Method, device and equipment for accurately extracting medical diagnosis material information and medium

Abstract

The invention relates to the technical field of medical treatment and image processing, and discloses a method, device and equipment for accurately extracting medical diagnosis material information and a medium, and the method comprises the steps: obtaining a medical material image, and preprocessing the medical material image; acquiring medical diagnosis material information based on double-link inspection; medical diagnosis material information subjected to double-link inspection is extracted and subjected to mutual verification; and extracting medical diagnosis material information with high confidence in the verification result and outputting the medical diagnosis material information. Based on the construction of the two independent text analysis links, the accuracy when the text values extracted from the two paths are the same is improved, automatic passing of the claim settlement case is realized, and the input efficiency of the whole medical material is improved. The method is applied to claim insurance under the medical background.

Description

Technical Field

[0001] The present invention relates to the fields of medical technology and image processing technology, and in particular to a method, device, equipment and medium for accurately extracting medical diagnosis material information. Background Art

[0002] In the process of entering medical claims materials, in addition to medical bills, it is also necessary to extract the main diagnosis and secondary diagnosis from medical materials such as outpatient medical records, prescriptions, and medical records. At the same time, the medical diagnosis materials are associated with the medical bills one by one through the relevant fields to complete the whole case entry. If the whole case entry is to be automated, the high confidence part of the main diagnosis and secondary diagnosis extracted by the model (the high confidence part refers to the part where the model task extraction result must be correct) needs to be saved without manual intervention. The accuracy rate can reach 99%. However, the accuracy rate of the high confidence extraction results output by the existing extraction model is basically around 95%, and the recall rate is within 10%, which is difficult to meet the requirements of automatic entry of the whole case. Therefore, the existing method can only use the extraction results as a reference, and it is necessary to manually enter the main diagnosis and secondary diagnosis. This method is not efficient, and the accuracy rate is also directly related to the professional ability of the business personnel, which is prone to errors. Summary of the invention

[0003] The present invention provides a method, device, computer equipment and medium for accurately extracting medical diagnosis material information to solve the technical problem that the semantic matching model is inaccurate, the matching ability is seriously insufficient, and the question-answering effect is poor.

[0004] In a first aspect, a method for accurately extracting medical diagnostic material information is provided, comprising:

[0005] Acquire medical material images and pre-process the medical material images;

[0006] Acquiring medical diagnostic material information based on dual-link testing;

[0007] Extract medical diagnostic material information from dual-link inspection and perform mutual verification;

[0008] Extract and output high-confidence medical diagnostic material information from the verification results.

[0009] In a second aspect, a device for accurately extracting medical diagnostic material information is provided, comprising:

[0010] A preprocessing module, used for acquiring medical material images and preprocessing the medical material images;

[0011] An inspection module, used to obtain medical diagnostic material information based on dual-link inspection;

[0012] A verification module, used to extract the medical diagnosis material information of the dual-link inspection and perform mutual verification;

[0013] The output module is used to extract and output the medical diagnostic material information with high confidence in the verification results.

[0014] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for accurately extracting medical diagnostic material information when executing the computer program.

[0015] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned method for accurately extracting medical diagnostic material information are implemented.

[0016] In the scheme implemented by the above-mentioned method, device, computer equipment and storage medium for accurately extracting medical diagnostic material information, the medical material image can be obtained, the medical material image can be preprocessed; the medical diagnostic material information can be obtained based on dual-link verification; the medical diagnostic material information of the dual-link verification can be extracted and mutually verified; the medical diagnostic material information with high confidence in the verification result can be extracted and output. In the present invention, in the medical background, insurance claims are settled, and the accuracy rate when the text values ​​extracted from the two paths are the same is improved by constructing two independent text parsing links, so as to realize the automatic passing of the claims case, improve the overall efficiency of medical material input, save labor costs, and extract and parse the relevant fields of the medical material image with the help of end-to-end models and pipeline tools, so as to improve the data cleaning efficiency and the quality of model training data, and at the same time, the end-to-end model can be quickly iterated to grasp the changes in input data in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0018] Figure 1 is a schematic diagram of an application environment of a method for accurately extracting medical diagnosis material information in one embodiment of the present invention;

[0019] Figure 2 is a flow chart of a method for accurately extracting medical diagnosis material information in one embodiment of the present invention;

[0020] Figure 3 yes Figure 2A schematic flow chart of a specific implementation of step S20;

[0021] Figure 4 yes Figure 2 Another specific implementation flow diagram of step S20;

