Underwriting method and device based on physical examination report, computer equipment, readable storage medium and program product

Abstract

The invention relates to an underwriting method and device based on a physical examination report, computer equipment, a computer readable storage medium and a computer program product, relates to the field of artificial intelligence, and can improve the underwriting efficiency of physical examination report underwriting. The method comprises the following steps: acquiring a physical examination report image, extracting report text content, and matching the report text content with preset examination items of a hospital department; if the number of the matched examination items is greater than or equal to a preset threshold value, extracting report text content fragments related to the matched examination items as department physical examination report texts; if the number of the matched examination items is smaller than a preset threshold value, medical keywords corresponding to the preset examination items are extracted, and report text content fragments related to the keywords are extracted to serve as department physical examination report texts. And obtaining a structured template, and inputting the department physical examination report text and the structured template into the semantic generation model to obtain structured physical examination information. And inputting the structured physical examination information into the underwriting strategy acquisition model to obtain a target underwriting strategy.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an underwriting method, apparatus, computer equipment, computer-readable storage medium, and computer program product based on a physical examination report. Background Art

[0002] During the insurance underwriting process, the insured typically undergoes a medical examination at a designated institution or hospital. The insurance company uses the results of the medical examination report to determine whether to reject, approve, or adjust the insurance premium. Traditionally, this process relies entirely on manual verification, which is inefficient, costly, and prone to errors. Manual verification is not only time-consuming and labor-intensive, but also prone to errors due to subjective judgment, which in turn affects the accuracy and reliability of the underwriting conclusion.

[0003] With the gradual development of image recognition and natural language processing technologies, automated underwriting processes have become a hot topic in the industry. Advanced image recognition and text parsing technologies enable automated processing and accurate recognition of medical examination reports, enabling automated underwriting decisions. However, existing automation technologies still face several challenges. First, a dedicated template must be created for each partner medical examination institution. As the number of partners increases, the cost of template production and maintenance rises dramatically. Second, each time a partner adjusts the format of its medical examination report, the template must be recreated, increasing operational complexity. Finally, because users often experience tilt and blur when taking photos and uploading medical examination reports, image quality varies, increasing the difficulty of template matching and thus leading to inefficiencies in the existing underwriting process. Summary of the Invention

[0004] Based on this, it is necessary to provide an underwriting method, device, computer equipment, computer-readable storage medium and computer program product based on a physical examination report to address the above technical issues.

[0005] In a first aspect, the present application provides an underwriting method based on a physical examination report, comprising:

[0006] Acquire a physical examination report image, extract the report text content corresponding to the physical examination report image, and match the report text content with the preset examination items of the hospital department to obtain matching examination items;

[0007] If the number of the matching examination items is greater than or equal to a preset threshold, extracting the report text content fragments related to the matching examination items as the department physical examination report text;

[0008] If the number of the matching examination items is less than the preset threshold, obtaining the medical keywords corresponding to the preset examination items, and extracting the report text content fragments related to the medical keywords from the report text content as the department physical examination report text;

[0009] Obtaining a structured template, and inputting the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model;

[0010] The structured physical examination information is input into an underwriting strategy acquisition model to obtain a target underwriting strategy output by the underwriting strategy acquisition model.

[0011] In one embodiment, before extracting the report text content corresponding to the physical examination report image, the method further includes:

[0012] Performing image quality detection on the physical examination report image to obtain a quality detection result;

[0013] If the quality inspection result indicates that the physical examination report image does not meet the preset quality condition, the user is prompted to re-upload the physical examination report image, and the re-uploaded physical examination report image is obtained.

[0014] In one embodiment, after performing image quality detection on the physical examination report image and obtaining the quality detection result, the method further includes:

[0015] If the quality inspection result indicates that the physical examination report image meets the preset quality condition, then obtaining the image file name corresponding to the physical examination report image for the physical examination report image that passes the quality inspection;

[0016] Based on a preset image file name sorting rule, the sorting order between the physical examination report images is determined according to the image file names, and the physical examination report images arranged according to the preset image file name sequence rule are obtained.

[0017] In one embodiment, the step of obtaining a structured template and inputting the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model includes:

[0018] Obtaining structured templates corresponding to preset examination items of different departments; the structured templates include preset field names and field value constraint rules;

[0019] Inputting the department physical examination report text and the structured template into the semantic generation model, identifying the department physical examination report text through the semantic generation model, and obtaining semantic generation information corresponding to the department physical examination report text;

[0020] The semantic generation model utilizes the structured template to perform field classification and field value constraints on the semantic generation information to obtain structured physical examination information corresponding to the physical examination report.

[0021] In one embodiment, before inputting the structured medical examination information into an underwriting rule model and using the output of the underwriting rule model as an underwriting strategy, the process further includes:

[0022] Obtaining a target input format corresponding to the underwriting rule model, and determining standardized templates corresponding to structured physical examination information of different departments based on the target input format;

[0023] The standardized template and the structured physical examination information are input into the semantic generation model to obtain standardized structured physical examination information output by the semantic generation model.

