Physical examination letter triggering and underwriting process optimization method and device, equipment and medium

Abstract

The invention relates to the technical field of natural language processing, can be applied to business system platforms of financial science and technology, medical health and the like, and discloses a physical examination letter triggering and underwriting process optimization method, device, equipment and medium. Carrying out definition identification and filling integrity identification on the questionnaire image to obtain a definition identification result and an integrity identification result, judging whether the questionnaire image meets requirements or not according to the definition identification result and the integrity identification result, and if the requirements are met, sending the questionnaire image to the server; if yes, performing question filling identification on the questionnaire image based on a natural language processing technology to obtain an identification result, dynamically generating a physical examination letter according to the identification result, and sending the physical examination letter to a user, otherwise, generating an auditing suggestion and a data supplement instruction according to the questionnaire image, and sending the auditing suggestion to the user. And sending the auditing suggestion and the data supplement description to the user. And the underwriting efficiency and safety are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of natural language processing, and in particular to a medical examination letter triggering and underwriting process optimization method, device, equipment and medium. BACKGROUND

[0002] In the field of medical health, health questionnaires are the key basis for medical examination reservation and health risk screening. The traditional process requires manual entry and review of questionnaire information, which is not only time-consuming and labor-intensive, but also prone to misjudgment due to illegible handwriting and missing information. Although existing automated systems have introduced image recognition technology, there are still significant limitations: first, general models have weak processing capabilities for low-definition questionnaires (such as shooting angle deviation and insufficient light), and edge feature extraction is not accurate, leading to recognition errors of key information such as "blood glucose value" and "allergy history"; second, there is a lack of deep association with medical knowledge graphs, which cannot accurately identify abnormal items in the questionnaire (such as "chronic nephritis without labeling of kidney function indicators" and "thyroid nodule without size annotation"), resulting in insufficient relevance of medical examination project recommendations; third, the identity verification mechanism is not perfect, and the matching degree of health information and user identity is low, which may lead to distorted health data due to others filling in the questionnaire, affecting the development of subsequent diagnosis and treatment plans. At the same time, information correlation between sub-specialties (such as collaboration between endocrinology and nephrology in screening for diabetic nephropathy) is not considered, further reducing the reliability of health risk assessment.

[0003] In the field of financial technology, health information verification is a core link in the insurance underwriting process. The traditional process relies on manual review of user-filled health questionnaires, which has low efficiency and large subjective errors. Although existing automated systems attempt to process questionnaires through image recognition technology, they still face multiple challenges: on the one hand, general image recognition models have rough judgment of the clarity of questionnaire images (such as not considering recognition errors caused by fuzzy text edges and reflections), and inaccurate detection of the completeness of key filling areas (such as medical history and surgery records), often missing mandatory items; on the other hand, there is insufficient semantic understanding of professional terms in the questionnaire (such as "exemption amount related diseases" and "relevant medical history of heavy illness insurance exemption clauses"), leading to biased identification of abnormal items and affecting the accuracy of medical examination letter triggering. In addition, there are vulnerabilities in the identity verification process, with traditional reliance on single certificate verification, which is prone to risks such as non-personal filling and certificate forgery. Moreover, due to incompatible data formats between different underwriting systems, the process efficiency is further reduced, making it difficult to meet the high requirements of the insurance industry for underwriting accuracy and security. SUMMARY

[0004] The present application provides a medical examination letter triggering and underwriting process optimization method, device, computer equipment and medium to solve the problem of low efficiency and low security of existing CRS cash machine identity verification in the market.

[0005] In a first aspect, a medical examination letter triggering and underwriting process optimization method is provided, comprising:

[0006] An image of a health questionnaire filled by a user is obtained, and a clarity recognition is performed on the image to obtain a clarity recognition result;

[0007] A completeness recognition is performed on the image to obtain a completeness recognition result;

[0008] It is determined whether the image meets requirements according to the clarity recognition result and the completeness recognition result;

[0009] If the image meets the requirements, a fill-in problem recognition is performed on the image based on a natural language processing technology to obtain a recognition result, a medical examination letter is dynamically generated according to the recognition result, and the medical examination letter is sent to the user;

[0010] If the image does not meet the requirements, an audit suggestion and a supplementary description of materials are generated according to the image, and the audit suggestion and the supplementary description of materials are sent to the user.

[0011] In a second aspect, a medical examination letter triggering and underwriting process optimization device is provided, comprising:

[0012] A clarity recognition module is configured to obtain an image of a health questionnaire filled by a user, and perform a clarity recognition on the image to obtain a clarity recognition result;

[0013] A completeness recognition module is configured to perform a completeness recognition on the image to obtain a completeness recognition result;

[0014] An image determination module is configured to determine whether the image meets requirements according to the clarity recognition result and the completeness recognition result;

[0015] A medical examination letter generation module is configured to perform a fill-in problem recognition on the image based on a natural language processing technology to obtain a recognition result, dynamically generate a medical examination letter according to the recognition result, and send the medical examination letter to the user;

[0016] An artificial audit module is configured to generate an audit suggestion and a supplementary description of materials according to the image, and send the audit suggestion and the supplementary description of materials to the user.