[0022] Figure 5 It is a schematic diagram of a structure of a device for accurately extracting medical diagnosis material information in one embodiment of the present invention;

[0023] Figure 6 is a schematic diagram of a structure of a computer device in one embodiment of the present invention;

[0024] Figure 7 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] The method for accurately extracting medical diagnostic material information provided by the embodiment of the present invention can be applied in the following aspects: Figure 1 In the application environment, the client communicates with the server through the network. The server can obtain medical material images, pre-process medical material images; obtain medical diagnostic material information based on dual-link inspection; extract medical diagnostic material information of dual-link inspection and perform mutual verification; extract and output medical diagnostic material information with high confidence in the verification results. In the present invention, in the medical background, insurance claims are settled, and the accuracy of the text values ​​extracted from the two paths is improved by constructing two independent text parsing links, so as to realize the automatic passing of claims cases, improve the overall efficiency of medical material input, save labor costs, and extract and parse the relevant fields of medical material images with the help of end-to-end models and pipeline tools to improve data cleaning efficiency and the quality of model training data. At the same time, the end-to-end model can be quickly iterated to grasp the changes in input data in time. Among them, the client can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices. The server can be implemented with an independent server or a server cluster composed of multiple servers. The present invention is described in detail below through specific embodiments.

[0027] See also Figure 2 As shown, Figure 2A flow chart of a method for accurately extracting medical diagnostic material information provided by an embodiment of the present invention includes the following steps:

[0028] S10: acquiring medical material images and preprocessing the medical material images;

[0029] The method for accurately extracting medical diagnostic material information provided by the present invention can be applied to claims data entry systems such as insurance claims or medical claims in various application scenarios. The insured party enters his or her own medical diagnostic material information in the claims data entry system, and then the method for accurately extracting medical diagnostic material information in the claims data entry system accurately confirms the diagnosis results in the medical diagnostic material information, so as to improve the accuracy of medical diagnostic material information extraction in medical claims.

[0030] Among them, the medical diagnosis material information at least includes medical receipts, outpatient medical records, prescription notes, medical record front pages, etc., and the main diagnosis, secondary diagnosis, material category, hospital name, campus, department, admission time, discharge time and other data are extracted from the medical diagnosis material information. Among them, the preset pre-information is the data extracted from the medical diagnosis material information except the main diagnosis and secondary diagnosis.

[0031] The insured party uploads the medical diagnostic material information to the claim data entry system in the form of images. After the claim data entry system obtains the medical material images, it pre-processes the medical material images. Pre-processing the medical material images is to convert the input medical material images into a format acceptable to the subsequent training model, thereby facilitating the extraction of effective features and ensuring that the input meets the input requirements of the model. Preprocessing medical material images is specifically as follows: Scaling medical material images: adjusting the image to a fixed size to ensure that the model can handle inputs of consistent size. For example, for adapting the Donut visual language model, scale the image to a predetermined size (such as 224x224 or 512x512); Standardizing medical material images: ensuring that the pixel values ​​of the medical material images are within a certain range, normalizing the pixel values ​​of the image, where the mean of the training data is subtracted and divided by the standard deviation. Standardization can help accelerate model training and improve performance. For example, assuming that the training data uses the mean [0.485, 0.456, 0.406] and standard deviation [0.229, 0.224, 0.225] of RGB values, these means and standard deviations can be used to standardize test images; Performing color space conversion on medical material images, converting the image from the original BGR format (such as in OpenCV) or RGB (such as PIL or other libraries) format to a standard RGB format. Perform data augmentation on medical material images and perform some random transformations on training images, such as rotation, cropping, flipping, color adjustment, etc., to improve the generalization ability of the model. Convert medical material images to tensors. Convert images to tensor form, usually converted to the shape of [batch_size, channels, height, width]. Deep learning frameworks such as PyTorch or TensorFlow require that the input must be a tensor. Batch process and convert medical material images. If the model processes multiple images in batches, the medical material images will be organized into a batch during preprocessing and converted appropriately.

[0032] In this embodiment, the specific steps of image preprocessing are automatically completed by AutoProcessor. Specifically, the Pillow (PIL) library is used to load the image; if the image is in other formats (such as BGR), it will be converted to RGB format; the image is scaled to a size suitable for the model and normalized (for example, the mean is subtracted and divided by the standard deviation); finally, the image is converted to a tensor format as the input of the model. Among them, AutoProcessor is a tool for image and text preprocessing, which can automatically process the input image according to the requirements of the model.