[0024] In one embodiment, before matching the report text content with the preset examination items of the hospital department, the method further includes:

[0025] Conducting a normative test on the medical text related to the report text content to obtain a normative test result;

[0026] If the normative test result indicates that the medical text contains confusion in terms, the medical text with confusion in terms is corrected according to a preset medical terminology library to obtain a corrected report text content.

[0027] In a second aspect, the present application further provides an underwriting device based on a medical examination report, comprising:

[0028] A report image extraction and matching module is used to obtain a physical examination report image, extract the report text content corresponding to the physical examination report image, and match the report text content with the preset examination items of the hospital department to obtain matching examination items;

[0029] A matching examination item text extraction module is used to extract report text content fragments related to the matching examination items as department physical examination report text if the number of the matching examination items is greater than or equal to a preset threshold;

[0030] A keyword extraction text module is used to obtain medical keywords corresponding to the preset examination items if the number of the matching examination items is less than the preset threshold, and extract report text content segments related to the medical keywords from the report text content as the department physical examination report text;

[0031] A structured physical examination information generation module obtains a structured template, and inputs the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model;

[0032] The underwriting strategy derivation module is used to input the structured physical examination information into the underwriting strategy acquisition model to obtain the target underwriting strategy output by the underwriting strategy acquisition model.

[0033] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0034] Acquire a physical examination report image, extract the report text content corresponding to the physical examination report image, and match the report text content with the preset examination items of the hospital department to obtain matching examination items;

[0035] If the number of the matching examination items is greater than or equal to a preset threshold, extracting the report text content fragments related to the matching examination items as the department physical examination report text;

[0036] If the number of the matching examination items is less than the preset threshold, obtaining the medical keywords corresponding to the preset examination items, and extracting the report text content fragments related to the medical keywords from the report text content as the department physical examination report text;

[0037] Obtaining a structured template, and inputting the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model;

[0038] The structured physical examination information is input into an underwriting strategy acquisition model to obtain a target underwriting strategy output by the underwriting strategy acquisition model.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0040] Acquire a physical examination report image, extract the report text content corresponding to the physical examination report image, and match the report text content with the preset examination items of the hospital department to obtain matching examination items;

[0041] If the number of the matching examination items is greater than or equal to a preset threshold, extracting the report text content fragments related to the matching examination items as the department physical examination report text;

[0042] If the number of the matching examination items is less than the preset threshold, obtaining the medical keywords corresponding to the preset examination items, and extracting the report text content fragments related to the medical keywords from the report text content as the department physical examination report text;

[0043] Obtaining a structured template, and inputting the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model;

[0044] The structured physical examination information is input into an underwriting strategy acquisition model to obtain a target underwriting strategy output by the underwriting strategy acquisition model.

[0045] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the following steps:

[0046] Acquire a physical examination report image, extract the report text content corresponding to the physical examination report image, and match the report text content with the preset examination items of the hospital department to obtain matching examination items;

[0047] If the number of the matching examination items is greater than or equal to a preset threshold, extracting the report text content fragments related to the matching examination items as the department physical examination report text;

[0048] If the number of the matching examination items is less than the preset threshold, obtaining the medical keywords corresponding to the preset examination items, and extracting the report text content fragments related to the medical keywords from the report text content as the department physical examination report text;

[0049] Obtaining a structured template, and inputting the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model;

[0050] The structured physical examination information is input into an underwriting strategy acquisition model to obtain a target underwriting strategy output by the underwriting strategy acquisition model.

[0051] The above-mentioned underwriting method, device, computer equipment, computer-readable storage medium and computer program product based on physical examination reports obtain a physical examination report image, extract the report text content corresponding to the physical examination report image, match the report text content with the preset examination items of the hospital department, and obtain matching examination items; if the number of matching examination items is greater than or equal to the preset threshold, extract the report text content fragments related to the matching examination items as the department physical examination report text; if the number of matching examination items is less than the preset threshold, obtain the medical keywords corresponding to the preset examination items, and extract the report text content fragments related to the medical keywords from the report text content as the department physical examination report text; obtain a structured template, and input the department physical examination report text and the structured template into the semantic generation model to obtain structured physical examination information output by the semantic generation model; input the structured physical examination information into the underwriting strategy acquisition model to obtain the target underwriting strategy output by the underwriting strategy acquisition model. In this application, the image information of the physical examination report is converted into processable text data and compared with the preset hospital department examination items to ensure that the extracted physical examination information has high accuracy and relevance; in different scenarios, the department physical examination report text is obtained directly from the text content of the matching examination item, or the department examination report text is indirectly extracted by medical keywords based on the matching examination item, so that the model input becomes accurate and concise, filling the shortcomings of the large model in matching limits and long text understanding; through the intelligent processing of the semantic generation model, the medical meaning of the report text is understood more accurately, and structured information that meets the standard requirements is output, which not only reduces the need for manual intervention, but also improves data processing efficiency and data quality; by inputting structured physical examination information into the underwriting strategy acquisition model, the underwriting strategy is automatically generated, which improves the accuracy of underwriting and thus improves the underwriting efficiency of the entire underwriting process. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 1 is a flow chart of an underwriting method based on a physical examination report in one embodiment;

