An auxiliary diagnosis system for orthodontic diseases based on deep learning

Abstract

The present invention relates to the technical field of orthodontic disease auxiliary diagnosis, and in particular to an oral orthodontic disease auxiliary diagnosis system based on deep learning, comprising: an image processing module, used to detect an image quality reference coefficient and perform image segmentation; a data processing module, used to determine a preliminary file set and determine the characteristic state of each preliminary file set; a data screening module, used to determine the processing method corresponding to each preliminary file set according to the characteristic state of each preliminary file set; a first analysis module, used to determine the matching method of target patient information and the initial file set according to the target patient category; a second analysis module, used to determine the matching method of target patient information and the initial file set according to the abnormal tooth state of the target patient; a report recommendation module, used to generate a recommendation report from the matching historical file and send it to a client; a data storage module, used to store patient files. The present invention improves the accuracy of file matching and the processing efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of orthodontic disease auxiliary diagnosis, and in particular to an oral orthodontic disease auxiliary diagnosis system based on deep learning. Background Art

[0002] The existing deep learning-based auxiliary diagnosis technology for oral orthodontic diseases fails to determine the targeted treatment method according to the characteristic status of the file in the actual matching process, which leads to the failure to select the matching method in a targeted manner in the subsequent file matching process, resulting in low accuracy and processing efficiency of file matching.

[0003] Chinese patent publication number CN111709959B discloses a digital intelligent diagnosis method for oral orthodontics, including: performing tooth segmentation through three-dimensional scanning and CBCT image data, and constructing a three-dimensional gum model, using three-dimensional features as case classification features, establishing a case database, realizing automatic diagnosis of symptoms based on a deep learning method, and establishing a three-dimensional visualized dynamic treatment process simulation method based on a collision detection mechanism of the tooth treatment process. However, the above technical solution failed to determine a targeted processing method according to the characteristic state of the file in the actual matching process, which resulted in the failure to select a matching method in a targeted manner in the subsequent file matching process, resulting in low accuracy and processing efficiency of file matching. Summary of the invention

[0004] To this end, the present invention provides an oral orthodontic disease auxiliary diagnosis system based on deep learning, which is used to overcome the problem that the prior art fails to determine a targeted processing method according to the characteristic state of the file in the actual matching process, which leads to the failure to select a matching method in a targeted manner in the subsequent file matching process, resulting in low accuracy of file matching and processing efficiency.

[0005] To achieve the above objectives, the present invention provides an oral orthodontic disease auxiliary diagnosis system based on deep learning, comprising:

[0006] An image processing module is used to detect the image quality reference coefficient of the historical detection images in the historical archives, and perform image segmentation on the target detection images, historical detection images and past detection images;

[0007] A data processing module is used to determine a preliminary selection file set according to the target patient and the object information of each historical file and the medical record information, and perform feature learning on each preliminary selection file set to determine the feature state of each preliminary selection file set;

[0008] A data screening module, which is connected to the image processing module and the data processing module respectively, and is used to determine the processing method corresponding to each preliminary selected file set according to the characteristic state of each preliminary selected file set, and the processing method is to determine the matching method of the target patient information and the initial file set according to the target patient category or the target patient's dental abnormality state, and perform secondary screening on the matching historical files according to the image quality reference coefficient of each matching historical file;

[0009] A first analysis module, which is connected to the data screening module, is used to determine a matching method between the target patient information and the initial archive set according to the target patient category, wherein the matching method includes performing archive matching according to image similarity between a target detection image of the target patient and a historical detection image of each historical archive in the initial archive set, or performing archive matching according to a matching coefficient of each historical archive in the initial archive set;

[0010] A second analysis module, which is connected to the data screening module, is used to determine a matching method between the target patient information and the initial archive set according to the target patient's tooth abnormality state, wherein the matching method includes performing archive matching according to the feature similarity of the abnormal morphology teeth in the target detection image and the historical detection image in the initial archive set, or sending the target patient's object information and medical record information to the user;

[0011] A report recommendation module, which is connected to the data screening module and is used to generate a recommendation report by matching historical archives and send it to the client;

[0012] The data storage module is respectively connected to the image processing module, the data processing module and the data screening module, and is used to store historical files, past files and newly added files.

[0013] Furthermore, the data processing module determines a preliminary selection file set based on the target patient and the object information and medical record information of each historical file;

[0014] The preliminary selected archive set is a historical archive set whose object information difference value is less than a preset information difference value, and the information similarity between historical archives in any single historical archive set is greater than the preset information similarity.

