WO2022194062A1 - 疾病标签检测方法、装置、电子设备及存储介质 - Google Patents

疾病标签检测方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2022194062A1
WO2022194062A1 PCT/CN2022/080470 CN2022080470W WO2022194062A1 WO 2022194062 A1 WO2022194062 A1 WO 2022194062A1 CN 2022080470 W CN2022080470 W CN 2022080470W WO 2022194062 A1 WO2022194062 A1 WO 2022194062A1
Authority
WO
WIPO (PCT)
Prior art keywords
disease
label
data
vector
labels
Prior art date
Application number
PCT/CN2022/080470
Other languages
English (en)
French (fr)
Inventor
李响
柳恭
Original Assignee
康键信息技术(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 康键信息技术(深圳)有限公司 filed Critical 康键信息技术(深圳)有限公司
Publication of WO2022194062A1 publication Critical patent/WO2022194062A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
    • G16H10/65ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records stored on portable record carriers, e.g. on smartcards, RFID tags or CD
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02ATECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Definitions

  • the present application relates to the field of artificial intelligence, and in particular, to a disease label detection method, device, electronic device, and computer-readable storage medium.
  • the prediction confidence use the disease regression module in the disease detection model to screen out the disease labels that meet the preset conditions from the candidate disease labels to obtain the predicted disease labels;
  • the present application also provides a disease label detection device, the device comprising:
  • the division module is used to obtain the historical inquiry form, divide the inquiry data in the historical inquiry form into structured data and unstructured data according to the data structure, and use the disease label to classify the structured data and the unstructured data. Labeling to get real disease labels;
  • the conversion module is used to convert the structured data and unstructured data into structured data vectors and unstructured data vectors through vector conversion operations, and obtain a structure composed of the structured data vectors and the unstructured data vectors. training vector;
  • a model training module for performing position coding on the training vector using the coding layer in the pre-built disease detection model to obtain an initial training vector
  • the model training module is further configured to calculate the candidate disease label of the initial training vector by using the disease classification module in the disease detection model, and calculate the prediction confidence of the candidate disease label;
  • the model training module is further configured to, according to the prediction confidence, use the disease regression module in the disease detection model to screen out disease labels that meet preset conditions from the candidate disease labels to obtain a predicted disease label;
  • the model training module is further configured to calculate the loss value of the disease detection model according to the real disease label and the predicted disease label;
  • the model training module is further configured to adjust the parameters of the disease detection model when the loss value does not meet the preset condition, and return the coding layer of the pre-built disease detection model to the training vector. The steps and subsequent steps of performing position encoding to obtain the initial training vector;
  • the model training module is further configured to obtain a trained disease detection model when the loss value satisfies a preset condition
  • the detection module is configured to use the trained disease detection model to perform disease detection on the information of the user to be consulted, obtain an initial disease label, and use a preset filter to screen the initial disease label to obtain a final disease label.
  • the present application also provides an electronic device, the electronic device comprising:
  • the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to implement the following steps:
  • the prediction confidence use the disease regression module in the disease detection model to screen out the disease labels that meet the preset conditions from the candidate disease labels to obtain the predicted disease labels;
  • the present application also provides a computer-readable storage medium, where at least one computer program is stored in the computer-readable storage medium, and the at least one computer program is executed by a processor in an electronic device to implement the following steps:
  • the prediction confidence use the disease regression module in the disease detection model to screen out the disease labels that meet the preset conditions from the candidate disease labels to obtain the predicted disease labels;
  • FIG. 1 is a schematic flowchart of a disease label detection method provided by an embodiment of the present application.
  • FIG. 2 is a detailed schematic flowchart of one of the steps of the disease label detection method provided in FIG. 1 in the first embodiment of the present application;
  • FIG. 3 is a schematic diagram of a module of a disease label detection device provided by an embodiment of the present application.
  • FIG. 4 is a schematic diagram of the internal structure of an electronic device for implementing a disease label detection method provided by an embodiment of the present application
  • the embodiments of the present application provide a disease label detection method.
  • the execution subject of the disease label detection method includes, but is not limited to, at least one of electronic devices that can be configured to execute the method provided by the embodiments of the present application, such as a server and a terminal.
  • the disease label detection method can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform.
  • the server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like.
  • the disease label detection method includes:
  • the historical consultation sheet refers to the user's offline diagnosis sheet, which includes: user basic data, doctor's prescription, diagnosis information, and user chief complaint data, etc.
  • the user basic data includes: name, age, and Gender, etc.
  • the doctor's diagnosis and treatment prescription includes: the type, dosage and time of the drug to be taken
  • the diagnosis information includes: the type of disease, the cause of the disease, etc.
  • the user's main complaint data includes: physical change state, mental change state, etc.
  • the historical medical questionnaire may be obtained by accessing a medical database.
  • the structured data refers to the data structure rules or complete, with a predefined data model, convenient to use the database two-dimensional logic table data, such as the user basic data
  • the unstructured data refers to the data structure Irregular or incomplete numbers without a predefined data model, which are inconvenient to be represented by a two-dimensional logical table in the database, such as the diagnostic information.
  • the division of the medical consultation data in the historical medical consultation form into structured data and unstructured data includes:
  • the feature extraction is used to filter out the useless data in the consultation data and improve data processing efficiency.
  • the feature extraction is implemented based on an actual business scenario, such as extracting the doctor's diagnosis and treatment prescription in the consultation data.
  • the data The two-dimensional table structure is used to filter data with standardized data formats and lengths, such as age, gender, and the like.
  • the present application uses disease labels to mark the structured data and unstructured data to obtain real disease labels, which are used as the comparison of the subsequent model prediction results, and improve the model's performance. robustness.
  • the marking of the disease label may be manually marked. For example, if the unstructured data is diagnostic information, which includes: dizziness, headache, and mental weakness, the corresponding disease label may be: fever. , colds, etc.
  • the real disease label can also be stored in a blockchain node.
  • the structured data and unstructured data will contain a large number of characters, and the neural network can only accept numerical input and cannot support the input of word characters. If the structured data and unstructured data are directly used If the constructed pronoun entity resolution model is trained, the corresponding disease label cannot be identified. Therefore, the embodiment of the present application performs vector transformation on the structured data and unstructured data to determine the structured data and unstructured data. The numerical information of each character in the data is converted to realize subsequent model training.
  • the vector conversion of the structured data can be implemented by the currently known one-hot algorithm
  • the vector conversion of the unstructured data can be implemented by the currently known word2vec algorithm.
  • the one-hot algorithm and the word2vec algorithm are currently relatively mature technologies, and will not be described further.
  • the characters in the structured data are "Ping”, “An”, “Medical”, “Healing”
  • the one-hot algorithm is used to convert the "Ping", “An”, “Medical”, “Medical” "treatment” into the corresponding character vector can be [1,0,0], [0,1,0], [0,0,1], [0,1,0]'.
  • the structured data vector and the unstructured data vector are used as training vectors to be used as input vectors for subsequent model training.
  • the pre-built disease detection model includes a Transformer network, which is used to output disease labels and corresponding confidence levels.
  • a Transformer network which is used to output disease labels and corresponding confidence levels.
