Model training, data processing and oral cavity detection methods and devices, and electronic equipment

Abstract

The invention provides a model training method and device, a data processing method and device, an oral cavity detection method and device and electronic equipment, and relates to the technical field of big data and oral cavity medical treatment. The model training method disclosed by the invention comprises the following steps: acquiring a detection sample in oral medical diagnosis and treatment, wherein the detection sample comprises detection data and a label; enhancing the detection sample to obtain a training sample; the training sample is processed through a self-supervised learning model until training is completed, and parameters of the self-supervised learning model are adjusted according to the output diagnosis result and the label.

Description

Technical Field

[0001] The present disclosure relates to the field of big data and oral medical technology, and in particular to a method, device and electronic equipment for model training, data processing and oral detection. Background Art

[0002] The mandible is a free bone, connected to the skull only by the temporomandibular joint. Its position is regulated and restricted by the temporomandibular joint, occlusal contact, jaw muscles, and the central nervous system. It is the only bone in the maxillofacial region that can move. Its movement is coordinated by the various components of the stomatognathic system. Through the contact between the cusps and fossa of the upper and lower teeth, and under the repeated regulation of proprioception at the periodontal, temporomandibular joint, and masticatory muscles, it is memorized by the nerves and muscles, and gradually forms a movement form with individual characteristics. Whether this movement form is normal is one of the important contents in evaluating the function of the stomatognathic system.

[0003] Mandibular movements are often simplified into opening and closing, lateral movement, and forward and backward movement. Tapping movement is a small unconscious habitual opening and closing movement, which is related to the repeated reinforcement of neuromuscular memory in the mandibular movement of the oral and maxillary system. The frequency, stability, and speed of the tapping movement reflect the coordination between the various components of the system and are the key factors in determining whether the mandibular movement is normal or not. One of the indicators of functional stability.

[0004] When factors affecting jaw movement, such as the temporomandibular joint and Any abnormal changes in the (teeth) will affect the mandibular movement, thus affecting the patient's oral health. Physicians with rich clinical experience will combine visual examination, palpation and auscultation to determine whether the mandibular movement is normal, so as to conduct targeted diagnosis and screening for temporomandibular joint disorders (such as clicking and murmurs) and tooth structure (such as early contact, interference) and other related diseases and their severity.

[0005] In addition, doctors may use diagnostic tools to contact the treatment site to assist in diagnosis and treatment, such as using metal instruments to percuss the root apex to determine whether and how severe the pain is, or using dental floss to pass through the adjacent area to check the tightness of the adjacent contact. Summary of the invention

[0006] One object of the present disclosure is to improve the efficiency and accuracy of oral medical treatment.

[0007] According to one aspect of some embodiments of the present disclosure, a model training method is proposed, including: obtaining test samples in oral medical diagnosis and treatment, the test samples including test data and labels; enhancing the test samples to obtain training samples; processing the training samples through a self-supervised learning model until the training is completed, wherein the parameters of the self-supervised learning model are adjusted according to the output diagnosis results and labels.

[0008] In some embodiments, the diagnosis result includes an oral health status, and the test sample includes a test sample under a healthy oral state and a test sample under at least one oral disease state.

[0009] In some embodiments, the diagnosis result includes the state of the restoration, and the test samples include test samples of the restoration in different states, wherein the restoration state includes at least one of the occlusal contact state of the restoration, the degree of implant bone integration or the state of the internal preparation.

[0010] In some embodiments, the oral health status includes health, or the diagnosis result of oral diseases, and oral diseases include temporomandibular joint disorder, tooth disease, occlusal disorder, periodontal tissue disease, restoration problems, and oral and maxillofacial tumors.

[0011] In some embodiments, the diagnosis result also includes at least one of the location, amount, and extent of the disease.

[0012] In some embodiments, the detection data includes one or more of optical data, sound data, terahertz data, or electromagnetic wave data.

[0013] In some embodiments, the optical data includes a three-dimensional image reconstructed based on the difference in reflection intensity after a sensor receives the optical signal by emitting a predetermined type of light to the soft and hard tissue surfaces of the oral cavity.

[0014] In some embodiments, the terahertz wave data includes transmission and reflection data obtained by transmitting and receiving electromagnetic waves to the soft and hard tissues of the oral cavity.

[0015] In some embodiments, the electromagnetic wave data includes data on the amount of X-rays absorbed by different parts of the oral tissue, obtained by emitting electromagnetic waves to the soft and hard tissues of the oral cavity.

[0016] In some embodiments, the sound data includes data generated by the patient's active sound source and data generated by a passive sound source, wherein the data generated by the active sound source includes the sound data generated by the mandibular movement process, and the data generated by the passive sound source includes the sound data generated by the contact between the diagnostic and treatment tools and the internal structure of the oral cavity.

[0017] In some embodiments, enhancing detection samples and obtaining training samples include: when the detection data includes sound data, processing the sound data through a filter to obtain a first spectrum map sample; processing the first spectrum map sample through at least one of waveform displacement, harmonic distortion or resampling to obtain a second spectrum map sample, wherein the training samples include the first spectrum map sample and the second spectrum map sample.

[0018] In some embodiments, the label corresponding to the second spectrogram sample is the same as the corresponding first spectrogram sample.

[0019] In some embodiments, processing the sound data through a filter to obtain a first spectrogram sample includes: processing the sound data through a Mel filter to obtain a Mel spectrogram as the first spectrogram sample.

[0020] In some embodiments, the first spectrum map sample is processed by at least one of waveform displacement, harmonic distortion or resampling to obtain the second spectrum map sample, which includes: extracting time domain features and frequency domain features based on the first spectrum map sample; obtaining time-frequency fusion features based on the time domain features and frequency domain features; extracting local features in the time-frequency fusion features, and obtaining dynamic position codes based on the time-frequency fusion features; extracting context-aware features based on the local features and dynamic position codes based on a self-attention mechanism; and fusing the context-aware features and the time-frequency fusion features to obtain the second spectrum map sample.

[0021] In some embodiments, the self-supervised learning model is a Vision Transformer model.

[0022] According to one aspect of some embodiments of the present disclosure, a method for processing oral diagnosis and treatment data is proposed, including: obtaining test data of a target patient during oral diagnosis and treatment; determining a diagnosis result based on a machine learning model according to the test data, wherein the machine learning model is generated by training according to any one of the model training methods mentioned above.

[0023] In some embodiments, the oral diagnosis and treatment data processing method further includes: when the detection data includes sound data, converting the sound data of the target patient in oral diagnosis and treatment into a sound spectrum graph; wherein the machine learning model determines the diagnosis result by processing the sound spectrum graph.

[0024] According to one aspect of some embodiments of the present disclosure, an oral detection method is proposed, comprising: collecting detection data of a target patient during oral diagnosis and treatment; determining a diagnosis result based on a machine learning model according to the detection data, wherein the machine learning model is generated by training based on detection samples in oral medical diagnosis and treatment.

[0025] In some embodiments, the detection data includes one or more of optical data, sound data, terahertz data, or electromagnetic wave data.

[0026] In some embodiments, the detection data includes one or more of optical data, sound data, terahertz data, or electromagnetic wave data.

[0027] In some embodiments, the optical data includes a three-dimensional image reconstructed based on the difference in reflection intensity after a sensor receives the optical signal by emitting a predetermined type of light to the soft and hard tissue surfaces of the oral cavity.

[0028] In some embodiments, the terahertz wave data includes transmission and reflection data obtained by transmitting and receiving electromagnetic waves to the soft and hard tissues of the oral cavity.

