Data extraction method, device, computer and readable storage medium

Abstract

The present invention discloses a data extraction method, which is applied to the field of data processing technology and is used to solve the technical problem that most existing technologies rely on manual extraction of case information, which is prone to errors. The method provided by the present invention includes: screening out hospitalization medical record images, physical examination report images, and surgical report images from the acquired case image set; extracting event elements from the hospitalization medical record images, physical examination report images, and surgical report images; if the event element is a paragraph, identifying the entity data corresponding to the paragraph, the entity data includes the disease type, treatment method, and examination results; if the event element is a table, obtaining the structured data in the table, the structured data includes the examination method, examination site, and surgery name. This method can accurately and automatically extract the required case data, thereby improving the efficiency of case review. At the same time, the present invention also relates to blockchain technology.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data extraction method, device, computer and readable storage medium. Background Art

[0002] Insurance refers to a commercial insurance activity in which the insured pays premiums to the insurer according to the contract, and the insurer bears the liability for compensating the insurance money for the property loss caused by the occurrence of an accident that may occur as stipulated in the contract, or bears the liability for paying insurance money when the insured becomes ill or reaches the age, term, and other conditions stipulated in the contract.

[0003] In the existing insurance field, whether in the underwriting or claims process, customers are required to upload case images in real time, and then draw underwriting conclusions through manual verification.

[0004] However, since existing insurance customers are widely distributed across provinces and cities across the country, and the hospitals providing services vary greatly, manually extracting key information is prone to errors and consumes high labor costs, which is not conducive to the large-scale promotion and use of insurance. Summary of the Invention

[0005] Based on this, the embodiments of the present invention provide a data extraction method, apparatus, computer equipment and storage medium to solve the technical problem that the existing technology relies solely on manual extraction of key information, which is prone to errors and consumes high labor costs.

[0006] A first aspect of the present invention provides a data extraction method, comprising:

[0007] Screen out hospitalization medical record pictures, physical examination report pictures, and operation report pictures from the acquired case picture set;

[0008] Extracting event elements from the hospitalization medical record picture, the physical examination report picture, and the operation report picture, wherein the event elements include paragraphs and tables;

[0009] If the event element is a paragraph, identifying entity data corresponding to the paragraph, the entity data including disease type, treatment method, and examination results;

[0010] If the event element is a table, then structured data in the table is acquired, where the structured data includes an examination method, an examination site, and a surgery name.

[0011] In one embodiment, the step of screening out hospitalization medical record images, physical examination report images, and surgical report images from the acquired case image set includes:

[0012] Train the TextCNN model using the preset training data, and use the features output by the TextCNN model as input features for the xgboost classification model;

[0013] Based on the xgboost classification model, high-frequency words in hospitalization medical record pictures, physical examination report pictures, and surgical report pictures are identified, and a corresponding high-frequency word dictionary is constructed;

[0014] Calculating the hit rate of high-frequency words in the current image according to the high-frequency word dictionary;

[0015] The hospitalization medical record pictures, the physical examination report pictures and the operation report pictures are screened out from the acquired case picture set according to the hit rate.

[0016] In one embodiment, the step of extracting event elements from the hospitalization medical record image, the physical examination report image, and the operation report image includes:

[0017] Identify the semantic labels of each paragraph or table in the hospitalization medical record image, the physical examination report image, and the operation report image based on the LayoutLm model;

[0018] Identifying the categories corresponding to the semantic tags, and determining whether the categories are within a preset range;

[0019] If so, the paragraph or table corresponding to the category is defined as an event element of the inspection event.

[0020] In one embodiment, if the event element is a paragraph, the step of identifying the entity data corresponding to the paragraph includes:

[0021] When the event element is identified as a paragraph, entity data related to medical clinics in the current paragraph is identified and extracted based on the Bert-lstm-crf model or the Crf++ model.

[0022] In one embodiment, if the event element is a table, the step of obtaining structured data in the table includes:

[0023] When the event element is identified as a table, structured data in the table is obtained based on the Xgboost table parsing model.

