Insurance excess dose identification digital intelligent system of multi-agent platform based on image-text multi-mode processing engine

Abstract

The invention belongs to the technical field of insurance industry, and particularly relates to a digital intelligent system for insurance excess dose identification of a multi-agent platform based on an image-text multi-modal processing engine, and the system comprises an image-text data collection module which is used for collecting a drug prescription document image and drug use record text data related to insurance claim settlement; and the image-text preprocessing module is used for preprocessing the collected image and text data. According to the invention, the multi-modal fusion module highlights important features by using an attention mechanism and improves the recognition accuracy and efficiency, the excess dose recognition agent is combined with CNN and RNN structures to further improve the recognition performance, the insurance claim decision module automatically executes the claim process, simplifies the operation and improves the efficiency, and the user interaction interface provides a feedback mechanism. On the whole, the system improves the claim settlement accuracy and efficiency through multi-modal data processing and intelligent decision support, and optimizes the user experience and the system performance.

Description

Technical Field

[0001] The present invention relates to the technical field of the insurance industry, and specifically to a digital intelligence system for identifying insurance overdose based on a multi-agent platform of a graphic and text multimodal processing engine. Background Art

[0002] In the insurance industry, monitoring of drug usage is a key link in ensuring the rationality of insurance claims. Traditional methods of identifying overdose rely on manual review, which is inefficient and prone to errors. With the development of artificial intelligence technology, it has become possible to use machine learning and image recognition technology to identify overdose, but there is currently no system that combines a multi-agent platform with a multi-modal processing engine to achieve this function. Therefore, it is necessary to develop a digital intelligence system for insurance overdose identification based on a multi-agent platform with a multi-modal processing engine. Summary of the invention

[0003] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and invention title of this application to avoid blurring the purpose of this section, specification abstract and invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0004] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:

[0005] A digital intelligence system for identifying overdose of insurance medication based on a multi-agent platform of a graphic and text multimodal processing engine, comprising:

[0006] The image and text data collection module is used to collect images of drug prescription documents and text data of drug use records related to insurance claims;

[0007] The image and text preprocessing module is used to preprocess the collected image and text data, including image denoising and text cleaning;

[0008] The image-text feature extraction module includes an image feature extraction unit and a text feature extraction unit. The image feature extraction unit is used to extract visual features of drug packaging and quantity from the preprocessed image, and the text feature extraction unit is used to extract information features of drug name and dosage from the preprocessed text.

[0009] Multimodal fusion processing module, used to fuse image features and text features to form a unified feature representation;

[0010] The overdose recognition agent includes a learning model and a decision unit. The learning model is used to train an overdose recognition model based on historical data. The decision unit is used to determine whether the overdose is present based on fusion features and the recognition model.

[0011] The insurance claim decision module is used to automatically execute the insurance claim process and trigger the manual review process based on the decision results of the overdose recognition agent;

[0012] The user interaction interface module is used to display the recognition results and claim status, and provide user feedback and manual review interface.

[0013] As a preferred solution of the digital intelligence system for insurance overdose identification based on a multi-agent platform of a graphic and text multimodal processing engine described in the present invention, the graphic and text data acquisition module realizes the recognition and text extraction of drug prescription document images through OCR technology.

[0014] As a preferred solution of the digital intelligence system for identifying insurance overdose based on a multi-agent platform of a graphic and text multimodal processing engine described in the present invention, the graphic and text preprocessing module further includes an image enhancement unit and a text correction unit, the image enhancement unit is used to improve the image quality, and the text correction unit is used to correct text recognition errors.

[0015] As a preferred solution of the digital intelligence system for identifying insurance overdose based on a multi-agent platform of a graphic and text multimodal processing engine described in the present invention, the graphic and text feature extraction module further includes a deep learning network for extracting deep visual features from images.

[0016] As a preferred solution of the digital intelligence system for identifying insurance overdose based on a multi-agent platform of a graphic and text multimodal processing engine described in the present invention, the multimodal fusion processing module adopts an attention mechanism to highlight important features and suppress irrelevant information.

[0017] As a preferred solution of the digital intelligence system for insurance overdose identification based on a multi-agent platform of a graphic and text multimodal processing engine described in the present invention, the learning model of the overdose identification agent adopts a hybrid structure of a convolutional neural network (CNN) and a recurrent neural network (RNN).

[0018] As a preferred solution of the digital intelligence system for insurance overdose identification based on a multi-agent platform of a graphic and text multimodal processing engine described in the present invention, the insurance claim decision module automatically executes the insurance claim process and triggers the manual review process according to the decision results of the overdose identification agent, including automatic calculation of the claim amount and automatic update of the claim status.