[0015] Furthermore, the data processing module performs feature learning for each preliminary selected file set. The feature learning includes determining the feature state of the preliminary selected file set according to the feature position difference value and feature morphology difference value corresponding to the single preliminary selected file set. The feature state includes:

[0016] A first preset characteristic state in which the characteristic position difference value is greater than a preset characteristic position difference value or the characteristic morphology difference value is greater than a preset characteristic morphology difference value;

[0017] A second preset characteristic state in which the characteristic position difference value is less than or equal to the preset characteristic position difference value and the characteristic morphology difference value is less than or equal to the preset characteristic morphology difference value.

[0018] Furthermore, the feature position difference value is determined according to formula (1):

[0019] (1)

[0020] In formula (1), m is the number of historical archives in each preliminary archive set, is the reference value of abnormal tooth distance of the jth historical file, is the average value of the abnormal tooth distance reference values ​​of each historical archive;

[0021] The characteristic morphological difference value is determined according to formula (2):

[0022] (2)

[0023] In formula (2), is the reference value of abnormal tooth morphology of the jth historical file, It is the average value of abnormal tooth morphology reference values ​​in each historical archive.

[0024] Furthermore, the data screening module determines the processing method corresponding to each preliminary selected file set according to the characteristic state of each preliminary selected file set;

[0025] If the preliminary selected archive set is in the first preset characteristic state, the first analysis module determines a matching method between the target patient information and the initial archive set according to the target patient category;

[0026] If the initially selected file set is in the second preset characteristic state, the second analysis module determines a matching method between the target patient information and the initial file set according to the target patient's dental abnormality state.

[0027] Furthermore, if the preliminary selected archive set is in the first preset characteristic state, the first analysis module determines a matching method between the target patient information and the initial archive set according to the target patient category, and the matching method includes:

[0028] If the target patient is a newly added patient, file matching is performed based on the image similarity between the target detection image of the target patient and the historical detection images of each historical file in the initial file set;

[0029] If the target patient is a past patient, file matching is performed based on the matching coefficients of each historical file in the initial file set;

[0030] The target patient category is determined based on whether there is target patient information in the data storage module. Past patients are target patients whose object information exists in the data storage module and whose correction process has not been completed. Target patients whose object information does not exist in the data storage module are recorded as new patients.

[0031] Furthermore, the first analysis module determines the matching coefficient of each historical file in the initial file set according to the image similarity between the target detection image of the target patient and each historical detection image in the corresponding initial file set and the corrected reference difference value with each corresponding historical file. The matching coefficient is determined according to formula (3):

[0032] (3)

[0033] In formula (3), P is the matching coefficient of each historical file in the initial file set, S is the image similarity between the target detection image of the target patient and each historical detection image in the initial file set, As is the influence coefficient corresponding to the image similarity, C is the corrected reference difference value between the target patient and each historical file in the initial file set, and Ac is the influence coefficient corresponding to the corrected reference difference value.

[0034] Furthermore, if the preliminary selected file set is in the second preset characteristic state, the second analysis module determines a matching method between the target patient information and the initial file set according to the target patient's dental abnormality state, and the matching method includes:

[0035] If the tooth abnormality state is in a class of tooth abnormality state, file matching is performed based on the feature similarity between the target detection image and the abnormal morphology teeth in the historical detection images in the initial file set;

[0036] If the tooth abnormality state is in the second category of tooth abnormality state, the object information and medical record information of the target patient are sent to the user.

[0037] Furthermore, the tooth abnormality is determined based on whether there are abnormally worn teeth among the target patient's abnormally shaped teeth. The tooth abnormality includes:

[0038] A type of dental abnormality in which abnormally shaped teeth do not have abnormally worn teeth;

[0039] Abnormal morphology teeth include abnormally worn teeth in the second category of abnormal tooth conditions;

[0040] The abnormally worn teeth are teeth with abnormal morphology whose tooth ratio difference value is greater than a preset difference value.

[0041] Further, if the number of matching historical files screened by the data screening module is greater than the preset number of files, the data screening module performs secondary screening on the matching historical files according to the image quality reference coefficient of each matching historical file, and selects the matching historical files of the preset number of files;

[0042] Among them, the image quality reference coefficients of the matching historical archives obtained from the secondary screening are all greater than the preset detection image quality coefficient. If the number of matching historical archives whose image quality reference coefficients are greater than the preset detection image quality coefficient is less than the preset number of archives, only the matching historical archives whose image quality reference coefficients are greater than the preset detection image quality coefficient are selected.

[0043] Compared with the prior art, the beneficial effect of the present invention lies in that the data processing module in the technical solution of the present invention determines the characteristic state of the preliminary file set according to the characteristic position difference value and the characteristic morphology difference value of each preliminary file set, and the data screening module determines the corresponding processing method of each preliminary file set according to the characteristic state of each preliminary file set, and then determines the corresponding matching method, so that the matching method meets the actual matching scenario, so as to improve the accuracy of the matching result and the efficiency of the matching process.