[0033] S20: Acquiring medical diagnostic material information based on dual-link inspection;

[0034] In this embodiment, the dual-link verification is two methods for extracting image data, so two methods for extracting text information in an image are selected as the dual-link verification scheme of the present invention. Among them, the optional schemes for extracting text information in an image are at least two of the following types: template matching, connected region analysis, character segmentation and classification, open source OCR engine, convolutional neural network, convolutional recurrent neural network, EAST (Efficient and Accurate Scene Text Detector), CTPN (Connectionist Text Proposal Network), end-to-end network for text detection and recognition, self-attention mechanism of Transformer network, visual language model based on Transformer, OCR library based on deep learning.

[0035] It should be noted that template matching matches the characters in the image through pre-defined character templates. Each character template represents a possible character form. The system recognizes the characters by calculating the similarity between the character template and the image. Connected region analysis recognizes characters by detecting connected regions in the image. By extracting regions with similar pixel values ​​in the image, the algorithm regards these regions as characters. Character segmentation and classification performs character segmentation on the entire image, extracts each character from the image, and then uses template matching or machine learning classifiers (such as support vector machines SVM) for recognition.

[0036] Tesseract OCR, Tesseract is an open source text extraction (OCR) engine that uses traditional character segmentation and classification methods and also combines some statistical-based techniques. It can recognize multiple languages ​​and support the positioning and analysis of text areas. Convolutional neural network, uses convolutional neural network to extract image features and further classifies through fully connected layers or RNN (Recurrent Neural Network, i.e. recurrent neural network). Convolutional recurrent neural network, combines convolutional neural network and RNN (usually using LSTM (Long Short-Term Memory Network)) for image text extraction. Convolutional neural network extracts image features, RNN captures the temporal features of text, and finally uses CTC (Connectionist Temporal Classification) loss for training. EAST (Efficient and Accurate Scene Text Detector) is a deep learning method for image text detection. It can automatically detect text areas in images and predict text bounding boxes. This method is end-to-end and supports efficient text detection. CTPN (Connectionist Text Proposal Network) uses convolutional neural networks to extract candidate boxes of text areas and uses RNN to perform sequence modeling to achieve text detection and recognition. The end-to-end network-enhanced text detection algorithm for text detection and recognition uses convolutional neural networks and RNN layers to achieve text detection and recognition. The self-attention mechanism of the Transformer network (a neural network based on the self-attention mechanism) is combined with the encoder-decoder structure for text extraction, and can adaptively focus on relevant areas in the image through the attention mechanism. The Transformer-based visual language model (Donut), which is specifically used for document understanding tasks, can automatically recognize and extract structured text information in documents without relying on traditional text extraction (OCR) tools. EasyOCR is a text extraction (OCR) library based on deep learning that can handle multiple fonts, styles, and complex text scenes. It uses convolutional recurrent neural networks for text extraction and is suitable for complex text and natural scenes.

[0037] In this embodiment, the dual link inspection uses end-to-end model link detection and pipeline tool detection.

[0038] Specifically, Figure 3 As shown, the end-to-end model link detection implementation steps include:

[0039] S21: Analyze medical material images based on visual language model to obtain preset key fields;

[0040] S22: Obtain the diagnosis results and preset pre-information in the historical data, and use the diagnosis results and the preset pre-information as label data for training to construct a deep learning model;

[0041] S23: Extract text information from preset key fields, and predict medical diagnostic material information corresponding to the text information based on a deep learning model.

[0042] For steps S21-S23, after preprocessing the medical material image, locate the area in the image that may contain text, that is, find the text box or text area, where the preset key field is the area that may contain text, and then pass the input image to the end-to-end convolutional neural network. Visual features in the medical material image are extracted by methods such as convolutional neural network (CNN). The convolutional neural network identifies the text area in the medical material image and locates the text box or polygonal area through a self-attention mechanism or a convolutional feature map. After the text area is detected, the end-to-end model further processes the text area and extracts the text content, usually using methods such as CRNN, RNN or LSTM for character sequence recognition. Finally, the detected text area and its corresponding text content are output. Among them, the construction of the end-to-end model is to obtain the diagnosis results and preset pre-information in the historical data, and use the diagnosis results and preset pre-information as label data for training to construct a deep learning model.