[0019] As a preferred solution of the digital intelligence system for identifying insurance overdose based on a multi-agent platform of a graphic and text multimodal processing engine described in the present invention, the user interaction interface module provides a real-time feedback mechanism, allowing users to confirm and raise objections to the recognition results, and use user feedback for subsequent model optimization.

[0020] As a preferred solution of the touch query integrated machine described in the present invention, it includes an integrated machine body, the integrated machine body internally stores a digital intelligence system for identifying insurance overdose based on a multi-agent platform of a graphic multimodal processing engine according to claims 1-8, the integrated machine body includes a touch screen arranged on the surface of its shell, and the user uses the digital intelligence system through the touch screen;

[0021] The integrated machine body is arranged on an operating table, and an adjusting mechanism for adjusting the angle, orientation and height of the touch screen when in use is arranged on the operating table.

[0022] As a preferred solution of the touch-controlled query machine described in the present invention, the adjustment mechanism includes a rotating table rotatably arranged on the top of the operating table, the machine body is rotatably arranged on one side of the top of the rotating table, a mounting groove is provided on the rotating table, an electric telescopic rod is hingedly arranged in the groove of the mounting groove, and the upper end of the electric telescopic rod is hinged on the rear side of the machine body;

[0023] Vertical electric lifting legs are arranged around the bottom of the operating table, and four of the electric lifting legs are arranged in series;

[0024] A vertical rotating column is rotatably provided at the bottom center of the operating table through a bearing sleeve, the upper end of the rotating column is fixed at the center of the bottom wall of the rotating table, the shaft of the rotating column is located below the operating table and a worm gear is provided, a worm is installed at the bottom of the operating table through a bracket, the worm gear is rotatably provided and drives with the worm gear, a servo motor is fixed at the bottom of the operating table, and the output shaft of the servo motor is connected to one end of the worm gear through a coupling;

[0025] A plurality of control switches are arranged on the top of the operating table, and different control switches are electrically connected to the electric telescopic rod, the electric lifting leg and the servo motor respectively.

[0026] The beneficial effects of the present invention are as follows: by collecting and preprocessing drug prescription images and usage record data, data quality is ensured, laying the foundation for feature extraction and recognition; the feature extraction module extracts key information from images and texts, such as drug packaging, quantity, name and dosage, to accurately identify overdose situations; the multimodal fusion module uses the attention mechanism to highlight important features and improve recognition accuracy and efficiency; the overdose recognition agent combines CNN and RNN structures to further improve recognition performance; the insurance claim decision module automatically executes the claim process, simplifies operations, and improves efficiency; the user interaction interface provides a feedback mechanism for model optimization and improving user satisfaction; overall, the system improves claim accuracy and efficiency, optimizes user experience and system performance through multimodal data processing and intelligent decision support, BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0028] Figure 1 is a system block diagram of the digital intelligence system of the present invention;

[0029] Figure 2 It is a flow chart of the steps of the graphic data acquisition module of the present invention;

[0030] Figure 3 is a flow chart of the steps of the image enhancement unit of the present invention;

[0031] Figure 4 is a flowchart of the steps of the text correction unit of the present invention;

[0032] Figure 5 It is a flowchart of the steps of the graphic feature extraction module of the present invention;

[0033] Figure 6 is a flowchart of the steps of the multimodal fusion processing module of the present invention;

[0034] Figure 7 This is a flowchart of the steps of the insurance claim decision module of the present invention;

[0035] Figure 8 is a flow chart of steps of the user interaction interface module of the present invention;

[0036] Fig. 9 It is a structural schematic diagram of the integrated machine body of the present invention;

[0037] Fig.10 For the present invention Fig. 9A schematic diagram of the structure in a side rear view direction;

[0038] Fig.11 For the present invention Fig. 9 Schematic diagram of the structure viewed from the side.

[0039] In the figure: an integrated machine body 100 , a touch screen 110 , an operating table 200 , a rotating table 210 , a mounting slot 220 , an electric telescopic rod 230 , an electric lifting leg 240 , a rotating column 250 , a worm gear 260 , a worm 270 , a servo motor 280 , and a control switch 290 . DETAILED DESCRIPTION

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0041] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0042] Secondly, the present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention in detail, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0043] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0044] Example 1

[0045] See also Figure 1-Figure 8 , shows a system block diagram of a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine according to the present invention;

[0046] A digital intelligence system for identifying overdose of insurance medication based on a multi-agent platform of a graphic and text multimodal processing engine, comprising:

[0047] The image and text data collection module is used to collect images of drug prescription documents and text data of drug use records related to insurance claims;

[0048] The image and text preprocessing module is used to preprocess the collected image and text data, including image denoising and text cleaning;

[0049] The image-text feature extraction module includes an image feature extraction unit and a text feature extraction unit. The image feature extraction unit is used to extract visual features of drug packaging and quantity from the preprocessed image, and the text feature extraction unit is used to extract information features of drug name and dosage from the preprocessed text.

