WO2018228151A1 - 分诊方法、装置和设备以及计算机可读存储介质 - Google Patents

分诊方法、装置和设备以及计算机可读存储介质 Download PDF

Info

Publication number
WO2018228151A1
WO2018228151A1 PCT/CN2018/087986 CN2018087986W WO2018228151A1 WO 2018228151 A1 WO2018228151 A1 WO 2018228151A1 CN 2018087986 W CN2018087986 W CN 2018087986W WO 2018228151 A1 WO2018228151 A1 WO 2018228151A1
Authority
WO
WIPO (PCT)
Prior art keywords
disease
patient
condition
case
vector
Prior art date
Application number
PCT/CN2018/087986
Other languages
English (en)
French (fr)
Inventor
张超
张振中
Original Assignee
京东方科技集团股份有限公司
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 京东方科技集团股份有限公司 filed Critical 京东方科技集团股份有限公司
Publication of WO2018228151A1 publication Critical patent/WO2018228151A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16ZINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
    • G16Z99/00Subject matter not provided for in other main groups of this subclass

Definitions

  • the present disclosure relates to a triage method, a triage device, a triage device, and a computer readable storage medium.
  • a method for triage comprising: acquiring condition characteristic data of a patient; and obtaining condition characteristic data of each case in the case database according to the condition data of the patient; Determining the likelihood that the patient will be suffering from each disease in the case database; outputting the results of the triage of the patient based on the likelihood that the patient is suffering from each disease in the case database.
  • condition characteristic data includes at least one of condition symptom information and a sign detection parameter.
  • the step of acquiring patient condition characteristic data further comprises: converting the input patient's condition characteristic into a patient according to a correspondence between a preset vector element position and a condition characteristic represented by the vector element The disease feature distribution vector, each element in the patient's condition feature distribution vector indicating whether a disease feature corresponding to the location of the element appears in the patient's condition feature of the input patient.
  • the determining, based on the patient's condition characteristic data, and the condition characteristic data of each case under all the diseases in the case database, determining the likelihood that the patient is suffering from each disease in the case database The steps further include:
  • the likelihood of each disease in the case database is determined based on the determined correlation coefficient vector X * .
  • the method further comprises: converting a condition characteristic of each case under each disease into a case condition feature distribution vector according to a correspondence between a position of the preset vector element and a condition characteristic represented by the vector element.
  • Each element in the condition feature distribution vector of the case indicates whether a condition characteristic corresponding to the location of the element appears in the condition feature of the case.
  • the predetermined condition includes a first predetermined condition
  • 2 ⁇ ⁇ and a second predetermined condition X * arg min
  • the step of determining the likelihood of each disease in the case database based on the determined correlation coefficient vector X * further comprises:
  • the probability calculation formula is: C i represents the probability that the patient is suffering from the disease i in the case database, M in the case indicates M diseases in the case database, and the ⁇ is an error vector, Is the square of the L2 paradigm.
  • the step of outputting the triage result of the patient according to the likelihood of each disease in the case database further comprises:
  • the step of outputting the triage result of the patient according to the likelihood of each disease in the case database further comprises outputting the most likely disease of all the diseases as the patient Triage results.
  • a triage device comprising: an acquirer configured to acquire patient condition characteristic data; a processor configured to be based on the patient's condition characteristic data, and all diseases in the case database The disease characteristic data of each case determines the possibility that the patient suffers from each disease in the case database; the output device is configured to output according to the possibility that the patient suffers from each disease in the case database The result of the triage of the patient.
  • condition characteristic data includes at least one of condition symptom information and a sign detection parameter.
  • the acquirer is further configured to: convert the input patient's condition characteristic into a patient's condition feature distribution vector according to a correspondence between a position of the preset vector element and a condition characteristic represented by the vector element, Each element in the patient's condition feature distribution vector indicates whether a condition characteristic corresponding to the location of the element appears in the patient's condition characteristic of the input patient.
  • the processor is further configured to:
  • the likelihood of each disease in the case database is determined based on the determined correlation coefficient vector X * .
  • the triage device may further include a converter, and the converter may be configured to place each disease type according to a correspondence between a position of the preset vector element and a condition characteristic represented by the vector element.
  • the condition characteristic of the case is converted into a case condition feature distribution vector, and each element in the case condition feature distribution vector indicates whether the condition characteristic corresponding to the position of the element appears in the condition feature of the case.
  • the predetermined condition includes a first predetermined condition
  • 2 ⁇ ⁇ and a second predetermined condition X * arg min
  • the processor when determining the likelihood of each disease in the case database according to the determined correlation coefficient vector X 0 , is specifically configured to:
  • the probability calculation formula is: C i represents the probability that the patient is suffering from the disease i in the case database, M in the case indicates M diseases in the case database, and the ⁇ is an error vector, Is the square of the L2 paradigm.
  • the outputter is further configured to:
  • the outputter is further configured to:
  • a triage device comprising: one or more processors; and a memory having computer executable instructions stored thereon, the computer executable instructions being configured to be When executed by a plurality of processors, one or more steps of any of the methods described above are performed.
  • a computer readable storage medium having computer-executable instructions that, when executed by one or more processors, cause the one or more processors to execute One or more steps of any of the methods described above.
  • FIG. 1 is a flowchart of a method for a triage method according to an embodiment of the present disclosure
  • FIG. 2 is a schematic diagram of a semantic space provided according to an embodiment of the present disclosure
  • FIG. 3 is a schematic structural diagram of a triage device according to an embodiment of the present disclosure.
  • FIG. 4 is a schematic structural diagram of a triage device according to an embodiment of the present disclosure.
  • the words “first”, “second”, etc. are used to distinguish the same or similar items whose functions or functions are substantially the same, in the field.
  • the skilled person will understand that the words “first”, “second” and the like do not limit the number and order of execution.
  • the executive body of the triage method may be a triage device.
  • the triage device may be a terminal or device for performing the above-described triage method, or may be a processor for executing the above-described triage method.
  • the terminal or device may be a terminal device such as a computer, a smart phone, a tablet computer, a notebook computer, an UMPC (Ultra-mobile Personal Computer), a netbook, a PDA (Personal Digital Assistant), and the like. Not limited to this.
  • FIG. 1 is a flow chart of a method for a triage method according to an embodiment of the present disclosure. As shown in FIG. 1, the method can include steps 101-103.
  • Step 101 Obtain patient condition characteristic data.
  • the patient's condition characteristic data is data for indicating the condition characteristics of the patient.
  • a patient or other person eg, a nurse
  • the voice recognition system installed on the patient recognizes the patient's oral information, and the triage system installed on the computer selects or ambiguously matches the disease characteristic keyword from the identified patient oral information (refers to the disease characteristics existing in the disease characteristic database)
  • the triage system installed on the computer can also identify the patient identity information (such as scanning ID card or medical card, etc.) through the identification function of the computer, and check the database from the hospital according to the patient identity information (can be configured to store the sign detection parameters)
  • the database, the physical examination parameters include the items and results of the patient's physical examination, and the physical examination parameters of the patient are retrieved.
  • condition characteristic data may include: condition symptom information and/or a sign detection parameter, wherein the condition symptom information is a symptom observed by the patient or a symptom perceived by the patient, for example, may be a patient's oral symptom or input.
  • Symptom texts such as: shortness of breath, numbness of the limbs, tinnitus, etc.
  • the physical examination detection parameter may include the detection value of each indicator of the patient, such as a blood pressure value, a blood sugar level, etc., and the condition characteristic reflected may be a high blood pressure, a high blood pressure, and the like.
  • condition characteristic data may be a set of numbers indicating the condition characteristic data, in addition to the above type, and may be, for example, a vector (or a matrix based on a vector).