[0017] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above medical examination letter triggering and underwriting process optimization method when executing the computer program.

[0018] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the steps of the medical examination letter triggering and underwriting process optimization method.

[0019] In the scheme implemented by the medical examination letter triggering and underwriting process optimization method, device, computer equipment and storage medium, the intelligent processing is performed on the health questionnaire image filled by the user, the automatic triggering of the medical examination letter and the optimization of the underwriting process are realized. The scheme first acquires the health questionnaire image filled by the user, performs the definition recognition on the image, scales the image to a preset fixed resolution, extracts the edge features to obtain an edge gradient matrix after the gray scale processing, obtains the definition recognition result by calculating the gradient variance and the average gradient and weighted fusion. At the same time, the completeness recognition is performed on the questionnaire image, the key filling area is recognized based on the target detection algorithm, the character filling area and the check area are distinguished, the character recognition and the check mark recognition are respectively performed, and then the completeness recognition result is generated. The definition and completeness recognition results are used to determine whether the questionnaire meets the requirements. If yes, the questionnaire image is processed based on the natural language processing technology, and the underwriting efficiency and safety are improved. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the description of the embodiments of the present application. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 is an application environment diagram of the medical examination letter triggering and underwriting process optimization method in an embodiment of the present application;

[0022] Figure 2 is a flow diagram of the medical examination letter triggering and underwriting process optimization method in an embodiment of the present application;

[0023] Figure 3 is a structure diagram of the medical examination letter triggering and underwriting process optimization device in an embodiment of the present application;

[0024] Figure 4 is a structure diagram of the computer equipment in an embodiment of the present application;

[0025] Figure 5 is another structure diagram of the computer equipment in an embodiment of the present application. DETAILED DESCRIPTION

[0026] Clearly, the described embodiments are part of the present application and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0027] The physical examination letter triggering and underwriting process optimization method provided by the embodiments of the present application can be applied in application environments such as Figure 1 The application environment, wherein the client communicates with the server through the network. The server can obtain the questionnaire image of the health questionnaire filled by the user through the client, identify the clarity and completeness of the questionnaire image, obtain the clarity identification result and the completeness identification result, judge whether the questionnaire image meets the requirements according to the clarity identification result and the completeness identification result, if it meets the requirements, identify the filling problem of the questionnaire image based on natural language processing technology, obtain the identification result, dynamically generate the physical examination letter according to the identification result, and send the physical examination letter to the user, otherwise, generate the audit suggestion and the material supplement according to the questionnaire image, and send the audit suggestion and the material supplement to the user. The efficiency and safety of underwriting are improved. The client can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers. The present application will be described in detail through specific embodiments.

[0028] Please refer to Figure 2 , which is a flowchart of the physical examination letter triggering and underwriting process optimization method provided by the embodiments of the present application, comprising the following steps: Figure 2

[0029] S1, obtaining the questionnaire image of the health questionnaire filled by the user, identifying the clarity of the questionnaire image, and obtaining the clarity identification result.

[0030] In the field of medical health, the user can upload the questionnaire image through the intelligent physical examination system.

[0031] In the field of financial technology, the user can upload the questionnaire image through the medical insurance underwriting system.

[0032] In the embodiments of the present application, the clarity identification result is a numerical value for indicating the clarity.

[0033] In the embodiments of the present application, the clarity identification of the questionnaire image includes:

[0034] ​scaling the questionnaire image to a preset fixed resolution to obtain a scaled image;

[0035] graying the scaled image to obtain a graying image;

[0036] extracting edge features from the graying image to obtain an edge gradient matrix;

[0037] calculating a gradient variance and an average gradient of the edge gradient matrix;

[0038] performing weighted fusion on the gradient variance and the average gradient to obtain the definition recognition result.

[0039] In detail, the scaling of the questionnaire image to a preset fixed resolution to obtain a scaled image is to eliminate the interference of the resolution difference of the original image caused by different shooting devices and angles on the subsequent definition analysis. By setting a fixed resolution (such as scaling to 1920*1080 pixels uniformly), the distribution of the pixel points and the calculation dimension are ensured to be consistent in the subsequent processing, thereby providing standardized input data for the subsequent steps such as graying and edge extraction, and outputting a scaled image meeting the preset resolution.

[0040] In detail, the graying of the scaled image is to convert the scaled color image (containing three color channels of RGB) into a single-channel gray image. This process removes the redundant influence of color information on the definition evaluation by weighted average of the pixel values of each channel (such as calculation according to the weights of R=0.299, G=0.587, and B=0.114), and only retains the brightness information.

[0041] In detail, the edge feature extraction from the graying image to obtain an edge gradient matrix can be performed by a gradient algorithm such as Laplace operator or Sobel operator, to calculate the gradient change intensity (i.e., edge sharpness) of each pixel point in the graying image, and finally generate an edge gradient matrix. The matrix records the edge intensity information of all pixels in the image, thereby providing a data basis for the subsequent definition quantification.