[0029] In some embodiments, the electromagnetic wave data includes data on the amount of X-rays absorbed by different parts of the oral tissue, obtained by emitting electromagnetic waves to the soft and hard tissues of the oral cavity.

[0030] In some embodiments, the sound data includes at least one of the sound data of the patient's active sound source and the sound data of the passive sound source. The sound data of the active sound source includes the sound data generated by the interlacing and sliding friction of the cusps of the upper and lower teeth during the movement of the mandible. The sound data of the passive sound source includes the sound data generated by the contact between the diagnostic and treatment tools and the internal structure of the oral cavity.

[0031] In some embodiments, the diagnosis result includes the state of the restoration, and the test samples include test samples of the restoration in different states, wherein the restoration state includes at least one of the occlusal contact state of the restoration, the degree of implant bone integration or the state of the internal preparation.

[0032] In some embodiments, the oral health status includes health, or the diagnosis result of oral diseases, and oral diseases include temporomandibular joint disorder, tooth disease, occlusal disorder, periodontal tissue disease, restoration problems, and oral and maxillofacial tumors.

[0033] In some embodiments, the machine learning model is generated by training according to any of the model training methods mentioned above.

[0034] According to one aspect of some embodiments of the present disclosure, a model training device is proposed, comprising: a sample acquisition unit, configured to acquire a detection sample, the detection sample including detection data and a label; a sample enhancement unit, configured to enhance the detection sample and acquire a training sample; a training unit, configured to process the training sample through a self-supervised learning model until the training is completed, wherein the parameters of the self-supervised learning model are adjusted according to the output diagnosis results and labels.

[0035] According to one aspect of some embodiments of the present disclosure, an oral sound data processing device is proposed, comprising: a data acquisition unit, configured to acquire detection data of a target patient during oral diagnosis and treatment; a state determination unit, configured to determine a diagnosis result based on a machine learning model according to the detection data, wherein the machine learning model is generated by training according to any one of the model training methods mentioned above.

[0036] According to one aspect of some embodiments of the present disclosure, a data processing device is provided, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any one of the above-mentioned methods based on instructions stored in the memory.

[0037] According to one aspect of some embodiments of the present disclosure, a non-transitory computer-readable storage medium is provided, on which computer program instructions are stored, and when the instructions are executed by a processor, any one of the methods mentioned above is implemented.

[0038] According to one aspect of some embodiments of the present disclosure, a computer program product is provided, including a computer program or instructions, and when the computer program or instructions are executed by a processor, any one of the methods mentioned above is implemented.

[0039] According to one aspect of some embodiments of the present disclosure, an electronic device is proposed, comprising: a data acquisition device, configured to acquire detection data of a target patient during oral diagnosis and treatment; and a data processor, configured to determine a diagnosis result based on a machine learning model according to the detection data, wherein the machine learning model is generated based on detection sample training.

[0040] In some embodiments, the data acquisition device includes at least one combination of: a laser light source, a laser detector, and a laser scanning system; an electromagnetic wave source and an electromagnetic wave receiver; a sound sensor; or a terahertz wave source and a terahertz wave detector. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0042] Figure 1 The present invention is a flowchart of some embodiments of the oral cavity detection method disclosed herein.

[0043] Figure 2 Flowcharts of some embodiments of the model training method disclosed herein.

[0044] Figure 3 Schematic diagrams of some embodiments of the model training method disclosed in the present invention.

[0045] Figure 4 Flowchart of some embodiments of sample enhancement in the model training method disclosed in the present invention.

[0046] Figure 5 Flowcharts of some embodiments of the oral diagnosis and treatment data processing method disclosed herein.

[0047] Figure 6 Schematic diagrams of some embodiments of the model training device disclosed in the present invention.

[0048] Figure 7 Schematic diagrams of some embodiments of the oral sound data processing device disclosed in the present invention.

[0049] Figure 8 Schematic diagrams of some embodiments of the data processing device disclosed herein.

[0050] Fig. 9 Schematic diagrams of other embodiments of the data processing device disclosed in the present invention.

[0051] Fig.10 Schematic diagrams of some embodiments of the electronic device disclosed herein. DETAILED DESCRIPTION

[0052] The technical solution of the present disclosure is further described in detail below through the accompanying drawings and embodiments.

[0053] With the high degree of specialization of stomatology, limited by theoretical knowledge and clinical skills, each oral sub-specialty has a parochial mindset during diagnosis and treatment, and lacks comprehensive inspection and treatment awareness. When patients visit doctors, doctors often choose the inspection and treatment plans within their own sub-specialty fields as the first choice, which is easy to miss, misdiagnose or fail to solve problems outside their professional scope, and cannot meet the patient's needs for overall, systematic and comprehensive treatment. Moreover, the clinical skill levels and diagnosis and treatment concepts of doctors in different departments may be different, which may lead to contradictions in the diagnosis and treatment plans of different departments, which not only reduces the efficiency of clinical diagnosis and treatment, but also reduces the patient's satisfaction and compliance with the treatment. Therefore, the oral medicine diagnosis and treatment model that is guided by patient needs and centered on disease diagnosis and treatment, with the coordinated development of general and specialist medicine, is proposed, but this model requires a higher comprehensive skill level for doctors, and requires long-term professional training and skill assessment to master the diagnosis and treatment methods of various specialties more comprehensively. In response to the above problems, the present invention proposes a model training, data processing and oral detection method, device and electronic equipment to improve the efficiency of clinical diagnosis and treatment, patient satisfaction and compliance.

[0054] In the method proposed in the present invention, one or more detection means are used to collect detection data, and oral medical diagnosis is performed through data analysis, which can locate the type and affected area of ​​specific oral problems, improve diagnostic efficiency, reduce the dependence of diagnostic operations on the doctor's sub-specialty fields, experience and manual operations, and improve the accuracy and comprehensiveness of diagnostic results.

[0055] In addition, most patients only seek medical treatment for the main complaint, and are prone to ignoring other hidden dangers of oral diseases, often waiting for them to cause obvious discomfort, affecting daily chewing or appearance, such as caries developing to the middle and deep layers, causing pain and discomfort; loose teeth to II or III degrees, significantly affecting chewing efficiency; temporomandibular joint disorder causes obvious pain and difficulty opening the mouth, only then will they actively seek medical treatment, often missing the best time for diagnosis and treatment. The method proposed in this disclosure is applied to the early intelligent detection and diagnosis of oral diseases to supplement the traditional oral specialization diagnosis and treatment model, which is conducive to optimizing the diagnosis process, improving the diagnosis and treatment efficiency and comprehensive diagnosis and treatment level of oral medical institutions, improving patient satisfaction and compliance, and achieving early screening and diagnosis.

[0056] There are many oral disease detection technologies in the related technologies, such as X-ray, CT, magnetic resonance imaging (MRI), ultrasound, etc. Each detection technology has its own advantages. Although it can provide a certain basis for diagnosis, these methods usually have certain limitations, such as radiation hazards, limited imaging accuracy, cumbersome detection process, and no real-time feedback.

[0057] With the development of new technologies such as acoustics and terahertz waves, the structural detection and disease diagnosis of oral soft and hard tissues will be more comprehensive and detailed, with the characteristics of non-invasive, high resolution, high sensitivity, real-time monitoring and rapid detection. The detection technologies involved in this disclosure include optical diagnosis, electromagnetic wave diagnosis, acoustic diagnosis and terahertz wave diagnosis.