[0050] Multimodal fusion processing module, used to fuse image features and text features to form a unified feature representation;

[0051] The overdose recognition agent includes a learning model and a decision unit. The learning model is used to train an overdose recognition model based on historical data. The decision unit is used to determine whether the overdose is present based on fusion features and the recognition model.

[0052] The insurance claim decision module is used to automatically execute the insurance claim process and trigger the manual review process based on the decision results of the overdose recognition agent;

[0053] User interaction interface module, used to display recognition results and claim status, and provide user feedback and manual review interface;

[0054] Wherein: the image and text data acquisition module realizes the recognition and text extraction of the drug prescription document image through OCR technology, including the following steps:

[0055] Step 1: The drug prescription document image needs to be preprocessed, including image denoising, binarization, and rotation correction, to improve the accuracy of subsequent recognition;

[0056] Step 2: Use the text positioning algorithm in OCR technology to locate the text area in the image and distinguish between text and non-text areas;

[0057] Step 3: Apply the OCR engine to segment and recognize characters in the located text area, and convert the text in the image into an editable text format;

[0058] Step 4: Post-process the recognized text, including proofreading and formatting, to ensure the accuracy and readability of the text information;

[0059] Wherein: the image and text preprocessing module further includes an image enhancement unit and a text correction unit, the image enhancement unit is used to improve the image quality, and the text correction unit is used to correct text recognition errors;

[0060] The steps of the image enhancement unit for improving the image quality include:

[0061] Step 1: Improve the contrast of the image to enhance the differences between different areas in the image, making the visual features of the drug packaging and quantity more obvious;

[0062] Step 2: Improve the clarity of the image. By sharpening the image, the details in the image are made clearer, which is helpful for subsequent feature extraction.

[0063] Step 3: Adjust the color saturation of the image and enhance the color vividness to make the color features of the drug packaging more prominent;

[0064] The steps of the text correction unit for correcting text recognition errors include:

[0065] Step 1: Proofread the text recognized by OCR technology to check and correct errors in the recognition process, such as typos and missing characters;

[0066] Step 2: Format the text to ensure the accuracy and readability of the text information, such as adjusting the layout and format of the text to make it easier to understand and process;

[0067] Wherein: the image-text feature extraction module further includes a deep learning network for extracting deep visual features from the image, including the following steps:

[0068] Step 1: Automatically learn and extract complex features in the image, including the shape, color, and texture of the drug packaging, through a multi-layer neural network structure;

[0069] Step 2: By training the deep learning network, the system identifies the key visual elements in the image and converts these elements into feature representations that can be used for subsequent processing;

[0070] Wherein: the multimodal fusion processing module adopts an attention mechanism to highlight important features and suppress irrelevant information, including the following steps:

[0071] Step 1: When fusing image features and text features, the attention mechanism automatically identifies and assigns higher weights to important features, while assigning lower weights or ignoring unimportant features;

[0072] Step 2: The attention mechanism is used to highlight the features most relevant to the overdose recognition task, while suppressing irrelevant information that causes interference, which helps improve the recognition accuracy and efficiency of the system;

[0073] Among them: the learning model of the overdose recognition agent adopts a hybrid structure of convolutional neural network (CNN) and recurrent neural network (RNN). Since CNN and RNN each have unique advantages in processing different types of data; CNN performs well in the field of image recognition and can effectively extract spatial features in images, such as the visual features of the shape, color and texture of drug packaging; while RNN is good at processing sequence data and can capture dynamic information in time series, such as the drug name and dosage information features in text data; the combination of CNN and RNN can simultaneously utilize the advantages of both in image and text feature extraction to improve the accuracy and efficiency of overdose recognition; in addition, the hybrid structure can better process multimodal data of images and texts, and form a unified feature representation by fusing the features of images and texts, thereby improving the model's ability to recognize overdose situations;

[0074] Wherein: the insurance claim decision module automatically executes the insurance claim process and triggers the manual review process according to the decision result of the overdose recognition agent, including automatic calculation of the claim amount and automatic update of the claim status, including the following steps:

