Automatic structural processing method and system for dental information by using part recording method

Abstract

The invention provides an automatic structural processing method and system for dental information through a part recording method, and the method comprises the steps: recognizing a quadrant symbol through an FDI information data set to obtain FDI information annotation data, obtaining tooth position annotation data of other data according to a tooth position information data set, carrying out the quadrant / region word and sentence description data verification of the FDI information annotation data, and carrying out the verification of the quadrant / region word and sentence description data. If yes, quadrant / region word and sentence description data, tooth position numbers and identifier recognition is conducted again, post-checking tooth position labeling data and FDI information labeling data are obtained, the post-checking FDI information labeling data are converted into tooth position labeling data, the tooth position labeling data are combined with the post-checking tooth position labeling data, and structured tooth position data are obtained; a dental information automatic structuring model is established according to the mapping relation between the historical patient dental medical record and the structured tooth position data, the patient dental medical record to be processed is processed to obtain the patient structured tooth position data to draw a tooth position part map, and the problem that the structured tooth position data generated by an existing OCR technology is inaccurate is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of dental information technology, in particular to a dental information automatic structured processing method and system based on position recording method. BACKGROUND

[0002] Dental position information is often recorded in unstructured form in medical records. In order to store patient dental position information, hospitals often use OCR technology to identify medical record data and store it in the hospital information system (HIS) and electronic health records (EHR). However, different doctors may use various expression methods to express dental position information, such as dental position recording method and FDI (Federation Dentaire Internationale) recording method. The two recording methods have highly similar layout structures and special symbols, which leads to errors when traditional OCR technology identifies medical record data and converts dental position information into structured dental position information data.

[0003] As shown in Figure 1 and Figure 2 , the dental position information stored in the medical record mixes the dental position recording method and the FDI recording method, resulting in inaccurate structured dental position information data obtained by using OCR technology. This in turn leads to large errors in calling structured dental position information data from the hospital information system (HIS) and electronic health records (EHR) when performing statistical analysis, AI-assisted diagnosis, and remote consultation, which limits the application scenarios of standard data formats. SUMMARY

[0004] The technical problem to be solved by the technical scheme of the present application is that the structured dental position information data obtained by using existing OCR technology is inaccurate.

[0005] The technical scheme of the present application provides a dental information automatic structured processing method based on position recording method, comprising the following steps:

[0006] Step 1: Obtain a historical patient dental medical record, which includes physical examination data, diagnosis data, examination and treatment data, and auxiliary examination data.

[0007] Step 2: According to the FDI information dataset, the quadrant symbol in the physical examination data, diagnosis data, auxiliary examination data and examination processing data in the historical patient dental medical record is identified, if the quadrant symbol exists, the tooth position number and tooth position state are continuously identified and segmented to obtain the FDI information labeled data with tooth position state corresponding to the historical patient dental medical record, if the quadrant symbol does not exist, the quadrant / region word description data, tooth position number, identifier and tooth position state are identified according to the tooth position information dataset to obtain the tooth position site labeled data with tooth position state corresponding to the historical patient dental medical record;

[0008] Step 3: According to the tooth position information dataset, the quadrant / region word description data in the FDI information labeled data is checked, if the corresponding content of the quadrant / region word description data is not found, it is determined that the tooth position site labeled data is correctly identified, the checked tooth position site labeled data and the checked FDI information labeled data are obtained, if the corresponding content of the quadrant / region word description data is found, the quadrant / region word description data, tooth position number, identifier are re-identified to obtain the checked tooth position site labeled data and the checked FDI information labeled data, according to the tooth position information and FDI information conversion set, the checked FDI information labeled data is converted into tooth position site labeled data and combined with the checked tooth position site labeled data to obtain the final tooth position site labeled data as the structured tooth position data;

[0009] Step 4: According to the mapping relationship between the historical patient dental medical record and the structured tooth position data, a dental information automatic structuring model is established;

[0010] Step 5: The patient dental medical record to be processed is input into the dental information automatic structuring model to obtain the patient structured tooth position data corresponding to the patient dental medical record to be processed, and a patient tooth position site map is drawn according to the patient structured tooth position data.

[0011] Preferably, the historical patient dental medical record forms include paper scanning image form, X-ray labeled image form, electronic document image form, handwritten image form and audio form.

[0012] Preferably, the tooth position site information and FDI information conversion set includes tooth position site information dataset, FDI information dataset, and corresponding relationship between tooth position site information dataset and FDI information dataset.