[0058] Optical diagnosis: The scanner emits a specific type of light to the surface of the oral soft and hard tissues, which causes reflection or scattering. Different tooth tissues have different optical properties. For example, carious tissue has more organic components and water content, absorbs the light source more strongly, and has lower reflectivity. After the sensor receives and analyzes the optical signal, the computer reconstructs a high-precision three-dimensional image based on the difference in reflection intensity of different tissues and regions to obtain the surface model of the oral soft and hard tissues and even the internal tissue structure, thereby realizing the detection of common oral problems such as ulcers, tartar, and interproximal caries. This detection technology is characterized by being non-invasive, non-ionizing radiation, and convenient, but has limited penetration and may be interfered by surface contaminants.

[0059] Electromagnetic wave diagnosis: The conduction characteristics of electromagnetic waves can reflect the conductivity and impedance of different tissues. When electromagnetic waves are used to detect oral tissues, hard tissues (such as enamel and dentin) absorb more X-rays, and the image shows brighter areas, while soft tissues (such as gums and pulp) absorb less X-rays, and the image shows darker areas. Analyzing these characteristic changes and further determining the lesion area can be used for early detection of oral tumors, inflammation, and soft tissue lesions. The characteristic of this detection technology is that it can clearly and accurately display the internal structure of the teeth, but there is a certain radiation risk.

[0060] Acoustic diagnosis: Physicians with rich clinical experience will judge whether the occlusal contact is abnormal based on the sound characteristics of the patient's mandibular movement and tapping teeth; use diagnostic tools to contact the diagnostic position to assist diagnosis and treatment, such as using metal instruments to perform percussion to determine whether the root apex is painful and the degree, or use dental floss to check the tightness of the interproximal contact through the adjacent area. Different diseases or different degrees of disease present different acoustic characteristics. This detection technology is characterized by being fast, convenient, non-invasive, low technical sensitivity, and comprehensive, but no visual image is generated, and the test results are directly output.

[0061] Terahertz wave diagnosis: Terahertz waves refer to electromagnetic waves in the range of 0.1 to 10 THz, which can be divided into two imaging optical paths: transmission and reflection. The transmission optical path can be used to determine the refractive index and absorption coefficient of enamel, dentin and caries; the reflection optical path can effectively detect early caries and other tissue types. Terahertz has a strong ability to penetrate biological tissues and can effectively detect structural changes inside oral soft and hard tissues; and terahertz waves are highly sensitive to tissue hydration and can be used to detect micro-lesions, tumors and inflammation in teeth, gums and bones. This detection technology has the advantages of high resolution, non-destructive, no ionization damage, and the ability to identify molecular feature fingerprints.

[0062] The detection technology for collecting detection data in the embodiments shown below can be one or more of the optical diagnosis, electromagnetic wave diagnosis, acoustic diagnosis, and terahertz wave diagnosis mentioned above. In the embodiments, acoustic diagnosis, obtaining sound data and applying them are mainly used as examples to describe the model training, data processing and oral detection methods, devices and electronic equipment of the present disclosure. The types of detection data and the types of detection technologies used in the embodiments are only examples of feasible implementation methods and do not constitute undue limitations on the present disclosure. Based on the solutions in the embodiments of the present disclosure, the model training, data processing and oral detection methods of the present disclosure can be applied to scenarios using optical diagnosis, electromagnetic wave diagnosis, and terahertz wave diagnosis, which are also within the scope of protection of the present disclosure.

[0063] The flowcharts of some embodiments of the oral cavity detection method disclosed in the present invention are as follows: Figure 1 shown.

[0064] In step S11, during the process of oral diagnosis and treatment of the patient, the detection data of the target patient in the oral diagnosis and treatment is collected. In some embodiments, the sound data of the target patient in the oral diagnosis and treatment can be collected by a sound sensor. The acoustic information in the oral medical diagnosis and treatment process comes from the patient's active sound source and passive sound source. The active sound source is mostly generated in the mandibular movement process, and the cusps and fossa of the upper and lower teeth are staggered and slid and rubbed. The mandibular movement is completed by the various components of the oral and maxillary system. Through the contact relationship between the cusps and fossa of the upper and lower teeth, under the repeated regulation of proprioception at the periodontal, temporomandibular joint and masticatory muscles, it is memorized by the nerves and muscles, and gradually forms a movement form with individual characteristics, including opening and closing mouth movement, lateral movement and forward and backward movement. Any abnormal changes in related factors affecting mandibular movement, such as the temporomandibular joint and (teeth), will affect mandibular movement, thereby affecting the patient's oral health. Doctors with rich clinical experience will combine visual examination, palpation and auscultation to determine whether the mandibular movement is normal, so as to conduct targeted diagnosis and screening for temporomandibular joint disorders (such as clicking and murmurs) and tooth problems (such as early contact, According to the specific performance of mandibular movement, reverse judgment is made on whether the occlusal contact of the restoration (inlay, full crown, full denture, etc.) is normal, and the clinician is guided to make targeted occlusal adjustments, thereby improving the quality of restoration and the efficiency of diagnosis and treatment.

[0065] Passive sound sources mainly come from the sounds produced by doctors using diagnostic tools, such as using metal instruments to percuss the root apex to determine whether it is painful and the degree of pain, or using dental floss to check the tightness of the interproximal contact. These acoustic signals contain and reflect a wealth of basic oral health conditions. However, usually only expert doctors with rich diagnostic and treatment experience can make full use of the above acoustic information to assist in the diagnosis of corresponding diseases, and the clinical technology has a high sensitivity.

[0066] The sound sensor is used to collect the sound data generated by the active sound source during the mandibular movement of the patient. In some embodiments, the patient can be guided to perform mandibular movement at will, and the sound sensor is used to collect the sound data during the mandibular movement within a predetermined distance range of the patient's oral cavity. This method can reduce the requirements for the patient's movement mode and improve user friendliness and feasibility.

[0067] In some embodiments, during the sound data collection process, the patient can be guided to perform specified mandibular movements, such as tooth tapping, lateral movement, and protrusive movement, collect sound data for each movement mode, and label the movement type. This method can increase the amount of effective information in the sound data for oral disease diagnosis, facilitate more and better exposure of patients' oral problems, reduce the difficulty of subsequent data processing, and help improve the accuracy of diagnosis.

[0068] The sound sensor collects the sound data generated by the patient's passive sound source, that is, the sound data generated when the doctor touches the internal structure of the oral cavity through the diagnosis and treatment tools, such as the sound data generated when the doctor taps the teeth with instruments, or uses dental floss to check the adjacent surface through the adjacent area. This method can extract rich sound data, provide more data support for subsequent diagnosis, and help improve the accuracy of the diagnosis results.

[0069] In step S12, the sound data is processed by a machine learning model to determine a diagnosis result. The machine learning model is generated based on sound sample training.

[0070] In some embodiments, based on the machine learning model architecture for processing sound data in the related art, the model can be trained using sound samples with labeled diagnosis results to obtain a machine learning model that can be used in step S12. In this way, model training can be performed on the basis of a machine learning model that processes sound data, thereby improving convenience and efficiency.

[0071] In some embodiments, the machine learning model can be a deep learning model. Deep learning relies on powerful and effective learning algorithms and end-to-end feature extraction capabilities. It is a subset of artificial intelligence. It uses a hierarchical algorithm similar to a human neural network in a computer system to simulate human cognitive abilities. It has the ability to process data including audio, images, text, etc. The deep learning model can assist doctors and researchers in achieving efficient and high-quality medical image recognition, electronic medical record analysis, drug development, genomics analysis, and acoustic information detection. The acoustic information detection method based on deep learning has the characteristics of high accuracy and non-invasiveness.