[0054] Figure 2 is a flowchart of an underwriting method based on a physical examination report in another embodiment;

[0055] Figure 3 This is a structural block diagram of an underwriting device based on a physical examination report in one embodiment;

[0056] Figure 4 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0058] In one embodiment, Figure 1 As shown, a medical examination report-based underwriting method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0059] Step S102: Acquire a physical examination report image, extract the report text content corresponding to the physical examination report image, match the report text content with the preset examination items of the hospital department, and obtain matching examination items.

[0060] The hospital department's pre-set examination items refer to a list of examination items pre-set by the insurance company based on the examination content of different medical departments. This list of examination items is used to compare with the extracted medical examination report text to determine which medical examination items in the medical examination report meet the pre-set examination item requirements.

[0061] For example, a medical examination report image uploaded by a user is first received. The medical examination report image can be uploaded to the terminal through various methods such as scanning and taking a photo. After the image is uploaded, OCR (Optical Character Recognition) technology is used to extract text from the medical examination report image by analyzing the characters and layout in the image.

[0062] The terminal first loads the department's pre-set standardized examination item lists. These lists contain the names of common examination items and their corresponding test items. The terminal then performs a text match between the extracted report text and these pre-set examination items. This matching process utilizes natural language processing technology and keyword matching algorithms to automatically identify examination items mentioned in the physical examination report and match them with the corresponding content in the pre-set items. The resulting list of matched examination items is a list of all examination items that were successfully matched in the report text.

[0063] Step S104: If the number of matching examination items is greater than or equal to a preset threshold, extract the report text content segments related to the matching examination items as the department physical examination report text.

[0064] Among them, the department physical examination report text refers to the text related to the physical examination items extracted from the report text content and in line with the standards of a specific department.

[0065] For example, based on the number of matching examination items obtained in the previous step, the terminal will determine whether to continue extracting relevant physical examination report content fragments as the department physical examination report text. If the number of matching examination items is greater than or equal to a preset threshold set by the system (for example, if the number of matching examination items reaches 5 or more), it can be considered that the physical examination content in the report is rich enough to represent the physical examination content of the corresponding department. At this time, the terminal will extract the content related to these matching items from the report text to form a new department physical examination report text.

[0066] Specifically, the terminal searches for corresponding text paragraphs in the report text content according to each matched examination item. These paragraphs include information such as the item name, examination results, reference values, etc., forming a complete department physical examination report text.

[0067] Optionally, a location-based text extraction method or context analysis technology is used to select an examination item as the center and obtain a regional text as the matched department physical examination report text, ensuring that the content extracted from the original report is highly relevant to the matched department examination item and covers the key physical examination data of the department.

[0068] Step S106: If the number of matching examination items is less than a preset threshold, obtain the medical keywords corresponding to the preset examination items, and extract the report text content segments related to the medical keywords from the report text content as the department physical examination report text.

[0069] Among them, medical keywords refer to professional medical terms related to preset examination items. These words are key diagnostic indicators, symptom descriptions or other medical information in the department or physical examination items. For example, for cervical TCT examination, NILM (no intraepithelial lesion or malignancy) is its medical keyword.

[0070] For example, the terminal system will first determine whether the number of matching examination items in the previous stage is less than the preset threshold. If the number of matching items is insufficient, it will automatically switch to another processing method, that is, use the medical keywords related to the preset examination items to continue to extract the relevant report content. Specifically, the terminal obtains the medical keyword set corresponding to these items based on the preset examination items of each department. Then, the terminal searches for these keywords in the report text content, and extracts paragraphs or sentences related to the keywords through a text matching algorithm. These extracted text fragments are the department physical examination report text.

[0071] Optionally, the terminal may use techniques such as regular expression matching, word frequency analysis, contextual semantic analysis, etc. to ensure that the extracted text content is indeed highly relevant to the medical keywords.

[0072] Step S108: Obtain a structured template, and input the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model.

[0073] A structured template is a set of predefined, standardized data formats used to calibrate the information in a medical examination report. A semantic generative model is an artificial intelligence-based technology used to understand and convert textual information into structured data. Trained using medical text and rules, the semantic generative model can identify semantic information in text and generate corresponding structured data based on predefined template formats.

[0074] For example, the terminal first needs to obtain a structured template corresponding to the department's physical examination report text. The structured template is a pre-set data format for a specific department or examination item, defining the fields of each physical examination item and the acceptable value range of each field.

[0075] The extracted department physical examination report text is input into the semantic generation model together with the structured template. By analyzing the input department physical examination report text content, the key information in the text is identified and mapped to the corresponding structured template field.