  • the following method is used to perform position encoding on the training vector:
  • PE(pos, 2i) represents the position of the even-numbered characters in the initial training vector
  • PE(pos, 2i+1) represents the position of the odd-numbered characters in the initial training vector
  • pos represents the position sequence of the characters in the training vector
  • i represents the first position of the training vector.
  • d model represents the character encoding function.
  • the disease classification module is used to detect the disease type of the initial training vector, so as to output the candidate disease label of the initial training vector, which includes: a feedforward attention mechanism, a fully connected layer, and an activation function , the candidate disease label refers to the disease category of the initial training vector, and the prediction confidence refers to the probability corresponding to the candidate disease label.
  • calculating the candidate disease labels of the initial training vector by using the disease classification module in the disease detection model includes: using a feedforward attention mechanism in the disease classification module to characterize the initial training vector Character extraction to obtain a characteristic character vector, use the fully connected layer in the disease classification module to detect the disease label information in the characteristic character vector, and use the activation function in the disease classification module to output the disease label information to obtain a candidate Disease labels.
  • the feature character extraction of the initial training vector is realized by a convolution kernel in the feedforward attention mechanism.
  • the following method is used to detect the disease label information in the feature character vector:
  • s' represents the disease label information
  • s represents the characteristic character vector
  • e represents an infinite non-repeating decimal.
  • the activation function includes a softmax function.
  • the following method is used to calculate the prediction confidence of the candidate disease label:
  • P j represents the prediction confidence of the jth candidate disease label
  • K represents the number of candidate disease labels
  • k represents the kth candidate disease label
  • x T represents the label regression function
  • W j represents the jth candidate disease label
  • Wk denotes the kth candidate disease label.
  • the disease regression module in the disease detection model to screen out disease labels that meet preset conditions from the candidate disease labels to obtain a predicted disease label.
  • the disease regression module in the disease detection model is used to screen out disease labels that meet preset conditions from the candidate disease labels, so as to improve the accuracy of disease label detection,
  • the disease regression module includes: a sampling layer and an output layer.
  • using a disease regression module in the disease detection model to screen out disease labels that meet preset conditions from the candidate disease labels to obtain a predicted disease label including: using the disease label
  • the sampling layer in the disease regression module up-samples the candidate disease labels to obtain the sampled candidate disease labels, selects the disease labels whose prediction confidence corresponding to the sampled candidate disease labels satisfies the preset condition, and uses the disease regression module
  • the output layer in outputs the selected disease label to obtain the predicted disease label.
  • the up-sampling refers to sampling the candidate disease labels to a specified dimension.
  • the dimension of the original disease label is (416, 416, 3).
  • a candidate disease label whose dimension is (13, 416, 3) is obtained. 13, 16), in order to compare the candidate disease label with the corresponding original disease label, the candidate disease label needs to be changed to the dimension size of (416, 416, 3), and this process is called upsampling.
  • the preset condition may be dynamically set according to the disease label during the actual screening process, for example, if the preset condition is set as the prediction confidence level is less than 0.6, then the prediction confidence level is less than 0.6 corresponding to Candidate disease signatures were screened.
  • the predicted disease label is output by the following method:
  • s(x) represents the predicted disease label
  • x represents the feature character vector of the candidate disease label
  • e represents an infinite non-repeating decimal.
  • the following method is used to calculate the loss value of the disease detection model:
  • L(s) represents the loss value
  • k represents the number of predicted disease labels
  • j represents the jth predicted disease label
  • yi represents the ith predicted disease label
  • y′ i represents the ith real disease label.
  • the preset condition includes that the loss value is less than a loss threshold, that is, when the loss value is less than the loss threshold, it means that the loss value satisfies the preset condition, and when the loss value is less than the loss threshold When the loss value is greater than or equal to the loss threshold, it means that the loss value does not meet the preset condition.
  • the loss threshold may be set to 0.1, or may be set according to actual scenarios.
  • parameter adjustment of the disease detection model may be implemented by a currently known stochastic gradient descent algorithm, which will not be described further herein.
  • S8 is performed to obtain a disease detection model that has been trained.
  • the user information to be consulted includes: basic user information and user chief complaint information
  • the preset filters include: positive filtering rules for disease complications rules and negative filtering rules for violating medical common sense
  • the detected initial disease labels include: fever, chest tightness, headache, and osteoporosis.
  • the initial disease rules for osteoporosis can be filtered through the disease complication rules. Based on the filter, the final generated disease label can be further guaranteed. accuracy.
  • the filter can be generated by compiling the Java language.
  • the embodiment of the present application first divides the data structure of the consultation data in the historical consultation form, generates structured data and unstructured data, marks the disease labels of the structured data and the unstructured data, obtains the real disease label, and analyzes the structured data and unstructured data.
  • Data and unstructured data are vectorized to obtain structured data vectors and unstructured data vectors, and the structured data vectors and unstructured data vectors are used as training vectors, which can improve the medical consultation data in subsequent historical medical consultation sheets.
  • the embodiment of the present application uses real disease labels and training vectors to train a pre-built disease detection model, and uses the trained disease detection model to perform disease detection on the user information to be consulted to obtain an initial disease label, which can ensure The confidence accuracy rate of a single disease label prediction, so that the number of candidate disease labels can be output more accurately, and the difficulty of disease label detection is reduced; further, the embodiment of the present application uses a preset filter to screen the initial disease label to obtain the final disease. The label can further ensure the accuracy of the final generated disease label and reduce the difficulty of disease label detection. Therefore, the present application can reduce the difficulty of disease label detection.
  • FIG. 3 it is a functional block diagram of the disease label detection device of the present application.
  • the disease label detection device 100 described in this application can be installed in an electronic device.
  • the disease label detection apparatus may include a division module 101 , a conversion module 102 , a model training module 103 and a detection module 104 .
  • the modules described in the present invention can also be called units, which refer to a series of computer program segments that can be executed by the electronic device processor and can perform fixed functions, and are stored in the memory of the electronic device.
  • each module/unit is as follows:
  • the division module 101 is used to obtain historical medical questionnaires, divide the medical consultation data in the historical medical questionnaires into structured data and unstructured data according to the data structure, and use disease labels to classify the structured data and unstructured data. Mark structured data to get real disease labels;
  • the conversion module 102 is configured to convert the structured data and unstructured data into structured data vectors and unstructured data vectors through vector conversion operations to obtain the structured data vectors and unstructured data vectors.
  • a training vector consisting of vectors;
  • the model training module 103 is configured to use the disease classification module in the disease detection model to calculate the candidate disease label of the initial training vector, and calculate the prediction confidence of the candidate disease label;
  • the model training module 103 is also used to detect the candidate disease label of the initial training vector using the disease classification module in the disease detection model, and calculate the prediction confidence of the candidate disease label;