[0043] It should be noted that the end-to-end model uses the donut visual language model, the encoder uses swin, and the decoder uses the first four layers of bart. The input only requires medical material images, and structured text can be directly output. Therefore, there are several innovations in the use of the end-to-end model: First, data cleaning uses the image_to_text image conversion capability of the large model to parse the image information of the hospital name, department, admission time and other fields, and performs dual-link verification with the current pipeline (pipeline tool) on the parsing results of the hospital name, department, admission time and other fields, and compares the unequal parsing results of the two to perform key data cleaning, thereby saving a lot of data annotation time, while improving data quality, and being able to construct a large amount of training data. Second, training data construction, in addition to adding the primary diagnosis and secondary diagnosis to the training label data, a number of pre-information is also added, such as material category, hospital name, campus, department, etc. Because the decoding part of donut uses autoregression to predict text output, the extraction of primary and secondary diagnoses used by departments in different hospital campuses has commonalities. Therefore, by allowing the model to learn this information, the prediction accuracy and generalization ability of the end-to-end model can be improved.

[0044] In order to reduce the difficulty of end-to-end model training, by obtaining the diagnostic results and preset pre-information in the historical data, the preset complex diagnostic logic in the diagnostic results is extracted; the extracted diagnostic results and preset pre-information are used as label data for training to build a deep learning model. That is, in the extraction of primary and secondary diagnoses, some complex diagnostic logic needs to be extracted. For example, the first Western medicine diagnosis should be the primary diagnosis, and the rest should be secondary diagnoses, and all Chinese medicine diagnoses should also be secondary diagnoses. Therefore, in order to reduce the learning difficulty of the end-to-end model for this type of material, the end-to-end model is allowed to directly learn the content of Western medicine diagnosis and Chinese medicine diagnosis instead of directly letting it learn the content of primary and secondary diagnoses. When the end-to-end model can correctly extract Western medicine diagnosis and Chinese medicine diagnosis, the first Western medicine diagnosis can be used as the primary diagnosis by writing relevant rules, and the other Western medicine diagnoses plus Chinese medicine diagnoses can be returned as secondary diagnoses. Optimize the construction logic of the training data, remove some of the complex association logic from the training samples, and summarize and summarize them by developing rules in the later stage, so as to reduce the difficulty of end-to-end model training, thereby improving the accuracy and recall rate of end-to-end model extraction.

[0045] Specifically, Figure 4 As shown, the pipeline tool detection implementation steps include:

[0046] S24: Detecting the position of text on the medical material image based on the text detection model;

[0047] S25: recognizing the image information at the position of the text on the medical material image into text information based on the text recognition model;

[0048] S26: Obtain medical diagnosis material information based on structured text information of the information extraction model.

[0049] For steps S24-S26, the pipeline tool first detects the position of the text on the material through the text detection model (dptext-deltr), and then recognizes the medical material image information framed by the text box on the medical material image as text information through the text recognition model (CRNN+CTC), and finally structures the text information through the information extraction model (BERT+rules). The method of locating the position of the diagnostic information on the medical diagnostic material, for example, uses regular expressions to extract keywords such as "clinical diagnosis" and "Western medicine diagnosis" from the full text, and then extracts the diagnostic information from the text behind these keywords. Specifically, the BERT (pre-trained language model) model is used to extract the primary diagnosis and secondary diagnosis information.

[0050] S30: extracting medical diagnosis material information of dual-link inspection and performing mutual verification;

[0051] Specifically, by extracting the medical diagnostic material information detected by the end-to-end model link and the medical diagnostic material information detected by the pipeline tool; comparing and verifying the medical diagnostic material information detected by the end-to-end model link and the medical diagnostic material information detected by the pipeline tool, and obtaining the verification result. Through the parsing ability of the end-to-end model and the parsing ability of the traditional pipeline tool, the relevant fields of the medical material data are extracted and parsed, and the extraction results of the two are mutually verified. The labeling personnel can focus on the data cleaning of the different parsed parts of the two, thereby improving the data cleaning efficiency and the quality of the model training data, and can also quickly iterate the end-to-end model and grasp the changes in the input data in time.

[0052] By using two independent image-to-text extraction methods for dual-link verification, the accuracy and recall of high-confidence extraction results are improved to meet business requirements and achieve the effect of 1+1 greater than 2. When the text information in medical material images is extracted by the pipeline tool (detection + recognition + information extraction) combined with the dual-link verification of the end-to-end model, and mutually verified, the experimental results show that the accuracy of high-confidence extraction results reaches about 99%, and the recall reaches about 50%.

