and the spatial position relationship with adjacent structures. Based on this prior knowledge, this application uses the three-dimensional segmentation mask image as the second channel of the input image to improve the accuracy of wisdom tooth recognition results.

[0070] In an exemplary embodiment, Figure 7As shown, the third deep learning model includes a first classification network and a second classification network.

[0071] According to the mandibular wisdom tooth interest region image and the three-dimensional segmentation mask image, a pre-trained first classification network is used to determine the impaction position, impaction direction, eruption depth, relationship with the second molar, and relationship with the mandibular ascending ramus.

[0072] The input image of the first classification network is a dual-channel image consisting of a mandibular wisdom tooth region of interest image and a three-dimensional segmentation mask image, wherein the three-dimensional segmentation mask image includes wisdom teeth, second molars and mandibles.

[0073] According to the mandibular wisdom tooth region of interest image and the three-dimensional segmentation mask image, a pre-trained second classification network is used to determine the root morphology, the number of roots, the relationship with the mandibular canal, and the thickness of the buccal and lingual bone plates.

[0074] The input image of the second classification network is a dual-channel image consisting of a mandibular wisdom tooth region of interest image and a three-dimensional segmentation mask image, wherein the three-dimensional segmentation mask image includes the wisdom tooth and the mandibular canal.

[0075] In a specific example, the first classification network and the second classification network are both three-dimensional densely connected networks (3D DenseNet) or any other deep learning models with similar functions, such as ResNet, EfficientNet, Transformer model, etc.

[0076] The present application can further perform three-dimensional reconstruction and display of the lesion or anatomical structure based on the mandibular wisdom tooth segmentation mask image and the three-dimensional segmentation mask image, and score the difficulty of wisdom tooth extraction based on the mandibular wisdom tooth recognition result.

[0077] The training process of the three deep neural models used in this application includes the following steps 301 to 303.

[0078] Step 301, retrospectively collect image data of 1600 CBCT scans of mandibular wisdom teeth of patients who underwent CBCT examinations in dental hospitals or dental clinics within a set period of time in the past, and include them in the database.

[0079] The CBCT brands and models included in the database include:

[0080] ①3D Accuitomo (JMoritaMfg., Japan).

[0081] ②NewTomVGi, NewTom VG (Quantitative Radiology, Italy).

[0082] ③Carestream 9300 (Carestream Dental LLC, Atlanta, USA).

[0083] ④PlanmecaProMax 3D (Planmeca, Helsinki, Finland).

[0084] ⑤LargeV Smart 3D, LargeV HiRes 3D, LargeV Smart3DX (Beijing Longshi Instrument Co., Ltd., China).

[0085] ⑥RAYSCANN AlphaPlus (Ray Co., Ltd., Hwaseong-si, South Korea).

[0086] ⑦PointNix 8003D Plus (PointNix Co., Ltd, Seoul, South Korea).

[0087] ⑧MEYERmDX-13SDKL1A (Hefei Meiya Optoelectronics Technology Co., Ltd., Hefei, China).

[0088] ⑨Implagraphy (Vatech, Hwaseong-si, South Korea).

[0089] ⑩DENTRI-C (HDXWILL, Seoul, South Korea).

[0090] - CAT FLX (Imaging Sciences International, USA).

[0091] Included in the scanning field of view: 6cm×6cm, 8cm×8cm, 9cm×9cm, 10cm×10cm, 12cm×8cm, 15cm×12cm, 16cm×5cm, 16cm×6cm, 16cm×7cm, 16cm×8cm, 16cm×9cm, 16cm×10cm, 16cm×13cm, 16cm×15cm, 16cm×16cm, 17cm×13cm.

[0092] Inclusion criteria: (1) clear images without motion artifacts or hardening artifacts that affect the observation of wisdom teeth; (2) complete scans of mandibular wisdom teeth, adjacent second molars, and the mandibular canal in the corresponding area; (3) no history of mandibular trauma, surgery, tumors, or systemic diseases.

[0093] Exclusion criteria: (1) the presence of artifacts that affect observation; (2) incomplete display of mandibular impacted wisdom teeth, second molars, and mandibular canals; (3) incomplete or missing mandibular impacted wisdom teeth and second molars; (4) the presence of space-occupying diseases, fractures, osteomyelitis, or post-orthognathic surgery in the segmented area.