  • the model training module 103 is further configured to, according to the prediction confidence, use the disease regression module in the disease detection model to screen out disease labels that meet preset conditions from the candidate disease labels to obtain a predicted disease label;
  • the model training module 103 is further configured to calculate the loss value of the disease detection model according to the real disease label and the predicted disease label;
  • the model training module 103 is further configured to adjust the parameters of the disease detection model when the loss value does not meet the preset conditions, and return the coding layer in the pre-built disease detection model to the training method.
  • the vector is subjected to position encoding to obtain the steps and subsequent steps of the initial training vector;
  • the model training module 103 is further configured to obtain a trained disease detection model when the loss value satisfies a preset condition
  • the detection module 104 is configured to use the trained disease detection model to perform disease detection on the information of the user to be consulted, obtain an initial disease label, and use a preset filter to screen the initial disease label to obtain a final disease label .
  • modules in the disease label detection device 100 in the embodiments of the present application use the same technical means as the disease label detection methods described in the above-mentioned FIG. 1 and FIG. 2 , and can generate the same The technical effect will not be repeated here.
  • FIG. 4 it is a schematic structural diagram of an electronic device implementing the disease label detection method of the present application.
  • the electronic device 1 may include a processor 10, a memory 11 and a bus, and may also include a computer program stored in the memory 11 and executable on the processor 10, such as a disease tag detection program 12.
  • the memory 11 includes at least one type of computer-readable storage medium, and the computer-readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, Disks, CDs, etc.
  • the memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1 .
  • the memory 11 may also be an external storage device of the electronic device 1, such as a pluggable mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital) equipped on the electronic device 1. , SD) card, flash memory card (Flash Card), etc.
  • the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device.
  • the memory 11 can not only be used to store application software installed in the electronic device 1 and various types of data, such as the codes of the disease label detection program 12, etc., but also can be used to temporarily store data that has been output or will be output.
  • the processor 10 may be composed of integrated circuits, for example, may be composed of a single packaged integrated circuit, or may be composed of multiple integrated circuits packaged with the same function or different functions, including one or more integrated circuits.
  • Central Processing Unit CPU
  • microprocessor digital processing chip
  • graphics processor and combination of various control chips, etc.
  • the processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect the various components of the entire electronic device, by running or executing the program or module (for example, executing the program) stored in the memory 11. Disease label detection program 12, etc.), and call data stored in the memory 11 to execute various functions of the electronic device 1 and process data.
  • the bus may be a peripheral component interconnect (PCI for short) bus or an extended industry standard architecture (Extended industry standard architecture, EISA for short) bus or the like.
  • PCI peripheral component interconnect
  • EISA Extended industry standard architecture
  • the bus can be divided into address bus, data bus, control bus and so on.
  • the bus is configured to enable connection communication between the memory 11 and at least one processor 10 and the like.
  • FIG. 4 only shows an electronic device with components. Those skilled in the art can understand that the structure shown in FIG. 4 does not constitute a limitation on the electronic device 1, and may include fewer or more components than those shown in the drawings. components, or a combination of certain components, or a different arrangement of components.
  • the electronic device 1 may also include a power supply (such as a battery) for powering the various components, preferably, the power supply may be logically connected to the at least one processor 10 through a power management device, so that the power management
  • the device implements functions such as charge management, discharge management, and power consumption management.
  • the power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and any other components.
  • the electronic device 1 may further include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.
  • the electronic device 1 may also include a network interface, optionally, the network interface may include a wired interface and/or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used in the electronic device 1 Establish a communication connection with other electronic devices.
  • a network interface optionally, the network interface may include a wired interface and/or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is usually used in the electronic device 1 Establish a communication connection with other electronic devices.
  • the electronic device 1 may further include a user interface, and the user interface may be a display (Display), an input unit (eg, a keyboard (Keyboard)), optionally, the user interface may also be a standard wired interface or a wireless interface.
  • the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, and the like.
  • the display may also be appropriately called a display screen or a display unit, which is used for displaying information processed in the electronic device 1 and for displaying a visualized user interface.
  • the disease label detection program 12 stored in the memory 11 in the electronic device 1 is a combination of multiple programs, and when running in the processor 10, it can realize:
  • Obtain the historical consultation sheet divide the data structure of the consultation data in the historical consultation sheet, generate structured data and unstructured data, and mark the disease labels of the structured data and unstructured data to obtain the real disease Label;
  • Vector conversion is performed on the structured data and the unstructured data to obtain a structured data vector and an unstructured data vector, and the structured data vector and the unstructured data vector are used as training vectors;
  • the disease regression module in the disease detection model to screen out the disease labels that meet the preset conditions from the predicted disease labels to obtain a standard predicted disease label
  • the loss value does not meet the preset condition, adjust the parameters of the disease detection model, and return to the coding layer in the pre-built disease detection model to perform position coding on the training vector, and obtain the initial training vector steps and subsequent steps;
  • the modules/units integrated in the electronic device 1 may be stored in a non-volatile computer-readable storage medium.
  • the computer-readable storage medium may be volatile or non-volatile.
  • the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a removable hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM, Read-Only). Memory).
  • the present application also provides a computer-readable storage medium, where the computer-readable storage medium may be volatile or non-volatile.
  • the computer-readable storage medium stores a computer program, and when executed by the processor of the electronic device, the computer program can realize:
  • Obtain the historical consultation sheet divide the data structure of the consultation data in the historical consultation sheet, generate structured data and unstructured data, and mark the disease labels of the structured data and unstructured data to obtain the real disease Label;
  • Vector conversion is performed on the structured data and the unstructured data to obtain a structured data vector and an unstructured data vector, and the structured data vector and the unstructured data vector are used as training vectors;
  • the disease regression module in the disease detection model to screen out the disease labels that meet the preset conditions from the predicted disease labels to obtain a standard predicted disease label
  • the loss value does not meet the preset condition, adjust the parameters of the disease detection model, and return to the coding layer in the pre-built disease detection model to perform position coding on the training vector, and obtain the initial training vector steps and subsequent steps;
  • modules described as separate components may or may not be physically separated, and components shown as modules may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.
  • each functional module in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units can be implemented in the form of hardware, or can be implemented in the form of hardware plus software function modules.
  • the blockchain referred to in this application is a new application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, and encryption algorithm.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with cryptographic methods. Each data block contains a batch of network transaction information to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • Epidemiology (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Primary Health Care (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Pathology (AREA)
  • Databases & Information Systems (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