  • step 101 may further include:
  • Step 101a Convert the input condition of the patient into a disease characteristic distribution vector of the patient according to the correspondence between the position of the preset vector element and the condition characteristic represented by the vector element, and each element in the patient's disease feature distribution vector Indicates whether the condition characteristic of the location of the element appears in the patient's condition characteristic of the input.
  • I j (1 ⁇ j ⁇ Q) represents the condition characteristic at the j-th position
  • the set I represents the condition characteristic from the disease condition at the first position to the Q-th position.
  • I 500 is the conditional feature at the 500th position.
  • the set I represents the condition characteristic from the disease condition at the first position to the 1000th position.
  • Step 102 Determine the possibility of each disease in the patient's case database according to the patient's condition characteristic data and the disease characteristic data of each case under all the disease records in the case database.
  • the probability of each disease in the patient's case database described above may refer to the probability that the patient has each disease, represented by a value between 0 and 1. Or it may be a value corresponding to the possibility of the patient suffering from each disease type (may be a value greater than 1), and a larger value indicates a greater possibility.
  • the condition characteristic data is the disease characteristic text
  • the patient's condition characteristic text is dizziness, nausea, and palpitations
  • the possibility of determining each disease in the patient's case database in the above step 102 is determined.
  • the specific process can refer to the following content:
  • the number of diseases in the case database here is taken as three cases, namely, disease type A, disease type B, and disease type C, wherein: disease type A takes three cases as an example.
  • the disease B is exemplified by the inclusion of 4 cases
  • the case C is exemplified by 5 cases
  • the patient has 3 cases as an example.
  • step 102 may further include the following steps:
  • D i [D i,1 , D i,2 ,...,D i,j ,...,D i,K ]
  • D i,j is a case condition feature distribution vector of the jth case of the disease type i in the case database
  • the K indicating that the disease type i in the case database includes K cases
  • the M represents the case database Includes M diseases.
  • Step 102b Determine, according to the determined correlation coefficient vector X * , the likelihood of each disease in the case database.
  • the method may further include the following steps:
  • A1 According to the correspondence between the position of the preset vector element and the condition of the disease indicated by the vector element, the condition of each case under each disease is converted into a case feature distribution vector. Wherein, each element in the condition feature distribution vector of the above case is used to indicate whether the condition characteristic corresponding to the location of the element appears in the condition feature of the case.
  • the Q feature elements are included in the disease feature distribution vector of each case under each disease in the case database.
  • the set I of the disease features in the database is a Q-dimensional vector
  • the disease feature distribution vector of any case under the corresponding disease A is also a Q-dimensional vector.
  • I j (1 ⁇ j ⁇ Q) represents a condition characteristic at the j-th position of any case under the disease A
  • D A1 represents the case from the first position in any case under the disease type A
  • the condition is characterized by the condition at the Q position.
  • the foregoing relationship model may be pre-established or may be established in real time according to requirements.
  • For the establishment process of the above relationship model reference may be made to the following contents:
  • Case Since some embodiments of the present disclosure are based on medical big data for disease prediction, a large number of cases (eg, confirmed cases of hospitals over the years) are required, which corresponds to each case in the case database in the flowchart.
  • the present disclosure uses the symbol D to represent a collection of disease species in a case database. Assuming that a total of M diseases (ie, M disease types) are included, D i (1 ⁇ i ⁇ M) represents the i-th disease of the case database. Assuming that K cases are included in the i-th disease, D ij (1 ⁇ i ⁇ M, 1 ⁇ j ⁇ K) represents the j-th case of the i-th disease.
  • Each case consists of a corresponding feature vector (such as symptom and sign detection parameters), and D constitutes the semantic space of a confirmed case.
  • the symptoms in case 1 are "dizziness, nausea, palpitations and shortness of breath”.
  • case 2 The symptoms in case 2 are "heart qi, shortness of breath, tinnitus, limb numbness”.
  • the symptoms in case 3 are "dizziness, nausea”.
  • each disease can be represented as a semantic subspace composed of the cases it contains, and a case belonging to the disease can be composed of a linear combination of corresponding subspaces (condition characteristics).
  • D [D 1 , D 2 , . . . , D M ]
  • D 1 , D 2 , ..., D M (1 ⁇ i ⁇ M) is included under the ith disease
  • D [D 11 , D 12 , D 21 , D 22 , ..., D M1 , D M2 ].
  • disease A contains 2 cases
  • disease B contains 3 cases
  • disease C contains 2 cases.
  • D [D A1 , D A2 , D B1 , D B2 , D B3 , D C1 , D C2 ].
  • D [D A1 , D A2 , D B1 , D B2 , D B3 , D C1 , D C2 ], assuming the disease characteristics based on the seven cases of the above three diseases
  • X [ ⁇ A1 , ⁇ A2 , ⁇ B1 , ⁇ B2 , ⁇ B3 , ⁇ C1 , ⁇ C2 ] T are expressed.
  • the number of diseases in the above case database is hundreds or thousands, and the cases under each disease may be hundreds or even more, based on The symptoms of the cases under each disease may be tens of thousands, and the above is only an example for explaining the present solution, and is not intended to limit the present disclosure.
  • the predetermined condition in step 102a comprises a first predetermined condition
  • 2 ⁇ ⁇ and a second predetermined condition X * arg min
  • 2 is the L2 normal form, the ⁇ is a preset parameter, and X * arg min
  • case disease feature distribution vectors of case 1, case 2, and case 3 can be expressed as [1, 1, 1, respectively. 0,0] T , [0,0,1,1,1] T , [1,1,1,1,0] T , and the new patient's condition feature distribution vector can be expressed as [0,0, 1,0,1] T .
  • a sparse solution method is adopted, that is, using a minimum number of cases to reconstruct the condition of the patient h under a certain precision condition, using a sparse solution method
  • step 102b may include the following steps:
  • Step 102b1 determining, from the determined correlation coefficient vector X * , a correlation coefficient vector ⁇ i (X * ) of each disease in the case database, and ⁇ i (X * ) indicating that the coefficient vector X * is The dimension belonging to the disease D i is multiplied by 1, and the remaining dimensions are multiplied by 0 to obtain a vector. That is to say, the correlation coefficient of each case under the i-th disease in X * is retained, and other elements are set to 0, and ⁇ i (X * ) is obtained.
  • Kind of probability
  • the probability calculation formula is:
  • C i represents the probability of a disease i in the patient's case database
  • M in the case represents M disease cases in the case database
  • is the error matrix
  • h h 1 + h 2 + ... + h M + ⁇
  • h the disease characteristic distribution vector of the new patient
  • h i is the semantic component (1 ⁇ i ⁇ M) corresponding to each disease i in the disease characteristic distribution vector h of the patient
  • the h i is the linearity of the disease characteristic distribution vector of all cases under the disease i combination. Is the square of the L2 paradigm.
  • the disease A contains 2 cases
  • the disease B contains 3 cases
  • the disease C contains 2 cases.
  • D [D A1 , D A2 , D B1 , D B2 , D B3 , D C1 , D C2 ].
  • C i reflects the possibility that the patient h belongs to the disease D i (C ⁇ reflects the possibility that the patient h does not belong to any of the preceding diseases D 1 -D M ) .
  • C i, h indicate the condition wherein the patient profile configuration vector comprising belong more cases of D i disease, i.e., patients h D i located more portions of semantic subspace belong disease of D i The greater the possibility.
  • FIG. 2 shows a schematic diagram of a semantic space provided according to an embodiment of the present disclosure.
  • Step 103 Output a triage result of the patient according to the possibility of each disease in the patient's case database.
  • the above-mentioned triage result may include a department assigned to the patient, and may further include a triage process, a doctor assigned to the patient, and a treatment guide referenced.
  • step 103 can be implemented in any of the following manners:
  • Mode A Output the most likely disease among all the diseases as the result of the patient's triage. For example, calculating the likelihood of a patient suffering from each disease type, and determining that the patient is most likely to have the disease type A, the triage result corresponding to the patient's disease type A is displayed on the interface of the triage system.