[0042] In detail, the gradient variance reflects the distribution dispersion degree of the edge intensity. The edge intensity difference of a clear image is large (the strong edge and the weak edge are obviously distinguished), and the variance value is high. The edge intensity of a blurred image is concentrated, and the variance value is low. The average gradient reflects the edge sharpness of the whole image. The higher the value, the clearer the whole image. By calculating the two indexes, the definition features of the image can be quantified from the global and distribution dimensions, and the output is a specific gradient variance value and an average gradient value.

[0043] In the embodiment of the present application, the edge feature extraction from the graying image to obtain an edge gradient matrix comprises:

[0044] smooth the gray image by Gaussian filtering to obtain a smooth image;

[0045] calculate horizontal gradient and vertical gradient of the smooth image by using gradient operator;

[0046] calculate gradient amplitude of the smooth image according to the horizontal gradient and the vertical gradient to obtain a gradient amplitude matrix;

[0047] standardize the gradient amplitude matrix to obtain an edge gradient matrix.

[0048] In detail, the smoothing of the gray image by Gaussian filtering is a convolution operation of the image by a Gaussian kernel (such as a 3x3 or 5x5 matrix), so that each pixel value becomes a weighted average of the pixel values around it (the closer the pixel, the higher the weight), thereby smoothing the salt and pepper noise, grain noise, etc. in the image caused by shooting or transmission. After processing, irrelevant fine fluctuations in the image are suppressed, while the main edge structure (such as the text edge in the questionnaire and the table lines) is retained, and the final output smooth image can more accurately reflect the true edge features.

[0049] In detail, the calculation of the horizontal gradient and the vertical gradient of the smooth image by using the gradient operator is an operation of two convolution kernels respectively for the horizontal direction (x-axis) and the vertical direction (y-axis) with the smooth image: the horizontal gradient operator detects the edge change in the horizontal direction (such as the horizontal stroke edge of the text), and the vertical gradient operator detects the edge change in the vertical direction (such as the vertical stroke edge of the text). The final output is a horizontal gradient matrix and a vertical gradient matrix.

[0050] In detail, the calculation of the gradient amplitude of the smooth image according to the horizontal gradient and the vertical gradient to obtain a gradient amplitude matrix is a comprehensive evaluation of the overall edge sharpness of each pixel by gradient amplitude. The calculation of the gradient amplitude uses the formula: gradient amplitude = √(horizontal gradient2 + vertical gradient2), the larger the value, the more intense the comprehensive edge change of the pixel point in the horizontal and vertical directions, i.e. the clearer the edge.

[0051] In detail, the standardization processing of the gradient amplitude matrix is a linear transformation that maps all the values of the gradient amplitude matrix to a fixed range of 0-255.

[0052] In the embodiment of the present application, the questionnaire image of the health questionnaire filled by the user is acquired, the clarity of the questionnaire image is identified, and a clarity identification result is obtained. The questionnaire image is standardized (scaled to a fixed resolution and grayed), combined with edge feature extraction and gradient calculation (gradient variance and average gradient weighted fusion), and the image clarity is accurately judged. The function is to eliminate the image quality interference caused by the differences in shooting equipment, angle and light, to ensure that the subsequent text recognition and content analysis are based on clear images, to avoid information misreading caused by blurred images, to provide a prerequisite guarantee for accurate extraction of questionnaire content, and to improve the reliability of the entire process.

[0053] S2, fill-in completeness of the questionnaire image is identified, and a completeness identification result is obtained.

[0054] In the embodiment of the present application, the fill-in completeness of the questionnaire image is identified, and a completeness identification result is obtained, comprising:

[0055] The key fill-in area in the questionnaire image is identified based on a target detection algorithm;

[0056] The text fill-in area and the check area in the key fill-in area are identified;

[0057] The character recognition result is obtained by character recognition on the text fill-in area;

[0058] The check mark recognition is performed on the check area;

[0059] The completeness identification result is generated according to the character recognition result and the check mark recognition.

[0060] In detail, the key fill-in area in the questionnaire image is identified based on a target detection algorithm (such as YOLO, Faster R-CNN, etc.), a pre-trained model (training data containing fill-in boxes, option areas and other samples of various questionnaires) is used to scan the questionnaire image and identify all key fill-in areas - these areas are usually parts that the user is required to fill in or select in the questionnaire design, such as personal information column, medical history option box, signature area, etc. The algorithm outputs the coordinate range (such as the pixel position of the upper left corner and the lower right corner) of each key area, so as to divide the questionnaire image into multiple independent detection units, to ensure that the subsequent identification is only performed on these core areas, and to avoid irrelevant background interference.

[0061] In detail, the text fill-in area and the check area in the key fill-in area are identified by image feature analysis (such as the text fill-in area is mostly a rectangular blank box, and the check area is mostly a regular small circle / square with option text around it), which can classify the key areas and prepare for subsequent targeted detection (text recognition or check state judgment).

[0062] In detail, the character recognition of the text filling area to obtain the character recognition result is to recognize the content in the area and extract character information through an OCR (Optical Character Recognition) technology (such as Tesseract, PaddleOCR, etc.). In the recognition process, meaningless messy handwriting or stains are filtered out, and only the content conforming to the text features (such as continuous letters, numbers, and Chinese characters) is retained. The output character recognition result includes two parts: one is "whether there is content" (such as the blank area recognition result is "no valid character", and the area filled with text is "character").