一种疾病标签检测方法,涉及人工智能领域,该方法包括:划分历史问诊单中问诊数据的数据结构,生成结构化数据和非结构化数据,标记结构化数据和非结构化数据的疾病标签,得到真实疾病标签;对结构化数据和非结构化数据进行向量转换,得到结构化数据向量和非结构化数据向量,将所述结构化数据向量和非结构化数据向量作为训练向量,利用真实疾病标签和训练向量对预构建的疾病检测模型进行训练,利用训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,利用预设的过滤器对初始疾病标签筛选,得到最终疾病标签。此外,该方法还涉及区块链技术,所述真实疾病标签可存储于区块链中。该方法可以降低疾病标签检测难度。

Description

疾病标签检测方法、装置、电子设备及存储介质
本申请要求于2020年3月16日提交中国专利局、申请号为CN202110293429.4、名称为“疾病标签检测方法、装置、电子设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能领域,尤其涉及一种疾病标签检测方法、装置、电子设备及计算机可读存储介质。
背景技术
随着人工智能技术的不断发展,线上医疗诊断逐渐成为人们研究的热点,相比于传统的线下问诊,线上问诊作为新兴业态逐渐渗透为常规医疗问诊的有益补充,通过线上问诊可以大大减少线下问诊的人力和时间成本。然而,发明人意识到,医疗传统的诊断分类模型,无论是专家系统或者先进的神经网络模型大多只支持一种最可能的疾病诊断,当用户病症状态可能同时存在一种或多种疾病时,精准定位不同层级的并发症存在难度,从而会带来一定的疾病标签检测难度。同时,医疗传统的诊断分类模型下诊断基于全部信息机型下诊断,并不能很好配合多轮问询,及病情关键信息层层递进的业务场景,从而也会带来一定的疾病标签检测难度。
发明内容
本申请提供的一种疾病标签检测方法,包括:
获取历史问诊单,根据数据结构将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
在所述损失值不满足预设条件时,调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤;
在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
本申请还提供一种疾病标签检测装置,所述装置包括:
划分模块,用于获取历史问诊单,根据数据结构将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
转换模块,用于通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
模型训练模块,用于利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
所述模型训练模块,还用于利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
所述模型训练模块,还用于根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
所述模型训练模块,还用于根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
所述模型训练模块,还用于在所述损失值不满足预设条件时,则调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤及后续步骤;
所述模型训练模块,还用于在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
检测模块,用于利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
本申请还提供一种电子设备,所述电子设备包括:
至少一个处理器;以及,
与所述至少一个处理器通信连接的存储器;其中,
所述存储器存储有可被所述至少一个处理器执行的计算机程序,所述计算机程序被所述至少一个处理器执行,以实现如下步骤:
获取历史问诊单,根据数据结构将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
在所述损失值不满足预设条件时,调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤;
在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
本申请还提供一种计算机可读存储介质,所述计算机可读存储介质中存储有至少一个计算机程序,所述至少一个计算机程序被电子设备中的处理器执行以实现如下步骤:
获取历史问诊单,根据数据结构将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练 向量;
利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
在所述损失值不满足预设条件时,调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤;
在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
附图说明
图1为本申请一实施例提供的疾病标签检测方法的流程示意图;
图2为本申请第一实施例中图1提供的疾病标签检测方法其中一个步骤的详细流程示意图;
图3为本申请一实施例提供的疾病标签检测装置的模块示意图;
图4为本申请一实施例提供的实现疾病标签检测方法的电子设备的内部结构示意图;
本申请目的的实现、功能特点及优点将结合实施例,参照附图做进一步说明。
具体实施方式
应当理解,此处所描述的具体实施例仅仅用以解释本申请,并不用于限定本申请。
本申请实施例提供一种疾病标签检测方法。所述疾病标签检测方法的执行主体包括但不限于服务端、终端等能够被配置为执行本申请实施例提供的该方法的电子设备中的至少一种。换言之,所述疾病标签检测方法可以由安装在终端设备或服务端设备的软件或硬件来执行,所述软件可以是区块链平台。所述服务端包括但不限于:单台服务器、服务器集群、云端服务器或云端服务器集群等。
参照图1所示,为本申请一实施例提供的疾病标签检测方法的流程示意图。在本申请实施例中,所述疾病标签检测方法包括:
S1、获取历史问诊单,将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签。
本申请实施例中,历史问诊单是指用户线下诊断单,其包括:用户基本数据、医生诊疗处方、诊断信息以及用户主诉数据等,其中,所述用户基本数据包括;姓名、年龄以及性别等,所述医生诊疗处方包括:服用药品的种类、剂量以及时间等,所述诊断信息包括:疾病种类、疾病发生原因等,所述用户主诉数据包括:身体变化状态、精神变化状态等。
一个可选实施例中,所述历史问诊单可以通过访问医疗数据库获取。
进一步地,应该了解,在获取的历史问诊单中会存在结构化数据和非结构化数据,为了更好的对所述历史问诊单进行数据分析,本申请实施例通过划分所述历史问诊单中问诊数据的数据结构,以提高后续数据处理的速度。其中,所述结构化数据是指数据结构规则或完整,有预定义的数据模型,方便用数据库二维逻辑表来的数据,如所述用户基本数据,所述非结构化数据是指数据结构不规则或不完整,没有预定义的数据模型,不方便用数据库二维逻辑表来表现的数,如所述诊断信息。
详细地,参阅图2所示,所述将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,包括:
S20、对所述问诊数据进行特征提取,生成特征问诊数据;
S21、将所述特征问诊数据中符合二维表结构的数据作为结构化数据;
S22、将所述特征问诊数据中不符合二维表结构的数据作为非结构化数据。
其中,所述特征提取用于筛选出所述问诊数据中无用数据,提高数据处理效率,可选的,所述特征提取基于实际业务场景实现,如提取所述问诊数据中的医生诊疗处方的数据。所述二维表结构用于筛选数据格式和长度规范的数据,如年龄、性别等。
进一步地,为了更好的监督后续模型的学习能力,本申请利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签,以作为后续模型预测结果的对照,提高模型的鲁棒性。一个可选实施例中,所述疾病标签的标记可以通过人工进行标记,比如,所述非结构化数据为诊断信息,其包括:眩晕、头痛、精神无力,则对应的疾病标签可以为:发烧、感冒等。
进一步地,为保障所述真实疾病标签的复用性和隐私性,所述真实疾病标签还可存储于一区块链节点中。
S2、通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量。
应该了解,所述结构化数据和非结构化数据中会包含大量的字符,而在神经网络只能接受数值输入,无法支持单词字符的输入,若直接利用所述结构化数据和非结构化数据对构建的代词实体消解模型进行训练,则无法识别出对应的疾病标签,因此,本申请实施例对所述结构化数据和非结构化数据进行向量转换,以确定所述结构化数据和非结构化数据中每个字符的数值信息,从而实现后续的模型训练。