[0053] S40: extracting and outputting the medical diagnosis material information with high confidence in the verification results.

[0054] Specifically, by cleaning the unequal data in the verification results based on the verification results; confirming the high-confidence medical diagnosis material information in the verification results and proposing output. The main diagnosis and secondary diagnosis are extracted by dual-link inspection. The accuracy and recall rate of the extracted diagnosis results with high confidence can meet the accuracy and recall rate requirements of the automatic passing of medical claims judgment, thereby realizing the automatic passing of claims cases, improving the overall efficiency of medical material image input, and saving labor costs.

[0055] It can be seen that in the above scheme, for complex medical insurance entities such as medical claims insurance, the medical material images are first obtained and preprocessed; medical diagnostic material information is obtained based on dual-link verification; the medical diagnostic material information of the dual-link verification is extracted and cross-verified; the medical diagnostic material information with high confidence in the verification result is extracted and output. In the present invention, in the medical background of claims insurance, two independent text parsing links are constructed to improve the accuracy when the text values ​​extracted from the two paths are the same, so as to realize the automatic passing of claims cases, improve the overall efficiency of medical material input, save labor costs, and extract and parse the relevant fields of medical material images with the help of end-to-end models and pipeline tools to improve the data cleaning efficiency and the quality of model training data. At the same time, the end-to-end model can be quickly iterated to grasp the changes in input data in time.

[0056] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0057] In one embodiment, a device for accurately extracting medical diagnosis material information is provided, and the device for accurately extracting medical diagnosis material information corresponds one-to-one with the method for accurately extracting medical diagnosis material information in the above embodiment. Figure 5 As shown, the device for accurately extracting medical diagnosis material information includes a receiving module 101, a recall screening module 102, a rough sorting module 103, a judgment module 104, an adjustment module 105, a selection module 106 and a response module 107. The detailed description of each functional module is as follows:

[0058] A preprocessing module 101 is used to obtain medical material images and preprocess the medical material images;

[0059] The inspection module 102 is used to obtain medical diagnosis material information based on dual-link inspection;

[0060] Verification module 103, used to extract medical diagnosis material information of dual link inspection and perform mutual verification;

[0061] The output module 104 is used to extract and output the medical diagnosis material information with high confidence in the verification results.

[0062] In one embodiment, the inspection module 102 is specifically configured to:

[0063] Analyze medical material images based on visual language models to obtain preset key fields;

[0064] Obtain the diagnostic results and preset pre-information in the historical data, and use the diagnostic results and preset pre-information as label data for training and building a deep learning model;

[0065] Extract text information from preset key fields and predict the medical diagnostic material information corresponding to the text information based on the deep learning model;

[0066] Obtain diagnostic results and preset pre-information from historical data, and remove complex diagnostic logic preset in diagnostic results;

[0067] The extracted diagnostic results and the preset prior information are used as label data for training and building a deep learning model.

[0068] In one embodiment, the inspection module 102 is further configured to:

[0069] Detect the position of text on the medical material image based on the text detection model;

[0070] Recognize the image information in the position of the text on the medical material image into text information based on the text recognition model;

[0071] Obtain medical diagnostic material information based on structured text information of information extraction model.

[0072] In one embodiment, the verification module 103 is specifically configured to:

[0073] Extracting medical diagnostic material information for end-to-end model link detection and medical diagnostic material information for pipeline tool detection;

[0074] Compare and verify the medical diagnostic material information detected by the end-to-end model link with the medical diagnostic material information detected by the pipeline tool, and obtain the verification result.

[0075] In one embodiment, the output module 104 is specifically configured to:

[0076] Clean the unequal data in the verification results based on the verification results;

[0077] Confirm high-confidence medical diagnostic material information in the verification results and provide output.

[0078] The present invention provides a device for accurately extracting medical diagnostic material information, which first obtains medical material images and pre-processes the medical material images; obtains medical diagnostic material information based on dual-link verification; extracts the medical diagnostic material information of the dual-link verification and performs mutual verification; extracts and outputs the medical diagnostic material information with high confidence in the verification results. In the present invention, in the medical background, insurance claims are settled, and the accuracy rate when the text values ​​extracted from the two paths are the same is improved by constructing two independent text parsing links, so as to realize the automatic passing of the claims case, improve the overall medical material input efficiency, save labor costs, and extract, parse and verify the relevant fields of the medical material images with the help of end-to-end models and pipeline tools, so as to improve the data cleaning efficiency and the quality of the model training data, and at the same time, the end-to-end model can be quickly iterated to grasp the changes in the input data in time.