  • the disease in which the possibility is not zero among all the diseases is output as the result of the triage of the patient in order of possibility.
  • the disease in which the possibility is not zero among all the diseases is output as the result of the triage of the patient in order of possibility. For example, to calculate the likelihood of a patient suffering from each disease, and to order the diseases of all diseases that are not likely to be zero, in order of probability, from large to small, at the interface of the triage system from large to large.
  • the small sequence shows the results of the triage of the patient's likelihood of each disease.
  • the patient's condition characteristic data is obtained; then, the patient's case is determined according to the patient's condition characteristic data and the disease characteristic data of each case under all the disease records in the case database.
  • the possibility of each disease in the database is output, thereby realizing intelligent triage of patients to reduce the hospital's triage pressure.
  • a triage device provided by some embodiments of the present disclosure will be described below based on the related description in the embodiment of the triage method of FIG.
  • the technical terms, concepts, and the like related to the above embodiments in the following embodiments reference may be made to the above embodiments, and details are not described herein again.
  • FIG. 3 is a schematic structural diagram of a triage device according to an embodiment of the present disclosure.
  • the apparatus may include an acquirer 31, a processor 32, and an outputter 33, wherein:
  • the acquirer 31 can be configured to acquire patient condition data.
  • the processor 32 can be configured to determine the likelihood of the patient suffering from each of the disease databases in the case database based on the patient's condition profile data and the condition profile data for each case under all disease in the case database.
  • the outputter 33 can be configured to output a triage result of the patient based on the likelihood of each disease in the patient's case database.
  • the patient's condition characteristic data may be a condition feature text or data indicating the condition feature text.
  • condition characteristic data may include: symptom symptom information and/or a sign detection parameter, wherein the condition symptom information may be a symptom observed by the patient or a symptom perceived by the patient, for example, may be a patient's oral symptom or input.
  • Symptom texts such as: shortness of breath, numbness of the limbs, tinnitus, etc.
  • the physical examination detection parameter may include the detection value of each indicator of the patient, such as a blood pressure value, a blood sugar level, etc., and the condition characteristic reflected may be a high blood pressure, a high blood pressure, and the like.
  • the probability of each disease in the patient's case database described above may refer to the probability that the patient has each disease, represented by a value between 0 and 1. Or it may be a value corresponding to the likelihood that the patient will suffer from each disease type (may be a value greater than 1), and a larger value indicates a greater likelihood.
  • the processor 32 determines the possibility of each disease in the patient's case database.
  • the specific process refers to the following contents:
  • the number of diseases in the case database here is taken as three cases, which are disease type A, disease type B, and disease type C, among which: disease type A includes three cases as an example, and the disease type B takes 4 cases as an example, and case C contains 5 cases as an example, and the patient has 3 cases as an example.
  • the patient's first condition is present in the first case under disease A in the case database
  • the second condition appears in the second case under disease A
  • the third condition appears in disease A.
  • the patient's three disease characteristics are not all present in the disease characteristics text of other diseases, the patient is most likely to have the disease A, but the two results are the smallest.
  • the acquirer 31 described above may be further configured to:
  • the input patient's condition feature is converted into the patient's disease feature distribution vector, and each element in the patient's disease feature distribution vector indicates the input patient. Whether the disease characteristics corresponding to the location of the element appear in the disease characteristic.
  • I [I 1 , I 2 , ..., I 1000 ] T .
  • I 500 is the condition characteristic at the 500th position, and thus the set I represents the condition characteristic from the disease condition at the first position to the 1000th position.
  • processor 32 described above may be further configured to:
  • the case database to determine the likelihood of each of the disease
  • the apparatus may further include, for example, a converter 34, wherein: the converter 34 may be configured to be based on a correspondence between a position of the preset vector element and a condition characteristic represented by the vector element, The disease characteristics of each case under each disease were converted into case disease feature distribution vectors. Wherein, each element in the condition feature distribution vector of the above case indicates whether the condition characteristic corresponding to the position of the element appears in the condition feature of the case.
  • the Q feature elements are included in the disease feature distribution vector of each case under each disease in the case database.
  • the set I of the disease features in the database is a Q-dimensional vector
  • the disease feature distribution vector of any case under the corresponding disease A is also a 0-dimensional vector.
  • I j (1 ⁇ j ⁇ Q) indicates a condition characteristic at the j-th position of any one of the cases A, and thus any one of the cases A of the disease A represented by D A1 is from the first position The condition of the condition to the condition of the condition at the Q position.
  • the foregoing relationship model may be pre-established or may be established in real time according to requirements.
  • the establishment process of the foregoing relationship model reference may be made to the content of the method part, and details are not described herein again.
  • the predetermined condition may include a first predetermined condition
  • 2 ⁇ ⁇ and a second predetermined condition X * arg min
  • 2 is the L2 paradigm, which is a preset parameter.
  • is introduced to reduce the influence of “noise”, and the value of this parameter depends on the precision required for implementation.
  • X * arg min
  • the processor 32 described above adopts a sparse solution method in determining the correlation coefficient vector X * by using the first predetermined condition and the second predetermined condition, that is, using a minimum number of cases to reconstruct the condition of the patient h under a certain precision condition.
  • the processor 32 described above determines that the likelihood of each disease in the case database is represented by a probability according to the determined correlation coefficient vector X * , the processor 32 may be further configured to:
  • C i represents the probability of a disease i in the patient's case database
  • M in the case represents M disease cases in the case database
  • is the error matrix
  • h h 1 + h 2 + ... ... + h M + ⁇
  • h is the condition of the new patient
  • the distribution vector h i is the semantic component (1 ⁇ i ⁇ M) corresponding to each disease i in the disease characteristic distribution vector h of the patient, and the h i is the disease characteristic distribution vector of all cases under the disease i Linear combination. Is the square of the L2 paradigm.
  • C i reflects the possibility that the patient h belongs to the disease D i (C ⁇ reflects the possibility that the patient h does not belong to any of the preceding diseases D 1 -D M ) .
  • C i, h indicate the condition wherein the patient profile configuration vector comprising belong more cases of D i disease, i.e., patients h D i located more portions of semantic subspace belong disease of D i The greater the possibility.
  • Figure 2 it is assumed that there are three disease types or subspaces in the known case space, which correspond to different shapes: four-pointed stars, triangles, and hexagonal stars, respectively corresponding to the first three dimensions of C, and the last dimension is the error.
  • C ⁇ The rounded node represents the new patient.
  • the two circles shown in Figure 2 represent two linear combinations for representing a new patient, respectively.
  • the above-mentioned triage results include information such as a treatment guide that can be referred to, a triage procedure, and the departments involved.
  • the outputter 33 described above may be further configured to be implemented in any of the following ways:
  • Mode A Output the most likely disease among all the diseases as the result of the patient's triage. For example, calculating the likelihood of a patient suffering from each disease type, and determining that the patient is most likely to have the disease type A, the triage result corresponding to the patient's disease type A is displayed on the interface of the triage system.
  • the disease in which the possibility is not zero among all the diseases is output as the result of the triage of the patient in order of possibility.
  • the disease in which the possibility is not zero among all the diseases is output as the result of the triage of the patient in order of possibility. For example, to calculate the likelihood of a patient suffering from each disease, and to order the diseases of all diseases that are not likely to be zero, in order of probability, from large to small, at the interface of the triage system from large to large.
  • the small sequence shows the results of the triage of the patient's likelihood of each disease.
  • the device obtains the patient's condition characteristic data; and then, according to the patient's condition characteristic data, and the disease characteristic data of each case under all the disease records in the case database, the patient is determined to be affected by the patient.
  • FIG. 4 shows a schematic structural view of the triage device 400.