[0063] In detail, the check mark recognition of the check area can determine whether each check area is effectively marked through pixel value comparison (such as calculating the proportion of non-background color pixels in the area) or contour detection (such as recognizing the shape contour of "√" and "×").

[0064] In detail, the generation of the completeness recognition result according to the character recognition result and the check mark recognition includes the following steps: first, the number of "unfilled" (no valid character) in the text filling area is counted, especially the mandatory items marked with "*"; second, the mandatory items (such as the questionnaire requires "at least one item of allergic history" but the corresponding check area is not marked) in the check area are counted; finally, whether there is a logical defect is judged according to the questionnaire rules (such as "if the text area is filled with'surgical history', the corresponding surgical time check area must be filled").

[0065] In the embodiment of the present application, the completeness recognition result is obtained by performing filling completeness recognition on the questionnaire image. Based on the target detection algorithm, the key filling area is located, the text filling area and the check area are distinguished, and the contents are recognized respectively, so that it can be accurately judged whether the user has completely filled in the mandatory items (such as personal medical history, symptom description, etc.) in the questionnaire. The role is to avoid the omission of manual review through structured analysis, quickly identify the key information (such as "surgical history time" and "allergic drug") that is not filled in, ensure the completeness of the questionnaire information, provide comprehensive data support for subsequent abnormal item recognition and medical letter generation, and reduce the process rework caused by information loss.

[0066] S3, judging whether the questionnaire image meets the requirements according to the clarity recognition result and the completeness recognition result.

[0067] In the embodiment of the present application, judging whether the questionnaire image meets the requirements according to the clarity recognition result and the completeness recognition result includes:

[0068] judging whether the clarity of the questionnaire image is greater than a preset clarity threshold according to the clarity recognition result.

[0069] If less, it is determined that the questionnaire image does not meet the requirements;

[0070] If greater, it is determined whether the text filling area and the check area in the questionnaire image are both filled completely according to the completeness recognition result;

[0071] If not filled completely, it is determined that the questionnaire image does not meet the requirements;

[0072] If both are filled completely, the questionnaire image meets the requirements.

[0073] In the embodiment of the application, the judgment of whether the questionnaire image meets the requirements according to the clarity recognition result and the completeness recognition result filters valid questionnaires through double standards (clarity meets the requirements and completeness is qualified), which functions to establish a strict pre-audit mechanism: for questionnaires that do not meet the requirements, audit suggestions and supplementary explanations are generated in time to guide users to improve information and avoid invalid data from entering the subsequent process; for questionnaires that meet the requirements, further content analysis is triggered to ensure that the subsequent medical examination letter is generated based on high-quality data, thereby improving process efficiency from the source and reducing erroneous decisions (such as missing key medical examination items) caused by low-quality questionnaires.

[0074] If it meets the requirements, S4 is performed, the filling problem recognition of the questionnaire image is performed based on natural language processing technology, a recognition result is obtained, a medical examination letter is dynamically generated according to the recognition result, and the medical examination letter is sent to the user.

[0075] In the embodiment of the application, the filling problem recognition of the questionnaire image based on natural language processing technology and the obtaining of the recognition result include the following steps.

[0076] Text content in the questionnaire image is extracted by using a text recognition algorithm to obtain original text;

[0077] Useless text filtering is performed on the original text to obtain filtered text;

[0078] Key entities in the filtered text are recognized, and the key entities are subjected to associated logic recognition and semantic classification to obtain semantic label data;

[0079] The semantic label data is subjected to medical problem recognition based on a preset medical knowledge graph to obtain a recognition result.

[0080] In detail, the text content in the questionnaire image can be located by using algorithms such as Tesseract and PaddleOCR in OCR technology (including user-filled content and questionnaire-provided question descriptions), and the image information of these areas is converted into original text in the form of a string.

[0081] In detail, the useless text filtering of the original text refers to removing redundant information in the original text that may be irrelevant to the filling content, such as decorative text of the questionnaire frame, repeated watermarks, garbled characters (such as “###” and “@￥”) generated by OCR recognition errors, or obviously meaningless handwriting (such as messy scribbles).

[0082] In detail, the key entity in the filtered text is identified, the key entity is subjected to associated logic recognition and semantic classification, and semantic label data is obtained. The key entity is extracted from the filtered text, and these entities are core information related to health, such as disease name (“thyroid abnormality”), symptoms (“dizziness”), numerical value (“blood pressure 140 / 90”), time (“surgery in 2022”), etc. These entities are accurately positioned and labeled by a named entity recognition model (such as a BERT-based NER model). Then, the associated logic between entities is analyzed, for example, “thyroid abnormality” and “surgery in 2022” may have a “disease-treatment time” association. Finally, the entity and the logical relationship are subjected to semantic classification, such as classifying “thyroid abnormality” as “past medical history” and “blood pressure 140 / 90” as “abnormal numerical value”, and assigning corresponding semantic labels.