[0013] Preferably, the tooth position site information dataset includes quadrant / region word description data, tooth position number and identifier.

[0014] Preferably, the FDI information dataset includes quadrant symbol and tooth position number.

[0015] Preferably, the content of the tooth site information data set and the content of the FDI information data set are in the form of a paper scan image, an X-ray film annotation image, an electronic document image, a handwriting image, or an audio form.

[0016] The embodiment of the present application also provides a site recording method dental information automatic structured processing system, which adopts the site recording method dental information automatic structured processing method as described above, and comprises:

[0017] A multi-modal input interface module is configured to acquire historical patient dental medical records, and the historical patient dental medical records comprise physical examination data, diagnosis data, examination and treatment data, and auxiliary examination data.

[0018] A dental special tooth site recognition module is configured to perform quadrant symbol recognition on the physical examination data, the diagnosis data, the auxiliary examination data, and the examination and treatment data in the historical patient dental medical records according to the FDI information data set, if there is a quadrant symbol, continue to recognize tooth site numbers and tooth site states and perform segmentation to obtain FDI information annotation data with tooth site states corresponding to the historical patient dental medical records, and if there is no quadrant symbol, perform quadrant / region word description data recognition, tooth site number recognition, identifier recognition, and tooth site state recognition according to the tooth site information data set to obtain tooth site annotation data with tooth site states corresponding to the historical patient dental medical records.

[0019] A dental semantic analysis module is configured to perform quadrant / region word description data checking on the FDI information annotation data according to the tooth site information data set, if no corresponding content of the quadrant / region word description data is found, determine that the tooth site annotation data recognition is correct, obtain checked tooth site annotation data and checked FDI information annotation data, if the corresponding content of the quadrant / region word description data is found, re-perform quadrant / region word description data recognition, tooth site number recognition, and identifier recognition, obtain checked tooth site annotation data and checked FDI information annotation data, convert the checked FDI information annotation data into tooth site annotation data according to the tooth site information and FDI information conversion set, and combine the converted tooth site annotation data with the checked tooth site annotation data to obtain final tooth site annotation data as structured tooth site data.

[0020] A tooth site information automatic structured model generation module is configured to establish a dental information automatic structured model according to a mapping relationship between the historical patient dental medical records and the structured tooth site data.

[0021] A tooth site structured data and tooth site atlas generation module is configured to input a to-be-processed patient dental medical record into the dental information automatic structured model to obtain patient structured tooth site data corresponding to the to-be-processed patient dental medical record, and draw a patient tooth site atlas according to the patient structured tooth site data.

[0022] Preferably, the multi-modal input interface module supports HTTP API interface docking hospital information system, electronic medical record system, integrated speech recognition, and provides a local file upload interface.

[0023] Preferably, the dental site position information dataset and the FDI information dataset are introduced into the text recognition model as a specific vocabulary for text detection and recognition.

[0024] Preferably, the patient dental site position map is drawn using HTML5+Canvas.

[0025] The technical scheme of the present application provides a dental information automatic structured processing method and system, which obtains FDI information annotation data by recognizing quadrant symbols according to the FDI information dataset, obtains dental site position annotation data by recognizing the remaining data according to the dental site position information dataset, checks quadrant / region sentence description data of the FDI information annotation data according to the dental site position information dataset, re-recognizes the quadrant / region sentence description data, dental site number, and identifier if there is quadrant / region sentence description data, obtains checked dental site position annotation data and checked FDI information annotation data, converts the checked FDI information annotation data into dental site position annotation data and combines it with the checked dental site position annotation data to obtain final dental site position annotation data as structured dental site data, establishes a dental information automatic structured model according to the mapping relationship between historical patient dental medical records and structured dental site data to process patient dental medical records to obtain patient structured dental site data and draw a patient dental site position map, and solves the problem of inaccurate structured dental site information data obtained by the existing OCR technology. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 A medical record schematic diagram of the FDI recording method in the prior art;

[0027] Figure 2 A medical record schematic diagram of the dental site position recording method in the prior art;

[0028] Figure 3 A patient dental site position annotation data structured format schematic diagram provided by the embodiment of the present application. DETAILED DESCRIPTION

[0029] The present application will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present application and not used to limit the scope of the present application. In addition, it should be understood that those skilled in the art can make various modifications or changes to the present application after reading the content taught by the present application, and these equivalent forms also fall within the scope defined by the appended claims of the present application.