[0076] Step S110: input the structured physical examination information into the underwriting strategy acquisition model to obtain the target underwriting strategy output by the underwriting strategy acquisition model.

[0077] The underwriting strategy acquisition model is a decision support system based on artificial intelligence algorithms that generates appropriate underwriting strategies based on structured medical examination information. The model uses historical data, medical rules, and insurance company-defined assessment criteria to assess the insured's health and recommends a corresponding underwriting strategy based on the assessment results.

[0078] For example, the terminal first extracts and structures the medical information from the medical report and inputs it into the underwriting strategy acquisition model. The underwriting strategy acquisition model combines the insurance company's preset underwriting standards and historical data to determine whether the insured meets the insurance requirements. By evaluating various health indicators, it generates a personalized underwriting decision for each insurance application.

[0079] In the above-mentioned underwriting method based on physical examination reports, by obtaining a physical examination report image, extracting the report text content corresponding to the physical examination report image, matching the report text content with the preset examination items of the hospital department, and obtaining matching examination items; if the number of matching examination items is greater than or equal to the preset threshold, extracting the report text content fragments related to the matching examination items as the department physical examination report text; if the number of matching examination items is less than the preset threshold, obtaining the medical keywords corresponding to the preset examination items, extracting the report text content fragments related to the medical keywords from the report text content as the department physical examination report text; obtaining a structured template, and inputting the department physical examination report text and the structured template into the semantic generation model to obtain structured physical examination information output by the semantic generation model; inputting the structured physical examination information into the underwriting strategy acquisition model to obtain the target underwriting strategy output by the underwriting strategy acquisition model. In this embodiment, the image information of the physical examination report is converted into processable text data and compared with the preset hospital department examination items to ensure that the extracted physical examination information has high accuracy and relevance; in different scenarios, the department physical examination report text is obtained directly from the text content of the matching examination item, or the department examination report text is indirectly extracted by medical keywords based on the matching examination item, so that the model input becomes accurate and concise, filling the shortcomings of the large model in matching limits and long text understanding; through the intelligent processing of the semantic generation model, the medical significance of the report text is understood more accurately, and structured information that meets the standard requirements is output, which not only reduces the need for manual intervention, but also improves data processing efficiency and data quality; by inputting structured physical examination information into the underwriting strategy acquisition model, the underwriting strategy is automatically generated, which improves the accuracy of underwriting and thus improves the underwriting efficiency of the entire underwriting process.

[0080] In an exemplary embodiment, before extracting the report text content corresponding to the physical examination report image, it also includes: performing image quality detection on the physical examination report image to obtain a quality detection result; if the quality detection result indicates that the physical examination report image does not meet the preset quality conditions, the user is prompted to re-upload the physical examination report image and obtain the re-uploaded physical examination report image.

[0081] Image quality testing involves the system analyzing uploaded medical examination report images to determine whether they meet preset quality standards, including but not limited to factors such as image resolution, clarity, and contrast. Preset quality conditions refer to image quality standards defined in advance by the system. These conditions can be specific parameters, such as a resolution of at least 300 DPI or image clarity of at least a certain threshold.

[0082] Specifically, the terminal evaluates the medical examination report image using a quality detection algorithm. This algorithm determines whether the image meets quality standards based on image clarity, resolution, contrast, and the presence of excessive noise (such as interfering lines generated during scanning). Applicable evaluation techniques include image sharpening, contrast enhancement, and edge detection.

[0083] After processing an image, the quality detection algorithm generates a quality check result. If the image fails to meet the preset quality criteria (e.g., blurry image, resolution below 300 DPI, or severe noise), the terminal automatically marks it as unacceptable and prompts the user to re-upload the image. Optionally, the user will be prompted to improve image clarity or adjust the shooting angle in order to upload an image that meets the requirements for the report.

[0084] After receiving the prompt, the user will upload a new medical examination report image. The terminal will perform a quality check on the new image again and repeat the above steps until it passes the test. Only then will the system send the image to the text recognition algorithm for text extraction.

[0085] By implementing this step, the risk of the text recognition model extracting erroneous information from low-quality images is effectively avoided, the success rate of subsequent processing steps is improved, the user experience is further optimized, repetitive work and erroneous information input caused by quality issues are reduced, and the accuracy of the entire physical examination report automation process is ensured.

[0086] In an exemplary embodiment, after performing image quality detection on the physical examination report image and obtaining the quality detection result, it also includes: if the quality detection result indicates that the physical examination report image meets the preset quality conditions, then obtaining the image file name corresponding to the physical examination report image for the physical examination report image that passes the quality detection; based on the preset image file name sorting rules, according to the image file names, determining the sorting order between the physical examination report images, and obtaining the physical examination report images arranged according to the preset image file name sequence rules.

[0087] The image file name refers to the file name assigned to each medical examination report image file. The file name can be automatically generated upon upload or specified by the user and can include information such as a timestamp, numeric number, or user-specific identifier. The preset image file name sorting rules are a set of predefined rules used to determine the sorting order of medical examination report images. Specifically, the sorting rules refer to sorting based on the date, number, or other formatted fields corresponding to the image file name.