  • the triage device 400 can include: one or more processors 401; and a memory 402 coupled to the processor 401 having computer executable instructions stored thereon, the computer executable instructions being configured to One or more steps of any of the methods described above are performed when executed by the one or more processors.
  • the triage device 400 can be implemented as a computer product structure of the local computing, that is, the triage method described in the above embodiment is implemented on the user side; and the computer product structure of the local and remote interaction can also be implemented, that is, the terminal on the user side is implemented.
  • Some steps of the triage method described in the above embodiments, such as input of condition characteristic data, output of triage results; other steps of the triage method described in the above embodiments are implemented on the network side connected to the user side terminal, for example
  • the patient has a calculation of the likelihood of each disease in the case database.
  • one or more processors of the triage device may be located in the same computer product or in different computer products. For example, a portion of the processor may be located on a user-side computer product, and a portion of the processor may be located at a remote or cloud-based server computer product to perform a portion of the steps of the triage method, respectively.
  • the processor 401 can be a central processing unit (CPU) or a field programmable logic array (FPGA) or a microcontroller (MCU) or a digital signal processor (DSP) or an application specific integrated circuit (ASIC) having data processing capabilities and/or program execution.
  • CPU central processing unit
  • FPGA field programmable logic array
  • MCU microcontroller
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • the memory 402 can be implemented in any of a variety of volatile or non-volatile storage devices, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Disk or Optical Disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read only memory
  • EPROM Erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • Disk Disk or Optical Disk.
  • Computer instructions include one or more processor operations defined by an instruction set architecture corresponding to a processor, which may be logically included and represented by one or more computer programs.
  • the triage device 400 can also connect various input devices 403 (eg, user interface, keyboard, etc.), various output devices 404 (eg, speakers, etc.), and display device 405 to implement interaction of computer products with other products or users. I will not repeat them here.
  • various input devices 403 eg, user interface, keyboard, etc.
  • various output devices 404 eg, speakers, etc.
  • display device 405 to implement interaction of computer products with other products or users. I will not repeat them here.
  • the connection may be through a network module 406, such as a wireless network, a wired network, and/or any combination of a wireless network and a wired network.
  • the network may include a local area network, the Internet, a telecommunications network, an Internet of Things based Internet and/or telecommunications network, and/or any combination of the above networks, and the like.
  • the wired network can communicate by, for example, twisted pair, coaxial cable or optical fiber transmission.
  • the wireless network can adopt a communication method such as a 3G/4G/5G mobile communication network, Bluetooth, Zigbee or Wi-Fi.
  • a computer readable storage medium comprising computer executable instructions for causing the one or more processes when executed by one or more processors
  • the apparatus performs one or more steps of any of the methods described above.
  • triage device and computer readable storage medium For the specific implementation of the above-mentioned triage device and computer readable storage medium, reference may be made to the previous description of the triage method, and details are not described herein again. Those of ordinary skill in the art will appreciate that the above-described triage device and computer readable storage medium can also achieve intelligent triage of patients, which can reduce the pressure of triage in hospitals.