[0083] In detail, the semantic label data is subjected to medical problem identification based on a preset medical knowledge graph. The semantic label data is compared with the knowledge graph: if “thyroid abnormality” is found to be not associated with necessary information such as “nodule size”, it is identified as “incomplete information”; if “blood pressure 160 / 100” exceeds the normal range in the knowledge graph, it is marked as “numerical value abnormality”; if the entity of a key mandatory item (such as “ID card number”) is missing, it is determined as “mandatory item omission”. Finally, these identified problems are sorted into structured identification results, including problem type, associated entity, specific description, etc., to provide direct basis for dynamically generating a physical examination letter.

[0084] In the embodiment of the application, the physical examination letter is dynamically generated according to the identification result, and the physical examination letter is sent to the user, comprising:

[0085] Extracting abnormal item information in the identification result, structuring and storing the abnormal item information based on the user ID to obtain a structured information list;

[0086] According to a preset physical examination item mapping rule library, the abnormal items contained in the structured information list are batch matched to obtain a supplementary examination item list;

[0087] Generating a physical examination letter according to the supplementary examination item list and a preset physical examination template;

[0088] Send the physical examination letter to the corresponding user based on the user ID contained in the structured information list.

[0089] In detail, the abnormal item information can refer to the body abnormal item described by the user, for example, the user fills in "thyroid has abnormality in the past", and the abnormal item is "thyroid".

[0090] In detail, the structured storage of the abnormal item information based on the user ID can refer to storing the abnormal item information according to a preset format, for example: user ID-abnormal item-preset abnormal item priority.

[0091] In detail, the abnormal item in the structured information list is matched in batches according to the preset physical examination item mapping rule library, and the abnormal item is associated with the targeted examination item through the rule library. The preset physical examination item mapping rule library is supported by medical professional knowledge, for example, the rule library can stipulate that "thyroid abnormality and incomplete description → match thyroid B-ultrasound examination" and "blood pressure value missing → match blood pressure measurement (including systolic and diastolic blood pressure)". The system traverses each abnormal item in the structured information list, automatically finds the corresponding supplementary examination item according to the rule library, and generates a list in batches.

[0092] In detail, the preset physical examination template includes a fixed framework, such as a user basic information column, a physical examination notice, and basic examination items (such as height, weight, blood routine, etc.).

[0093] In detail, the generation of the physical examination letter according to the supplementary examination item list and the preset physical examination template is to call the basic content in the template, and then insert the items in the supplementary examination item list into the template according to logical classification (such as according to body parts or examination types), while supplementing the associated abnormal items of each supplementary item (such as "thyroid B-ultrasound: because you fill in that the thyroid has abnormality in the past, this item is used to further clarify the situation"), and finally generate the physical examination letter.

[0094] In the embodiment of the application, when the questionnaire image does not meet the requirements, the filling problem recognition of the questionnaire image is performed based on a natural language processing technology to obtain a recognition result, a physical examination letter is dynamically generated according to the recognition result, and the physical examination letter is sent to the user, the unstructured questionnaire content is converted into structured abnormal item information, the accurate matching of physical examination items is realized (such as automatically associating a blood glucose detection item with "diabetes history"), and subjective bias of manual underwriting is avoided.

[0095] If it does not meet the requirements, S5 is performed, an audit suggestion and a material supplement are generated according to the questionnaire image, and the audit suggestion and the material supplement are sent to the user.

[0096] In the embodiment of the present application, the generation of the review suggestion and the supplementary description of the material according to the questionnaire image is achieved by sending the questionnaire image to manual review, and giving the corresponding review suggestion and the supplementary description of the material through manual review.

[0097] In the medical health field, the method can be applied to the automatic physical examination reservation and health risk screening scene of the health management platform. For example, when a user makes an annual physical examination reservation through the platform, the user needs to fill in a health questionnaire (including past medical history, symptom description, etc.) and upload an image. The system first processes the questionnaire image: the clarity is judged through grayscale and edge gradient matrix calculation (such as whether the text edge of the diabetes history filled area is clear), and then the key area (such as “blood glucose value” and “allergy history” which are mandatory items) is identified through target detection. If it is found that the “fasting blood glucose” is not filled or the checked area is not marked, it is determined that the completeness is not up to standard, and the supplementary description (“please supplement the fasting blood glucose value in the past 1 month”) is sent.

[0098] In the field of financial technology, the method can be applied to the health information verification and automatic triggering of physical examination letter in the insurance underwriting process. For example, when a user applies for critical illness insurance, the user needs to fill in a health questionnaire and upload a questionnaire image. The system first identifies the clarity of the questionnaire image: the image is scaled to a fixed resolution, and after grayscale and edge feature extraction, the clarity is judged through the weighted fusion of gradient variance and average gradient (if the gradient variance is less than a preset threshold, it is determined to be fuzzy); at the same time, the completeness is identified, the key filling area (such as past medical history, operation history, blood pressure value, etc.) is located based on the target detection algorithm, the text filling content is identified through OCR, and the checked mark (such as the checked state of “whether there is hypertension”) is detected, and the number of mandatory items not filled is counted.