[0088] Specifically, the terminal extracts the image file name from the medical examination report image that has passed the quality inspection, and extracts the information contained therein, such as the upload time, digital number, user ID, etc. The obtained image file names are sorted based on the preset image file name sorting rules. In some specific embodiments, if the file name contains a timestamp, the image file name can be sorted in ascending or descending order according to the upload time in the file name. If the file name contains a digital number, it can be sorted according to the size of the number, such as "report001.jpg", "report002.jpg", etc. If the image file name contains a mixture of letters and numbers (user ID), it can be sorted according to the character order of the letters and numbers. After the sorting is completed, the terminal will generate a sorted list of the medical examination report images.

[0089] In this embodiment, by sorting the medical examination report images, it is ensured that each report image is processed in a predetermined order. Furthermore, by sorting based on the image file name, potential problems caused by disordered processing order can be avoided, and the order of each report is guaranteed.

[0090] In an exemplary embodiment, a structured template is obtained, and the department physical examination report text and the structured template are input into a semantic generation model to obtain structured physical examination information output by the semantic generation model, including: obtaining structured templates corresponding to preset examination items of different departments; the structured template includes preset field names and field value constraint rules; the department physical examination report text and the structured template are input into the semantic generation model, and the department physical examination report text is identified by the semantic generation model to obtain semantic generation information corresponding to the department physical examination report text; the semantic generation model uses the structured template to perform field classification and field value constraints on the semantic generation information to obtain structured physical examination information corresponding to the physical examination report.

[0091] First, the corresponding structured template is retrieved from the preset database based on the provided department information. Each department's physical examination report has a set of predefined structured templates, which specify the field names and field value constraints corresponding to each examination item in the report. For example, the predefined structured template specifies the field names and field value constraints corresponding to each examination item in the report. For example, the structured template for a routine urine test may include fields such as "urine specific gravity," "urine pH," "urine white blood cells," "urine microscopic white blood cells," "urine microscopic red blood cells," "urine microalbumin / creatinine," "urine occult blood," "urine protein qualitative," "urine glucose qualitative," "urine ketone body qualitative," "urobilinogen," "urine bilirubin," "urine nitrite," and "epithelial cells." The format and value range of each field are set. For example, the "urine specific gravity" field requires an integer to be entered, and the value range should be within a certain interval.

[0092] The terminal inputs the report text and the corresponding structured template into the semantic generation model. Based on the report text and the field names in the structured template, the semantic generation model uses text analysis techniques (such as lexical analysis, syntactic analysis, and semantic parsing) to extract the data in the report and obtain preliminary semantic generation information. After obtaining the preliminary semantic generation information, the semantic generation model needs to further classify the fields and constrain the field values ​​of the information according to the structured template. Specifically, the model will classify the parsed data into the corresponding fields based on the name of each field and verify the data according to the field value constraint rules. If the field value does not meet the constraint rules, the system can prompt the user or automatically adjust it. After completing the field classification and value constraints, the semantic generation model will generate structured physical examination information that meets the template requirements.

[0093] In this embodiment, a combination of semantic generation models and structured templates is used to convert traditional unstructured physical examination report text into structured physical examination information. The automated process greatly reduces the need for manual intervention, reduces data problems caused by manual input errors, and further improves the system reliability and user experience.

[0094] In an exemplary embodiment, the structured physical examination information is input into the underwriting policy acquisition model to obtain the target underwriting policy output by the underwriting policy acquisition model, and also includes: obtaining the target input format corresponding to the underwriting policy acquisition model, and based on the target input format, determining the standardized template corresponding to the structured physical examination information of different departments; inputting the standardized template and the structured physical examination information into the semantic generation model to obtain the standardized structured physical examination information output by the semantic generation model.

[0095] A standardized template is a data format template used to convert medical examination report data from different departments and types into a unified structured format. Because medical examination reports often present different data across different test items, standardized templates use clear rules to standardize these disparate data into an input format acceptable to the underwriting strategy acquisition model.

[0096] The terminal obtains the target input format required by the underwriting policy acquisition model. This format is usually defined by the insurance company based on specific business needs and model design to ensure that different types of physical examination reports can be processed under the same framework. After the terminal obtains the target input format of the underwriting policy acquisition model, it next determines how the structured physical examination information of each department should be standardized according to the target format. After the standardized template is determined, it is input into the semantic generation model together with the actual structured physical examination information. The semantic generation model performs the necessary classification, formatting, and field value verification by matching the input structured physical examination information with the fields in the standardized template. After completing the above steps, the output of the semantic generation model is the standardized structured physical examination information.