[0075] Step 1: The decision unit of the overdose identification agent determines whether the drug dosage is over-dose based on the fusion features and the recognition model; if the judgment result is over-dose, the insurance claim decision module will automatically execute the insurance claim process; if the judgment result is not over-dose, the manual review process will be triggered;

[0076] Step 2: In the case of automatic insurance claim processing, the automatic calculation of the claim amount will be carried out according to the terms and conditions of the insurance contract, involving the calculation of the drug cost of the excess amount, and the determination of the final claim amount according to the deductible and compensation ratio parameters of the insurance contract;

[0077] Step 3: Automatic update of claim status includes recording the current status of the claim process in the system to facilitate tracking of claim progress; including status updates at each stage of claim application submission, review and processing results;

[0078] Step 4: If the manual review process is triggered, the calculation of the claim amount and the update of the claim status require manual intervention, and the claim reviewer will make judgments and handle the case based on the specific circumstances and the terms of the insurance contract;

[0079] Wherein: the user interaction interface module provides a real-time feedback mechanism, allowing the user to confirm the recognition results, raise objections, and use the user feedback for subsequent model optimization, including the following steps:

[0080] Step 1: The user views the recognition results provided by the system through the user interaction interface module, including information features of the drug name and dosage;

[0081] Step 2: The user confirms the recognition result and raises an objection. If the user confirms that the result is correct, no further operation is required. If the user raises an objection, the system will record the user's feedback information.

[0082] Step 3: The system collects and stores the user's feedback information for subsequent analysis and processing;

[0083] Step 4: By analyzing the user feedback data, the system can identify the deficiencies and errors in the model and adjust and optimize the model accordingly;

[0084] Step 5: The optimized model will be reapplied to the image-text feature extraction module and the overdose recognition agent to improve recognition accuracy and efficiency.

[0085] Example 2

[0086] See also Figure 9-11 , shows a schematic diagram of the structure of a touch query all-in-one machine of the present invention, including an all-in-one machine body 100, wherein the all-in-one machine body 100 stores a digital intelligence system for identifying over-dose insurance drug based on a multi-agent platform of a graphic multimodal processing engine according to claims 1-8, and the all-in-one machine body 100 includes a touch screen 110 arranged on the surface of its shell, and a user uses the digital intelligence system through the touch screen 110;

[0087] The integrated machine body 100 is disposed on an operating table 200 , and an adjustment mechanism for adjusting the angle, orientation and height of the touch screen 110 when in use is disposed on the operating table 200 ;

[0088] The adjustment mechanism includes a rotating table 210 rotatably arranged on the top of the operating table 200, the integrated machine body 100 is rotatably arranged on one side of the top of the rotating table 210, the rotating table 210 is provided with a mounting groove 220, an electric telescopic rod 230 is hingedly arranged in the groove of the mounting groove 220, and the upper end of the electric telescopic rod 230 is hingedly connected to the rear side of the integrated machine body 100;

[0089] Vertical electric lifting legs 240 are arranged around the bottom of the operating table 200, and four electric lifting legs 240 are arranged in series;

[0090] A vertical rotating column 250 is rotatably provided at the bottom center of the operating table 200 through a bearing sleeve, and the upper end of the rotating column 250 is fixed at the center of the bottom wall of the rotating table 210. The shaft of the rotating column 250 is located below the operating table 200 and is provided with a worm gear 260. A worm 270 is installed at the bottom of the operating table 200 through a bracket. The worm 270 is rotatably provided and drives with the worm gear 260. A servo motor 280 is fixed at the bottom of the operating table 200, and the output shaft of the servo motor 280 is connected to one end of the worm 270 through a coupling.

[0091] A plurality of control switches 290 are disposed on the top of the operating platform 200, and different control switches 290 are electrically connected to the electric telescopic rod 230, the electric lifting leg 240 and the servo motor 280 respectively;

[0092] When the use angle of the touch screen 110 needs to be adjusted, the electric telescopic rod 230 is started, and the electric telescopic rod 230 extends and retracts to push the integrated machine body 100 to adjust the angle between the touch screen 110 and the operating table 200;

[0093] When the use orientation of the touch screen 110 needs to be adjusted, the servo motor 280 is started to drive the worm 270 to rotate, and the rotating column 250 drives the operating table 200 to rotate through the worm gear 260, thereby adjusting the orientation of the touch screen 110 facing the user;

[0094] When the height of the touch screen 110 needs to be adjusted, since the four electric lifting legs 240 are connected in series, when the four electric lifting legs 240 are started at the same time, the height of the touch screen 110 can be adjusted by extending and retracting the electric lifting legs 240.