一个可选实施例中,所述结构化数据的向量转换可以通过当前已知的one-hot算法实现,所述非结构化数据的向量转换可以通过当前已知的word2vec算法实现,需要说明的是,所述one-hot算法和word2vec算法属于当前较为成熟的技术,再次不做进一步赘述。比如,所述结构化数据中字符为“平”,“安”,“医”,“疗”,则利用所述one-hot算法将所述“平”,“安”,“医”,“疗”转换成对应的字符向量可以为[1,0,0],[0,1,0],[0,0,1],[0,1,0]’。
进一步地,本申请实施例将所述结构化数据向量和非结构化数据向量作为训练向量,以作为后续模型训练的输入向量。
S3、利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量。
本申请实施例中,所述预构建的疾病检测模型包括Transformer网络,其用于输出疾病标签及对应的置信度,为了更好的了解所述训练向量中字符在模型中对应的位置信息,本申请实施例利用所述疾病检测模型中编码层对所述训练向量进行位置编码。
一个可选实施例中,利用下述方法对所述训练向量进行位置编码:
Figure PCTCN2022080470-appb-000001
Figure PCTCN2022080470-appb-000002
其中,PE(pos,2i)表示初始训练向量中偶数字符位置,PE(pos,2i+1)表示初始训练向量中奇数字符位置,pos表示训练向量中字符的位置序列,i表示训练向量的第i个维度,d model表示字符编码函数。
S4、利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度。
本申请实施例中,所述疾病分类模块用于检测所述初始训练向量的疾病种类,以输出所述初始训练向量的候选疾病标签,其包括:前馈注意力机制、全连接层以及激活函数,所述候选疾病标签是指所述初始训练向量的疾病种类,所述预测置信度指的是对应候选疾病标签的概率。
详细地,所述利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,包括:利用所述疾病分类模块中的前馈注意力机制对所述初始训练向量进行特征字符提取,得到特征字符向量,利用所述疾病分类模块中的全连接层检测所述特征字符向量中的疾病标签信息,利用所述疾病分类模块中的激活函数输出所述疾病标签信息,得到候选疾病标签。
一个可选实施例中,所述初始训练向量的特征字符提取通过所述前馈注意力机制中的卷积核实现。
一个可选实施例中,利用下述方法检测所述检测所述特征字符向量中的疾病标签信息:
Figure PCTCN2022080470-appb-000003
其中,s′表示疾病标签信息,s表示特征字符向量,e表示无限不循环小数。
一个可选实施例中,所述激活函数包括softmax函数。
进一步地,在本申请的一个可选实施例中,利用下述方法计算所述候选疾病标签的预测置信度:
Figure PCTCN2022080470-appb-000004
其中,P j表示第j个候选疾病标签的预测置信度,K表示候选疾病标签的数量,k表示第k个候选疾病标签,x T表示标签回归函数,W j表示第j个候选疾病标签,W k表示第k个候选疾病标签。
S5、根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签。
本申请实施例中,根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,以提高疾病标签检测的准确率,其中,所述疾病回归模块包括:采样层和输出层。
详细地,所述根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签,包括:利用所述疾病回归模块中的采样层对所述候选疾病标签进行上采样,得到采样候选疾病标签,选取所述采样候选疾病标签对应预测置信度满足所述预设条件的疾病标签,利用所述疾病回归模块中的输出层输出选取的所述疾病标签,得到预测疾病标签。
其中,所述上采样是指将候选疾病标签采样到指定维度大小,比如原始疾病标签的维度为(416,416,3)经过一系列卷积池化操作后,得到一个候选疾病标签维度为(13,13,16),为了把这个候选疾病标签和对应原始疾病标签进行比较,需要将这个候选疾病标签变成(416,416,3)的维度大小,该过程就称为上采样。
一个可选实施例中,所述预设条件可以根据疾病标签在实际筛选过程中进行动态设置,比如设置所述预设条件为预测置信度小于0.6,则将所述预测置信度小于0.6对应的候选疾病标签筛选出来。
一个可选实施例中,利用下述方法输出所述预测疾病标签:
Figure PCTCN2022080470-appb-000005
其中,s(x)表示预测疾病标签,x表示候选疾病标签的特征字符向量,e表示无限不循环小数。
S6、根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值。
本申请实施例中,利用下述方法计算所述疾病检测模型的损失值:
Figure PCTCN2022080470-appb-000006
其中,L(s)表示损失值,k表示预测疾病标签的数量,j表示第j个预测疾病标签,y i表示第i个预测疾病标签,y′ i表示第i个真实疾病标签。
若所述损失值不满足预设条件时,则执行S7、调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量的步骤及后续步骤。
在本申请实施例中,所述预设条件包括所述损失值小于损失阈值,即当所述损失值小于所述损失阈值时,则表示所述损失值满足所述预设条件时,当所述损失值大于或者等于所述损失阈值时,则表示所述损失值不满足所述预设条件时。其中,所述损失阈值可以设置为0.1,也可以根据实际场景设置。
进一步地,所述疾病检测模型的参数调整可以通过当前已知的随机梯度下降算法实现,在此不做进一步赘述。
若所述损失值满足预设条件时,则执行S8、得到训练完成的疾病检测模型。
S9、利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
本申请实施例中,所述待问诊用户信息包括:用户基本信息以及用户主诉信息,所述预设的过滤器包括:疾病并发症规则的正向过滤规则和违背医学常识的负向过滤规则,比如检测的初始疾病标签包括:发烧、胸闷、头痛、骨质疏松,通过疾病并发症规则可以过滤所述骨质疏松的初始疾病规则,基于所述过滤器可以进一步保障最终生成的疾病标签的准确性。
一个可选实施例中,所述过滤器可以通过Java语言编译生成。
本申请实施例首先划分历史问诊单中问诊数据的数据结构,生成结构化数据和非结构化数据,标记结构化数据和非结构化数据的疾病标签,得到真实疾病标签,并对结构化数据和非结构化数据进行向量转换,得到结构化数据向量和非结构化数据向量,将所述结构化数据向量和非结构化数据向量作为训练向量,可以提高后续历史问诊单中问诊数据处理的速度;其次,本申请实施例利用真实疾病标签和训练向量对预构建的疾病检测模型进行训练,利用训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,可以保证单个疾病标签预测的置信度准确率,从而更准确输出多少个候选疾病标签数量,降低疾病标签检测的难度;进一步地,本申请实施例利用预设的过滤器对初始疾病标签筛选,得到最终疾病标签,可以进一步保障最终生成的疾病标签的准确性,实现疾病标签检测难度的降低。因此,本申请可以降低疾病标签检测难度。
如图3所示,是本申请疾病标签检测装置的功能模块图。
本申请所述疾病标签检测装置100可以安装于电子设备中。根据实现的功能,所述疾病标签检测装置可以包括划分模块101、转换模块102、模型训练模块103以及检测模块104。本发所述模块也可以称之为单元,是指一种能够被电子设备处理器所执行,并且能够完成固定功能的一系列计算机程序段,其存储在电子设备的存储器中。
在本实施例中,关于各模块/单元的功能如下:
所述划分模块101,用于获取历史问诊单,根据数据结构将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
所述转换模块102,用于通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
所述模型训练模块103,用于利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
所述模型训练模块103,还用于利用所述疾病检测模型中的疾病分类模块检测所述初 始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
所述模型训练模块103,还用于根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
所述模型训练模块103,还用于根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
所述模型训练模块103,还用于在所述损失值不满足预设条件时,则调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤及后续步骤;
所述模型训练模块103,还用于在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
所述检测模块104,用于利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