[0044] Furthermore, the data processing module in the present invention determines a preliminary file set based on the object information and medical record information of the target patient and each historical file, obtains a historical file set based on the age range, growth stage and plaque index, keratinized gingival width reference value and tooth looseness reference value of the historical patients corresponding to each historical file, and determines a preliminary file set based on the difference value of the object information between each historical file set and the target patient, thereby narrowing the matching range for the subsequent file matching process. The present invention improves the efficiency of the file matching process, thereby improving the processing efficiency of auxiliary diagnosis.

[0045] Furthermore, the first analysis module in the present invention determines the matching method between the target patient information and the initial file set according to the target patient category. When the target patient is a past patient, file matching is performed according to the matching coefficients of each historical file in the initial file set. The matching coefficient of each historical file is determined based on the image similarity between the target detection image of the target patient and the corresponding historical detection images in the initial file set and the correction reference difference value with the corresponding historical files, so as to match the historical files with similar correction processes, thereby further optimizing the file matching results. The present invention improves the accuracy of the file matching results.

[0046] Furthermore, the second analysis module in the present invention determines the abnormal state of the teeth according to whether there are abnormally worn teeth among the abnormally morphed teeth of the target patient, and determines the matching method of the target patient information and the initial archive set according to the abnormal state of the teeth of the target patient. When there are abnormally worn teeth among the abnormally morphed teeth of the target patient, the user is reminded in time. When there are no abnormally worn teeth among the abnormally morphed teeth of the target patient, archive matching is performed based on the feature similarity between the target detection image and the abnormally morphed teeth in the historical detection images in the initial archive set, thereby avoiding the influence of the presence of abnormally worn teeth on the accuracy of the matching results. The present invention improves the accuracy of the archive matching results. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a module connection diagram of the deep learning-based oral orthodontic disease auxiliary diagnosis system of the present invention;

[0048] Figure 2 This is a flow chart of the present invention for determining the processing method corresponding to each preliminary selected file set according to the characteristic state of each preliminary selected file set;

[0049] Figure 3 A flowchart of the present invention for determining a matching method between target patient information and the initial file set according to the target patient category;

[0050] Figure 4 The present invention is a flow chart of determining a matching method between target patient information and the initial file set according to the target patient's dental abnormality status. DETAILED DESCRIPTION

[0051] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0052] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0053] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0054] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0055] See also Figures 1 to 4 As shown, the present invention provides an auxiliary diagnosis system for oral orthodontic diseases based on deep learning, comprising:

[0056] An image processing module is used to detect the image quality reference coefficient of the historical detection images in the historical archives, and perform image segmentation on the target detection images, historical detection images and past detection images;

[0057] A data processing module is used to determine a preliminary selection file set according to the target patient and the object information of each historical file and the medical record information, and perform feature learning on each preliminary selection file set to determine the feature state of each preliminary selection file set;

[0058] A data screening module, which is connected to the image processing module and the data processing module respectively, and is used to determine the processing method corresponding to each preliminary selected file set according to the characteristic state of each preliminary selected file set, and the processing method is to determine the matching method of the target patient information and the initial file set according to the target patient category or the target patient's dental abnormality state, and perform secondary screening on the matching historical files according to the image quality reference coefficient of each matching historical file;

[0059] A first analysis module, which is connected to the data screening module, is used to determine a matching method between the target patient information and the initial archive set according to the target patient category, wherein the matching method includes performing archive matching according to image similarity between a target detection image of the target patient and a historical detection image of each historical archive in the initial archive set, or performing archive matching according to a matching coefficient of each historical archive in the initial archive set;

[0060] A second analysis module, which is connected to the data screening module, is used to determine a matching method between the target patient information and the initial archive set according to the target patient's tooth abnormality state, wherein the matching method includes performing archive matching according to the feature similarity of the abnormal morphology teeth in the target detection image and the historical detection image in the initial archive set, or sending the target patient's object information and medical record information to the user;

[0061] A report recommendation module, which is connected to the data screening module and is used to generate a recommendation report by matching historical archives and send it to the client;

[0062] The data storage module is respectively connected to the image processing module, the data processing module and the data screening module, and is used to store historical files, past files and newly added files.