[0024] In one embodiment, the method further comprises:

[0025] The text characters in the entity data and the structured data are converted into corresponding display signals, and the display information is transmitted to a display terminal, so that the entity data and the structured data are displayed in real time on the display terminal.

[0026] In one embodiment, the results of the above data extraction method can be uploaded to the blockchain, so that the blockchain can encrypt and store the results of the data extraction method.

[0027] A second aspect of the present invention provides a data extraction device, comprising:

[0028] A screening module is used to screen out hospitalization medical record images, physical examination report images, and operation report images from the acquired case image set;

[0029] An extraction module, configured to extract event elements from the hospitalization medical record image, the physical examination report image, and the operation report image, wherein the event elements include paragraphs and tables;

[0030] An execution module is used to identify the entity data corresponding to the paragraph if the event element is a paragraph, and the entity data includes the disease type, treatment method and examination results; if the event element is a table, obtain the structured data in the table, and the structured data includes the examination method, examination site and operation name.

[0031] Wherein, in the above-mentioned data extraction device, the screening module is specifically used for:

[0032] Train the TextCNN model using the preset training data, and use the features output by the TextCNN model as input features for the xgboost classification model;

[0033] Based on the xgboost classification model, high-frequency words in hospitalization medical record pictures, physical examination report pictures, and surgical report pictures are identified, and a corresponding high-frequency word dictionary is constructed;

[0034] Calculating the hit rate of high-frequency words in the current image according to the high-frequency word dictionary;

[0035] The hospitalization medical record pictures, the physical examination report pictures and the operation report pictures are screened out from the acquired case picture set according to the hit rate.

[0036] Wherein, in the above-mentioned data extraction device, the extraction module is specifically used for:

[0037] Identify the semantic labels of each paragraph or table in the hospitalization medical record image, the physical examination report image, and the operation report image based on the LayoutLm model;

[0038] Identifying the categories corresponding to the semantic tags, and determining whether the categories are within a preset range;

[0039] If so, the paragraph or table corresponding to the category is defined as an event element of the inspection event.

[0040] Wherein, in the above-mentioned data extraction device, the execution module is specifically used for:

[0041] When the event element is identified as a paragraph, entity data related to medical clinics in the current paragraph is identified and extracted based on the Bert-lstm-crf model or the Crf++ model.

[0042] Wherein, in the above-mentioned data extraction device, the execution module is specifically used for:

[0043] When the event element is identified as a table, structured data in the table is obtained based on the Xgboost table parsing model.

[0044] A third aspect of the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the data extraction method described above when executing the computer program.

[0045] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the data extraction method described above are implemented.

[0046] The data extraction method, device, computer equipment and storage medium proposed in the present invention screen out hospitalized case images, physical examination report images and surgical report images from the acquired case image set, and further extract event elements from the hospitalized case images, the physical examination report images and the surgical report images, wherein the event elements include paragraphs and tables. Finally, if the event element is a paragraph, the entity data corresponding to the paragraph is identified; if the event element is a table, the structured data in the table is obtained, thereby accurately and automatically extracting the required case data. Through the above method, a system solution for extracting structured information of medical examination reports in hospitalized cases and physical examination reports can be proposed. The images can be automatically generated into structured data through OCR+NLP technology and written into the background database. This reduces the labor costs caused by manual input, review and browsing, and greatly improves the efficiency of case review in the underwriting and claims settlement process in the insurance industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. 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 paying any creative labor.

[0048] Figure 1 is a schematic diagram of an application environment of a data extraction method according to an embodiment of the present invention;

[0049] Figure 2 is a flow chart of a data extraction method according to an embodiment of the present invention;

[0050] Figure 3 is a structural diagram of a data extraction device in one embodiment of the present invention;

[0051] Figure 4 FIG. 1 is a schematic diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0053] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. The drawings illustrate several embodiments of the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the present invention.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0055] The data extraction method provided in this application can be applied to Figure 1In an application environment, the computer device can communicate with the server via a network. The computer device includes but is not limited to various personal computers, laptops, smartphones, tablet computers, and portable wearable devices. The server can be implemented as a standalone server or a server cluster consisting of multiple servers.