[0072] In some embodiments, the diagnosis result includes a diagnosis result for an oral disease, such as an oral health status. The oral health status includes health, or a diagnosis result of an oral disease. The oral diseases mentioned in the present disclosure include temporomandibular joint disorder, tooth disease (such as supernumerary teeth, caries, cracked teeth, pulpitis, apical periodontitis, impacted teeth, etc.), occlusal disorder disease (such as the presence of early contact, the presence of In some embodiments, the diagnosis result can be specific to the type of disease, such as caries, apical periodontitis, etc. In some embodiments, the diagnosis result can also include the location, quantity, and degree of the disease, such as the target tooth position, several diseases or several locations with diseases, and the degree of the disease, so as to facilitate the doctor to locate the disease and make subsequent diagnosis and treatment operations to avoid missed diagnosis.

[0073] In some embodiments, the diagnosis results include the occlusal contact status of the restoration (e.g., inlay, full crown, full denture, etc.), whether the occlusal contact of the patient after using the restoration is normal, for example, requiring the patient with full denture restoration to perform tooth tapping, and collecting sound data, based on the sound, it can be determined whether the occlusal contact of the denture is uniform, guide the doctor whether to continue grinding, and determine the grinding position based on the sound. This method can assist the doctor in adjusting the restoration, improve the adjustment efficiency, avoid adjustment errors caused by unclear expression and inaccurate feeling of the patient, and improve the accuracy of restoration adjustment.

[0074] Through the method in the above embodiment, it is possible to use sound data to determine whether there is temporomandibular joint disorder, tooth-related diseases (such as supernumerary teeth, caries, cracked teeth, pulpitis, apical periodontitis, impacted teeth, etc.), occlusal contact conditions (such as early contact, Interference, etc.), periodontal tissue diseases (alveolar bone resorption, loose teeth, food impaction, etc.) and restoration conditions (implant bone integration, internal preparations) and other related oral problems, and accurately analyze the specific location, quantity, type and degree to improve the efficiency of early diagnosis of the system comprehensive.

[0075] In some embodiments, the sound data may be first converted into other types of data, such as text data or image data, and then based on the machine learning model architecture for processing text and image data in the relevant technology, the model may be trained using the text and image data corresponding to the labeled sound samples to obtain a machine learning model that can be used in step S12. This method can increase the range of optional machine learning model architectures, thereby facilitating the selection of more suitable and accurate models for training and improving the accuracy of oral cavity detection.

[0076] In some embodiments, the sound samples include sound samples in a healthy oral state and sound samples in at least one oral disease state. For example, the sound samples include sound samples in a healthy oral state and at least one of the following sound samples: sound samples in a temporomandibular joint disorder state, sound samples in various tooth disease states (e.g., sound samples in a loose tooth state, sound samples in a carious tooth state), sound samples in various occlusal disorder disease states (e.g., sound samples in a state with early contact, sound samples in a state with The above are only examples, and the corresponding sound sample data can be collected according to the specific type of oral disease that needs to be checked, and the sound samples can be expanded. The above examples do not constitute an undue limitation on the present disclosure. The number of sound samples in each state is greater than a predetermined value (for example, more than 10) to ensure that the model fully learns the sound features in the corresponding state and improves the accuracy of the model.

[0077] In some embodiments, the sound samples include sound samples when the occlusal contact of the restoration is in a normal state, and sound samples when the occlusal contact of at least one restoration is in an abnormal state.

[0078] The sound sample consists of two parts: sound sample data and labels. The labels include the oral health status of the person being tested or the occlusal contact status of the restoration that generates the sound sample data. The oral health status can be healthy or the diagnosis result of an oral disease. Specifically, oral diseases include temporomandibular joint disorder, tooth disease (such as supernumerary teeth, caries, cracked teeth, pulpitis, apical periodontitis, impacted teeth, etc.), occlusal disorder disease (such as the presence of early contact, the presence of The occlusal contact state of the restoration includes whether the occlusal contact state of the restoration is normal or abnormal. The machine learning model trained based on such samples can not only determine whether the person being tested has oral diseases and whether the occlusal contact state of the restoration is normal, but also locate the specific type of oral diseases, thereby improving the efficiency of oral detection.

[0079] In some embodiments, the label also includes the severity of the disease, for example, multiple severity levels can be set. Using such sound samples can facilitate carrying the severity of the disease in the output diagnosis results, giving doctors more information and further improving the efficiency of oral detection.

[0080] In some embodiments, the label also includes at least one of the location, quantity, and degree of the disease, for example, specific to the target tooth position, several diseases or several locations with diseases, and the degree of the disease, so that the trained model has the ability to diagnose the type, location, and degree of the disease, thereby improving the richness of the diagnostic results and further improving the efficiency of oral detection.

[0081] In some embodiments, the above-mentioned machine learning model can be trained and generated according to any of the following model training methods.

[0082] Based on the method in the above-mentioned embodiment of the present disclosure, it is possible to collect the test data generated by the person being tested during the oral diagnosis and treatment process, process the test data using a machine learning model, and identify the oral health status or restoration status of the person being tested, thereby improving the diagnostic efficiency and reducing the diagnostic operation's dependence on the doctor's sub-professional fields, experience, and manual operations, thereby improving the accuracy and comprehensiveness of the diagnostic results.

[0083] By using the oral detection method in the above embodiment, patients can be given preventive and comprehensive diagnosis and treatment suggestions in a timely manner, the consultation process can be optimized, the consultation time can be shortened, and the diagnosis and treatment efficiency can be improved. In some embodiments, the diagnosis results of whether there are various common oral diseases or abnormal occlusal contact problems of restorations can be accurately determined to the specific location, quantity and degree of the disease, reducing the requirements for the doctor's diagnostic workload and comprehensive skill level, and improving the comprehensiveness of the diagnosis.

[0084] In some embodiments, Figure 1 The detection data in the illustrated embodiment may be one or more of optical data, sound data, terahertz data or electromagnetic wave data. Optical data includes a three-dimensional image reconstructed based on the difference in reflection intensity after a predetermined type of light is emitted to the surface of the oral soft and hard tissues and the optical signal is received by the sensor. Terahertz wave data includes transmission and reflection data received by emitting electromagnetic waves to the oral soft and hard tissues. Electromagnetic wave data includes data on the amount of X-rays absorbed by different parts of the oral tissues obtained by emitting electromagnetic waves to the oral soft and hard tissues.

[0085] In some embodiments, a variety of detection technologies can be used to obtain detection results based on a variety of detection data. As mentioned above, each detection technology has its own scope of application, as well as its own advantages and disadvantages. For example, optical diagnosis has the advantages of non-invasiveness, no ionizing radiation, and convenience, but its penetration is limited and may be interfered by surface contaminants; electromagnetic wave diagnosis has the advantage of being able to clearly and accurately display the internal structure of teeth, but there is a certain radiation risk; acoustic diagnosis has the advantages of being fast, convenient, non-invasive, low technical sensitivity, and comprehensive, but there is no visible image generation and the detection results are directly output; terahertz wave diagnosis is suitable for detecting micro-lesions, tumors, and inflammation in teeth, gums, and bones, and has the advantages of high resolution, non-destructiveness, no ionization damage, and identifiable molecular feature fingerprints.

[0086] Different imaging detection technologies have different scopes of application and characteristics. By integrating different imaging technologies, information of different levels and dimensions can be obtained simultaneously, such as anatomical structure, functional status, blood flow dynamics, etc.; the scope, size and location of the lesion can be clearly displayed, and the accuracy of diagnosis and early diagnosis rate can be improved, precise and personalized medical treatment can be provided, the patient's medical experience and treatment effect can be improved, and trauma can be reduced. With the continuous development of technology, multimodal biomedical imaging technology will be more accurate, convenient and low-cost. Integrated and intelligent imaging systems will be able to provide more comprehensive, real-time and dynamic diagnosis and treatment support for oral medicine.