[0097] In a specific embodiment, the structured physical examination information is a data set in JSON format, for example, "urinalysis": {"urinalysis": {"urinalysis": {"urinalysis specific gravity":"1.015","urinalysis pH":"6.5","urinalysis leukocytes":"negative","urinalysis leukocytes":"0","urinalysis erythrocytes":"0","urinalysis microscopy erythrocytes":"0","urinalysis microalbumin / creatinine":"","urinalysis occult blood":"negative","urinalysis protein qualitative":"negative","urinalysis glucose qualitative":"negative","urinalysis ketone bodies qualitative":"negative","urinalysis urobilinogen":"negative","urinalysis bilirubin":"negative","urinalysis nitrites":"negative","epithelial cells":"0",}}. According to the underwriting policy, the target input format of the model is obtained to obtain the corresponding standardized template, for example, {'urine leukocytes': ['03020004H01$positive', '03020004H02$negative'], 'urine occult blood': ['03020013H01$trace (+-, 0-2 HP)', '03020013H02$mild (1+, 3-5 HP)', '03020013H03$moderate (2+, 6-10 HP)', '03020013H04$significant (3+, 6-10 HP)', '03020013H05$severe (4+, >20 HP)', '03020013H06$negative'] '03020014H01$ trace (+- to +)', '03020014H02$ mild (+ to ++)', '03020014H03$ moderate (++ to +++)', '03020014H04$ significant (+++ to ++++)', '03020014H05$ negative (-)'], 'Urobilinogen': ['03020019H01$ positive', '03020019H02$ weakly positive', '03020019H03$ negative'], 'Urine bilirubin': ['03020020H01$ positive', '03020020H02$ negative']. In the standardized template, the structured physical examination information is re-constrained. It is only necessary to obtain the {"urine leukocytes","urine occult blood","urine protein qualitative","urine bilitis","urine bilirubin"} fields in the structured physical examination information and the corresponding value constraints of the fields, so as to obtain standardized results. For example, {"urine leukocytes":"03020004H02$negative","urine occult blood":"03020013H06$negative","urine protein qualitative":"03020014H05$negative","urine bilitis":"03020019H03$negative","urine bilirubin":"03020020H02$negative"}. In this embodiment, the data from different departments are converted into a unified format through the standardized template, thereby ensuring the consistency of data input and avoiding the failure of model parsing due to inconsistent formats.

[0098] In an exemplary embodiment, before matching the report text content with the preset examination items of the hospital department, it also includes: performing a standardization test on the medical text related to the report text content to obtain a standardization test result; if the standardization test result indicates that there is confusion in the wording of the medical text, then according to the preset medical terminology library, the medical text with confusion in the wording is corrected to obtain the corrected report text content.

[0099] Specifically, each medical text item in the report is scanned and potential errors or non-standard sections are flagged. For example, if a report states that the result of "urine glucose qualitative" is "positive" but uses the non-standard phrase "urine glucose test result: +", the standardization test will identify this phrase and flag it as confusing.

[0100] The report text is corrected using a pre-set medical terminology library. This library contains a wide range of standardized medical terms and specifications, enabling the system to automatically identify and replace non-compliant terms. For example, if the description "microalbuminuria" appears in the report text, but the standardized term should be "qualitative urine protein," the system will correct the term using the medical terminology library and replace it with a standard term.

[0101] In this embodiment, standardization detection and correction can eliminate confusion, errors or ambiguous expressions in the text, making the medical information in the physical examination report clearer and more accurate, reducing ambiguity in understanding, and ensuring that the underwriting model and other subsequent analysis tools can accurately read the physical examination data, thereby improving the accuracy of risk assessment and avoiding underwriting decision-making errors caused by non-standard text.

[0102] In order to enable those skilled in the art to better understand the above steps, the embodiment of the present application is illustrated below by using an example, but it should be understood that the embodiment of the present application is not limited to this.

[0103] In an exemplary embodiment, Figure 2 As shown, the terminal first receives the user's uploaded medical examination report image, a scanned image file containing the examination items and results. The terminal then uses image processing algorithms to test image quality parameters such as clarity, resolution, and noise. If the image quality does not meet the preset standards (for example, the resolution is too low or there is blur), the terminal will prompt the user to re-upload an image that meets the standards. If the image quality is acceptable, the terminal proceeds to the next step.

[0104] Next, the terminal sorts the qualified medical examination report images. This process sorts the images based on their file names or other pre-defined sorting rules (such as upload time or report number) to ensure the reports are in the correct order. This sorting operation helps ensure that the contents of each medical examination report are processed in the correct logical order.

[0105] The terminal uses optical character recognition (OCR) technology to extract text from medical report images. The OCR engine recognizes the text in the image and generates raw text data. However, due to the professionalism and complexity of medical reports, the recognized text may contain errors or confusion. To address this issue, the terminal performs text standardization correction.

[0106] The terminal uses a preset medical terminology library to correct the recognized text, fix possible terminology errors, spelling errors or formatting irregularities, and ensure the accuracy and consistency of the report text.

[0107] Next, the terminal matches the report text content with the hospital department's preset examination items to obtain matching examination items. If the number of matching examination items is greater than or equal to a preset threshold, the report text content segments related to the matching examination items are extracted as the department's physical examination report text. If the number of matching examination items is less than the preset threshold, the terminal obtains the medical keywords corresponding to the preset examination items and extracts the report text segments related to the medical keywords from the report text content as the department's physical examination report text.