[0056] Since existing insurance customers are widely distributed across provinces and cities across the country, and the hospitals providing services vary greatly, manually extracting key information is prone to errors and consumes high labor costs, which is not conducive to the large-scale promotion and use of insurance.

[0057] In one embodiment, if Figure 2 As shown, a data extraction method is provided, which is applied in Figure 1 The computer device in the embodiment is used as an example to illustrate, which includes the following steps S101 to S103.

[0058] This data extraction method is mainly used to solve the technical problem that the existing technology relies solely on manual extraction of key information, which is prone to errors and consumes high labor costs.

[0059] Specifically, the data extraction method provided in this embodiment includes the following steps:

[0060] S101: Filter out hospitalization case images, physical examination report images, and operation report images from the acquired case image set.

[0061] Specifically, in this embodiment, it should be noted that the data extraction method provided in this embodiment is currently applicable to document types including: hospitalization records, or single examination, test, or operation report sheets in physical examination reports.

[0062] Furthermore, the types of medical events extracted by the data extraction method provided in this embodiment include: examinations (MRI, CT, X-ray, endoscopy, electroencephalogram, electrocardiogram, pathology, ultrasound), tests (blood routine, urine routine, biochemistry, liver function, hepatitis B five items, etc.), and surgery.

[0063] In this embodiment, scanned copies of various types of medical records will be pre-stored in the database, that is, pictures of various types of medical records will be stored, and corresponding case picture sets can be generated from the pictures of various types of medical records. Furthermore, the required hospitalization case pictures, physical examination report pictures and surgical report pictures can be screened out from the case picture set.

[0064] S102: extracting event elements from the hospitalization case picture, the physical examination report picture, and the operation report picture, wherein the event elements include paragraphs and tables.

[0065] It should be noted that in this step, the above-mentioned hospitalized case pictures, physical examination report pictures and surgical report pictures will be extracted and processed. Specifically, this step will extract the event elements in the above-mentioned hospitalized case pictures, physical examination report pictures and surgical report pictures for the next step of extraction and processing, wherein the above-mentioned event elements include paragraphs and tables.

[0066] In addition, it should be noted that this step will only be further performed if the case picture meets one of the above-mentioned hospitalization case pictures, physical examination report pictures and operation report pictures.

[0067] S103: If the event element is a paragraph, identifying entity data corresponding to the paragraph, the entity data including disease type, treatment method, and examination results;

[0068] If the event element is a table, then structured data in the table is acquired, where the structured data includes an examination method, an examination site, and a surgery name.

[0069] In this step, it should be noted that when the event elements in the above-mentioned inpatient case picture, the physical examination report picture and the surgical report picture are obtained through the above-mentioned step S102, the event elements will be immediately analyzed, that is, it will be determined whether the current event element belongs to a paragraph or a table.

[0070] Furthermore, if the event element is a paragraph, entity data corresponding to the current paragraph is identified, and the entity data includes disease type, treatment method, and examination result;

[0071] If the event element is a table, the structured data in the current table is obtained, where the structured data includes the examination method, the examination site, and the name of the operation.

[0072] Finally, in this step, the required entity data and structured data can be quickly and automatically obtained through the above steps and written into the background database, eliminating the manual verification process and greatly improving the efficiency of the review.

[0073] In addition, in this embodiment, the data extraction method further includes:

[0074] The text characters in the entity data and the structured data are converted into corresponding display signals, and the display information is transmitted to a display terminal, so that the entity data and the structured data are displayed in real time on the display terminal.

[0075] In summary, the data extraction method provided in this embodiment provides a systematic solution for extracting structured information from medical examination reports such as hospitalization records and physical examination reports. Structured data can be automatically generated from images using OCR and NLP technology and written to a backend database. This reduces the labor costs associated with manual data entry, review, and review, significantly improving the efficiency of case review during underwriting and claims processing in the insurance industry.

[0076] In another embodiment, the data extraction method includes steps S201 to S203.