[0030] The embodiment of the present application provides a dental information automatic structured processing method based on a site recording method, comprising the following steps.

[0031] Step 1: obtaining a historical patient dental record, wherein the historical patient dental record comprises physical examination data, diagnosis data, examination and treatment data and auxiliary examination data; and the historical patient dental record has a form of a paper scanning image, a form of an X-ray film annotation image, a form of an electronic document image, a form of a handwriting image and a form of an audio.

[0032] Step 2: performing quadrant symbol identification on the physical examination data, the diagnosis data, the auxiliary examination data and the examination and treatment data in the historical patient dental record according to an FDI information data set; if there is a quadrant symbol, then continuing to identify a tooth site number and a tooth site state and cutting to obtain FDI information annotation data corresponding to the historical patient dental record and having the tooth site state; and if there is no quadrant symbol, then identifying quadrant / region word description data, a tooth site number, an identifier and a tooth site state according to a tooth site information data set to obtain tooth site annotation data corresponding to the historical patient dental record and having the tooth site state.

[0033] The tooth site information data set for recording tooth site information by using a tooth site recording method comprises quadrant / region word description data, a tooth site number and an identifier; wherein the quadrant / region word description data represents four regions of a patient, i.e., a right upper region, a left upper region, a right lower region and a left lower region, by using ABCD / 1234; the tooth site number comprises a central incisor 1, a lateral incisor 2, a canine 3, a first premolar 4, a second premolar 5, a first molar 6, a second molar 7 and a third molar 8; and the identifier comprises a symbol "-" representing all adjacent tooth sites, for example, 12-21 represents all adjacent tooth sites from 12 to 21, and specifically comprises 12, 11, 21 and 22.

[0034] The FDI information data set for recording tooth site information by using an FDI recording method comprises a quadrant symbol and a tooth site number; wherein the quadrant symbol comprises a right maxilla ┛, a left maxilla ┗, a right mandible ┓ and a left mandible ┏; and the tooth site number comprises a central incisor 1, a lateral incisor 2, a canine 3, a first premolar 4, a second premolar 5, a first molar 6, a second molar 7 and a third molar 8.

[0035] The content forms in the tooth site information data set and the FDI information data set comprise a paper scanning image form, an X-ray film annotation image form, an electronic document image form, a handwriting image form and an audio form.

[0036] Step 3: According to the quadrant / region word description data of the tooth position information data set, the FDI information annotation data is checked, if the corresponding content of the quadrant / region word description data is not found, it is determined that the tooth position annotation data is correct, and the checked tooth position annotation data and the checked FDI information annotation data are obtained, if the corresponding content of the quadrant / region word description data is found, the quadrant / region word description data, tooth position number and identifier are re-identified, and the checked tooth position annotation data and the checked FDI information annotation data are obtained, according to the tooth position information conversion set and the FDI information conversion set, the checked FDI information annotation data is converted into tooth position annotation data and combined with the checked tooth position annotation data, and the final tooth position annotation data is obtained as structured tooth position data.

[0037] The tooth position information conversion set and the FDI information conversion set include a tooth position information data set, an FDI information data set, and a corresponding relationship between the tooth position information data set and the FDI information data set.

[0038] Step 4: According to the mapping relationship between the historical patient dental medical record and the structured tooth position data, a dental information automatic structured model is established.

[0039] Step 5: The patient dental medical record to be processed is input into the dental information automatic structured model, and the patient structured tooth position data corresponding to the patient dental medical record to be processed is obtained, and a patient tooth position map is drawn according to the patient structured tooth position data.

[0040] The embodiment of the application also provides a dental information automatic structured processing system based on a position recording method, which comprises:

[0041] A multi-modal input interface module is configured to obtain a historical patient dental medical record, wherein the historical patient dental medical record comprises physical examination data, diagnosis data, examination and treatment data, and auxiliary examination data. The multi-modal input interface module supports an HTTP API interface to connect a hospital information system (HIS) and an electronic medical record system (EMR), integrates a speech recognition SDK (such as Xunfei speech and Google Speech-to-Text), and provides a local file upload interface (in PDF, JPG, PNG, DOCX and other formats).

[0042] The dental special tooth site recognition module is used for quadrant symbol recognition of physical examination data, diagnosis data, auxiliary examination data and examination processing data in the historical patient dental medical record according to the FDI information data set, if the quadrant symbol exists, the tooth site number and tooth site state are continuously recognized and divided to obtain the FDI information marked data corresponding to the historical patient dental medical record with the tooth site state, and if the quadrant symbol does not exist, the quadrant / region word description data, tooth site number, identifier and tooth site state are recognized according to the tooth site part information data set to obtain the tooth site part marked data corresponding to the historical patient dental medical record with the tooth site state.