  • the disclosed triage device can be implemented in other ways.
  • the embodiments of the apparatus described above are merely illustrative.
  • the division of the modules or units is only one logical function division, and the actual implementation may have another division manner, such as multiple units or components. It can be combined or integrated into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.
  • the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e., may be located in one place, or may be distributed over multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.
  • each functional unit in various embodiments of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.
  • the integrated unit if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium.
  • the technical solution of the present disclosure may contribute to the prior art or all or part of the technical solution may be embodied in the form of a software product stored in a storage medium.
  • a number of instructions are included to cause a computer device (which may be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the methods described in various embodiments of the present disclosure.
  • the foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like. .
  • any reference signs placed in parentheses shall not be construed as limiting the claim.
  • the word “comprising” does not exclude the presence of the elements or the The word “a” or “an” or “an”
  • the present disclosure may be implemented by means of hardware comprising several discrete elements, or by suitably programmed software or firmware, or by any combination thereof.

Landscapes

  • Medical Treatment And Welfare Office Work (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

本公开实施例提供一种分诊方法、分诊装置、分诊设备以及计算机可读存储介质。该方法包括:获取患者的病情特征数据;根据患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定患者患病例数据库中每个病种的可能性;根据患者患病例数据库中每个病种的可能性,输出患者的分诊结果。

Description

分诊方法、装置和设备以及计算机可读存储介质
相关申请
本申请要求2017年6月13日提交、申请号为201710444355.3的中国专利申请的优先权,该申请的全部内容通过引用并入本文。
技术领域
本公开涉及一种分诊方法、分诊装置、分诊设备和计算机可读存储介质。
背景技术
随着生活条件的不断改善,对于健康的需求也越来越旺盛。近年来,各大医院的门急诊量急剧增长,由此造成医疗分诊压力大,流程滞后、等候时间长,继而带来医疗质量难以保证,医患矛盾增加等一系列问题。
发明内容
根据本公开的一个方面,提供了一种分诊方法,所述方法包括:获取患者的病情特征数据;根据所述患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定所述患者患所述病例数据库中每个病种的可能性;根据所述患者患所述病例数据库中每个病种的可能性,输出所述患者的分诊结果。
在一些实施例中,所述病情特征数据包括病情症状信息和体征检测参数的至少一者。
在一些实施例中,所述获取患者的病情特征数据的步骤进一步包括:根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将输入的患者的病情特征转换成患者的病情特征分布向量,所述患者的病情特征分布向量中的每个元素指示所述输入的患者的病情特征中是否出现该元素所在位置对应的病情特征。
在一些实施例中,所述根据所述患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定所述患者患所述病例数据库中每个病种的可能性的步骤进一步包括:
基于患者的病情特征分布向量,通过对关系模型h=DX求解X来确定满足预定条件的相关系数向量X *,其中,h为所述患者的病情特征分布向量,h和X均为列向量,D为由所述病例数据库中每个病种下各 病例的病情特征分布向量组成的矩阵,D=[D 1,D 2,......,D i,......,D M],其中:D i=[D i,1,D i,2,...,D i,j,...,D i,K],D i,j为所述病例数据库中病种i的第j个病例的病例病情特征分布向量,所述K表示所述病例数据库中病种i包括K个病例,所述M表示病例数据库中包括M种疾病;
根据所述确定的相关系数向量X *,确定所述病例数据库中每个病种的可能性。
在一些实施例中,所述方法还包括:根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将每个病种下各病例的病情特征转换成病例病情特征分布向量;所述病例的病情特征分布向量中的每个元素指示所述病例的病情特征中是否出现该元素所在位置对应的病情特征。
在一些实施例中,所述预定条件包括第一预定条件||DX-h|| 2≤ε以及第二预定条件X *=arg min||X|| 1,其中:||·|| 1是L1范式,||·|| 2是L2范式,所述ε为预设参数。
在一些实施例中,根据所述确定的相关系数向量X *确定所述病例数据库中每个病种的可能性的步骤进一步包括:
从所述确定的相关系数向量X *中确定出所述病例数据库中每个病种的相关系数向量δ i(X *),δ i(X *)表示通过将系数向量X *中属于病种D i的维度乘以1,其余维度乘以0所得到的一个向量;
根据所述病例数据库中每个病种的相关系数向量δ i(X *)确定所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *);
将所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *)代入到概率计算公式中,得到所述患者患所述病例数据库中每个病种的概率;
其中,所述概率计算公式为:
Figure PCTCN2018087986-appb-000001
C i表示所述患者患所述病例数据库中的病种i的概率,
Figure PCTCN2018087986-appb-000002
中的M表示所述病例数据库中的M个病种,所述η为误差向量,所述
Figure PCTCN2018087986-appb-000003
是L2范式的平方。
在一些实施例中,所述根据所述病例数据库中每个病种的可能性,输出所述患者的分诊结果的步骤进一步包括:
将所有病种中可能性不为零的病种,按照可能性大小顺序输出作为所述患者的分诊结果。
在一些实施例中,所述根据所述病例数据库中每个病种的可能性,输出所述患者的分诊结果的步骤进一步包括输出所有病种中可能性最大的病种作为所述患者的分诊结果。
根据本公开另一个方面,提供了一种分诊装置,包括:获取器,配置成获取患者的病情特征数据;处理器,配置成根据所述患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定所述患者患所述病例数据库中每个病种的可能性;输出器,配置成根据所述患者患所述病例数据库中每个病种的可能性,输出所述患者的分诊结果。
在一些实施例中,所述病情特征数据包括病情症状信息和体征检测参数的至少一者。
在一些实施例中,所述获取器进一步配置成:根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将输入的患者的病情特征转换成患者的病情特征分布向量,所述患者的病情特征分布向量中的每个元素指示所述输入的患者的病情特征中是否出现该元素所在位置对应的病情特征。
在一些实施例中,所述处理器进一步配置成:
基于患者的病情特征分布向量,通过对关系模型h=DX求解X来确定满足预定条件的相关系数向量X *,其中,h为所述患者的病情特征分布向量,h和X均为列向量,D为由所述病例数据库中每个病种下各病例的病情特征分布向量组成的矩阵,D=[D 1,D 2,......,D i,......,D M], 其中:D i=[D i,1,D i,2,...,D i,j,...,D i,K],D i,j为所述病例数据库中病种i的第j个病例的病例病情特征分布向量,所述K表示所述病例数据库中病种i包括K个病例,所述M表示病例数据库中包括M种疾病;