[0099] In the embodiment of the present application, when the questionnaire image does not meet the requirements, the review suggestion and the supplementary description of the material are generated according to the questionnaire image, and the review suggestion and the supplementary description of the material are sent to the user, which provides clear modification guidance for the user, reduces repeated submission caused by understanding deviation of the user, and shortens the information supplement period. At the same time, invalid questionnaires are intercepted in time, low-quality data is avoided from occupying subsequent processing resources, the overall process operation efficiency is indirectly improved, and it is ensured that the information entering the underwriting link meets the analysis requirements.

[0100] It can be seen that in the above scheme, by intelligently processing the health questionnaire image filled by the user, the automatic triggering of the physical examination letter and the optimization of the underwriting process are realized. First, the health questionnaire image filled by the user is obtained, and the clarity of the image is identified: the image is scaled to a preset fixed resolution, the edge features are extracted after grayscale processing to obtain an edge gradient matrix, the gradient variance and average gradient are calculated and weighted fusion is performed to obtain the clarity identification result. At the same time, the completeness of the questionnaire image is identified, the key filling area is identified based on a target detection algorithm, the text filling area and the check area are distinguished, character recognition and check mark recognition are performed respectively, and then the completeness identification result is generated. According to the clarity and completeness identification results, it is judged whether the questionnaire meets the requirements: if it meets the requirements, the questionnaire image is processed based on natural language processing technology, which improves the efficiency and security of underwriting.

[0101] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0102] In an embodiment, a physical examination letter triggering and underwriting process optimization device is provided, which corresponds to the physical examination letter triggering and underwriting process optimization method in the above embodiment. As shown in the figure, the physical examination letter triggering and underwriting process optimization device includes a clarity identification module 101, a completeness identification module 102, an image judgment module 103, a physical examination letter generation module 104, and an artificial audit module 105. The detailed description of each functional module is as follows: Figure 3 The clarity identification module 101 is configured to obtain the questionnaire image of the health questionnaire filled by the user, identify the clarity of the questionnaire image, and obtain a clarity identification result.

[0103] The completeness identification module 102 is configured to identify the completeness of the filling of the questionnaire image and obtain a completeness identification result.

[0104] The image judgment module 103 is configured to judge whether the questionnaire image meets the requirements according to the clarity identification result and the completeness identification result.

[0105] The physical examination letter generation module 104 is configured to identify the filling questions of the questionnaire image based on natural language processing technology, obtain an identification result, dynamically generate a physical examination letter according to the identification result, and send the physical examination letter to the user.

[0106] The artificial audit module 105 is configured to generate audit suggestions and supplementary information based on the questionnaire image, and send the audit suggestions and the supplementary information to the user.

[0107]

[0108] ​In an embodiment, the clarity identification module 101, when performing the clarity identification on the questionnaire image obtained by filling the health questionnaire by the user, obtains a clarity identification result, and specifically is configured to:

[0109] scaling the questionnaire image to a preset fixed resolution to obtain a scaled image;

[0110] performing a grayscale processing on the scaled image to obtain a grayscale image;

[0111] performing an edge feature extraction on the grayscale image to obtain an edge gradient matrix;

[0112] calculating a gradient variance and an average gradient of the edge gradient matrix;

[0113] performing a weighted fusion on the gradient variance and the average gradient to obtain the clarity identification result.

[0114] In an embodiment, the clarity identification module 101, when performing the edge feature extraction on the grayscale image to obtain the edge gradient matrix, specifically is configured to:

[0115] performing a smoothing processing on the grayscale image by a Gaussian filter to obtain a smoothed image;

[0116] calculating a horizontal gradient and a vertical gradient of the smoothed image by a gradient operator;

[0117] calculating a gradient amplitude of the smoothed image according to the horizontal gradient and the vertical gradient to obtain a gradient amplitude matrix;

[0118] performing a standardization processing on the gradient amplitude matrix to obtain the edge gradient matrix.

[0119] In an embodiment, the completeness identification module 102, when performing the filling completeness identification on the questionnaire image to obtain a completeness identification result, specifically is configured to:

[0120] identifying a key filling area in the questionnaire image based on a target detection algorithm;

[0121] identifying a character filling area and a check area in the key filling area;

[0122] performing a character recognition on the character filling area to obtain a character recognition result;

[0123] performing a check mark recognition on the check area;

[0124] generating the completeness identification result according to the character recognition result and the check mark recognition.

[0125] In an embodiment, the image judgment module 103, when performing the judgment on whether the questionnaire image meets the requirements according to the clarity identification result and the completeness identification result, is specifically used for:

[0126] judging whether the clarity of the questionnaire image is greater than a preset clarity threshold according to the clarity identification result;

[0127] if less than, determining that the questionnaire image does not meet the requirements;

[0128] if greater than, judging whether the text filling area and the check area in the questionnaire image are both filled completely according to the completeness identification result;

[0129] if not filled completely, determining that the questionnaire image does not meet the requirements;

[0130] if both filled completely, the questionnaire image meets the requirements.