[0094] Step 302, randomly divide the database into training set, validation set, and test set according to the ratio of 70%, 10%, and 20%, with 1120 cases of data as training samples, 160 cases as validation samples, and 320 cases as test samples; 100 cases are randomly selected from the test samples as the clinical validation test set for comparison of results with radiologists and clinical physicians.

[0095] (1) Annotation of training samples for segmentation of wisdom teeth and surrounding structures: First, a dental imaging specialist used ITK-SNAP 3.2 software to delineate the mandibular wisdom teeth, mandibular second molars, and mandibular canals in the CBCT images. Then, another senior dental imaging specialist modified and confirmed the delineated images.

[0096] (2) Annotation of the test samples for segmentation of wisdom teeth and surrounding structures: First, a dental imaging specialist used ITK-SNAP 3.2 software to delineate the mandibular wisdom teeth, mandibular second molars, and mandibular canals in the CBCT images. Two other senior dental imaging specialists then modified and confirmed the delineated images. This served as the gold standard for image segmentation evaluation.

[0097] The above-annotated training samples and test samples are used as training samples and test samples for the first deep learning model and the second deep learning model.

[0098] (3) Annotation of wisdom tooth identification training samples: First, a dental imaging specialist used ITK-SNAP 3.2 software to annotate the mandibular wisdom teeth in the CBCT images. The annotation indicators included: impaction position, impaction direction, eruption depth, relationship with the second molar, mandibular ramus, and mandibular canal, root morphology, number of roots, and buccal and lingual bone plate thickness. Then, another senior dental imaging specialist reviewed and modified the annotation results.

[0099] (4) Annotation of wisdom tooth identification test samples: First, a dental imaging specialist used ITK-SNAP 3.2 software to annotate the mandibular wisdom teeth in the CBCT images. The annotated indicators included: impaction location, impaction direction, eruption depth, relationship with the second molar, mandibular ramus, and mandibular canal, root morphology, number of roots, and buccal and lingual bone plate thickness. Two other senior dental imaging specialists then reviewed and modified the annotated results. This was used as the gold standard for image segmentation evaluation.

[0100] The above-annotated training samples and test samples are used as training samples and test samples for the third deep learning model.

[0101] Step 303: deep learning model testing and comparison with radiologists and clinical physicians.

[0102] ① A deep learning model was used to automatically identify mandibular wisdom teeth in a total of 320 test samples, and the accuracy and time required for mandibular wisdom teeth identification were calculated.

[0103] ② Radiologists used ITK-SNAP to identify wisdom teeth on 100 test samples in the clinical validation test set, compared the radiologists' recognition results with the gold standard of the test data, and calculated the accuracy of classification diagnosis and the time required.

[0104] ③ Oral surgeons used ITK-SNAP to identify wisdom teeth on 100 test samples in the clinical validation test set. The identification results of oral surgeons were compared with the gold standard of test data, and the accuracy and time required for classification diagnosis were calculated.

[0105] ④ The recognition accuracy uses confusion matrix, accuracy, precision, sensitivity, recall rate, F1 score and AUC as evaluation indicators.

[0106] This application integrates the steps of automatic registration, automatic segmentation, and automatic identification of CBCT images to establish a complete set of AI-assisted automatic diagnosis software for mandibular impacted wisdom teeth. After rigorous data training, the deep learning model can quickly process a large number of CBCT images, helping clinicians to efficiently and accurately segment wisdom teeth and surrounding tissue structures, and accurately determine the position, shape, and relationship of mandibular impacted wisdom teeth with surrounding tissues, thereby greatly improving work efficiency, saving work time, and increasing the accuracy and consistency of disease diagnosis.

[0107] The method for automatic recognition of mandibular wisdom teeth provided in this application can be applied to the following scenarios:

[0108] (1) Dental imaging diagnosis: Deep learning technology is used to automatically segment and 3D reconstruct images of mandibular impacted wisdom teeth, second molars, mandibular canals, and mandibular bones in CBCT images to improve the accuracy and efficiency of diagnosis.

[0109] (2) Wisdom tooth extraction difficulty score: Based on the results of automatic diagnosis, such as the relationship between the wisdom tooth and surrounding structures, the shape and number of tooth roots, etc., the difficulty of wisdom tooth extraction is scored to help doctors assess the difficulty and risk of the operation.

[0110] (3) Oral and maxillofacial surgery planning: By automatically identifying the position, direction, eruption depth and other characteristics of mandibular impacted wisdom teeth, it assists doctors in making detailed surgical plans before surgery to reduce surgical risks, improve surgical success rates, and reduce complications such as nerve damage that may occur after wisdom tooth extraction.