根据所述确定的相关系数向量X *,确定所述病例数据库中每个病种的可能性。
在一些实施例中,所述分诊装置还可以包括转换器,该转换器可以配置成根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将每个病种下各病例的病情特征转换成病例病情特征分布向量,所述病例病情特征分布向量中的每个元素指示病例的病情特征中是否出现该元素所在位置对应的病情特征。
在一些实施例中,所述预定条件包括第一预定条件||DX-h|| 2≤ε以及第二预定条件X *=arg min||X|| 1,其中:||·|| 1是L1范式,||·|| 2是L2范式,所述ε为预设参数。
在一些实施例中,所述处理器在根据所述确定的相关系数向量X 0,确定所述病例数据库中每个病种的可能性时,具体配置成:
从所述确定的相关系数向量X *中确定出所述病例数据库中每个病种的相关系数向量δ i(X *),δ i(X *)表示通过将系数向量X *中属于病种D i的维度乘以1,其余维度乘以0所得到的一个向量;
根据所述病例数据库中每个病种的相关系数向量δ i(X *)确定所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *);
将所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *)代入到概率计算公式中,得到所述患者患所述病例数据库中每个病种的概率;
其中,所述概率计算公式为:
Figure PCTCN2018087986-appb-000004
C i表示所述患者患所述病例数据库中的病种i的概率,
Figure PCTCN2018087986-appb-000005
中的M表示所述病例数据库中的M个病种,所述η为误差向量,所述
Figure PCTCN2018087986-appb-000006
是L2范式的平方。
在一些实施例中,所述输出器进一步配置成:
将所有病种中可能性不为零的病种,按照可能性大小顺序输出作为所述患者的分诊结果。
在一些实施例中,所述输出器进一步配置成:
输出所有病种中可能性最大的病种作为所述患者的分诊结果。
根据本公开另一个方面,提供了一种分诊设备,包括:一个或多个处理器;和存储器,其上存储有计算机可执行指令,所述计算机可执行指令被配置为当被所述一个或多个处理器执行时,执行如上所述的任何一种方法的一个或多个步骤。
根据本公开另一个方面,提供了一种计算机可读存储介质,其上包含有计算机可执行指令,所述指令在由一个或多个处理器执行时,使所述一个或多个处理器执行如上所述的任何一种方法的一个或多个步骤。
上述以简化的形式介绍了本公开的一些构思,这些构思在下面的具体实施方式中进一步加以描述。发明内容部分并非要给出要求保护的主题的必要特征或实质特征,也不是要限制要求保护的主题的范围。此外,正如本文所描述的,各种各样的其他特征和优点也可以根据需要结合到这些技术中。
附图说明
为了更清楚地说明本公开一些实施例的技术方案,本公开提供了下列附图以便在实施例描述时使用,这些附图构成说明书的一部分,与本公开的实施例一起用于解释本公开一些实施例的技术方案。应当意识到,下面描述中的附图仅仅涉及一些实施例,并不构成对本公开技术方案的限制,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其它的附图,所述其它的附图 也在本公开的范围内。
图1为根据本公开一个实施例提供的一种分诊方法的方法流程图;
图2为根据本公开一个实施例提供的一种语义空间的示意图;
图3为根据本公开一个实施例提供的一种分诊装置的结构示意图;
图4为根据本公开一个实施例提供的一种分诊设备的结构示意图。
具体实施方式
为了能够更清楚地理解一些实施例的目的、技术方案和优点,下面结合附图和具体实施方式对这些实施例作进一步详细描述。本领域普通技术人员能够理解,所描述的实施例仅仅是本公开的一部分实施例,而不是全部的实施例。基于本公开的实施例,本领域普通技术人员在没有做出创造性劳动前提下能够获得其它的实施例,所获得的所有其他实施例都属于本公开保护的范围。
为了便于清楚描述本公开实施例的技术方案,在本公开的实施例中,采用了“第一”、“第二”等字样对功能或作用基本相同的相同项或相似项进行区分,本领域技术人员可以理解“第一”、“第二”等字样并不对数量和执行次序进行限定。
本文中术语“和/或”,仅仅是一种描述关联对象的关联关系,表示可以存在三种关系,例如,A和/或B可以表示:单独存在A,同时存在A和B,单独存在B这三种情况。另外,本文中字符“/”,一般表示前后关联对象是一种“或”的关系。
本公开的一些实施例提供的分诊方法的执行主体可以为分诊装置。例如,该分诊装置可以是用于执行上述分诊方法的终端或设备,也可以是用于执行上述分诊方法的处理器。其中:该终端或设备可以为计算机、智能手机、平板电脑、笔记本电脑、UMPC(Ultra-mobile Personal Computer,超级移动个人计算机)、上网本、PDA(Personal Digital Assistant,个人数字助理)等终端设备,且不限于此。
图1示出了根据本公开一个实施例提供的一种分诊方法的方法流程图。如图1所示,该方法可以包括步骤101-103。
步骤101、获取患者的病情特征数据。
该患者的病情特征数据是用于表示患者具有的病情特征的数据。示例性的,患者或者其他人(例如护士)可以通过计算机上安装的分 诊系统的输入界面,输入病情特征文本;还可以通过计算机的语音采集器(例如麦克风),采集患者声音,并且由计算机上安装的语音识别系统识别出患者口述信息,计算机上安装的分诊系统从识别出的患者口述信息中选择出、或模糊匹配出病情特征关键词(是指病情特征数据库中所存在的病情特征);计算机上安装的分诊系统还可通过计算机的身份识别功能,识别患者身份信息(例如扫描身份证或就诊卡等),根据患者身份信息从医院检查数据库(可以配置成存储体征检测参数的数据库,体征检测参数包括患者做身体检查的项目和结果)中调取该患者的体征检查参数。
示例性的,上述的病情特征数据可以包括:病情症状信息和/或体征检测参数,其中,病情症状信息为观察到患者的症状或患者感受到的症状,例如可以是患者的口述症状或输入的症状文本等,例如:心悸气短、肢体麻木、耳鸣等。而体征检测参数可以包括患者的各项指标检测值,例如血压值、血糖值等,其反映出的病情特征可以是血压微高、血压过高等。
当然,病情特征数据除了可以是上述类型以外,还可以是表示病情特征数据的数字集合,例如可以是一个向量(或基于向量构成的矩阵)。此时,步骤101可以进一步包括:
步骤101a、根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将输入的患者的病情特征转换成患者的病情特征分布向量,患者的病情特征分布向量中的每个元素指示输入的患者的病情特征中是否出现该元素所在位置对应的病情特征。
例如,数据库中有Q个病情特征,该预设的向量元素的位置与向量元素所表示的病情特征的对应关系用集合I表示,该集合I为Q维向量,I=[I 1,I 2,...,I Q,] T。其中,I j(1≤j≤Q)表示第j个位置处的病情特征,从而集合I表示的是从第1个位置处的病情特征到第Q个位置处的病情特征。
例如,数据库中有1000个病情特征,上述的集合I为:I=[I 1,I 2,...,I 1000,] T。例如,I 500为第500个位置处的病情特征。集合I表示从第1个位置处的病情特征到第1000个位置处的病情特征。
步骤102、根据患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定患者患病例数据库中每个病种的可能性。
示例性的,上述的患者患病例数据库中每个病种的可能性可以是指患者患每个病种的概率,用0至1间的数值进行表示。或者也可以是患者患每个病种的可能性对应的数值(可以为包含大于1的数值),数值越大表示可能性越大。
示例性的,上述的病情特征数据为病情特征文本时,例如:患者的病情特征文本为眩晕、恶心以及心悸气短时,上述的步骤102中确定患者患病例数据库中每个病种的可能性的具体过程可以参照以下内容:这里病例数据库中的病种个数以3个为例,分别为病种A、病种B以及病种C,其中:病种A以包含3个病例为例,病种B以包含4个病例为例,病种C以包含5个病例为例,而患者所具有的病情特征以3个为例。将患者的病情特征与病例数据库中的每个病种下各病例中的病情特征文本进行匹配,若患者的3个病情特征均出现在病例数据库中病种A下的同一个病例中,且该患者的3个病情特征没有全部出现在其他病种下的病情特征文本中,则该患者患病种A的可能性最大;若患者的2个病情特征出现在病例数据库中病种A下的第一病例中,剩下的1个病情特征出现在病种A下的第二病例中,且该患者的3个病情特征没有全部出现在其他病种下的病情特征文本中,则该患者患病种A的可能性最大,但是相对于上面的结果较小;若患者的第1个病情特征出现在病例数据库中病种A下的第一病例中,第2个病情特征出现在病种A下的第二病例中,第3个病情特征出现在病种A下的第三病例中,且该患者的3个病情特征没有全部出现在其他病种下的病情特征文本中,则该患者患病种A的可能性最大,但是相对于上面的两种结果是最小的。当然,通过匹配病情特征得到患者患每个病种的可能性大小的规则,可以根据实际需要进行设置。
示例性的,上述的病情特征数据为向量时,上述的步骤102可以进一步包括如下步骤:
步骤102a、将所述患者的病情特征分布向量代入到关系模型h=DX中,通过对所述关系模型求解X来确定满足预定条件的相关系数向量X *,其中,h为所述患者的病情特征分布向量,h和X均为列向量,D为 由所述病例数据库中每个病种下各病例的病情特征分布向量组成的矩阵,D=[D 1,D 2,......,D i,......,D M],其中:D i=[D i,1,D i,2,...,D i,j,...,D i,K],D i,j为所述病例数据库中病种i的第j个病例的病例病情特征分布向量,所述K表示所述病例数据库中病种i包括K个病例,所述M表示病例数据库中包括M种疾病。