[0131] In an embodiment, the medical letter generation module 104, when performing the filling problem identification on the questionnaire image based on the natural language processing technology to obtain an identification result, is specifically used for:

[0132] extracting the text content in the questionnaire image by using a text recognition algorithm to obtain original text;

[0133] filtering useless text from the original text to obtain filtered text;

[0134] identifying key entities in the filtered text, and performing associated logic identification and semantic classification on the key entities to obtain semantic tag data;

[0135] performing medical problem identification on the semantic tag data based on a preset medical knowledge graph to obtain an identification result.

[0136] In an embodiment, the medical letter generation module 104, when performing the dynamic generation of a medical letter according to the identification result and sending the medical letter to a user, is specifically used for:

[0137] extracting abnormal item information in the identification result, and structurally storing the abnormal item information based on a user ID to obtain a structured information list;

[0138] performing batch matching on abnormal items contained in the structured information list according to a preset medical project mapping rule library to obtain a supplementary examination project list;

[0139] generating a medical letter according to the supplementary examination project list and a preset medical template;

[0140] Send the medical examination letter to the corresponding user based on the user ID contained in the structured information list.

[0141] The application provides a medical examination letter triggering and underwriting process optimization device. The device realizes automatic triggering of the medical examination letter and optimization of the underwriting process by intelligently processing the health questionnaire image filled by the user. The device first acquires the health questionnaire image filled by the user, identifies the definition of the image, scales the image to a preset fixed resolution, extracts edge features to obtain an edge gradient matrix after grayscale processing, calculates the gradient variance and average gradient and performs weighted fusion to obtain the definition identification result. At the same time, the device identifies the completeness of the questionnaire image, identifies the key filling area based on a target detection algorithm, distinguishes the text filling area and the check area, respectively performs character recognition and check mark recognition, and further generates the completeness identification result. The device determines whether the questionnaire meets the requirements according to the definition and completeness identification results. If the questionnaire meets the requirements, the device processes the questionnaire image based on natural language processing technology, thereby improving the underwriting efficiency and security. The specific limitations of the medical examination letter triggering and underwriting process optimization device can be referred to the limitations of the medical examination letter triggering and underwriting process optimization method in the foregoing, which will not be described herein. Each module in the medical examination letter triggering and underwriting process optimization device can be realized by software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to each module.

[0142] In one embodiment, a computer device can be a server, and its internal structure diagram can be as shown in Figure 4 The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. The computer program is executed by the processor to implement the functions or steps of the medical examination letter triggering and underwriting process optimization method on the server side.

[0143] In one embodiment, a computer device can be a client, and its internal structure diagram can be as shown in Figure 5As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is used to communicate with the external server through the network connection. The computer program is executed by the processor to realize the functions or steps of the client side of the physical examination letter triggering and underwriting process optimization method

[0144] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor executing the computer program to implement the following steps:

[0145] Obtaining a questionnaire image of a health questionnaire filled by a user, performing clarity recognition on the questionnaire image to obtain a clarity recognition result;

[0146] Performing completeness recognition on the questionnaire image to obtain a completeness recognition result;

[0147] According to the clarity recognition result and the completeness recognition result, it is judged whether the questionnaire image meets the requirements;

[0148] If it meets the requirements, performing fill-in question recognition on the questionnaire image based on natural language processing technology to obtain a recognition result, dynamically generating a physical examination letter according to the recognition result, and sending the physical examination letter to the user;

[0149] If it does not meet the requirements, generating an audit suggestion and a supplementary explanation of the materials according to the questionnaire image, and sending the audit suggestion and the supplementary explanation of the materials to the user.

[0150] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the following steps:

[0151] Obtaining a questionnaire image of a health questionnaire filled by a user, performing clarity recognition on the questionnaire image to obtain a clarity recognition result;

[0152] Performing completeness recognition on the questionnaire image to obtain a completeness recognition result;

[0153] According to the clarity recognition result and the completeness recognition result, it is judged whether the questionnaire image meets the requirements;

[0154] If the requirements are met, the questionnaire image is filled in question recognition based on natural language processing technology to obtain a recognition result, a physical examination letter is dynamically generated according to the recognition result, and the physical examination letter is sent to the user.

[0155] If the requirements are not met, an audit suggestion and a supplementary description of materials are generated according to the questionnaire image, and the audit suggestion and the supplementary description of materials are sent to the user.

[0156] It should be noted that the functions or steps described above with respect to the computer-readable storage medium or the computer device can correspond to the related descriptions of the server side and the client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0157] Those skilled in the art can understand that all or part of the processes in the foregoing method embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the foregoing method embodiments. In the embodiments provided in the present application, any reference to a memory, storage, database or other medium can include a non-volatile and / or volatile memory. The non-volatile memory can include a read-only memory (ROM), a programmable ROM (PROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM) or a flash memory. The volatile memory can include a random access memory (RAM) or an external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM) and the like.

[0158] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified. In actual applications, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0159] Finally, it should be noted that if the application in the embodiment appears non-company software tools or components, only for example, and does not represent the actual use. The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the foregoing embodiments of the present application are described in detail, those skilled in the art should understand: its still can be modified to the technical scheme recorded in the foregoing embodiments, or part of the technical features are replaced; and these modifications or replacements, and the essence of the corresponding technical scheme does not deviate from the spirit and scope of the embodiments of the present application technical scheme, should be included in the protection scope of the present application.