详细地,本申请实施例中所述疾病标签检测装置100中的所述各模块在使用时采用与上述的图1和图2中所述的疾病标签检测方法一样的技术手段,并能够产生相同的技术效果,这里不再赘述。
如图4所示,是本申请实现疾病标签检测方法的电子设备的结构示意图。
所述电子设备1可以包括处理器10、存储器11和总线,还可以包括存储在所述存储器11中并可在所述处理器10上运行的计算机程序,如疾病标签检测程序12。
其中,所述存储器11至少包括一种类型的计算机可读存储介质,所述计算机可读存储介质包括闪存、移动硬盘、多媒体卡、卡型存储器(例如:SD或DX存储器等)、磁性存储器、磁盘、光盘等。所述存储器11在一些实施例中可以是电子设备1的内部存储单元,例如该电子设备1的移动硬盘。所述存储器11在另一些实施例中也可以是电子设备1的外部存储设备,例如电子设备1上配备的插接式移动硬盘、智能存储卡(Smart Media Card,SMC)、安全数字(Secure Digital,SD)卡、闪存卡(Flash Card)等。进一步地,所述存储器11还可以既包括电子设备1的内部存储单元也包括外部存储设备。所述存储器11不仅可以用于存储安装于电子设备1的应用软件及各类数据,例如疾病标签检测程序12的代码等,还可以用于暂时地存储已经输出或者将要输出的数据。
所述处理器10在一些实施例中可以由集成电路组成,例如可以由单个封装的集成电路所组成,也可以是由多个相同功能或不同功能封装的集成电路所组成,包括一个或者多个中央处理器(Central Processing unit,CPU)、微处理器、数字处理芯片、图形处理器及各种控制芯片的组合等。所述处理器10是所述电子设备的控制核心(Control Unit),利用各种接口和线路连接整个电子设备的各个部件,通过运行或执行存储在所述存储器11内的程序或者模块(例如执行疾病标签检测程序12等),以及调用存储在所述存储器11内的数据,以执行电子设备1的各种功能和处理数据。
所述总线可以是外设部件互连标准(peripheral component interconnect,简称PCI)总线或扩展工业标准结构(extended industry standard architecture,简称EISA)总线等。该总线可以分为地址总线、数据总线、控制总线等。所述总线被设置为实现所述存储器11以及至少一个处理器10等之间的连接通信。
图4仅示出了具有部件的电子设备,本领域技术人员可以理解的是,图4示出的结构并不构成对所述电子设备1的限定,可以包括比图示更少或者更多的部件,或者组合某些部件,或者不同的部件布置。
例如,尽管未示出,所述电子设备1还可以包括给各个部件供电的电源(比如电池),优选地,电源可以通过电源管理装置与所述至少一个处理器10逻辑相连,从而通过电源管理装置实现充电管理、放电管理、以及功耗管理等功能。电源还可以包括一个或一个以 上的直流或交流电源、再充电装置、电源故障检测电路、电源转换器或者逆变器、电源状态指示器等任意组件。所述电子设备1还可以包括多种传感器、蓝牙模块、Wi-Fi模块等,在此不再赘述。
进一步地,所述电子设备1还可以包括网络接口,可选地,所述网络接口可以包括有线接口和/或无线接口(如WI-FI接口、蓝牙接口等),通常用于在该电子设备1与其他电子设备之间建立通信连接。
可选地,该电子设备1还可以包括用户接口,用户接口可以是显示器(Display)、输入单元(比如键盘(Keyboard)),可选地,用户接口还可以是标准的有线接口、无线接口。可选地,在一些实施例中,显示器可以是LED显示器、液晶显示器、触控式液晶显示器以及OLED(Organic Light-Emitting Diode,有机发光二极管)触摸器等。其中,显示器也可以适当的称为显示屏或显示单元,用于显示在电子设备1中处理的信息以及用于显示可视化的用户界面。
应该了解,所述实施例仅为说明之用,在专利申请范围上并不受此结构的限制。
所述电子设备1中的所述存储器11存储的疾病标签检测程序12是多个程序的组合,在所述处理器10中运行时,可以实现:
获取历史问诊单,划分所述历史问诊单中问诊数据的数据结构,生成结构化数据和非结构化数据,并标记所述结构化数据和非结构化数据的疾病标签,得到真实疾病标签;
对所述结构化数据和非结构化数据进行向量转换,得到结构化数据向量和非结构化数据向量,并将所述结构化数据向量和非结构化数据向量作为训练向量;
利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
利用所述疾病检测模型中的疾病分类模块检测所述初始训练向量的预测疾病标签,并计算所述预测疾病标签的预测置信度;
根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述预测疾病标签中筛选出满足预设条件的疾病标签,得到标准预测疾病标签;
根据所述真实疾病标签和所述标准预测疾病标签,计算所述疾病检测模型的损失值;
若所述损失值不满足预设条件时,则调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤及后续步骤;
若所述损失值满足预设条件时,得到训练完成的疾病检测模型;
利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
具体地,所述处理器10对上述程序的具体实现方法可参考图1对应实施例中相关步骤的描述,在此不赘述。
进一步地,所述电子设备1集成的模块/单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时,可以存储在一个非易失性计算机可读取存储介质中。所述计算机可读存储介质可以是易失性的,也可以是非易失性的。例如,所述计算机可读介质可以包括:能够携带所述计算机程序代码的任何实体或装置、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM,Read-Only Memory)。
本申请还提供一种计算机可读存储介质,所述计算机可读存储介质可以是易失性的,也可以是非易失性的。所述计算机可读存储介质存储有计算机程序,所述计算机程序在被电子设备的处理器所执行时,可以实现:
获取历史问诊单,划分所述历史问诊单中问诊数据的数据结构,生成结构化数据和非结构化数据,并标记所述结构化数据和非结构化数据的疾病标签,得到真实疾病标签;
对所述结构化数据和非结构化数据进行向量转换,得到结构化数据向量和非结构化数 据向量,并将所述结构化数据向量和非结构化数据向量作为训练向量;
利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
利用所述疾病检测模型中的疾病分类模块检测所述初始训练向量的预测疾病标签,并计算所述预测疾病标签的预测置信度;
根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述预测疾病标签中筛选出满足预设条件的疾病标签,得到标准预测疾病标签;
根据所述真实疾病标签和所述标准预测疾病标签,计算所述疾病检测模型的损失值;
若所述损失值不满足预设条件时,则调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤及后续步骤;
若所述损失值满足预设条件时,得到训练完成的疾病检测模型;
利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
在本申请所提供的几个实施例中,应该理解到,所揭露的设备,装置和方法,可以通过其它的方式实现。例如,以上所描述的装置实施例仅仅是示意性的,例如,所述模块的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式。
所述作为分离部件说明的模块可以是或者也可以不是物理上分开的,作为模块显示的部件可以是或者也可以不是物理单元,即可以位于一个地方,或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
另外,在本申请各个实施例中的各功能模块可以集成在一个处理单元中,也可以是各个单元单独物理存在,也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现,也可以采用硬件加软件功能模块的形式实现。
对于本领域技术人员而言,显然本申请不限于上述示范性实施例的细节,而且在不背离本申请的精神或基本特征的情况下,能够以其他的具体形式实现本申请。
因此,无论从哪一点来看,均应将实施例看作是示范性的,而且是非限制性的,本申请的范围由所附权利要求而不是上述说明限定,因此旨在将落在权利要求的等同要件的含义和范围内的所有变化涵括在本申请内。不应将权利要求中的任何附关联图标记视为限制所涉及的权利要求。
本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
此外,显然“包括”一词不排除其他单元或步骤,单数不排除复数。系统权利要求中陈述的多个单元或装置也可以由一个单元或装置通过软件或者硬件来实现。第二等词语用来表示名称,而并不表示任何特定的顺序。
最后应说明的是,以上实施例仅用以说明本申请的技术方案而非限制,尽管参照较佳实施例对本申请进行了详细说明,本领域的普通技术人员应当理解,可以对本申请的技术方案进行修改或等同替换,而不脱离本申请技术方案的精神和范围。