步骤102b、根据所述确定的相关系数向量X *,确定所述病例数据库中每个病种的可能性。
在一些实施例中,在上述的步骤102之前,该方法例如还可以包括以下步骤:
A1、根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将每个病种下各病例的病情特征转换成病例病情特征分布向量。其中,上述的病例的病情特征分布向量中的每个元素用于指示病例的病情特征中是否出现该元素所在位置对应的病情特征。
例如,若病例数据库中有Q个病情特征,则病例数据库中每个病种下各病例的病情特征分布向量中包含Q个元素。
示例性的,病种A下的任一个病例的病情特征分布向量为D A1=[I 1,I 2,...,I Q,] T。由于数据库中的病情特征的集合I为Q维向量,相应的这里的病种A下的任一个病例的病情特征分布向量也为Q维向量。其中,I j(1≤j≤Q)表示病种A下的任一个病例的第j个位置处的病情特征,从而D A1表示病种A下的任一个病例中从第1个位置处的病情特征到第Q个位置处的病情特征。
示例性的,上述的关系模型可以是预先建立好的,也可以是根据需要实时进行建立的,对于上述的关系模型的建立过程可以参考以下的内容:
由于本公开的一些实施例是基于医学大数据进行疾病的病种预测,因此需要大量的案例(例如各医院历年的确诊病例),这对应着流程图中的病例数据库中每个病种下各病例。本公开使用符号D来表示病例数据库中病种集合,假设其中一共包含M种疾病(即M个病种), 则D i(1≤i≤M)表示病例数据库的第i种疾病。假设第i种疾病中包含K个病例,则D ij(1≤i≤M,1≤j≤K)表示第i种疾病中的第j个病例。每一个病例由对应的特征向量(如症状和体征检测参数)构成,则D构成了一个确诊病例的语义空间。
对于新来的患者h(其含义是指:患者的病情特征分布向量用h表示),假设其患有疾病D i,依据本公开的基本思想:患有同一疾病的患者极有可能出现相似的特征(如症状和体征检测参数),则患者h可以表示为D i中所包含病例的线性组合,即h=α i,1×D i,1i,2×D i,2+......+α i,K×D i,K,其中,α ij是相关系数。例如,对于疾病“高血压”,病例1中的症状有“眩晕、恶心、心悸气短”,病例2中的症状有“心悸气短、耳鸣、肢体麻木”,病例3中的症状有“眩晕、恶心、耳鸣、心悸气短”,新来患者的症状有“心悸气短、肢体麻木”,则有“新来患者=病例1+病例2-病例3”。
为了表示简洁和方便,上面的表达形式可以用矩阵表示。假设D i=[D i1,D i2,.....,D iK],X i=[α i1,α i2,......,α iK] T,其中上标T表示矩阵的转置,则有h=D iX i
通过上面的讨论,可以看到每一个病种可以表示成由其包含的病例所构成的语义子空间,属于该病种的某一病例可以由相应子空间(病情特征)的线性组合构成。
上面所讨论的是新来的患者h,假设其患有疾病D i所做的讨论,那么对于新来的患者在不知道所患病种的前提下,类比于上述的过程,当给定所有病种的确诊病例语义空间矩阵D,可以通过寻找患者h在D中的语义子空间来确定其所患疾病。令D=[D 1,D 2,......,D M],则患者h与给定病例间的关系模型为:h=DX。
具体的,对于上面的D=[D 1,D 2,......,D M],由于病种D 1,D 2,......,D M这M个病种的每个病种下可能包含多个病例,因此,这里D 1,D 2,......,D M中的D i(1≤i≤M)为由第i个病种下所包含的各病例的病情特征分布向量构成的集合。例如,假设M个病种的每个病种下均包含两个病例,则D=[D 11,D 12,D 21,D 22,......,D M1,D M2]。
示例性的,假设病例数据库中有3个病种,分别为病种A、病种B以及病种C,该病种A包含2个病例,病种B包含3个病例,病种C包含2个病例,则D=[D A1,D A2,D B1,D B2,D B3,D C1,D C2]。
基于上面的病种矩阵D,D=[D A1,D A2,D B1,D B2,D B3,D C1,D C2],假设基于上述的3个病种的7个病例统计出的病情特征数据有100个,那么,上面的患者h与给定病例间的关系模型:h=DX中的病种矩阵D是一个100*7的矩阵,而相关系数向量X为7维列向量,可以用X=[α A1,α A2,α B1,α B2,α B3,α C1,α C2] T来表示。
需要说明的是,在实际的应用中,上述的病例数据库中的病种的个数为成百上千个,而每个病种下的病例相应的可能也是成百上千乃至更多,基于每个病种下的病例所抽出的病情特征可能是成千上万个,因此,上面的内容仅仅是一种示例,用于解释说明本方案,而非用于限制本公开。
在一些实施例中,其中,步骤102a中的所述预定条件包括第一预定条件||DX-h|| 2≤ε以及第二预定条件X *=arg min||X|| 1,其中:||·|| 1是L1范式,||·|| 2是L2范式,所述ε为预设参数,X *=arg min||X|| 1表示使得||X|| 1取最小值时的X,记为X *
需要指出的是,在理想情况下h=DX,即||DX-h|| 2=0。但是在 现实中由于计算精度的限制,可能出现最后计算得到的DX不等于h的情况。例如在上面提到的“新来患者=病例1+病例2-病例3”的例子中,病例1、病例2、病例3的病例病情特征分布向量可以被分别表示为[1,1,1,0,0] T、[0,0,1,1,1] T、[1,1,1,1,0] T,而新来患者的病情特征分布向量可以被表示为[0,0,1,0,1] T。假设通过计算得到DX=0.8*[1,1,1,0,0] T+[0,0,1,1,1] T-[1,1,1,1,0] T=[0.2,0.2,0.8,0,1]。那么此时DX-h=[0.2,0.2,0.8,0,1] T-[0,0,1,0,1] T=[0.2,0.2,-0.2,0,0] T,则||DX-h|| 2=0.12。如果我们设定ε=0.2,则||DX-h|| 2≤ε依然成立。引入ε就是为了降低“噪音”的影响。
需要说明的是,上述的L1范式||·|| 1的运算是:范式中变量所包含的每个元素的绝对值之和,例如,若X=[α 11,α 12,......,α MK],则||X|| 1=|α 11|+|α 12|+...+|α MK|。而上述的L2范式||·|| 2的运算是:范式中变量所包含的每个元素的平方之和的平方根,例如,若X=[α 11,α 12,......,α MK],则
Figure PCTCN2018087986-appb-000007
上述采用第一预定条件和第二预定条件来确定相关系数向量X *的过程中采用的是稀疏解法,即在一定精度的条件下使用最少的病例去重构患者h的病情特征,采用稀疏解法能够降低“噪音”数据的影响,使得上述的关系模型h=DX具有良好的鲁棒性。
示例性的,当步骤102b中的可能性用概率来表示时,上述的步骤102b可以包括如下步骤:
步骤102b1、从所述确定的相关系数向量X *中确定出所述病例数据库中每个病种的相关系数向量δ i(X *),δ i(X *)表示通过将系数向量X *中属于病种D i的维度乘以1,其余维度乘以0所得到的一个向量。也就是说,将X *中第i个病种下各病例的相关系数保留,其他元素置为0,得到δ i(X *)。
示例性的,假设所确定的相关系数向量X *=[α A1,α A2,α B1,α B2,α B3,α C1,α C2] T,则病种A的相关系数向量为:δ A(X *)=[α A1,α A2,0,0,0,0,0] T;病种B的相关系数向量为:δ B(X *)=[0,0,α B1,α B2,α B3,0,0] T;病种C的相关系数向量为:δ C(X *)=[0,0,0,0,0,α C1,α C2] T
步骤102b2、根据所述病例数据库中每个病种的相关系数向量δ i(X *)确定所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *)。
步骤102b3、将所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *)代入到概率计算公式中,得到患者患病例数据库中每个病种的概率。
示例性的,概率计算公式为:
Figure PCTCN2018087986-appb-000008
其中,C i表示患者患病例数据库中的病种i的概率,
Figure PCTCN2018087986-appb-000009
中的M表示病例数据库中的M个病种,η为误差矩阵,h=h 1+h 2+......+h M+η,其中:h为新来患者的病情特征分布向量,h i为所述患者的病情特征分布向量h中对应各病种i的语义成分(1≤i≤M),而该h i是由病种i下的所有病例的病情特征分布向量的线性组合。
Figure PCTCN2018087986-appb-000010
是L2范式的平方。
具体的,对于上述的h i=D*δ i(X *),仍然以上文所列举的例子进行说明。假设病例数据库中有3个病种,分别为病种A、病种B以及 病种C,该病种A包含2个病例,病种B包含3个病例,病种C包含2个病例,则D=[D A1,D A2,D B1,D B2,D B3,D C1,D C2]。假设基于上述的3个病种的7个病例统计出的病情特征数据有100个,且假设所确定的相关系数向量为X *=[α A1,α A2,α B1,α B2,α B3,α C1,α C2] T
基于上面的内容,考虑到上述的3个病种的7个病例统计出的病情特征数据有100个,则对应的D为100*7的矩阵,所确定出h A=D*δ A(X *)中的δ A(X *)为7维列向量,δ A(X *)=[α A1,α A2,0,0,0,0,0] T;h B=D*δ B(X *)中的δ B(X *)为7维列向量,δ B(X *)=[0,0,α B1,α B2,α B3,0,0] T;h C=D*δ C(X *)中的δ C(X *)为7维列向量,δ C(X *)=[0,0,0,0,0,α C1,α C2] T。这样上述的h A、h B以及h C中的矩阵运算才满足矩阵乘法的准则。然后,将的h A、h B以及h C的内容带入到上面的公式1中可以得到患者患病种A、病种B以及病种C的概率。