[0063] Among them, the patient files in the present invention include historical files, new files and past files. The historical files include the object information, medical record information and several historical detection images of historical patients. The past files include the object information, medical record information and several past detection images of past patients. The historical patients are patients whose correction process has been completed. The target patients are objects that need auxiliary diagnosis. The target patients include new patients and past patients. If the target patient is a new patient, a new file is created based on the object information, medical record information and target detection image of the target patient and recorded as a new file. If the target patient is a past patient, the target detection image is saved in the corresponding past file. The object information is the name, age and number and volume of erupted wisdom teeth of the target patient or historical patient. The medical record information is the name of the disease and tooth characteristics of the target patient or historical patient. The medical record information of historical patients and past patients also includes the incentives of the correction process and the correction reference values ​​at different correction stages. The target detection images, past detection images and historical detection images in the present invention The measured images are all full-mouth curved surface tomograms, which are two-dimensional images showing structures such as the upper and lower jaws, temporomandibular joints, full-mouth teeth and periodontal tissues. The image processing module performs image segmentation on the target detection image, the past detection image and the historical detection image, including preprocessing the image to enhance image contrast and remove image noise, extracting image features and determining different segmentation algorithms according to the difficulty of image segmentation. How to use X-rays to obtain full-mouth curved surface tomograms and perform image segmentation to obtain single tooth images is easy for technicians in this field to understand and will not be elaborated here. In the present invention, the user performs abnormal morphological feature detection on the teeth in each single tooth image obtained by segmentation of the target detection image, the past detection image and the historical detection image, and determines the abnormal morphological teeth and their abnormal morphological features. The abnormal morphological features include but are not limited to abnormal growth position, abnormal tooth contour and abnormal axial inclination angle. The abnormal morphological feature detection results of each single tooth image are stored together with the corresponding image in the data storage module.

[0064] Specifically, the data processing module determines the preliminary selection file set based on the target patient and the object information and medical record information of each historical file;

[0065] The preliminary selected archive set is a historical archive set whose object information difference value is less than a preset information difference value, and the information similarity between historical archives in any single historical archive set is greater than the preset information similarity.

[0066] Among them, the disease name and tooth characteristics are extracted from the medical records of the target patient and the historical patients, and the disease name and tooth characteristics in the medical records of the target patient are recorded as key information. The historical archives with key information in the medical records are divided according to the age range, growth stage and plaque index, keratinized gingival width reference value and tooth looseness reference value of the historical patients corresponding to each historical archive, so as to obtain a historical archive set. For a single historical archive set, each historical archive in the historical archive set is in the same age range and the same growth stage. The growth stage is determined according to the growth parameters of wisdom teeth. The growth parameters of wisdom teeth = the number of erupted wisdom teeth × 50% + the average eruption volume of wisdom teeth × 50%. The average eruption volume of wisdom teeth = the sum of the volumes of each erupted wisdom tooth / the number of erupted wisdom teeth. Users can perform adaptive division according to age range and growth stage according to actual needs;

[0067] The information similarity = 1 / (plaque difference value × 30% + keratinized gingival width difference value × 30% + tooth looseness difference value × 40%), the plaque difference value is the absolute value of the difference between the plaque index of the historical patients corresponding to each historical file and the target patient, the plaque index = the sum of the plaque thickness on the surface of each tooth with plaque / the number of teeth with plaque, the keratinized gingival width difference value is the absolute value of the difference between the keratinized gingival width reference value of the historical patients corresponding to each historical file and the target patient, the keratinized gingival width reference value is the average value of the keratinized gingival width of each tooth, and the tooth looseness difference value is the average value of the keratinized gingival width of each tooth. The absolute value of the difference between the tooth looseness reference value of the historical patient corresponding to the historical file and the target patient, the tooth looseness reference value = the sum of the looseness amplitudes of each loose tooth / the number of loose teeth, the value of the preset information similarity, the user can set it according to actual needs and historical records, the higher the user's requirements for the accuracy of the file matching results, the greater the preset information similarity, and a method for setting the value of the preset information similarity is provided, and the minimum value of the information similarity of the historical files in each historical file set in the historical record of the user's requirements for the accuracy of the file matching results is recorded as the preset information similarity;

[0068] The object information difference value is the sum of the age difference value and the growth stage difference value. The age difference value is the absolute value of the difference between the target patient's age and the average age in each historical archive set. The growth stage difference value is the absolute value of the difference between the wisdom tooth growth parameter of the target patient and the average wisdom tooth growth parameter in each historical archive set. The value of the preset object information difference value can be set by the user according to actual needs and historical records. The higher the user's requirements for the accuracy of the archive matching results, the smaller the preset object information difference value. A method for setting the preset object information difference value is provided, and the average value of the object information difference values ​​of each preliminary archive set and the target patient in the historical records of the user's requirements for the accuracy of the archive matching results is recorded as the preset object information difference value.

[0069] Specifically, the data processing module performs feature learning for each preliminary selected file set. Feature learning includes determining the feature state of the preliminary selected file set according to the feature position difference value and feature morphology difference value corresponding to the single preliminary selected file set. The feature state includes:

[0070] A first preset characteristic state in which the characteristic position difference value is greater than a preset characteristic position difference value or the characteristic morphology difference value is greater than a preset characteristic morphology difference value;

[0071] A second preset characteristic state in which the characteristic position difference value is less than or equal to the preset characteristic position difference value and the characteristic morphology difference value is less than or equal to the preset characteristic morphology difference value.