Claims (20)

  1. 一种疾病标签检测方法,其中,所述方法包括:
    获取历史问诊单,将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
    通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
    利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
    利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
    根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
    根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
    在所述损失值不满足预设条件时,调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤;
    在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
    利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
  2. 如权利要求1所述的疾病标签检测方法,其中,所述将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,包括:
    对所述问诊数据进行特征提取,生成特征问诊数据;
    将所述特征问诊数据中符合二维表结构的数据作为结构化数据;
    将所述特征问诊数据中不符合二维表结构的数据作为非结构化数据。
  3. 如权利要求1所述的疾病标签检测方法,其中,所述利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量,包括:
    利用下述方法对所述训练向量进行位置编码:
    Figure PCTCN2022080470-appb-100001
    Figure PCTCN2022080470-appb-100002
    其中,PE(pos,2i)表示初始训练向量中偶数字符位置,PE(pos,2i+1)表示初始训练向量中奇数字符位置,pos表示训练向量中字符的位置序列,i表示训练向量的第i个维度,d model表示字符编码函数。
  4. 如权利要求1所述的疾病标签检测方法,其中,所述利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,包括:
    利用所述疾病分类模块中的前馈注意力机制对所述初始训练向量进行特征字符提取,得到特征字符向量;
    利用所述疾病分类模块中的全连接层检测所述特征字符向量中的疾病标签信息;
    利用所述疾病分类模块中的激活函数输出所述疾病标签信息,得到候选疾病标签。
  5. 如权利要求1中所述的疾病标签检测方法,其中,所述计算所述候选疾病标签的预测置信度,包括:
    利用下述公式计算所述候选疾病标签的预测置信度:
    Figure PCTCN2022080470-appb-100003
    其中,P j表示第j个候选疾病标签的预测置信度,K表示候选疾病标签的数量,k表示第k个候选疾病标签,x T表示标签回归函数,W j表示第j个候选疾病标签,W k表示第k个候选疾病标签。
  6. 如权利要求1所述的疾病标签检测方法,其中,所述根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签,包括:
    利用所述疾病回归模块中的采样层对所述候选疾病标签进行上采样,得到采样候选疾病标签;
    选取所述采样候选疾病标签对应预测置信度满足所述预设条件的疾病标签;
    利用所述疾病回归模块中的输出层输出选取的所述疾病标签,得到预测疾病标签。
  7. 如权利要求1至6中任一项所述的疾病标签检测方法,其中,所述根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值,包括:
    利用下述公式计算所述疾病检测模型的损失值:
    Figure PCTCN2022080470-appb-100004
    其中,L(s)表示损失值,k表示预测疾病标签的数量,j表示第j个预测疾病标签,y i表示第i个预测疾病标签,y′ i表示第i个真实疾病标签。
  8. 一种疾病标签检测装置,其中,所述装置包括:
    划分模块,用于获取历史问诊单,根据数据结构将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
    转换模块,用于通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
    模型训练模块,用于利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
    所述模型训练模块,还用于利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
    所述模型训练模块,还用于根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
    所述模型训练模块,还用于根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
    所述模型训练模块,还用于在所述损失值不满足预设条件时,则调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤及后续步骤;
    所述模型训练模块,还用于在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
    检测模块,用于利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
  9. 一种电子设备,其中,所述电子设备包括:
    至少一个处理器;以及,
    与所述至少一个处理器通信连接的存储器;其中,
    所述存储器存储有可被所述至少一个处理器执行的计算机程序,所述计算机程序被所述至少一个处理器执行,以使所述至少一个处理器能够执行如下步骤:
    获取历史问诊单,将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
    通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
    利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
    利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
    根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
    根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
    在所述损失值不满足预设条件时,调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤;
    在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
    利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
  10. 如权利要求9所述的电子设备,其中,所述将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,包括:
    对所述问诊数据进行特征提取,生成特征问诊数据;
    将所述特征问诊数据中符合二维表结构的数据作为结构化数据;
    将所述特征问诊数据中不符合二维表结构的数据作为非结构化数据。
  11. 如权利要求9所述的电子设备,其中,所述利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量,包括:
    利用下述方法对所述训练向量进行位置编码:
    Figure PCTCN2022080470-appb-100005
    Figure PCTCN2022080470-appb-100006
    其中,PE(pos,2i)表示初始训练向量中偶数字符位置,PE(pos,2i+1)表示初始训练向量中奇数字符位置,pos表示训练向量中字符的位置序列,i表示训练向量的第i个维度,d model表示字符编码函数。
  12. 如权利要求9所述的电子设备,其中,所述利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,包括:
    利用所述疾病分类模块中的前馈注意力机制对所述初始训练向量进行特征字符提取,得到特征字符向量;
    利用所述疾病分类模块中的全连接层检测所述特征字符向量中的疾病标签信息;
    利用所述疾病分类模块中的激活函数输出所述疾病标签信息,得到候选疾病标签。
  13. 如权利要求9中所述的电子设备,其中,所述计算所述候选疾病标签的预测置信度,包括:
    利用下述公式计算所述候选疾病标签的预测置信度:
    Figure PCTCN2022080470-appb-100007
    其中,P j表示第j个候选疾病标签的预测置信度,K表示候选疾病标签的数量,k表示第k个候选疾病标签,x T表示标签回归函数,W j表示第j个候选疾病标签,W k表示第k个候选疾病标签。
  14. 如权利要求9所述的电子设备,其中,所述根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签,包括:
    利用所述疾病回归模块中的采样层对所述候选疾病标签进行上采样,得到采样候选疾病标签;
    选取所述采样候选疾病标签对应预测置信度满足所述预设条件的疾病标签;
    利用所述疾病回归模块中的输出层输出选取的所述疾病标签,得到预测疾病标签。
  15. 如权利要求9至14中任一项所述的电子设备,其中,所述根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值,包括:
    利用下述公式计算所述疾病检测模型的损失值:
    Figure PCTCN2022080470-appb-100008
    其中,L(s)表示损失值,k表示预测疾病标签的数量,j表示第j个预测疾病标签,y i表示第i个预测疾病标签,y′ i表示第i个真实疾病标签。
  16. 一种计算机可读存储介质,存储有计算机程序,其中,所述计算机程序被处理器执行时实现如下步骤:
    获取历史问诊单,将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,利用疾病标签对所述结构化数据和非结构化数据进行标记,得到真实疾病标签;
    通过向量转换操作,将所述结构化数据和非结构化数据转换为结构化数据向量和非结构化数据向量,得到由所述结构化数据向量和非结构化数据向量组成的训练向量;
    利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量;
    利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,并计算所述候选疾病标签的预测置信度;
    根据所述预测置信度,利用所述疾病检测模型中的疾病回归模块从所述候选疾病标签中筛选出满足预设条件的疾病标签,得到预测疾病标签;
    根据所述真实疾病标签和所述预测疾病标签,计算所述疾病检测模型的损失值;
    在所述损失值不满足预设条件时,调整所述疾病检测模型的参数,并返回所述利用预构建的疾病检测模型中编码层对所述训练向量进行位置编码,得到初始训练向量的步骤;
    在所述损失值满足预设条件时,得到训练完成的疾病检测模型;
    利用所述训练完成的疾病检测模型对待问诊用户信息进行疾病检测,得到初始疾病标签,并利用预设的过滤器对所述初始疾病标签筛选,得到最终疾病标签。
  17. 如权利要求16所述的计算机可读存储介质,其中,所述将所述历史问诊单中问诊数据划分为结构化数据和非结构化数据,包括:
    对所述问诊数据进行特征提取,生成特征问诊数据;
    将所述特征问诊数据中符合二维表结构的数据作为结构化数据;
    将所述特征问诊数据中不符合二维表结构的数据作为非结构化数据。
  18. 如权利要求16所述的计算机可读存储介质,其中,所述利用预构建的疾病检测模型中的编码层对所述训练向量进行位置编码,得到初始训练向量,包括:
    利用下述方法对所述训练向量进行位置编码:
    Figure PCTCN2022080470-appb-100009
    Figure PCTCN2022080470-appb-100010
    其中,PE(pos,2i)表示初始训练向量中偶数字符位置,PE(pos,2i+1)表示初始训练向量中奇数字符位置,pos表示训练向量中字符的位置序列,i表示训练向量的第i个维度,d model表示字符编码函数。
  19. 如权利要求16所述的计算机可读存储介质,其中,所述利用所述疾病检测模型中的疾病分类模块计算所述初始训练向量的候选疾病标签,包括:
    利用所述疾病分类模块中的前馈注意力机制对所述初始训练向量进行特征字符提取,得到特征字符向量;
    利用所述疾病分类模块中的全连接层检测所述特征字符向量中的疾病标签信息;
    利用所述疾病分类模块中的激活函数输出所述疾病标签信息,得到候选疾病标签。
  20. 如权利要求16中所述的计算机可读存储介质,其中,所述计算所述候选疾病标签的预测置信度,包括:
    利用下述公式计算所述候选疾病标签的预测置信度:
    Figure PCTCN2022080470-appb-100011
    其中,P j表示第j个候选疾病标签的预测置信度,K表示候选疾病标签的数量,k表示第k个候选疾病标签,x T表示标签回归函数,W j表示第j个候选疾病标签,W k表示第k个候选疾病标签。
PCT/CN2022/080470 2021-03-16 2022-03-11 疾病标签检测方法、装置、电子设备及存储介质 WO2022194062A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110293429.4 2021-03-16
CN202110293429.4A CN112885423A (zh) 2021-03-16 2021-03-16 疾病标签检测方法、装置、电子设备及存储介质