[0077] S201: training a TextCNN model using preset training data, and using features output by the TextCNN model as input features for an xgboost classification model;

[0078] Identify high-frequency words in the hospitalized case pictures, the physical examination report pictures, and the surgical report pictures based on the xgboost classification model, and construct a corresponding high-frequency word dictionary;

[0079] Calculating the hit rate of high-frequency words in the current image according to the high-frequency word dictionary;

[0080] The hospitalization case pictures, the physical examination report pictures and the operation report pictures are screened out from the acquired case picture set according to the hit rate.

[0081] In this embodiment, it should be noted that the main idea of ​​the above TextCNN model is to take short sentences of different lengths as matrix input, use multiple filters of different sizes to extract key information from the sentences (similar to ngrams with multiple window sizes), and use them for the final classification.

[0082] Therefore, in this embodiment, the TextCNN model is first trained using the preset training data. Further, the features output by the TextCNN model are used as the input features of the preset xgboost classification model. Specifically, the xgboost classification model is also a type of gradient boosting tree model. It also generates models serially and takes the sum of all models as output. XGBoost performs a second-order Taylor expansion on the loss function, optimizes the loss function using the second-order derivative information of the loss function, and chooses whether to split the node based on whether the loss function is reduced. At the same time, XGBoost has added regularization, learning rate, column sampling, approximate optimal split point and other means to prevent overfitting. Certain optimizations have also been made in dealing with missing values.

[0083] Therefore, when the features output by the TextCNN model are input to the xgboost classification model, the xgboost classification model traverses all features and split points, and selects the feature and split point that minimizes the squared error for splitting. When a CART decision tree is used for classification, the xgboost classification model traverses all features and split points, and selects the feature and split point that minimizes the Gini coefficient for splitting.

[0084] Furthermore, this step will use the above-mentioned xgboost classification model to identify high-frequency words in the above-mentioned hospitalized case pictures, physical examination report pictures and surgical report pictures, and construct a corresponding high-frequency word dictionary, and calculate the high-frequency word hit rate in the current picture based on the high-frequency word dictionary. Finally, according to the hit rate, the hospitalized case pictures, the physical examination report pictures and the surgical report pictures are screened out from the acquired case picture set.

[0085] Specifically, this step calculates the frequency of each dictionary word in the current text as a keyword frequency feature. It also constructs regularizations for each image title type and calculates the number of hits for each image type, assigning a 1 if the title is a hit and a 0 if it is not. This is used to calculate the regularization feature. The XGBoost model complements the CNN model, further improving accuracy.

[0086] It should be noted that this embodiment will pre-set a hit rate threshold in the background server, for example, ≥75%. When the above-mentioned xgboost classification model recognizes that the hit rate of high-frequency words in the above-mentioned inpatient case pictures, physical examination report pictures and surgical report pictures is ≥75%, the current picture can be automatically filtered out, and the picture is one of the above-mentioned inpatient case pictures, physical examination report pictures and surgical report pictures, which greatly improves work efficiency.

[0087] S202: Identify semantic labels of each paragraph or table in the hospitalization case image, the physical examination report image, and the operation report image based on the LayoutLm model;

[0088] Identifying the categories corresponding to the semantic tags, and determining whether the categories are within a preset range;

[0089] If so, the paragraph or table corresponding to the category is defined as an event element of the inspection event.

[0090] In this embodiment, it should be noted that when the inpatient case picture, the physical examination report picture and the surgical report picture are obtained through the above step S201, this step will immediately identify the semantic labels of each paragraph or table in the inpatient case picture, the physical examination report picture and the surgical report picture based on the LayoutLm model.

[0091] It's worth noting that the aforementioned LayoutLM model is capable of pre-training language models on a large scale through self-supervised tasks. It effectively captures the semantic information contained in text during the pre-training phase, and effectively improves model performance after fine-tuning on downstream tasks. However, existing pre-trained language models primarily target a single modality, neglecting the visual structure of the document itself, which naturally aligns with the text. In actual use, the LayoutLM model requires only minimal annotated data to achieve industry-leading performance.

[0092] Therefore, in this step, the LayoutLm model is used to identify the semantic labels of each paragraph or table in the hospitalization case image, the physical examination report image, and the surgical report image. Furthermore, the category corresponding to each semantic label is identified, and a determination is made as to whether the category falls within a preset range. If so, the paragraph or table corresponding to the category is defined as an event element of the inspection event.