[0108] The extracted department report text and structured template are input into the semantic generation model, which recognizes the department physical examination report text and obtains the semantic generation information corresponding to the department physical examination report text. The semantic generation model uses the structured template to classify the semantic generation information and constrain the field values, thereby obtaining the structured physical examination information corresponding to the physical examination report. The core purpose of this process is to convert unstructured text information into structured data for subsequent operations.

[0109] After obtaining the structured medical information, the terminal obtains the target input format required by the underwriting policy acquisition model and determines how each department's structured medical information should be standardized according to the target format. After determining the standardized template, it is input into the semantic generation model along with the actual structured medical information. The semantic generation model matches the input structured medical information with the fields in the standardized template, performs the necessary classification, formatting, and field value validation, and generates the standardized structured medical information.

[0110] Finally, the standardized, structured medical examination information is fed into the underwriting strategy acquisition model, which generates an underwriting strategy based on this information. An underwriting strategy is an underwriting recommendation or decision made by the insurance company after conducting a risk assessment based on the medical examination data and pre-defined rules. The underwriting rules model analyzes the medical examination information to determine the insured's health status and outputs a corresponding underwriting strategy.

[0111] In this example, a complete automatic underwriting process based on a large model (semantic generative model) is proposed. The general generalization ability of the large model is used to perform structured extraction and standardization of medical examination text. By extracting medical examination report texts from different departments, the large model's control over the input data volume is reduced. Combined with the text recognition model, the generalization ability in different application scenarios is enhanced, thereby improving the efficiency of the entire underwriting process.

[0112] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0113] Based on the same inventive concept, the embodiments of the present application also provide a medical report-based underwriting device for implementing the aforementioned medical report-based underwriting method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the medical report-based underwriting device provided below can be found in the above-mentioned limitations of the medical report-based underwriting method, and will not be repeated here.

[0114] In an exemplary embodiment, Figure 3 As shown, an underwriting device based on a physical examination report is provided, comprising: a report image extraction and matching module 310, a matching examination item text extraction module 320, a keyword text extraction module 330, a structured physical examination information generation module 340, and an underwriting strategy derivation module 350, wherein:

[0115] The report image extraction and matching module 310 is used to obtain a physical examination report image, extract the report text content corresponding to the physical examination report image, and match the report text content with the preset examination items of the hospital department to obtain matching examination items;

[0116] A matching examination item text extraction module 320 is configured to extract report text content segments related to the matching examination items as department physical examination report texts if the number of the matching examination items is greater than or equal to a preset threshold;

[0117] The keyword extraction text module 330 is configured to obtain medical keywords corresponding to the preset examination items if the number of the matching examination items is less than the preset threshold, and extract report text content segments related to the medical keywords from the report text content as the department physical examination report text;

[0118] The structured physical examination information generating module 340 obtains a structured template, and inputs the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model;

[0119] The underwriting strategy derivation module 350 is configured to input the structured physical examination information into an underwriting strategy acquisition model to obtain a target underwriting strategy output by the underwriting strategy acquisition model.

[0120] In one embodiment, the report image extraction and matching module 310 is also used before extracting the report text content corresponding to the physical examination report image, and also includes: performing image quality detection on the physical examination report image to obtain a quality detection result; if the quality detection result indicates that the physical examination report image does not meet the preset quality conditions, prompting the user to re-upload the physical examination report image, and obtaining the re-uploaded physical examination report image.

[0121] In one embodiment, the report image extraction and matching module 310 is also used to perform image quality detection on the physical examination report image. After obtaining the quality detection result, it also includes: if the quality detection result indicates that the physical examination report image meets the preset quality conditions, then for the physical examination report image that passes the quality detection, obtain the image file name corresponding to the physical examination report image; based on the preset image file name sorting rules, determine the sorting order between the physical examination report images according to the image file name, and obtain the physical examination report image arranged according to the preset image file name sequence rules.

[0122] In one embodiment, the structured physical examination information generation module 340 is also used to obtain the structured template, and input the department physical examination report text and the structured template into the semantic generation model to obtain the structured physical examination information output by the semantic generation model, including: obtaining the structured template corresponding to the preset examination items of different departments; the structured template includes preset field names and field value constraint rules; inputting the department physical examination report text and the structured template into the semantic generation model, and identifying the department physical examination report text through the semantic generation model to obtain the semantic generation information corresponding to the department physical examination report text; the semantic generation model uses the structured template to perform field classification and field value constraints on the semantic generation information to obtain the structured physical examination information corresponding to the physical examination report.

[0123] In one embodiment, the structured physical examination information generation module 340 is also used to input the structured physical examination information into the underwriting policy acquisition model to obtain the target underwriting policy output by the underwriting policy acquisition model, and also includes: obtaining the target input format corresponding to the underwriting policy model, and based on the target input format, determining the standardized template corresponding to the structured physical examination information of different departments; inputting the standardized template and the structured physical examination information into the semantic generation model to obtain the standardized structured physical examination information output by the semantic generation model.