[0072] Specifically, the feature position difference value is determined according to formula (1):

[0073] (1)

[0074] In formula (1), m is the number of historical archives in each preliminary archive set, is the reference value of abnormal tooth distance of the jth historical file, is the average value of the abnormal tooth distance reference values ​​of each historical archive;

[0075] The characteristic morphological difference value is determined according to formula (2):

[0076] (2)

[0077] In formula (2), is the reference value of abnormal tooth morphology of the jth historical file, It is the average value of abnormal tooth morphology reference values ​​in each historical archive.

[0078] Among them, the abnormal tooth distance reference value is the average value of the interval distance between adjacent abnormal teeth in each historical file, the abnormal tooth morphology reference value f=the total contact area of ​​the whole dentition of each historical file×50%+the tooth offset reference value of each historical file×50%, the total contact area of ​​the whole dentition is the sum of the contact areas of the whole dentition during the occlusion process, the present invention uses the deformation of silicone rubber to determine the contact area of ​​the whole dentition, and the contact area of ​​the whole dentition is determined by measuring the deformation area of ​​the silicone rubber after the occlusion is completed, the tooth offset reference value is the average value of the tooth offset value of each tooth in each historical file, the tooth offset value is the absolute value of the difference between the axial inclination angle of each tooth and its standard axial inclination angle, how to use silicone rubber and determine the contact area of ​​the whole dentition according to the deformation of silicone rubber and set the standard axial inclination angle of each tooth is easy to understand for those skilled in the art, and will not be elaborated here;

[0079] The values ​​of the preset feature position difference value and the preset feature morphology difference value can be set by the user according to actual needs and historical records. The higher the user's requirements for the accuracy of the file matching results, the smaller the value of the preset feature position difference value is, and the smaller the value of the preset feature morphology difference value is. A method for determining the preset feature position difference value is provided, and the average value of the feature position difference values ​​in each preliminary file set in the historical records that meet the user's requirements for the accuracy of the file matching results is recorded as the preset feature position difference value. A method for determining the preset feature morphology difference value is provided, and the average value of the feature morphology difference values ​​in each preliminary file set in the historical records that meet the user's requirements for the accuracy of the file matching results is recorded as the preset feature morphology difference value.

[0080] Specifically, the data screening module determines a number of preliminary selected file sets according to the target patient information and determines the corresponding processing method for each preliminary selected file set according to the characteristic state of each preliminary selected file set;

[0081] If the initial selected file set is in the first preset characteristic state, the first analysis module determines a matching method between the target patient information and the initial file set according to the target patient category;

[0082] If the initially selected file set is in the second preset characteristic state, the second analysis module determines a matching method between the target patient information and the initial file set according to the target patient's dental abnormality state.

[0083] Specifically, if the preliminary selected file set is in the first preset characteristic state, the first analysis module determines the matching method between the target patient information and the initial file set according to the target patient category, and the matching method includes:

[0084] If the target patient is a newly added patient, file matching is performed based on the image similarity between the target detection image of the target patient and the historical detection images of each historical file in the initial file set;

[0085] If the target patient is a past patient, file matching is performed based on the matching coefficients of each historical file in the initial file set;

[0086] The target patient category is determined based on whether there is target patient information in the data storage module. Past patients are target patients whose object information exists in the data storage module and whose correction process has not been completed. Target patients whose object information does not exist in the data storage module are recorded as new patients.

[0087] Among them, the image similarity , n is the number of evaluation parameters used to determine image similarity, is the parameter reference value of the i-th evaluation parameter, is the weight coefficient corresponding to the i-th evaluation parameter. The user can determine the type of evaluation parameter and its corresponding weight coefficient according to actual needs and set them. In the present invention, the image similarity is determined according to the contour similarity reference value, the axis tilt angle difference reference value and the edge similarity reference value. The contour similarity reference value is the similarity of the overall tooth contours of the target detection image and each historical detection image. The axis tilt angle difference reference value is the average value of the absolute values ​​of the axis tilt angle differences of the target detection image and each historical detection image of the same category of teeth. The edge similarity reference value is the average value of the edge similarity of the target detection image and each historical detection image of the same category of teeth. A contour similarity reference value, an axis tilt angle difference reference value and a weight coefficient corresponding to the edge similarity reference value are provided. The weight coefficient corresponding to the contour similarity reference value is 40%, the weight coefficient corresponding to the axis tilt angle difference reference value is 30%, and the weight coefficient corresponding to the edge similarity reference value is 30%;

[0088] The historical archives corresponding to the historical detection images whose image similarity is greater than the preset image similarity are recorded as matching historical archives. The user can set the value of the preset image similarity according to actual needs and historical records. The higher the user's requirements for the accuracy of the matching results, the smaller the value of the preset image similarity. A method for setting the value of the preset image similarity is provided, and the average value of the image similarity between the historical detection images and the target detection images in each matching historical archive in the historical records that meet the user's requirements for the accuracy of the matching results is recorded as the preset image similarity.