[0093] It should be noted that, in this embodiment, the above semantic tags are divided into 10 categories, namely: 1. Event title, such as xx hospital xx report, or blood routine, urine routine; 2. Event process, such as examination description, examination findings, ultrasound description, etc.; 3. Conclusion, such as ultrasound diagnosis, examination conclusion, summary, impression, etc. (non-clinical diagnosis); 4. Basic information of the event, such as other information related to the current event (such as: name, surgical method, date, etc.); 5. Examination site, such as the examination site / examination item paragraph or related text fragment; 6. Examination method, such as the examination method / examination method paragraph or related text fragment; 7. Operation name, such as the operation name paragraph; 8. Preoperative diagnosis, such as the preoperative diagnosis paragraph; 9. Intraoperative diagnosis, such as the intraoperative diagnosis paragraph; 10. Postoperative diagnosis, such as the postoperative diagnosis paragraph.

[0094] Therefore, in this step, when the above-mentioned LayoutLm model identifies that the semantic labels of each paragraph or table in the inpatient case picture, the physical examination report picture and the surgical report picture belong to one of the above-mentioned 10 categories, the current paragraph or table will be defined as an event element of the inspection event.

[0095] S203: When the event element is identified as a paragraph, identifying and extracting entity data related to medical clinics in the current paragraph based on the Bert-lstm-crf model or the Crf++ model;

[0096] When the event element is identified as a table, structured data in the table is obtained based on the Xgboost table parsing model.

[0097] Finally, in this embodiment, it should be noted that when the current event element is identified as a paragraph in the above step S202, entity data related to medical clinical practice of the current paragraph will be identified and extracted based on the Bert-lstm-crf model or the Crf++ model;

[0098] It should be noted that the above-mentioned Bert-LSTM-CRF model first generates word vectors based on contextual information through BERT model preprocessing, and then inputs the trained word vectors into the BiLSTM-CRF model for further training processing, so as to accurately extract the entity data related to medical clinical practice in the current paragraph. The entity data includes disease type, treatment method, and examination results.

[0099] In addition, when the current event element is identified as a table through the above step S202, the structured data in the table will be obtained based on the Xgboost table parsing model.

[0100] It's important to note that the XgBoost table parsing model optimizes the loss function using its second-order derivative by performing a second-order Taylor expansion on the loss function, then decides whether to split a node based on whether the loss function decreases. XGBoost also incorporates regularization, learning rates, column sampling, and approximate optimal split points to prevent overfitting, enabling it to accurately capture the structured data in the table above, including the examination method, examination site, and surgical procedure name.

[0101] In summary, the data extraction method proposed in this application selects hospitalized case pictures, physical examination report pictures and surgical report pictures from the acquired case picture set, and further extracts event elements from the hospitalized case pictures, the physical examination report pictures and the surgical report pictures, wherein the event elements include paragraphs and tables. Finally, if the event element is a paragraph, the entity data corresponding to the paragraph is identified; if the event element is a table, the structured data in the table is obtained, thereby accurately and automatically extracting the required case data. Through the above method, a system solution for extracting structured information of medical examination reports in hospitalized cases and physical examination reports can be proposed. Structured data can be automatically generated from pictures using OCR+NLP technology and written into the background database. This reduces the labor costs caused by manual entry, review and browsing, and greatly improves the efficiency of case review in the underwriting and claims process in the insurance industry.

[0102] In an optional embodiment, the results of the above data extraction method can also be uploaded to the blockchain.

[0103] Specifically, corresponding summary information is obtained based on the results of the data extraction method. Specifically, the summary information is obtained by hashing the results of the data extraction method, such as using the sha256s algorithm. Uploading the summary information to the blockchain ensures its security and fairness and transparency to users. Users can download the summary information from the blockchain to verify whether the results of the data extraction method have been tampered with. The blockchain referred to in this example is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a string of data blocks generated using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity of the information (to prevent counterfeiting) and generate the next block. Blockchain can include the blockchain underlying platform, the platform product service layer, and the application service layer.