Claims

1. A method for optimizing the medical examination letter triggering and underwriting process, characterized in that, The method comprises the following steps: obtaining a questionnaire image of a health questionnaire filled by a user, and performing clarity identification on the questionnaire image to obtain a clarity identification result; performing filling completeness identification on the questionnaire image to obtain a completeness identification result; judging whether the questionnaire image meets the requirements according to the clarity identification result and the completeness identification result; if the questionnaire image meets the requirements, performing filling problem identification on the questionnaire image based on a natural language processing technology to obtain an identification result, dynamically generating a physical examination letter according to the identification result, and sending the physical examination letter to the user; if the questionnaire image does not meet the requirements, generating an audit suggestion and a supplementary description of materials according to the questionnaire image, and sending the audit suggestion and the supplementary description of materials to the user.

2. The method of claim 1, wherein the medical examination letter triggers an optimization of the underwriting and approval process. The clarity identification on the questionnaire image comprises the following steps: scaling the questionnaire image to a preset fixed resolution to obtain a scaled image; performing grayscale processing on the scaled image to obtain a grayscale image; performing edge feature extraction on the grayscale image to obtain an edge gradient matrix; calculating the gradient variance and the average gradient of the edge gradient matrix; performing weighted fusion on the gradient variance and the average gradient to obtain the clarity identification result.

3. The method of claim 2, wherein the medical examination letter triggers an optimization of the underwriting and approval process. The edge feature extraction on the grayscale image comprises the following steps: performing smoothing processing on the grayscale image by a Gaussian filter to obtain a smoothed image; calculating the horizontal gradient and the vertical gradient of the smoothed image by using a gradient operator; calculating the gradient amplitude of the smoothed image according to the horizontal gradient and the vertical gradient to obtain a gradient amplitude matrix; performing standardization processing on the gradient amplitude matrix to obtain an edge gradient matrix.

4. The method for optimizing the medical examination letter triggering and underwriting process as described in claim 1, characterized in that, The filling completeness identification on the questionnaire image comprises the following steps: identifying a key filling area in the questionnaire image based on a target detection algorithm; identifying a character filling area and a check area in the key filling area; performing character recognition on the character filling area to obtain a character recognition result; performing check mark recognition on the check area; generating a completeness identification result according to the character recognition result and the check mark recognition.

5. The method of claim 1, wherein the medical examination letter triggers an optimization of the underwriting and approval process. The judgment of whether the questionnaire image meets the requirements according to the clarity identification result and the completeness identification result comprises the following steps: judging whether the clarity of the questionnaire image is greater than a preset clarity threshold according to the clarity identification result; if the clarity is less than the preset clarity threshold, determining that the questionnaire image does not meet the requirements; if the clarity is greater than the preset clarity threshold, judging whether the character filling area and the check area in the questionnaire image are both filled completely according to the completeness identification result; if the character filling area and the check area are not both filled completely, determining that the questionnaire image does not meet the requirements; if the character filling area and the check area are both filled completely, the questionnaire image meets the requirements.

6. The method of claim 1, wherein the medical examination letter triggers an optimization of the underwriting and approval process. The filling problem identification on the questionnaire image based on the natural language processing technology comprises the following steps: extracting character content in the questionnaire image by using a character recognition algorithm to obtain an original text; performing useless text filtering on the original text to obtain a filtered text; Identify the key entities in the filtered text, perform associated logical recognition and semantic classification on the key entities, and obtain semantic label data; Based on the preset medical knowledge graph, perform medical question recognition on the semantic label data to obtain an identification result.

7. The method of claim 1, wherein the medical examination letter triggers an optimization of the underwriting and approval process. The method comprises the following steps: Extracting abnormal item information from the identification result, structuring the abnormal item information based on a user ID, and obtaining a structured information list; According to the preset medical examination item mapping rule library, the abnormal items contained in the structured information list are batch matched to obtain a supplementary examination item list; According to the supplementary examination item list and the preset medical examination template, a medical examination letter is generated; Based on the user ID contained in the structured information list, the medical examination letter is sent to the corresponding user.

8. A medical examination letter triggering and underwriting process optimization apparatus, characterized by, Comprise: The clarity recognition module is used for obtaining the questionnaire image of the health questionnaire filled by the user, performing clarity recognition on the questionnaire image, and obtaining a clarity recognition result; The completeness recognition module is used for performing filling completeness recognition on the questionnaire image to obtain a completeness recognition result; The image judgment module is used for judging whether the questionnaire image meets the requirements according to the clarity recognition result and the completeness recognition result; The medical examination letter generation module is used for performing filling problem recognition on the questionnaire image based on natural language processing technology to obtain an identification result, dynamically generating a medical examination letter according to the identification result, and sending the medical examination letter to the user; The artificial review module is used for generating review suggestions and data supplement instructions according to the questionnaire image, and sending the review suggestions and the data supplement instructions to the user.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the medical examination letter triggering and underwriting process optimization method according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to realize the steps of the medical examination letter triggering and underwriting process optimization method according to any one of claims 1-7.

Images