[0043] The dental special tooth site recognition module selects DBNet++ or PP-YOLOE as a text detector and selects CRNN, Vision Transformer (ViT) or PP-OCRv3 series model of PaddleOCR as a text recognition model. The tooth site part information data set and the FDI information data set are introduced into the text recognition model as a specific vocabulary table for text detection and recognition, and the dental commonly used number and score format are preferentially matched. A regularization loss function is designed to encourage the text recognition model to preferentially output expressions conforming to the specific vocabulary table, and an attention mechanism is used to guide the attention to the context related to the tooth site number during the recognition process.

[0044] The dental semantic analysis module is used for quadrant / region word description data checking of the FDI information marked data according to the tooth site part information data set, if the corresponding content of the quadrant / region word description data is not found, it is determined that the tooth site part marked data recognition is correct, and the checked tooth site part marked data and the checked FDI information marked data are obtained, if the corresponding content of the quadrant / region word description data is found, the quadrant / region word description data, tooth site number, identifier are re-recognized, and the checked tooth site part marked data and the checked FDI information marked data are obtained, the checked FDI information marked data is converted into tooth site part marked data according to the tooth site part information and FDI information conversion set, and is combined with the checked tooth site part marked data, and the final tooth site part marked data is obtained as structured tooth site data.

[0045] The tooth site information automatic structuring model generation module is used for establishing a dental information automatic structuring model according to the mapping relationship between the historical patient dental medical record and the structured tooth site data.

[0046] The tooth site structured data and tooth site part atlas generation module is used for inputting the patient dental medical record to be processed into the dental information automatic structuring model to obtain patient structured tooth site data corresponding to the patient dental medical record to be processed, and drawing a patient tooth site part atlas according to the patient structured tooth site data. The patient tooth site part atlas is drawn using HTML5+Canvas.

[0047] The structured format of the structured tooth position data of the patient is as shown in Figure 3 The structured format of the structured tooth position data of the patient is as shown in

[0048] The visual interactive interface module is used to display a graphical patient tooth position atlas for doctors to confirm or correct the recognition result, the patient tooth position atlas is distinguished by color according to tooth position states, for example, missing, residual root, loose, repair, pigmentation, swelling, deep pocket and the like, different colors can be displayed through different layers to ensure that the patient tooth position atlas is clear and intuitive to display the tooth position state, supports click modification, users can click a tooth to pop up an editing box for manual correction, modify the tooth position state, supports responsive layout, adapts to PC and mobile terminals, and provides a printing / export atlas function.

[0049] The method and system for automatically structuring dental information according to a tooth position recording method provided by the embodiment of the application are suitable for medical institutions such as dental hospitals and clinics, and aim to solve the problem of inaccurate recognition caused by the non-standard and difficult-to-unify expression of tooth position information in clinical records, and improve the accuracy and management efficiency of dental diagnosis and treatment information.

[0050] The embodiment of the application unifies unstructured dental information into tooth position labeling structured data, significantly reduces manual input errors and reduces the workload of clerical work, so that doctors can focus on diagnosis and treatment themselves, provides a high-quality data basis for AI-assisted diagnosis, treatment recommendation, remote consultation and other applications, and is more easily audited, shared and long-term preserved.

Claims

1. A method for automatically structuring dental information using a part recording method, characterized in that: The following steps are involved: Step 1: Obtain historical patient dental medical records, which include physical examination data, diagnosis data, examination and processing data, and auxiliary examination data; Step 2: Based on the FDI information dataset, quadrant symbol recognition is performed on the physical examination data, diagnosis data, auxiliary examination data, and examination processing data in the historical patient's dental medical records. If the quadrant symbol exists, the tooth position number and tooth position status are further identified and segmented to obtain the FDI information annotated data with the tooth position status corresponding to the historical patient's dental medical records. If the quadrant symbol does not exist, the quadrant / region word description data, tooth position number, identifier, and tooth position status are identified based on the tooth position information dataset to obtain the tooth position annotated data with the tooth position status corresponding to the historical patient's dental medical records; Step 3: Verify the quadrant / region word description data of the FDI information annotation data according to the tooth position information data set. If the corresponding content of the quadrant / region word description data is not found, it is determined that the tooth position annotation data is correctly identified, and the verified tooth position annotation data and the verified FDI information annotation data are obtained. If the corresponding content of the quadrant / region word description data is found, the quadrant / region word description data, tooth position number, and identifier are re-identified to obtain the verified tooth position annotation data and the verified FDI information annotation data. According to the tooth position information and FDI information conversion set, the verified FDI information annotation data is converted into tooth position annotation data and combined with the verified tooth position annotation data to obtain the final tooth position annotation data as structured tooth position data. Step 4: Establish an automatic structuring model of dental information based on the mapping relationship between historical patient dental records and structured tooth position data; Step 5: Input the dental medical records of the patient to be processed into the automatic structuring model of dental information to obtain the patient's structured tooth position data corresponding to the dental medical records of the patient to be processed, and draw a patient's tooth position map based on the patient's structured tooth position data.