Publications (1)

Publication Number Publication Date
WO2022194062A1 true WO2022194062A1 (zh) 2022-09-22

Family

ID=76041272

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/080470 WO2022194062A1 (zh) 2021-03-16 2022-03-11 疾病标签检测方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN112885423A (zh)
WO (1) WO2022194062A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112885423A (zh) * 2021-03-16 2021-06-01 康键信息技术(深圳)有限公司 疾病标签检测方法、装置、电子设备及存储介质
CN113488159A (zh) * 2021-08-11 2021-10-08 中国医学科学院阜外医院 一种基于神经网络的医疗科室推荐方法及装置
CN116072298B (zh) * 2023-04-06 2023-08-15 之江实验室 一种基于层级标记分布学习的疾病预测系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180046773A1 (en) * 2016-08-11 2018-02-15 Htc Corporation Medical system and method for providing medical prediction
CN110827929A (zh) * 2019-11-05 2020-02-21 中山大学 疾病分类编码识别方法、装置、计算机设备及存储介质
CN110910976A (zh) * 2019-10-12 2020-03-24 平安国际智慧城市科技股份有限公司 病历检测方法、装置、设备和存储介质
CN112397199A (zh) * 2020-11-30 2021-02-23 广东德澳智慧医疗科技有限公司 一种基于5g和区块链的大数据疾病预测系统
CN112435755A (zh) * 2020-11-23 2021-03-02 平安科技(深圳)有限公司 疾病分析方法、装置、电子设备及存储介质
CN112885423A (zh) * 2021-03-16 2021-06-01 康键信息技术(深圳)有限公司 疾病标签检测方法、装置、电子设备及存储介质

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11984206B2 (en) * 2018-02-16 2024-05-14 Google Llc Automated extraction of structured labels from medical text using deep convolutional networks and use thereof to train a computer vision model
CN109447183B (zh) * 2018-11-27 2020-10-16 东软集团股份有限公司 预测模型训练方法、装置、设备以及介质
CN109754852A (zh) * 2019-01-08 2019-05-14 中南大学 基于电子病历的心血管疾病风险预测方法
CN110867231A (zh) * 2019-11-18 2020-03-06 中山大学 基于文本分类的疾病预测方法、装置、计算机设备及介质
CN111127165B (zh) * 2019-12-26 2024-03-26 纪信智达(广州)信息技术有限公司 基于自注意力自编码器的序列推荐方法
CN111368210B (zh) * 2020-05-27 2020-11-27 腾讯科技(深圳)有限公司 基于人工智能的信息推荐方法、装置以及电子设备

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180046773A1 (en) * 2016-08-11 2018-02-15 Htc Corporation Medical system and method for providing medical prediction
CN110910976A (zh) * 2019-10-12 2020-03-24 平安国际智慧城市科技股份有限公司 病历检测方法、装置、设备和存储介质
CN110827929A (zh) * 2019-11-05 2020-02-21 中山大学 疾病分类编码识别方法、装置、计算机设备及存储介质
CN112435755A (zh) * 2020-11-23 2021-03-02 平安科技(深圳)有限公司 疾病分析方法、装置、电子设备及存储介质
CN112397199A (zh) * 2020-11-30 2021-02-23 广东德澳智慧医疗科技有限公司 一种基于5g和区块链的大数据疾病预测系统
CN112885423A (zh) * 2021-03-16 2021-06-01 康键信息技术(深圳)有限公司 疾病标签检测方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN112885423A (zh) 2021-06-01

Similar Documents

Publication Publication Date Title
WO2022194062A1 (zh) 疾病标签检测方法、装置、电子设备及存储介质
WO2021151345A1 (zh) 识别模型的参数获取方法、装置、电子设备及存储介质
WO2021208696A1 (zh) 用户意图分析方法、装置、电子设备及计算机存储介质
WO2022222943A1 (zh) 科室推荐方法、装置、电子设备及存储介质
WO2022121157A1 (zh) 语音合成方法、装置、电子设备及存储介质
CN113704429A (zh) 基于半监督学习的意图识别方法、装置、设备及介质
WO2021189903A1 (zh) 基于音频的用户状态识别方法、装置、电子设备及存储介质
CN112559687A (zh) 问题识别及查询方法、装置、电子设备及存储介质
WO2023178978A1 (zh) 基于人工智能的处方审核方法、装置、设备及介质
CN113064994A (zh) 会议质量评估方法、装置、设备及存储介质
WO2023137906A1 (zh) 文档标题生成方法、装置、设备及存储介质
CN113313211B (zh) 文本分类方法、装置、电子设备及存储介质
CN113434542B (zh) 数据关系识别方法、装置、电子设备及存储介质
CN111930963B (zh) 知识图谱生成方法、装置、电子设备及存储介质
CN111429085A (zh) 合同数据生成方法、装置、电子设备及存储介质
CN116741358A (zh) 问诊挂号推荐方法、装置、设备及存储介质
WO2023178979A1 (zh) 问题标注方法、装置、电子设备及存储介质
CN116720525A (zh) 基于问诊数据的疾病辅助分析方法、装置、设备及介质
WO2022141867A1 (zh) 语音识别方法、装置、电子设备及可读存储介质
CN116844711A (zh) 基于深度学习的疾病辅助识别方法及装置
CN113419951B (zh) 人工智能模型优化方法、装置、电子设备及存储介质
WO2021151303A1 (zh) 命名实体对齐方法、装置、电子设备及可读存储介质
WO2022141838A1 (zh) 模型置信度分析方法、装置、电子设备及计算机存储介质
CN113822215A (zh) 设备操作指引文件生成方法、装置、电子设备及存储介质
CN114492446A (zh) 法律文书处理方法、装置、电子设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22770417

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 16/01/2024)