[0111] (4) Education and training: In oral medicine education, it can be used as a teaching tool to help students and young doctors better master wisdom tooth extraction surgery.

[0112] (5) Patient communication and education: Through intuitive 3D reconstructed images, patients can better understand their condition and surgical procedures, thereby improving their understanding and acceptance of treatment.

[0113] In summary, this application has at least the following beneficial effects:

[0114] 1. Improved the recognition accuracy of mandibular wisdom teeth: Artificial intelligence technology based on deep learning can automatically analyze CBCT images, reduce missed diagnoses and misdiagnoses caused by human factors, and thus improve the diagnostic accuracy of mandibular impacted wisdom teeth.

[0115] 2. Improved recognition efficiency of mandibular wisdom teeth: The ability to automatically segment and reconstruct images in 3D enables doctors to quickly obtain the required information, significantly shortens the time for image analysis and diagnosis, and improves clinical work efficiency.

[0116] 3. Reduced workload of doctors: Automation reduces the workload of doctors in imaging diagnosis, allowing doctors to focus more on patient treatment and care.

[0117] 4. Optimized surgical planning: By providing detailed wisdom tooth position, direction, eruption depth and relationship with surrounding structures, it assists doctors in making more accurate surgical plans and reducing surgical risks.

[0118] 5. Reduced incidence of complications: Accurately identifying the relationship between wisdom teeth and important structures such as the mandibular canal can help avoid complications such as nerve damage during surgery and improve patient safety.

[0119] 6. Support personalized treatment: Through a comprehensive analysis of the characteristics of wisdom teeth, doctors can develop personalized treatment plans based on the patient's specific situation and improve treatment outcomes.

[0120] 7. Promote education and training: It can be used as a teaching tool to help medical students and young doctors better understand and master the relevant knowledge and skills of wisdom tooth extraction.

[0121] Based on the same inventive concept, the embodiment of the present application also provides a mandibular wisdom tooth automatic recognition device for implementing the mandibular wisdom tooth automatic recognition method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more mandibular wisdom tooth automatic recognition device embodiments provided below can refer to the limitations of the mandibular wisdom tooth automatic recognition method above, and will not be repeated here.

[0122] In an exemplary embodiment, Figure 8 As shown, a mandibular wisdom tooth automatic recognition device is provided, including: an image acquisition module 801, a first segmentation module 802, a cutting module 803, a second segmentation module 804 and an image recognition module 805.

[0123] The image acquisition module 801 is used to acquire oral and maxillofacial cone-beam computed tomography images.

[0124] The first segmentation module 802 is used to use a pre-trained first deep learning model to segment and locate the mandibular wisdom tooth interest area in the oral and maxillofacial cone-beam computed tomography image to obtain a mandibular wisdom tooth segmentation mask image; the mandibular wisdom tooth segmentation mask image includes wisdom teeth and second molars.

[0125] The cropping module 803 is used to crop the mandibular wisdom tooth region of interest image from the oral and maxillofacial cone-beam computed tomography image according to the mandibular wisdom tooth segmentation mask image.

[0126] The second segmentation module 804 is used to perform three-dimensional segmentation on the mandibular wisdom tooth region of interest image using a pre-trained second deep learning model to obtain a three-dimensional segmentation mask image. The three-dimensional segmentation mask image includes wisdom teeth, second molars, mandibles and mandibular canals.

[0127] The image recognition module 805 is used to obtain the mandibular wisdom tooth recognition result according to the mandibular wisdom tooth region of interest image and the three-dimensional segmentation mask image using a pre-trained third deep learning model. The mandibular wisdom tooth recognition result includes the impaction position, impaction direction, eruption depth, relationship with the second molar, relationship with the mandibular ascending ramus, root morphology, number of roots, relationship with the mandibular canal, and buccal and lingual bone plate thickness.

[0128] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig. 9As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store oral and maxillofacial cone beam computed tomography images. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for automatically identifying mandibular wisdom teeth is implemented.