[0089] Specifically, the first analysis module determines the matching coefficient of each historical file in the initial file set according to the image similarity between the target detection image of the target patient and each historical detection image in the corresponding initial file set and the corrected reference difference value with each corresponding historical file. The matching coefficient is determined according to formula (3):

[0090] (3)

[0091] In formula (3), P is the matching coefficient of each historical file in the initial file set, S is the image similarity between the target detection image of the target patient and each historical detection image in the initial file set, As is the influence coefficient corresponding to the image similarity, C is the corrected reference difference value between the target patient and each historical file in the initial file set, and Ac is the influence coefficient corresponding to the corrected reference difference value.

[0092] Among them, the correction monitoring module in the present invention detects the correction reference value of historical patients, and the correction reference value = the correction value of the target correction tooth / the expected correction value of the target correction tooth, and the correction value is the movement distance or the axis inclination angle change value of the target correction tooth compared with the oral detection before the current detection. The correction value is determined according to the abnormal morphological characteristics of the target correction tooth according to the movement distance or the axis inclination angle change value, and the expected correction value is set by the user according to actual needs. In addition, the technical personnel in this field can easily understand the content and will not be repeated here. For a single historical file, the absolute value of the difference between the abnormal morphological teeth of the same type of each target correction tooth in the same correction stage and the correction reference value of the corresponding target correction tooth is recorded as the correction difference value, and the average value of each correction difference value is recorded as the correction reference difference value of the historical file. The target correction tooth is the abnormal morphological tooth corrected in the medical records of historical patients and past patients;

[0093] The historical archives whose matching coefficients are greater than the preset matching coefficients are recorded as matching historical archives. The user can set the value of the preset matching coefficient according to actual needs and historical records. The higher the user's requirements for the accuracy of the matching results, the smaller the value of the preset matching coefficient. A method for setting the value of the preset matching coefficient is provided, and the average value of the matching coefficients of each matching historical archive in the historical records that meet the user's requirements for the accuracy of the matching results is recorded as the preset image similarity.

[0094] Specifically, if the preliminary selected file set is in the second preset characteristic state, the second analysis module determines the matching method between the target patient information and the initial file set according to the target patient's dental abnormality state, and the matching method includes:

[0095] If the tooth abnormality state is in a class of tooth abnormality state, file matching is performed based on the feature similarity between the target detection image and the abnormal morphology teeth in the historical detection images in the initial file set;

[0096] If the tooth abnormality state is in the second category of tooth abnormality state, the object information and medical record information of the target patient are sent to the user.

[0097] Among them, the feature similarity , x is the number of similar teeth between each historical detection image and the target detection image, μ1 is the influence coefficient corresponding to the number of similar teeth, y is the difference value of misaligned teeth between each historical detection image and the target detection image, μ2 is the influence coefficient corresponding to the difference value of misaligned teeth, the number of similar teeth is the number of teeth whose tooth width difference value and tooth length difference value are both less than the preset size difference value, the tooth width difference value is the absolute value of the difference in length of the widest part of the crown of the two teeth, and the tooth length difference value is the absolute value of the difference in vertical distance from the incisal edge or tooth tip to the gingival margin of the two teeth. Users can set the value of the preset size difference value according to actual needs and historical records The higher the user's requirement for the accuracy of the matching result, the smaller the preset size difference value is. A method for determining the preset size difference value is provided. The maximum value of the tooth width difference value and the tooth length difference value of each similar tooth in the historical records that meet the user's requirement for the accuracy of the matching result is recorded as the preset size difference value. The misaligned tooth difference value is the absolute value of the difference between the number of misaligned teeth in each historical detection image and the target detection image. The misaligned tooth is a tooth whose growth direction is inconsistent with the preset growth direction of the tooth category. How to set the preset growth direction for each category of teeth is easy for technicians in this field to understand and will not be elaborated here.

[0098] The historical archives corresponding to the historical detection images whose feature similarity is greater than the preset feature similarity are recorded as matching historical archives. The user can set the value of the preset feature similarity according to actual needs and historical records. The higher the user's requirement for the accuracy of the matching results, the smaller the preset feature similarity. A method for determining the value of the preset image similarity is provided. The average value of the feature similarity between the historical detection images and the target detection images in each matching historical archive in the historical records that meet the user's requirement for the accuracy of the matching results is recorded as the preset feature similarity. A value of the preset feature similarity is provided, and the value of the preset feature similarity is 0.75.