[0124] In one embodiment, the report image extraction and matching module 310 is also used for, before matching the report text content with the preset examination items of the hospital department, further including: performing a normative test on the medical text related to the report text content to obtain a normative test result; if the normative test result indicates that the medical text has confusion in terms, then correcting the medical text with confusion in terms according to a preset medical terminology library to obtain the corrected report text content.

[0125] Each module in the aforementioned medical examination report-based underwriting device may be implemented in whole or in part through software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor within a computer device in the form of hardware, or may be stored in a computer device memory in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0126] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 4As shown. The computer device includes a processor, memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals via wired or wireless means. The wireless means can be implemented via Wi-Fi, mobile cellular networks, near-field communication (NFC), or other technologies. When executed by the processor, the computer program implements an underwriting method based on a medical examination report. The display unit of the computer device is used to produce a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse.

[0127] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0128] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0129] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0130] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0131] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0132] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0133] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0134] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. An underwriting method based on a physical examination report, characterized in that: The method comprises: Acquire a physical examination report image, extract the report text content corresponding to the physical examination report image, and match the report text content with the preset examination items of the hospital department to obtain matching examination items; If the number of the matching examination items is greater than or equal to a preset threshold, extracting the report text content fragments related to the matching examination items as the department physical examination report text; If the number of the matching examination items is less than the preset threshold, obtaining the medical keywords corresponding to the preset examination items, and extracting the report text content fragments related to the medical keywords from the report text content as the department physical examination report text; Obtaining a structured template, and inputting the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model; The structured physical examination information is input into an underwriting strategy acquisition model to obtain a target underwriting strategy output by the underwriting strategy acquisition model.

2. The method according to claim 1, characterized in that Before extracting the report text content corresponding to the physical examination report image, the method further includes: Performing image quality detection on the physical examination report image to obtain a quality detection result; If the quality inspection result indicates that the physical examination report image does not meet the preset quality condition, the user is prompted to re-upload the physical examination report image, and the re-uploaded physical examination report image is obtained.

3. The method according to claim 2, characterized in that After performing image quality detection on the physical examination report image and obtaining the quality detection result, the method further includes: If the quality inspection result indicates that the physical examination report image meets the preset quality condition, then obtaining the image file name corresponding to the physical examination report image for the physical examination report image that passes the quality inspection; Based on a preset image file name sorting rule, the sorting order between the physical examination report images is determined according to the image file names, and the physical examination report images arranged according to the preset image file name sequence rule are obtained.

4. The method according to claim 1, wherein The step of obtaining a structured template and inputting the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model includes: Obtaining structured templates corresponding to preset examination items of different departments; the structured templates include preset field names and field value constraint rules; Inputting the department physical examination report text and the structured template into the semantic generation model, identifying the department physical examination report text through the semantic generation model, and obtaining semantic generation information corresponding to the department physical examination report text; The semantic generation model utilizes the structured template to perform field classification and field value constraints on the semantic generation information to obtain structured physical examination information corresponding to the physical examination report.

5. The method according to claim 1, characterized in that Before inputting the structured physical examination information into the underwriting strategy acquisition model to obtain the target underwriting strategy output by the underwriting strategy acquisition model, the method further includes: Obtaining a target input format corresponding to the underwriting strategy model, and determining standardized templates corresponding to structured physical examination information of different departments based on the target input format; The standardized template and the structured physical examination information are input into the semantic generation model to obtain standardized structured physical examination information output by the semantic generation model.

6. The method according to claim 1, characterized in that Before matching the report text content with the preset examination items of the hospital department, the method further includes: Conducting a normative test on the medical text related to the report text content to obtain a normative test result; If the normative test result indicates that the medical text contains confusion in terms, the medical text with confusion in terms is corrected according to a preset medical terminology library to obtain a corrected report text content.

7. An underwriting device based on a physical examination report, characterized in that: The device comprises: A report image extraction and matching module is used to obtain a physical examination report image, extract the report text content corresponding to the physical examination report image, and match the report text content with the preset examination items of the hospital department to obtain matching examination items; A matching examination item text extraction module is used to extract report text content fragments related to the matching examination items as department physical examination report text if the number of the matching examination items is greater than or equal to a preset threshold; A keyword extraction text module is used to obtain medical keywords corresponding to the preset examination items if the number of the matching examination items is less than the preset threshold, and extract report text content segments related to the medical keywords from the report text content as the department physical examination report text; A structured physical examination information generation module obtains a structured template, and inputs the department physical examination report text and the structured template into a semantic generation model to obtain structured physical examination information output by the semantic generation model; The underwriting strategy derivation module is used to input the structured physical examination information into the underwriting strategy acquisition model to obtain the target underwriting strategy output by the underwriting strategy acquisition model.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images