[0095] Although the present invention has been described above with reference to the embodiments, various modifications may be made thereto and parts thereof may be replaced by equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the various features in the embodiments disclosed in the present invention may be used in combination with each other in any manner, and the fact that these combinations are not exhaustively described in this specification is only for the sake of omitting space and saving resources. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine, characterized in that: include: The image and text data collection module is used to collect images of drug prescription documents and text data of drug use records related to insurance claims; The image and text preprocessing module is used to preprocess the collected image and text data, including image denoising and text cleaning; The image-text feature extraction module includes an image feature extraction unit and a text feature extraction unit. The image feature extraction unit is used to extract visual features of drug packaging and quantity from the preprocessed image, and the text feature extraction unit is used to extract information features of drug name and dosage from the preprocessed text. Multimodal fusion processing module, used to fuse image features and text features to form a unified feature representation; The overdose recognition agent includes a learning model and a decision unit. The learning model is used to train an overdose recognition model based on historical data. The decision unit is used to determine whether the overdose is present based on fusion features and the recognition model. The insurance claim decision module is used to automatically execute the insurance claim process and trigger the manual review process based on the decision results of the overdose recognition agent; The user interaction interface module is used to display the recognition results and claim status, and provide user feedback and manual review interface.

2. According to claim 1, a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine, characterized in that: The image and text data acquisition module realizes the recognition and text extraction of drug prescription document images through OCR technology.

3. According to claim 1, a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine, characterized in that: The image and text preprocessing module further includes an image enhancement unit and a text correction unit, wherein the image enhancement unit is used to improve image quality, and the text correction unit is used to correct text recognition errors.

4. According to claim 1, a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine, characterized in that: The image-text feature extraction module further includes a deep learning network for extracting deep visual features from images.

5. According to claim 1, a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine, characterized in that: The multimodal fusion processing module adopts an attention mechanism to highlight important features and suppress irrelevant information.

6. According to claim 1, a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine is characterized by: The learning model of the overdose recognition agent adopts a hybrid structure of a convolutional neural network (CNN) and a recurrent neural network (RNN).

7. According to claim 1, a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine, characterized in that: The insurance claim decision module automatically executes the insurance claim process and triggers the manual review process according to the decision result of the overdose identification intelligent agent, including automatic calculation of the claim amount and automatic update of the claim status.

8. According to claim 1, a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine, characterized in that: The user interaction interface module provides a real-time feedback mechanism, allowing the user to confirm and raise objections to the recognition results, and use the user feedback for subsequent model optimization.

9. A touch-screen query machine, characterized in that: The invention comprises an integrated machine body (100), wherein the integrated machine body (100) internally stores a digital intelligence system for identifying excessive drug dosage based on a multi-agent platform of a graphic and text multimodal processing engine according to claims 1 to 8, and the integrated machine body (100) comprises a touch screen (110) arranged on the surface of its shell, and a user uses the digital intelligence system through the touch screen (110); The integrated machine body (100) is arranged on an operating table (200), and an adjustment mechanism for adjusting the angle, orientation and height of the touch screen (110) when in use is arranged on the operating table (200).

10. The touch-control query integrated machine according to claim 9, characterized in that: The adjustment mechanism comprises a rotating platform (210) rotatably arranged on the top of the operating table (200); the integrated machine body (100) is rotatably arranged on one side of the top of the rotating platform (210); a mounting groove (220) is provided on the rotating platform (210); an electric telescopic rod (230) is hingedly arranged in the groove of the mounting groove (220); and the upper end of the electric telescopic rod (230) is hingedly connected to the rear side of the integrated machine body (100); Vertical electric lifting legs (240) are arranged around the bottom of the operating table (200), and four of the electric lifting legs (240) are arranged in series; A vertical rotating column (250) is rotatably provided at the center of the bottom of the operating table (200) through a bearing sleeve, the upper end of the rotating column (250) is fixed at the center of the bottom wall of the rotating table (210), the shaft of the rotating column (250) is located below the operating table (200) and is provided with a worm wheel (260), a worm (270) is installed at the bottom of the operating table (200) through a bracket, the worm (270) is rotatably provided and drives with the worm wheel (260), a servo motor (280) is fixed at the bottom of the operating table (200), and the output shaft of the servo motor (280) is connected to one end of the worm (270) through a coupling; A plurality of control switches (290) are arranged on the top of the operating platform (200), and different control switches (290) are electrically connected to the electric telescopic rod (230), the electric lifting leg (240) and the servo motor (280) respectively.

Images