[0099] Specifically, the tooth abnormality is determined based on whether there are abnormally worn teeth among the abnormally morphological teeth of the target patient, and the tooth abnormality includes:

[0100] A type of dental abnormality in which abnormally shaped teeth do not have abnormally worn teeth;

[0101] Abnormal morphology teeth include abnormally worn teeth in the second category of abnormal tooth conditions;

[0102] The abnormally worn teeth are teeth with abnormal morphology whose tooth ratio difference value is greater than a preset difference value.

[0103] Among them, the tooth ratio value = the length of the widest part of the crown of each tooth / the vertical distance from the incisal edge or the tip of each tooth to the gingival edge. For a single tooth, the absolute value of the difference between the tooth ratio value and the preset tooth ratio value corresponding to the tooth type is recorded as the tooth ratio difference value. The value of the preset difference value can be set by the user according to actual needs and historical records. The higher the user's requirement for the accuracy of the matching result, the smaller the value of the preset difference value. The maximum value of each tooth ratio difference value in each matching historical archive in the historical record that meets the user's requirement for the accuracy of the matching result is recorded as the preset difference value.

[0104] Specifically, if the number of matching historical files screened by the data screening module is greater than the preset number of files, the data screening module performs secondary screening on the matching historical files according to the image quality reference coefficients of the matching historical files, and selects the preset number of matching historical files;

[0105] Among them, the image quality reference coefficients of the matching historical archives obtained from the secondary screening are all greater than the preset detection image quality coefficient. If the number of matching historical archives whose image quality reference coefficients are greater than the preset detection image quality coefficient is less than the preset number of archives, only the matching historical archives whose image quality reference coefficients are greater than the preset detection image quality coefficient are selected.

[0106] For a single historical archive, the image quality reference coefficient is the average value of the image quality coefficients of each historical detection image in the historical archive. , D is the contrast of the historical detection image, is the weight coefficient corresponding to the contrast of the historical detection image, G is the interference area of ​​the historical detection image, The influence coefficient corresponding to the interference area of ​​the historical detection image. The user can set the value of the influence coefficient corresponding to the contrast of the historical detection image and the interference area according to actual needs and historical records. The greater the influence of the parameter on the image quality, the greater the influence coefficient corresponding to the parameter. Provide a value of the influence coefficient corresponding to the contrast of the historical detection image and the interference area. The value of the influence coefficient corresponding to the contrast of the historical detection image is 0.4, and the value of the influence coefficient corresponding to the interference area of ​​the historical detection image is 0.6;

[0107] The user can set the value of the preset detection image quality coefficient according to actual needs and historical records. The higher the user's requirements for the matching results, the larger the value of the preset detection image quality coefficient. A method for determining the value of the preset detection image quality coefficient is provided, and the average value of the detection image quality coefficients of historical detection images in the matching history archives in the historical records that meet the user's higher requirements for the matching results is recorded as the preset detection image quality coefficient.

[0108] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0109] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A deep learning-based auxiliary diagnosis system for orthodontic diseases, characterized in that: include: An image processing module is used to detect the image quality reference coefficient of the historical detection images in the historical archives, and perform image segmentation on the target detection images, historical detection images and past detection images; A data processing module is used to determine a preliminary selection file set according to the target patient and the object information of each historical file and the medical record information, and perform feature learning on each preliminary selection file set to determine the feature state of each preliminary selection file set; A data screening module, which is connected to the image processing module and the data processing module respectively, and is used to determine the processing method corresponding to each preliminary selected file set according to the characteristic state of each preliminary selected file set, and the processing method is to determine the matching method of the target patient information and the initial file set according to the target patient category or the target patient's dental abnormality state, and perform secondary screening on the matching historical files according to the image quality reference coefficient of each matching historical file; A first analysis module, which is connected to the data screening module, is used to determine a matching method between the target patient information and the initial archive set according to the target patient category, wherein the matching method includes performing archive matching according to image similarity between a target detection image of the target patient and a historical detection image of each historical archive in the initial archive set, or performing archive matching according to a matching coefficient of each historical archive in the initial archive set; A second analysis module, which is connected to the data screening module, is used to determine a matching method between the target patient information and the initial archive set according to the target patient's tooth abnormality state, wherein the matching method includes performing archive matching according to the feature similarity of the abnormal morphology teeth in the target detection image and the historical detection image in the initial archive set, or sending the target patient's object information and medical record information to the user; A report recommendation module, which is connected to the data screening module and is used to generate a recommendation report by matching historical archives and send it to the client; The data storage module is respectively connected to the image processing module, the data processing module and the data screening module, and is used to store historical files, past files and newly added files.

2. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 1, characterized in that: The data processing module determines the preliminary selection file set based on the target patient and the object information of each historical file and the medical record information; The preliminary selected archive set is a historical archive set whose object information difference value is less than a preset information difference value, and the information similarity between historical archives in any single historical archive set is greater than the preset information similarity.

3. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 2, characterized in that: The data processing module performs feature learning on each set of preliminary selected archives. The method includes determining the characteristic state of the primary selected file set according to the characteristic position difference value and the characteristic morphology difference value corresponding to the single primary selected file set, and the characteristic state includes: A first preset characteristic state in which the characteristic position difference value is greater than a preset characteristic position difference value or the characteristic morphology difference value is greater than a preset characteristic morphology difference value; A second preset characteristic state in which the characteristic position difference value is less than or equal to the preset characteristic position difference value and the characteristic morphology difference value is less than or equal to the preset characteristic morphology difference value.

4. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 3, characterized in that: The feature position difference value is determined according to formula (1): （1） In formula (1), m is the number of historical archives in each preliminary archive set, is the reference value of abnormal tooth distance of the jth historical file, is the average value of the abnormal tooth distance reference values ​​of each historical archive; The characteristic morphological difference value is determined according to formula (2): （2） In formula (2), is the reference value of abnormal tooth morphology of the jth historical file, It is the average value of abnormal tooth morphology reference values ​​in each historical archive.

5. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 4, characterized in that: The data screening module determines the processing method corresponding to each preliminary selected file set according to the characteristic status of each preliminary selected file set; If the preliminary selected archive set is in the first preset characteristic state, the first analysis module determines a matching method between the target patient information and the initial archive set according to the target patient category; If the initially selected file set is in the second preset characteristic state, the second analysis module determines a matching method between the target patient information and the initial file set according to the target patient's dental abnormality state.

6. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 5, characterized in that: If the preliminary selected file set is in the first preset characteristic state, the first analysis module determines a matching method between the target patient information and the initial file set according to the target patient category, and the matching method includes: If the target patient is a newly added patient, file matching is performed based on the image similarity between the target detection image of the target patient and the historical detection images of each historical file in the initial file set; If the target patient is a past patient, file matching is performed based on the matching coefficients of each historical file in the initial file set; The target patient category is determined based on whether there is target patient information in the data storage module. Past patients are target patients whose object information exists in the data storage module and whose correction process has not been completed. Target patients whose object information does not exist in the data storage module are recorded as new patients.

7. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 6, characterized in that: The first analysis module determines the matching coefficient of each historical file in the initial file set according to the image similarity between the target detection image of the target patient and each historical detection image in the corresponding initial file set and the corrected reference difference value with each corresponding historical file. The matching coefficient is determined according to formula (3): （3） In formula (3), P is the matching coefficient of each historical file in the initial file set, S is the image similarity between the target detection image of the target patient and each historical detection image in the initial file set, As is the influence coefficient corresponding to the image similarity, C is the corrected reference difference value between the target patient and each historical file in the initial file set, and Ac is the influence coefficient corresponding to the corrected reference difference value.

8. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 7, characterized in that: If the preliminary selected file set is in the second preset characteristic state, the second analysis module determines a matching method between the target patient information and the initial file set according to the target patient's dental abnormality state, and the matching method includes: If the tooth abnormality state is in a class of tooth abnormality state, file matching is performed based on the feature similarity between the target detection image and the abnormal morphology teeth in the historical detection images in the initial file set; If the tooth abnormality state is in the second category of tooth abnormality state, the object information and medical record information of the target patient are sent to the user.

9. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 8, characterized in that: The tooth abnormality is determined based on whether there are abnormally worn teeth among the abnormally shaped teeth of the target patient. The tooth abnormality includes: A type of dental abnormality in which abnormally shaped teeth do not have abnormally worn teeth; Abnormal morphology teeth include abnormally worn teeth in the second category of abnormal tooth conditions; The abnormally worn teeth are teeth with abnormal morphology whose tooth ratio difference value is greater than a preset difference value.

10. The deep learning-based oral orthodontic disease auxiliary diagnosis system according to claim 1, characterized in that: If the number of matching historical files screened by the data screening module is greater than the preset number of files, the data screening module performs a secondary screening on the matching historical files according to the image quality reference coefficient of each matching historical file, and selects the matching historical files of the preset number of files; Among them, the image quality reference coefficients of the matching historical archives obtained from the secondary screening are all greater than the preset detection image quality coefficient. If the number of matching historical archives whose image quality reference coefficients are greater than the preset detection image quality coefficient is less than the preset number of archives, only the matching historical archives whose image quality reference coefficients are greater than the preset detection image quality coefficient are selected.

Images