[0079] The specific definition of the device for accurately extracting medical diagnostic material information can be found in the definition of the method for accurately extracting medical diagnostic material information above, which will not be repeated here. Each module in the above-mentioned device for accurately extracting medical diagnostic material information can be implemented in whole or in part by software, hardware, and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0080] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the service side of a method for accurately extracting medical diagnostic material information.

[0081] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium 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 the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server through a network connection. When the computer program is executed by the processor, it realizes the functions or steps of the client side of a method for accurately extracting medical diagnostic material information

[0082] 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, wherein the processor implements the following steps when executing the computer program:

[0083] Acquire medical material images and pre-process the medical material images;

[0084] Acquiring medical diagnostic material information based on dual-link testing;

[0085] Extract medical diagnostic material information from dual-link inspection and perform mutual verification;

[0086] Extract and output high-confidence medical diagnostic material information from the verification results.

[0087] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0088] Acquire medical material images and pre-process the medical material images;

[0089] Acquiring medical diagnostic material information based on dual-link testing;

[0090] Extract medical diagnostic material information from dual-link inspection and perform mutual verification;

[0091] Extract and output high-confidence medical diagnostic material information from the verification results.

[0092] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0093] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and 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. As an illustration and not limitation, RAM is available in many 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).

[0094] Those skilled in the art can clearly understand that for the convenience and simplicity 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.

[0095] 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 the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for accurately extracting medical diagnostic material information, characterized in that: include: Acquire medical material images and pre-process the medical material images; Acquiring medical diagnostic material information based on dual-link testing; Extract medical diagnostic material information from dual-link inspection and perform mutual verification; Extract and output high-confidence medical diagnostic material information from the verification results.

2. The method for accurately extracting medical diagnostic material information as claimed in claim 1, characterized in that: The dual link test includes: end-to-end model link detection and pipeline tool detection.

3. The method for accurately extracting medical diagnostic material information as claimed in claim 2, characterized in that: The step of end-to-end model link detection includes: Analyze medical material images based on visual language models to obtain preset key fields; Obtain the diagnostic results and preset pre-information in the historical data, and use the diagnostic results and preset pre-information as label data for training and building a deep learning model; Extract text information from preset key fields and predict the medical diagnostic material information corresponding to the text information based on the deep learning model.

4. The method for accurately extracting medical diagnostic material information as claimed in claim 3, characterized in that: The step of obtaining the diagnosis results and preset pre-information in the historical data, and using the diagnosis results and the preset pre-information as label data for training and constructing a deep learning model includes: Obtain diagnostic results and preset pre-information from historical data, and remove complex diagnostic logic preset in diagnostic results; The extracted diagnostic results and the preset prior information are used as label data for training and building a deep learning model.

5. The method for accurately extracting medical diagnostic material information as claimed in claim 4, characterized in that: The steps of pipeline tool detection include: Detect the position of text on the medical material image based on the text detection model; Recognize the image information in the position of the text on the medical material image into text information based on the text recognition model; Obtain medical diagnostic material information based on structured text information of information extraction model.

6. The method for accurately extracting medical diagnostic material information as claimed in claim 2, characterized in that: The step of extracting the medical diagnosis material information of the dual-link test and performing verification comprises: Extracting medical diagnostic material information for end-to-end model link detection and medical diagnostic material information for pipeline tool detection; Compare and verify the medical diagnostic material information detected by the end-to-end model link with the medical diagnostic material information detected by the pipeline tool, and obtain the verification result.

7. The method for accurately extracting medical diagnostic material information as claimed in claim 2, characterized in that: The step of extracting and outputting high-confidence medical diagnostic material information from the verification results comprises: Clean the unequal data in the verification results based on the verification results; Confirm high-confidence medical diagnostic material information in the verification results and provide output.

8. A device for accurately extracting medical diagnostic material information, characterized in that: include: A preprocessing module, used for acquiring medical material images and preprocessing the medical material images; An inspection module, used to obtain medical diagnostic material information based on dual-link inspection; A verification module, used to extract the medical diagnosis material information of the dual-link inspection and perform mutual verification; The output module is used to extract and output the medical diagnostic material information with high confidence in the verification results.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for accurately extracting medical diagnostic material information as described in any one of claims 1 to 7 are implemented.

10. 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 method for accurately extracting medical diagnostic material information as described in any one of claims 1 to 7 are implemented.

Images