[0104] In one embodiment, a data extraction device 100 is provided, which corresponds one-to-one to the data extraction method in the above embodiment. Figure 3 As shown, the data extraction device 100 includes: a screening module 11, an extraction module 12 and an execution module 13. The detailed description of each functional module is as follows:

[0105] The screening module 11 is used to screen out hospitalization medical record pictures, physical examination report pictures and operation report pictures from the acquired case picture set;

[0106] An extraction module 12 is used to extract event elements from the hospitalization medical record image, the physical examination report image, and the operation report image, wherein the event elements include paragraphs and tables;

[0107] The execution module 13 is used to identify the entity data corresponding to the paragraph if the event element is a paragraph, and the entity data includes the disease type, treatment method and examination results; if the event element is a table, obtain the structured data in the table, and the structured data includes the examination method, examination site and operation name.

[0108] In the above-mentioned data extraction device 100, the screening module 11 is specifically used for:

[0109] Train the TextCNN model using the preset training data, and use the features output by the TextCNN model as input features for the xgboost classification model;

[0110] Based on the xgboost classification model, high-frequency words in hospitalization medical record pictures, physical examination report pictures, and surgical report pictures are identified, and a corresponding high-frequency word dictionary is constructed;

[0111] Calculating the hit rate of high-frequency words in the current image according to the high-frequency word dictionary;

[0112] The hospitalization medical record pictures, the physical examination report pictures and the operation report pictures are screened out from the acquired case picture set according to the hit rate.

[0113] In the above-mentioned data extraction device 100, the extraction module 12 is specifically used for:

[0114] Identify the semantic labels of each paragraph or table in the hospitalization medical record image, the physical examination report image, and the operation report image based on the LayoutLm model;

[0115] Identifying the categories corresponding to the semantic tags, and determining whether the categories are within a preset range;

[0116] If so, the paragraph or table corresponding to the category is defined as an event element of the inspection event.

[0117] In the above-mentioned data extraction device 100, the execution module 13 is specifically used to:

[0118] When the event element is identified as a paragraph, entity data related to medical clinics in the current paragraph is identified and extracted based on the Bert-lstm-crf model or the Crf++ model.

[0119] In the above-mentioned data extraction device 100, the execution module 13 is specifically used to:

[0120] When the event element is identified as a table, structured data in the table is obtained based on the Xgboost table parsing model.

[0121] Another embodiment of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the data extraction method described above when executing the computer program.

[0122] Another embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the data extraction method described above are implemented.

[0123] The meaning of "first" and "second" in the above modules / units is only to distinguish different modules / units, and is not used to limit which module / unit has a higher priority or other limiting meanings. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The division of modules in this application is only a logical division, and other division methods may be used in actual applications.

[0124] The specific definition of the data extraction device can be found in the definition of the data extraction method above and will not be repeated here. The various modules in the above-mentioned living body data extraction device can be implemented in whole or in part through software, hardware, or a combination thereof. The above-mentioned modules can be embedded in or independent of the processor of the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each of the above modules.

[0125] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown. The computer device includes a processor, memory, a network interface, a display screen, and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements a data extraction method.

[0126] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the data extraction method in the above embodiment are implemented, such as Figure 2 Alternatively, when the processor executes the computer program, the functions of each module / unit of the device for detecting living face images in the above embodiment are realized, for example, Figure 3 The functions of modules 11 to 13 are shown in FIG.

[0127] The processor may be a central processing unit (CPU), other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the computer device, connecting various parts of the entire computer device using various interfaces and lines.

[0128] The memory may be used to store the computer programs and / or modules. The processor implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory and calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.); and the data storage area may store data created based on the use of the mobile phone (such as audio data, video data, etc.).

[0129] The memory may be integrated into the processor or may be provided separately from the processor.

[0130] 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 of the data extraction method in the above embodiment are implemented, such as Figure 2 Alternatively, when the computer program is executed by a processor, the functions of each module / unit of the device for detecting living face images in the above embodiment are realized, for example, Figure 3 The functions of modules 11 to 13 are shown in FIG.