[0087] In some embodiments, one or more of the optical data, terahertz data, or electromagnetic wave data can be processed using the model training and data processing methods disclosed herein, or can be processed using methods and devices in related technologies, and then combined with the model training and data processing methods disclosed herein based on acoustic data. For example, after collecting high-resolution optical images of oral soft and hard tissues, a spectral analysis algorithm is used to analyze the scattering and absorption characteristics of light to diagnose the health status of the tissue; after obtaining the electromagnetic response data of the oral tissue, the parameters such as tissue conductivity and impedance are obtained by combining calculation and analysis to perform a health assessment of the oral tissue structure, such as the degree of alveolar bone absorption and the degree of implant bone integration; after obtaining high-resolution terahertz images, the physical properties of the oral tissue are evaluated through the propagation characteristics of the terahertz wave, and potential lesions inside the oral cavity, such as lichen planus and squamous cell carcinoma, are detected and identified from a pathological level.

[0088] Through the method in this embodiment, the burden of early model training can be reduced, and the system deployment efficiency can be improved while improving diagnostic accuracy and early diagnosis rate, precise personalized medicine, improving patient medical experience and treatment effects, and reducing trauma.

[0089] In some embodiments, the oral detection method of the present disclosure can be used in the early stages of a patient's visit to assist clinicians in conducting a comprehensive examination of the oral and maxillofacial system, thereby achieving early detection and early intervention treatment, thereby improving the oral health level of the entire population.

[0090] In order to improve the ability of the machine learning model to process the detection data in the oral diagnosis and treatment process and improve the accuracy of the diagnosis results, the present disclosure also proposes a model training method. In some embodiments, Figure 2 The data processing in this embodiment is also mainly based on sound data as an example. For optical data, terahertz data or electromagnetic wave data, the data processing process is implemented based on the technical concept mentioned in this disclosure and in combination with the processing method for corresponding types of data in the relevant technology.

[0091] In step S21, a test sample in oral medical diagnosis and treatment is obtained, and the test sample includes test data and labels. In some embodiments, the test sample can be collected and labeled in daily work; in some embodiments, volunteers can be called to collect sound data, and the test sample can be labeled according to the oral health status diagnosis results of the volunteers or the status of the restoration to form a test sample.

[0092] In some embodiments, a sample database may be constructed to store test samples.

[0093] The test samples include test samples in a healthy oral state and test samples in at least one oral disease state. In some embodiments, the test samples include test samples in different states of the restoration, wherein the restoration state includes at least one of the occlusal contact state of the restoration, the degree of implant bone integration, or the state of the internal preparation.

[0094] Oral health status includes health or oral disease diagnosis results. Oral diseases include temporomandibular joint disorder, tooth disease (such as supernumerary teeth, caries, cracked teeth, pulpitis, apical periodontitis, impacted teeth, etc.), occlusal disorder (such as the presence of early contact, the presence of Interference, etc.), periodontal tissue diseases (such as alveolar bone absorption, loose teeth, food impaction, etc.), restoration conditions (implant bone integration degree, internal preparations), oral and maxillofacial tumors (ameloblastoma, pleomorphic adenoma, etc.), etc. The above types of oral diseases are only examples and can be extended to various types of oral diseases known in the art. The above examples do not constitute an undue limitation on the present disclosure. The diagnosis results will be specific to the name of the disease, such as supernumerary teeth, premature contact, alveolar bone absorption, etc., so as to provide sufficient information.

[0095] The occlusal contact state of the restoration includes a normal occlusal contact state of the restoration or an abnormal occlusal contact state of the restoration. For example, the abnormal occlusal contact state of the restoration includes uneven occlusal contact. In some embodiments, when the occlusal contact is uneven, the occlusal contact state of the restoration also includes grinding adjustment site information.

[0096] In step S22, the detection samples are enhanced to obtain training samples.

[0097] In some embodiments, enhancing the detection sample may include expanding the number of samples. Considering that the collection of detection samples is time-consuming, and model training requires a large amount of sample data. Sample expansion can increase the number of samples and improve the accuracy of model training. In some embodiments, sample expansion can be performed by at least one of waveform displacement, harmonic distortion, or resampling.

[0098] In some embodiments, enhancing the detection samples may include denoising the detection sample data, etc., to reduce the amount of interference information and improve the accuracy of model training.

[0099] In some embodiments, enhancing the detection sample may also include converting the detection sample data into text or image data, for example, converting the sound sample data into a spectrogram to achieve visualization of the audio data and reduce the difficulty of subsequent processing.

[0100] In some embodiments, taking a sound sample as an example, the method for enhancing the sound sample is as follows: Figure 4 As shown in , it includes step 421 and step 422.

[0101] In step 421, the sound data is processed by a filter to obtain a first spectrogram sample.

[0102] The filter can be a Mel filter, such as Figure 3 As shown in , the sound data is passed through a Mel filter and the output Mel spectrogram is used as the first spectrogram sample.

[0103] In some embodiments, the process of converting sound data into a spectrogram includes pre-emphasis, framing, windowing, STFT (Short-Time Fourier Transform), conversion according to the correspondence between frequency and Mel value, etc. In some embodiments, pre-emphasis can be regarded as passing the sound data through a high-pass filter, which can not only reduce the influence of partial amplitude-frequency distortion and frequency response changes, but also reduce low-frequency noise in the audio signal. The pre-emphasized audio waveform is framed, windowed and STFTed to obtain a spectrum matrix. The process is as follows:

[0104]

[0105] Among them, y(t) is the time domain signal, x(f,r) is the frequency domain signal, w(tr) represents the Hamming window with the center position at r, f is the frequency, and r is the frame length.

[0106] In some embodiments, the window hop size is 256 and the window of the STFT is 512.

[0107] The Mel filter is designed according to the characteristics of the human ear to improve the absorption of the lower frequencies of the sound. In some embodiments, the number of Mel filters per second in the sound audio segment is set to 128, and the formula of the Mel filter is expressed as follows:

[0108]

[0109] Among them, f mel is the calculated Mel scale frequency and f is the Hertz frequency.

[0110] The Mel filter bank imitates the human ear to filter speech. M (e.g. 512) triangular filters are set within the frequency range of an audio segment. The filter layout changes from dense to sparse. As the Hertz frequency increases, the filter width changes from small to large. There is a 50% overlap between each filter to avoid information loss. These filters are displayed as equal width on the Mel scale. Finally, the output matrix is ​​converted into a spectrogram.

[0111] In some embodiments, the sampling rate of the sound data is set to 16000 Hz. Due to the different lengths of data, the sound sample data is unified into 4 seconds, the size of the generated spectrum graph is 128*251, and then it is scaled to 224*224 in the same proportion, which is conducive to the subsequent processing of spectrum data of the same size and reduces the difficulty of data processing.

[0112] Considering that the sound characteristics produced by tooth collision are relatively unique and the signal varies greatly over time, directly performing STFT on the sound data will cause some audio waveforms to be distorted. The method in the embodiment shown above can reduce waveform loss and increase the amount of effective information in the converted spectrum, which is beneficial to improving the accuracy of the model.

[0113] In step 422, the first spectrogram sample is processed by at least one of waveform displacement, harmonic distortion or resampling to obtain a second spectrogram sample, and the first and second spectrogram samples are combined to obtain a training sample. Through this method, the sample is expanded on the basis of realizing audio data visualization, thereby improving the accuracy of model training.