[0129] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0130] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0131] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0133] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0134] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0135] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0136] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0137] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for automatically identifying mandibular wisdom teeth, characterized in that: The method for automatically identifying mandibular wisdom teeth comprises: Oral and maxillofacial cone-beam computed tomography images were obtained; Using a pre-trained first deep learning model, segment and locate the mandibular wisdom tooth interest region in the oral and maxillofacial cone beam computed tomography image to obtain a mandibular wisdom tooth segmentation mask image; the mandibular wisdom tooth segmentation mask image includes wisdom teeth and second molars; According to the mandibular wisdom tooth segmentation mask image, cutting out a mandibular wisdom tooth region of interest image from the oral and maxillofacial cone beam computed tomography image; Using a pre-trained second deep learning model, three-dimensionally segment the mandibular wisdom tooth region of interest image to obtain a three-dimensional segmentation mask image; the three-dimensional segmentation mask image includes wisdom teeth, second molars, mandibles, and mandibular canals; According to the mandibular wisdom tooth area of ​​interest image and the three-dimensional segmentation mask image, a pre-trained third deep learning model is used to obtain the mandibular wisdom tooth recognition result; the mandibular wisdom tooth recognition result includes the impaction position, impaction direction, eruption depth, relationship with the second molar, relationship with the mandibular ascending ramus, root morphology, number of roots, relationship with the mandibular canal and buccal and lingual bone plate thickness.

2. The method for automatically identifying mandibular wisdom teeth according to claim 1, characterized in that: The first deep learning model is a 3D U-Net model.

3. The method for automatically identifying mandibular wisdom teeth according to claim 1, characterized in that: According to the mandibular wisdom tooth segmentation mask image, cutting out a mandibular wisdom tooth region of interest image from the oral and maxillofacial cone beam computed tomography image specifically includes: generating a three-dimensional bounding box based on the mandibular wisdom tooth segmentation mask image; The oral and maxillofacial cone-beam computed tomography image is cropped based on the three-dimensional bounding box to obtain an image of the mandibular wisdom tooth interest area.

4. The method for automatically identifying mandibular wisdom teeth according to claim 1, characterized in that: The second deep learning model is a 3D U-Net model.

5. The method for automatically identifying mandibular wisdom teeth according to claim 1, characterized in that: The third deep learning model includes a first classification network and a second classification network; According to the mandibular wisdom tooth interest area image and the three-dimensional segmentation mask image, a pre-trained third deep learning model is used to obtain a mandibular wisdom tooth recognition result, which specifically includes: According to the mandibular wisdom tooth interest region image and the three-dimensional segmentation mask image, a pre-trained first classification network is used to determine the impaction position, impaction direction, eruption depth, relationship with the second molar, and relationship with the mandibular ascending ramus; According to the mandibular wisdom tooth region of interest image and the three-dimensional segmentation mask image, a pre-trained second classification network is used to determine the root morphology, the number of roots, the relationship with the mandibular canal, and the thickness of the buccal and lingual bone plates.

6. The method for automatically identifying mandibular wisdom teeth according to claim 5, characterized in that: The first classification network and the second classification network are both three-dimensional densely connected networks.

7. A mandibular wisdom tooth automatic recognition device, applied to the mandibular wisdom tooth automatic recognition method according to any one of claims 1 to 6, characterized in that: The mandibular wisdom tooth automatic identification device comprises: An image acquisition module, used for acquiring oral and maxillofacial cone beam computed tomography images; A first segmentation module is used to segment and locate the mandibular wisdom tooth interest region in the oral and maxillofacial cone beam computed tomography image using a pre-trained first deep learning model to obtain a mandibular wisdom tooth segmentation mask image; the mandibular wisdom tooth segmentation mask image includes wisdom teeth and second molars; A cutting module, used for cutting out an image of a mandibular wisdom tooth region of interest from the oral and maxillofacial cone-beam computed tomography image according to the mandibular wisdom tooth segmentation mask image; A second segmentation module is used to perform three-dimensional segmentation on the mandibular wisdom tooth region of interest image using a pre-trained second deep learning model to obtain a three-dimensional segmentation mask image; the three-dimensional segmentation mask image includes wisdom teeth, second molars, mandibles, and mandibular canals; The image recognition module is used to obtain the mandibular wisdom tooth recognition result according to the mandibular wisdom tooth area of ​​interest image and the three-dimensional segmentation mask image by using a pre-trained third deep learning model; the mandibular wisdom tooth recognition result includes the impaction position, impaction direction, eruption depth, relationship with the second molar, relationship with the mandibular ascending ramus, root morphology, number of roots, relationship with the mandibular canal and buccal and lingual bone plate thickness.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for automatic identification of mandibular wisdom teeth according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for automatically identifying mandibular wisdom teeth described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for automatically identifying mandibular wisdom teeth described in any one of claims 1 to 6 is implemented.

Images