[0131] Those skilled in the art will appreciate 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. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0132] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0133] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A data extraction method, characterized in that: The method comprises: Screening out hospitalization medical record pictures, physical examination report pictures and operation report pictures from the acquired case picture set; the step of screening out hospitalization medical record pictures, physical examination report pictures and operation report pictures from the acquired case picture set includes: training a TextCNN model through preset training data, and using the features output by the TextCNN model as input features of an xgboost classification model; identifying high-frequency words in the hospitalization medical record pictures, physical examination report pictures and operation report pictures based on the xgboost classification model, and constructing a corresponding high-frequency word dictionary; calculating the high-frequency word hit rate in the current picture according to the high-frequency word dictionary; screening out the hospitalization medical record pictures, physical examination report pictures and operation report pictures from the acquired case picture set according to the hit rate; screening out the hospitalization medical record pictures, physical examination report pictures and operation report pictures from the acquired case picture set according to the hit rate includes screening out the hospitalization medical record pictures, physical examination report pictures and operation report pictures when the hit rate of the high-frequency words in the hospitalization medical record pictures, physical examination report pictures and operation report pictures is greater than a preset threshold; Extracting event elements from the hospitalization medical record picture, the physical examination report picture, and the operation report picture, wherein the event elements include paragraphs and tables; If the event element is a paragraph, identifying entity data corresponding to the paragraph, the entity data including disease type, treatment method, and examination results; If the event element is a table, then structured data in the table is acquired, where the structured data includes an examination method, an examination site, and a surgery name.

2. The data extraction method according to claim 1, characterized in that: The step of extracting event elements from the hospitalization medical record picture, the physical examination report picture, and the operation report picture comprises: Identify the semantic labels of each paragraph or table in the hospitalization medical record image, the physical examination report image, and the operation report image based on the LayoutLm model; Identifying the categories corresponding to the semantic tags, and determining whether the categories are within a preset range; If so, the paragraph or table corresponding to the category is defined as an event element of the inspection event.

3. The data extraction method according to claim 1, wherein: If the event element is a paragraph, the step of identifying the entity data corresponding to the paragraph includes: When the event element is identified as a paragraph, entity data related to medical clinics in the current paragraph is identified and extracted based on the Bert-lstm-crf model or the Crf++ model.

4. The data extraction method according to claim 1, wherein: If the event element is a table, the step of obtaining structured data in the table includes: When the event element is identified as a table, structured data in the table is obtained based on the Xgboost table parsing model.

5. The data extraction method according to claim 1, wherein: The method further comprises: The text characters in the entity data and the structured data are converted into corresponding display signals, and the display signals are transmitted to a display terminal, so that the entity data and the structured data are displayed in real time on the display terminal.

6. A data extraction device, characterized in that: The device comprises: A screening module is used to screen out hospitalization medical record images, physical examination report images, and operation report images from the acquired case image set; The screening module is specifically used to train a TextCNN model through preset training data, and use the features output by the TextCNN model as input features of an xgboost classification model; identify high-frequency words in hospitalized medical record pictures, physical examination report pictures and surgical report pictures based on the xgboost classification model, and construct a corresponding high-frequency word dictionary; calculate the hit rate of high-frequency words in the current picture according to the high-frequency word dictionary; screen out the hospitalized medical record pictures, the physical examination report pictures and the surgical report pictures from the acquired case picture set according to the hit rate; screening out the hospitalized medical record pictures, the physical examination report pictures and the surgical report pictures from the acquired case picture set according to the hit rate includes: when the hit rate of high-frequency words in the hospitalized medical record pictures, the physical examination report pictures and the surgical report pictures is greater than a preset threshold, screening out the hospitalized medical record pictures, the physical examination report pictures and the surgical report pictures; An extraction module, configured to extract event elements from the hospitalization medical record image, the physical examination report image, and the operation report image, wherein the event elements include paragraphs and tables; An execution module is used to identify the entity data corresponding to the paragraph if the event element is a paragraph, and the entity data includes the disease type, treatment method and examination results; if the event element is a table, obtain the structured data in the table, and the structured data includes the examination method, examination site and operation name.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the data extraction method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the data extraction method according to any one of claims 1 to 5 are implemented.

Images