[0114] In some embodiments, a time-frequency feature enhancement module TFAM may be provided, such as Figure 3 As shown in FIG. 1 , more sample data are expanded based on the spectrum output by the Mel filter as the subsequent model input.

[0115] In some embodiments, a time-frequency feature enhancement module TFAM is used to extract feature information from the spectrum graph from the time domain and frequency domain respectively, to obtain time domain features and frequency domain features, and then obtain time-frequency fusion features based on the time domain features and frequency domain features. For example, horizontal and vertical convolution operations are performed on the spectrum graph, and the frequency domain features are extracted through a 1x3 convolution block and the time domain features are extracted through a 3x1 convolution block, and then the frequency domain features and the time domain features are added and summed to obtain fusion features in the time and frequency domains. The information in the time and frequency domains is used to guide the learning of the attention matrix, enhance the fusion and representation capabilities of feature information, and distinguish between valid and invalid sound data (such as mandibular movement sounds and non-mandibular movement sounds), so as to reduce the interference of irrelevant information and improve the learning of valid information.

[0116] Furthermore, local features are extracted from the fused features, and dynamic position encoding is obtained based on the time-frequency fusion features. For example, a convolutional layer is used to extract local features to enhance the spatial feature extraction capability of the model. Let the input feature be X∈R C×H×W , where C is the number of channels, H and W are the height and width respectively. Local feature aggregation can be expressed as:

[0117] X local =Conv(X) (3)

[0118] Among them, Conv represents the convolution operation.

[0119] Dynamic position encoding can be achieved through learnable convolution kernels, expressed as:

[0120] P=Conv pos (X) (4)

[0121] Among them, Conv pos is a convolution operation used to generate positional encodings.

[0122] According to the local features and dynamic position encoding, context-aware features are extracted based on the self-attention mechanism. For example, this step combines local features and position encoding to extract context information through the self-attention mechanism. First, the local features and position encoding are merged to obtain the merged features:

[0123] X merged =X local +P (5)

[0124] Then, the self-attention mechanism is applied to obtain context-aware features:

[0125]

[0126] Among them, Q, K, V are the conversion functions of query, key and value respectively, and d. is the dimension of the key.

[0127] The context-aware features and the time-frequency fusion features are fused to obtain the second spectrogram sample. The context-aware features are fused with the original features, and the feature representation is:

[0128] X out =X+FFN(X cot ) (7)

[0129] Among them, FFN represents a feed-forward neural network, which is used to further transform contextual features.

[0130] Different convolutional layers are combined in parallel, and the result matrices processed by different convolutional layers are spliced ​​together in the depth dimension to form a deeper matrix.

[0131] In this way, the input image can be effectively processed and richer and more diverse features can be extracted. It can efficiently expand the depth and width of the network, improving the accuracy of the deep learning network while preventing overfitting.

[0132] In some embodiments, the above operations can be implemented using a machine learning model with relevant functions (such as CotNet) to improve implementation efficiency.

[0133] In step S23, the training samples are processed by the self-supervised learning model until the training is completed. During the training process, the parameters of the self-supervised learning model are adjusted according to the output diagnosis results and labels.

[0134] In some embodiments, a self-supervised learning model can be pre-deployed, such as Figure 3 As shown, the training sample data is input into the input end, and after encoding and decoding operations, the diagnosis result is outputted into the output end. By comparing the result with the label corresponding to the training sample data, the parameters of the self-supervised learning model are adjusted to realize the learning of the self-supervised learning model.

[0135] In some embodiments, the self-supervised learning model may be ViT (Vision transformer). This model is built on the Transformer architecture and can be used to detect complex patterns in images and understand global context. After the above processing, the sound data has been converted into a spectrogram, i.e., image data. Using the ViT model for processing images to process image data reduces the difficulty of processing sound data.

[0136] For example, suppose the size of the input image is H×W×C, where H and W represent the height and width of the image, respectively, and C represents the number of color channels. ViT first divides the image into N patches of size P×P, where N and P are positive integers. Each patch can be regarded as a "word", N = HW / P 2 Each small block is then flattened and converted into a D-dimensional vector. Specifically, if each small block is regarded as a vector Through a learnable linear projection (weight matrix ), then we get the embedding representation z of each small block p =x p E.

[0137] In addition, ViT introduces position embedding to encode the relative or absolute position of the image block in the original image. Considering that Transformer itself does not have the ability to process sequence position information, for the embedding of each image block, a position embedding vector is added The final embedded input z is obtained p ′=z p +pos.

[0138] The embedded image blocks are fed into a series of Transformer encoders for processing. Each encoder consists of two main parts: MSA (Multi-Head Self-Attention) and FFN (Feed-Forward Network). For each layer in the encoder, the input is The output is also a matrix of the same shape. This process can be expressed as:

[0139] MSA(Z)=Concat(head1,head2,…,head h ,)W O (8)

[0140] Among them, each head Implemented the self-attention mechanism, and is the learnable parameter matrix, h is the number of heads, and d k is the dimension of each head.

[0141] FFN(Z)=max(0,ZW1+b1)W2,+b2 (9)

[0142] Among them, W1, b1, W2, b2 are the parameters of the FFN layer,

[0143] After processing the embedded image patches, ViT uses a “category” (CLS) labeled embedding as the output for the image classification task. The final embedding of the CLS label is passed through a fully connected layer (linear layer) to predict the category of the image.

[0144] In some embodiments, ViT with a small skeleton size is used, such as ViT_B16. The ViT_B16 model divides the input image into smaller patches and converts each patch into a vector representation using linear embedding. These patch embeddings, along with the position embeddings that provide spatial information, are then passed through multiple layers of the Transformer encoder. In these layers, the model uses a self-attention mechanism to weigh the importance of each patch relative to other patches, effectively capturing long-range dependencies. The resulting transformed feature representations are then used for tasks such as image classification and other visual processing objectives. In this way, computational efficiency is improved while maintaining performance.

[0145] Through the method in the above embodiment of the present disclosure, it is possible to reduce the dependence of model training on collected real sample data through sample enhancement, reduce the burden of sample collection, improve the efficiency of model training, and improve the accuracy of oral cavity detection of the trained model.

[0146] In addition, the present disclosure also proposes a method for processing oral diagnosis and treatment data. The method for processing oral diagnosis and treatment data is based on the model training method mentioned above and processes the test data in the oral medical diagnosis of the person to be tested. In some embodiments, Figure 5 As shown in , the oral diagnosis and treatment data processing method includes steps S51 and S52.

[0147] In step S51, the detection data of the target patient in the oral diagnosis and treatment is obtained. The detection data includes one or more of optical data, sound data, terahertz data or electromagnetic wave data. In some embodiments, the sound data can come from a sound sensor; the optical data can come from a laser scanning system; the terahertz data can come from a terahertz wave detector; and the electromagnetic wave data can come from an electromagnetic wave receiver. In some embodiments, the detection data generated remotely can be obtained through a network; in some embodiments, the detection data pre-stored in a memory can be extracted.

[0148] In step S52, a diagnosis result is determined according to the detection data based on a machine learning model, where the machine learning model is generated by training according to any one of the model training methods described above.

[0149] In some embodiments, the machine learning model can process the detection data to obtain corresponding diagnostic results.

[0150] In some embodiments, the detection data can also be converted into a detection spectrum diagram between steps S51 and 52. The machine learning model can process the image data and obtain the corresponding oral state information by processing the sound spectrum diagram.

[0151] Through the method in the above embodiment, the trained machine learning model can be used to detect the test data of the person to be tested during the oral diagnosis and treatment process, so as to obtain the oral health status or the status of the restoration of the person to be tested, reduce the dependence of the diagnostic operation on the doctor's sub-professional field, experience and manual operation, and improve the efficiency, accuracy and comprehensiveness of the diagnosis.