2. The method for automatically structuring dental information using a part registration method according to claim 1, wherein: The historical patient dental medical records may be in the form of paper scanned images, X-ray annotated images, electronic document images, handwritten images, and audio.

3. The method for automatically structuring dental information using a part registration method according to claim 1, wherein: The tooth position information and FDI information conversion set includes a tooth position information data set, an FDI information data set, and a correspondence between the tooth position information data set and the FDI information data set.

4. The method for automatically structuring dental information using a part registration method according to claim 1, wherein: The tooth position information data set includes: quadrant / region word description data, tooth position numbers, and identifiers.

5. The method for automatically structuring dental information using a part registration method according to claim 1, wherein: The FDI information data set includes: quadrant symbols and tooth position numbers.

6. The method for automatically structuring dental information using a part registration method according to claim 1, wherein: The content representation forms of the tooth position information dataset and the FDI information dataset include paper scan image form, X-ray film annotated image form, electronic document image form, handwritten image form, and audio form.

7. A dental information automatic structured processing system using a part recording method, characterized in that: The method for automatically structuring dental information using a part recording method as claimed in claim 1 comprises: A multimodal input interface module is used to obtain historical patient dental medical records, which include physical examination data, diagnosis data, examination and processing data, and auxiliary examination data; A dental tooth position identification module is used to perform quadrant symbol recognition on the physical examination data, diagnosis data, auxiliary examination data, and examination processing data in the historical patient dental medical records based on the FDI information data set. If the quadrant symbol is present, the tooth position number and tooth position status are further recognized and segmented to obtain FDI information annotated data with the tooth position status corresponding to the historical patient dental medical records. If the quadrant symbol is not present, the quadrant / region word description data, tooth position number, identifier, and tooth position status are recognized based on the tooth position position information data set to obtain tooth position annotated data with the tooth position status corresponding to the historical patient dental medical records. The dental semantic parsing module is used to verify the quadrant / region word description data of the FDI information annotation data according to the tooth position information data set. If the corresponding content of the quadrant / region word description data is not found, it is determined that the tooth position annotation data is correctly identified, and the verified tooth position annotation data and the verified FDI information annotation data are obtained. If the corresponding content of the quadrant / region word description data is found, the quadrant / region word description data, tooth position number, and identifier are re-identified to obtain the verified tooth position annotation data and the verified FDI information annotation data. According to the tooth position information and FDI information conversion set, the verified FDI information annotation data is converted into tooth position annotation data and combined with the verified tooth position annotation data to obtain the final tooth position annotation data as structured tooth position data; A tooth position information automatic structured model generation module is used to establish a dental information automatic structured model based on the mapping relationship between historical patient dental medical records and structured tooth position data; The tooth position structured data and tooth position map generation module is used to input the dental medical records of the patient to be processed into the automatic structuring model of dental information, obtain the patient's structured tooth position data corresponding to the patient's dental medical records to be processed, and draw the patient's tooth position map based on the patient's structured tooth position data.

8. The automatic structured processing system for dental information using the part registration method according to claim 7, characterized in that: The multimodal input interface module supports HTTP API interface docking with hospital information system and electronic medical record system, integrates voice recognition, and provides a local file upload interface.

9. The automatic structured processing system for dental information using the part registration method according to claim 7, characterized in that: The text recognition model introduces a tooth position information dataset and an FDI information dataset as specific vocabulary tables for text detection and recognition.

10. The automatic structured processing system for dental information using the part registration method according to claim 7, characterized in that: The patient's tooth position map is drawn using HTML5+Canvas.

Images