[0152] In some embodiments, multiple detection technologies can be used, and corresponding equipment can be used to obtain multiple detection data, and corresponding detection results can be obtained based on the detection data. The detection results are combined to obtain the final result of oral diagnosis and treatment.

[0153] Through the method in this embodiment, information at different levels and dimensions can be obtained simultaneously, thereby improving diagnostic accuracy and early diagnosis rate, achieving precise personalized medical treatment, improving patient medical experience and treatment effects, and reducing trauma.

[0154] Schematic diagrams of some embodiments of the model training device 61 of the present disclosure are as follows Figure 6 as shown in .

[0155] The sample acquisition unit 611 can acquire a test sample in oral medical diagnosis and treatment, and the test sample includes test data and a label. In some embodiments, the sample acquisition unit 611 can execute the method in any embodiment of the above step S21.

[0156] The sample enhancement unit 612 can enhance the detection sample and obtain the training sample. In some embodiments, the sample enhancement unit 612 can execute the method in any embodiment of the above step S22.

[0157] The training unit 613 can process the training samples through the self-supervised learning model until the training is completed, wherein the parameters of the self-supervised learning model are adjusted according to the output diagnosis results and labels. In some embodiments, the training unit 613 can perform the method in any embodiment of step S23 above.

[0158] The model training device in the above-mentioned embodiment of the present disclosure can reduce the dependence of model training on collected real sample data through sample enhancement, reduce the burden of sample collection, improve the efficiency of model training, and improve the accuracy of oral cavity detection of the trained model.

[0159] Schematic diagrams of some embodiments of the oral diagnosis and treatment data processing device 72 of the present disclosure are as follows Figure 7 as shown in .

[0160] The data acquisition unit 721 can acquire the detection data of the target patient in the oral diagnosis and treatment. In some embodiments, the data acquisition unit 721 can execute the method in any embodiment of the above step S51.

[0161] The diagnosis result determination unit 722 can determine the diagnosis result based on the detection data and the machine learning model, wherein the machine learning model is generated by training according to any one of the model training methods described above. The above diagnosis result includes the oral health status or the restoration status. In some embodiments, the status determination unit 722 can execute the method in any of the embodiments of step S52 described above.

[0162] In some embodiments, the oral diagnosis and treatment data processing device 72 also includes a conversion unit 723, which can convert the detection data obtained by the data acquisition unit 721 into a spectrum graph, and then the state determination unit 722 can process the spectrum graph through a machine learning model to obtain a diagnosis result.

[0163] The oral diagnosis and treatment data processing device in the above-mentioned embodiment of the present disclosure can use the trained machine learning model to detect the diagnosis and treatment data of the person to be tested during the oral diagnosis and treatment process, obtain the oral health status or restoration status of the person to be tested, reduce the diagnosis operation's dependence on the doctor's sub-professional field, experience and manual operation, and improve the efficiency, accuracy and comprehensiveness of the detection.

[0164] A structural diagram of an embodiment of the data processing device disclosed in the present invention is shown in FIG. Figure 8As shown. The data processing device includes a memory 801 and a processor 802. Among them: the memory 801 can be a disk, a flash memory or any other non-volatile storage medium. The memory is used to store the instructions in the corresponding embodiments of the model training method or the oral diagnosis and treatment data processing method mentioned above. The processor 802 is coupled to the memory 801 and can be implemented as one or more integrated circuits, such as a microprocessor or a microcontroller. The processor 802 is used to execute the instructions stored in the memory, which can improve the accuracy of oral medicine diagnosis and treatment using a machine learning model.

[0165] In one embodiment, it is also possible to Fig. 9 As shown, the data processing device 900 includes a memory 901 and a processor 902. The processor 902 is coupled to the memory 901 via a BUS 903. The data processing device 900 can also be connected to an external storage device 905 via a storage interface 904 to call external data, and can also be connected to a network or another computer system (not shown) via a network interface 906. No further detailed description will be given here.

[0166] In this embodiment, the accuracy of oral medical diagnosis and treatment can be improved by storing data instructions in a memory and processing the instructions through a processor.

[0167] In another embodiment, a computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the steps of the method in the corresponding embodiment of the model training method or the oral diagnosis and treatment data processing method. It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, devices, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transitory storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0168] Schematic diagrams of some embodiments of the electronic device disclosed herein are as follows Fig.10 As shown in , it includes data acquisition equipment and data processor.

[0169] The data acquisition device 1010 can collect the detection data of the target patient during the oral diagnosis and treatment of the patient. In some embodiments, the data acquisition process of the data acquisition device 1010 can be as shown in any embodiment of step S11 above.

[0170] The data processor 1020 can process the test data through a machine learning model to determine a diagnosis result. The machine learning model is generated based on the training of test samples in oral medicine diagnosis and treatment. In some embodiments, the data processor 1020 can execute the method in any embodiment of step S12 above.

[0171] The electronic device in the above-mentioned embodiment of the present disclosure can collect the test data generated by the person being tested during the lower oral diagnosis and treatment process, use the machine learning model to process the test data, identify the oral health status of the person being tested or the occlusal contact status of the restoration, thereby improving the diagnosis efficiency and reducing the dependence of the diagnosis operation on the doctor's sub-professional field, experience and manual operation, thereby improving the accuracy and comprehensiveness of the diagnosis results.

[0172] By using the electronic device in the above embodiment, it is possible to give patients preventive and comprehensive treatment suggestions in a timely manner, optimize the treatment process, shorten the treatment time, and improve the diagnosis and treatment efficiency. In some embodiments, the diagnosis results of whether there are various common oral diseases or abnormal occlusal contact problems of restorations can be accurately determined to the specific location, quantity and degree of the disease.

[0173] In some embodiments, the data acquisition device 1010 may include a laser light source, a laser detector, and a laser scanning system, capable of acquiring high-resolution optical images of oral soft and hard tissues. The data processor 1020 uses a spectral analysis algorithm to analyze the scattering and absorption characteristics of light to diagnose the health status of the tissue.

[0174] In some embodiments, the data acquisition device 1010 may include an electromagnetic wave source and an electromagnetic wave receiver. The data processor 1020 calculates and analyzes parameters such as tissue conductivity and impedance to perform health assessment of oral tissue structures, such as the degree of alveolar bone absorption and the degree of implant bone integration.

[0175] In some embodiments, the data acquisition device 1010 may include a sound sensor. After the data processor 1020 converts the acoustic signal into a spectrogram, a self-supervised neural network is used to extract and train dual feature information in the time domain and the frequency domain.

[0176] In some embodiments, the data acquisition device 1010 may include a terahertz wave source and a terahertz wave detector. The data processor 1020 evaluates the physical properties of oral tissues through the propagation characteristics of terahertz waves, and even detects and identifies potential lesions inside the oral cavity from a pathological level, such as lichen planus, squamous cell carcinoma, etc.

[0177] Based on the electronic devices in the above embodiments, one or more of the optical, electromagnetic wave, and terahertz wave detection technologies can be used to process data using methods in related technologies, thereby reducing the burden of early model training, improving diagnostic accuracy and early diagnosis rate, providing precise personalized medicine, improving patient experience and treatment effects, and reducing trauma, while improving system deployment efficiency.

[0178] In some embodiments, the electronic device disclosed herein can be used by patients at the beginning of their consultation to assist clinicians in conducting a comprehensive examination of the oral and maxillofacial system, thereby achieving early detection and early intervention treatment, thereby improving the oral health level of the entire population.

[0179] In some embodiments, the data acquisition device 1010 can be connected to the data processor 1020 by wire or wireless means, so that the collected detection data can reach the data processor in time to obtain the diagnosis result, thereby improving the processing efficiency. In some embodiments, the data processor 1020 can be centrally arranged on the server side, and the detection data is transmitted to the server for processing via the network. The server sends the processing result to the client bound to the data acquisition device via the network, and displays the diagnosis result on the client. Such electronic equipment facilitates the effective use of server resources and improves the computing efficiency; it reduces the cost of equipment purchased by the client (e.g., hospital, doctor), which is conducive to the promotion and application.

[0180] In some embodiments, the data acquisition device 1010 and the data processor 1020 are arranged in the same housing. The data processor 1020 is a computing chip that can use a machine learning model to process sound data and determine the diagnosis results, thereby reducing the size of the device and improving the ease of use of the electronic device.

[0181] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0182] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0183] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0184] So far, the present disclosure has been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

[0185] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0186] It should be noted that the terms "first", "second", etc. in the specification, claims, and drawings of the present disclosure 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 interchangeable where appropriate, so that the embodiments of the present disclosure 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.

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present disclosure rather than to limit it. Although the present disclosure has been described in detail with reference to the preferred embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present disclosure can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present disclosure, which should be included in the scope of the technical solution for protection of the present disclosure.

Claims

1. A model training method, comprising: Obtaining test samples in oral medical diagnosis and treatment, wherein the test samples include test data and labels; Enhance the detection sample to obtain a training sample; The training samples are processed by a self-supervised learning model until the training is completed, wherein the parameters of the self-supervised learning model are adjusted according to the output diagnosis result and the label.

2. The model training method according to claim 1, wherein: The method meets at least one of the following requirements: The diagnosis result includes the oral health status, and the test samples include test samples under healthy oral status and test samples under at least one oral disease status; The diagnosis result includes the state of the restoration, and the test samples include test samples of the restoration in different states, wherein the restoration state includes at least one of the occlusal contact state of the restoration, the degree of implant bone integration or the state of the internal preparation.

3. The model training method according to claim 2, wherein: The oral health status includes health, or the diagnosis result of oral diseases, and the oral diseases include temporomandibular joint disorder, tooth disease, occlusal disorder, periodontal tissue disease, restoration problems, and oral and maxillofacial tumors.

4. The model training method according to claim 3, wherein: The diagnosis result also includes at least one of the location, quantity, and degree of the disease.

5. The model training method according to any one of claims 1 to 3, wherein: The detection data includes one or more of optical data, sound data, terahertz data or electromagnetic wave data.

6. The model training method according to claim 5, which meets at least one of the following requirements: The optical data includes a three-dimensional image reconstructed based on the difference in reflection intensity after emitting a predetermined type of light to the surface of the soft and hard tissues of the oral cavity and receiving the optical signal through the sensor; The terahertz wave data includes transmission and reflection data obtained by transmitting and receiving electromagnetic waves to the soft and hard tissues of the oral cavity; The electromagnetic wave data includes X-ray quantity data absorbed by different parts of the oral tissues obtained by emitting electromagnetic waves to the soft and hard tissues of the oral cavity; The sound data includes data generated by the patient's active sound source and data generated by the passive sound source, wherein: The data generated by the active sound source includes the sound data generated during the mandibular movement process, and the data generated by the passive sound source includes the sound data generated by the contact between the diagnosis and treatment tool and the internal structure of the oral cavity.

7. The model training method according to claim 5, wherein: The step of enhancing the detection sample and obtaining the training sample comprises: In the case where the detection data includes the sound data, Processing the sound data through a filter to obtain a first spectrogram sample; The first spectrogram samples are processed by at least one of waveform shift, harmonic distortion or resampling to obtain second spectrogram samples, wherein the training samples include the first spectrogram samples and the second spectrogram samples.

8. The model training method according to claim 7, wherein: The label corresponding to the second spectrogram sample is the same as the corresponding first spectrogram sample.

9. The model training method according to claim 7 or 8, wherein: Processing the sound data through a filter to obtain a first spectrum sample includes: The sound data is processed by a Mel filter to obtain a Mel spectrum graph as the first spectrum graph sample.

10. The model training method according to claim 7 or 8, wherein: The obtaining of the second spectrum map sample by processing the first spectrum map sample by at least one of waveform displacement, harmonic distortion or resampling comprises: Extracting time domain features and frequency domain features according to the first spectrum sample; Acquire time-frequency fusion features according to the time domain features and the frequency domain features; Extracting local features from the time-frequency fusion features, and obtaining dynamic position coding according to the time-frequency fusion features; Extracting context-aware features based on the local features and the dynamic position encoding based on a self-attention mechanism; The context-aware feature and the time-frequency fusion feature are fused to obtain the second spectrogram sample.

11. The model training method according to claim 1, wherein: The self-supervised learning model is a vision transformer model.

12. A method for processing oral diagnosis and treatment data, comprising: Obtain test data of target patients during oral diagnosis and treatment; According to the detection data, a diagnosis result is determined based on a machine learning model, wherein the machine learning model is generated by training according to the method according to any one of claims 1 to 11.

13. The oral diagnosis and treatment data processing method according to claim 12, further comprising: In the case where the detection data includes sound data, converting the sound data of the target patient during oral diagnosis and treatment into a sound spectrum graph; Wherein, the machine learning model determines the diagnosis result by processing the sound spectrogram.

14. An oral cavity detection method, comprising: Collect test data of target patients during oral diagnosis and treatment; According to the detection data, a diagnosis result is determined based on a machine learning model, wherein the machine learning model is generated by training based on detection samples in oral medicine diagnosis and treatment.

15. The oral cavity detection method according to claim 14, wherein: The detection data includes one or more of optical data, sound data, terahertz data or electromagnetic wave data.

16. The oral cavity detection method according to claim 14 or 15, wherein: The machine learning model is generated by training according to the model training method according to any one of claims 1 to 11.

17. A model training device, comprising: A sample acquisition unit, configured to acquire a test sample, wherein the test sample includes test data and a label; A sample enhancement unit, configured to enhance the detection sample and obtain a training sample; A training unit is configured to process the training samples through a self-supervised learning model until the training is completed, wherein the parameters of the self-supervised learning model are adjusted according to the output diagnosis result and the label.

18. An oral diagnosis and treatment data processing device, comprising: A data acquisition unit, configured to acquire detection data of a target patient during oral diagnosis and treatment; A state determination unit is configured to determine a diagnosis result based on the detection data and a machine learning model, wherein the machine learning model is generated by training according to the model training method according to any one of claims 1 to 11.

19. A data processing device comprising: Memory; as well as A processor coupled to the memory, the processor being configured to execute the method according to any one of claims 1 to 13 based on instructions stored in the memory.

20. A non-transitory computer-readable storage medium having computer program instructions stored thereon, which implement the method of any one of claims 1 to 13 when executed by a processor.

21. A computer program product, comprising a computer program or instructions, wherein when the computer program or instructions are executed by a processor, the method according to any one of claims 1 to 13 is implemented.

22. An electronic device, comprising: A data collection device configured to collect test data of a target patient during oral diagnosis and treatment; A data processor is configured to determine a diagnosis result based on the detection data and a machine learning model, wherein the machine learning model is generated based on detection sample training.

23. The electronic device according to claim 22, wherein: The data acquisition device includes at least one of the following combinations: Laser sources, laser detectors and laser scanning systems; Electromagnetic wave source and electromagnetic wave receiver; Sound sensor; or Terahertz wave source and terahertz wave detector.

Images