示例性的,通过将h i代入公式1可以得到C=[C 1,C 2,...,C M,C η,],由上述的公式1中可以得知C i满足C 1+C 2+...+C M+C η=1,其中,C η的计算公式如下:
Figure PCTCN2018087986-appb-000011
通过上述的公式1和公式2可以得知,C i反映了患者h属于病种D i可能性的大小(C η反映了患者h不属于前面任一病种D 1-D M的可能性)。这是因为C i越大,表明构成患者h的病情特征分布向量中包含属于病种 D i的病例越多,即患者h位于D i语义子空间的部分越多,则属于病种D i的可能性越大。图2示出了根据本公开一个实施例提供的一种语义空间的示意图。例如在图2中,假设已知病例空间中一共有三个病种或子空间,分别对应不同的形状:四角星、三角形、六角星,分别对应C的前三个维度,最后一个维度为误差C η。圆形节点表示新来的患者。图2所示的两个圆圈分别表示用来表示新来患者的两种线性组合。第一种仅仅用四角星节点所代表的子空间就可以表示新来患者,第二种需要用所有的三个子空间来表示新来的患者。可以清楚地看到,对于左边的图有C=[1,0,0,0],即患者可能处于四角星所代表的子空间。对于右边的图有C=[0.25,0.375,0.375,0],则很难分析患者处于哪个子空间或属于哪个病种。
步骤103、根据患者患病例数据库中每个病种的可能性,输出患者的分诊结果。
其中,上述的分诊结果可以包括为患者分配的科室、进一步还可以包括分诊流程、为患者所分配的医生、以及可参考的治疗指南等。
示例性的,上述的步骤103可以采用以下任一种方式实现:
方式A、输出所有病种中可能性最大的病种作为患者的分诊结果。例如,计算患者患每个病种的可能性大小,且所确定出患者患病种A的可能性最大,则在分诊系统的界面上显示出患者患病种A所对应的分诊结果。
方式B、将所有病种中可能性不为零的病种,按照可能性大小顺序输出作为患者的分诊结果。例如,计算出患者患每个病种的可能性,并将所有病种中可能性不为零的病种按可能性从大到小的顺序排序,则在分诊系统的界面上从大到小依次显示出患者患每个病种的可能性的分诊结果。
在以上实施例提供的分诊方法中,首先,通过获取患者的病情特征数据;然后,根据患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定患者患病例数据库中每个病种的可能性;最后,根据患者患病例数据库中每个病种的可能性,输出患者的分诊结果,从而实现了对患者的智能分诊,以减少医院的分诊压力。
下面将基于图1的分诊方法的实施例中的相关描述对本公开一些实施例提供的一种分诊装置进行介绍。以下实施例中与上述实施例相关的技术术语、概念等的说明可以参照上述的实施例,这里不再赘述。
图3示出了根据本公开一个实施例提供的一种分诊装置的结构示意图。如图3所示,该装置可以包括:获取器31、处理器32以及输出器33,其中:
获取器31可以配置成获取患者的病情特征数据。
处理器32可以配置成根据患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定患者患所述病例数据库中每个病种的可能性。
输出器33可以配置成根据患者患病例数据库中每个病种的可能性,输出患者的分诊结果。
示例性的,该患者的病情特征数据可以是病情特征文本,也可以是表示该病情特征文本的数据。
示例性的,上述的病情特征数据可以包括:病情症状信息和/或体征检测参数,其中,病情症状信息可以为观察到患者的症状或患者感受到的症状,例如可以是患者的口述症状或输入的症状文本等,例如:心悸气短、肢体麻木、耳鸣等。而体征检测参数可以包括患者的各项指标检测值,例如血压值、血糖值等,其反映出的病情特征可以是血压微高、血压过高等。
示例性的,上述的患者患病例数据库中每个病种的可能性可以是指患者患每个病种的概率,用0至1间的数值进行表示。或者可以是是患者患每个病种的可能性对应的数值(可以为包含大于1的数值),数值越大表示可能性越大。
示例性的,上述的病情特征数据为病情特征文本时,例如:患者的病情特征文本为眩晕、恶心以及心悸气短;上述的处理器32确定患者患病例数据库中每个病种的可能性的具体过程参照以下内容:这里病例数据库中的病种个数以3个为例,分别为病种A、病种B以及病种C,其中:病种A以包含3个病例为例,病种B以包含4个病例为例,病种C以包含5个病例为例,而患者所具有的病情特征以3个为例。将患者的病情特征与病例数据库中的每个病种下各病例中的病情特征文本进行匹配,若患者的3个病情特征均出现在病例数据库中病 种A下的同一个病例中,且该患者的3个病情特征没有全部出现在其他病种下的病情特征文本中,则该患者患病种A的可能性最大;若患者的2个病情特征出现在病例数据库中病种A下的第一病例中,剩下的1个病情特征出现在病种A下的第二病例中,且该患者的3个病情特征没有全部出现在其他病种下的病情特征文本中,则该患者患病种A的可能性最大,但是相对于上面的结果较小。若患者的第1个病情特征出现在病例数据库中病种A下的第一病例中,第2个病情特征出现在病种A下的第二病例中,第3个病情特征出现在病种A下的第三病例中,且该患者的3个病情特征没有全部出现在其他病种下的病情特征文本中,则该患者患病种A的可能性最大,但是相对于上面的两种结果是最小的。
示例性的,上述的获取器31可以进一步配置成:
根据预设的向量元素的位置与元素所表示的病情特征的对应关系,将输入的患者的病情特征转换成患者的病情特征分布向量,患者的病情特征分布向量中的每个元素指示输入的患者的病情特征中是否出现该元素所在位置对应的病情特征。
具体的,假设数据库中有Q个病情特征,该预设的向量元素的位置与向量元素所表示的病情特征的对应关系用集合I表示,该集合I为Q维向量,I=[I 1,I 2,...,I Q] T。其中,I j(1≤j≤Q)表示第j个位置处的病情特征,从而集合I表示的是从第1个位置处的病情特征到第Q个位置处的病情特征。
示例的,假设数据库中有1000个病情特征,上述的集合I为:I=[I 1,I 2,...,I 1000] T。其中,I 500为第500个位置处的病情特征,从而集合I表示从第1个位置处的病情特征到第1000个位置处的病情特征。
示例性的,上述的处理器32具体可以进一步配置成:
将所述患者的病情特征数据代入到关系模型h=DX中,通过对所述关系模型求解X来确定满足预定条件的相关系数向量X *,其中,h为所述患者的病情特征分布向量,h和X均为列向量,D为由所述病例数据库中每个病种下各病例的病情特征分布向量组成的矩阵,D=[D 1, D 2,......,D i,......,D M],其中:D i=[D i,1,D i,2,...,D i,j,...,D i,K],D i,j为所述病例数据库中病种i的第j个病例的病例病情特征分布向量,所述K表示所述病例数据库中病种i包括K个病例,所述M表示病例数据库中包括M种疾病;
根据所述确定的相关系数向量X *,确定所述病例数据库中每个病种的可能性
在一些实施例中,如图3所示,该装置例如还可以包括转换器34,其中:转换器34可以配置成根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将每个病种下各病例的病情特征转换成病例病情特征分布向量。其中,上述的病例的病情特征分布向量中的每个元素指示病例的病情特征中是否出现该元素所在位置对应的病情特征。
具体的,若病例数据库中有Q个病情特征,则病例数据库中每个病种下各病例的病情特征分布向量中包含Q个元素。
示例性的,病种A下的任一个病例的病情特征分布向量为D A1=[I 1,I 2,...,I Q,] T。由于数据库中的病情特征的集合I为Q维向量,相应的这里的病种A下的任一个病例的病情特征分布向量也为0维向量。其中,I j(1≤j≤Q)表示病种A下的任一个病例的第j个位置处的病情特征,从而D A1表示的病种A下的任一个病例中从第1个位置处的病情特征到第Q个位置处的病情特征。
示例性的,上述的关系模型可以是预先建立好的,也可以是根据需要实时进行建立的,对于上述的关系模型的建立过程具体可以参考方法部分的内容,这里不再详细赘述。
示例性的,所述预定条件可以包括第一预定条件||DX-h|| 2≤ε以及第二预定条件X *=arg min||X|| 1,其中||·|| 1是L1范式,||·|| 2是L2范式,所述ε为预设参数,正如之前所提到的,引入ε是为了降低“噪声”的影 响,该参数的值取决于实现所需要达到的精度,X *=arg min||X|| 1表示使得||X|| 1取最小值时的X,记为X *
需要说明的是,上述的L1范式||·|| 1的运算是:范式中变量所包含的每个元素的绝对值之和,例如,若X=[α 11,α 12,......,α MK],则||X|| 1=|α 11|+|α 12|+...+|α MK|。而上述的L2范式||·|| 2的运算是:范式中变量所包含的每个元素的平方之和的平方根,例如,若X=[α 11,α 12,......,α MK],则
Figure PCTCN2018087986-appb-000012
上述的处理器32在采用第一预定条件和第二预定来确定相关系数向量X *的过程中所采用的是稀疏解法,即在一定精度条件下使用最少的病例去重构患者h的病情特征,采用稀疏解法能够降低“噪音”数据的影响,使得上述的关系模型h=DX具有良好的鲁棒性。
示例性的,当上述的处理器32在根据确定的相关系数向量X *,确定病例数据库中每个病种的可能性用概率来表示时,该处理器32可以进一步配置成:
从所述确定的相关系数向量X *中确定出所述病例数据库中每个病种的相关系数向量δ i(X *),δ i(X *)表示通过将系数向量X *中属于病种D i的维度乘以1,其余维度乘以0所得到的一个向量。也就是说,将X *中第i个病种下各病例的相关系数保留,其他元素置为0,得到δ i(X *)。
示例性的,假设所确定的相关系数向量X *=[α A1,α A2,α B1,α B2,α B3,α C1,α C2] T,则病种A的相关系数向量为:δ A(X *)=[α A1,α A2,0,0,0,0,0] T;病种B的相关系数向量为:δ B(X *)=[0,0,α B1,α B2,α B3,0,0] T;病种C的相关系数向量为: δ C(X *)=[0,0,0,0,0,α C1,α C2] T
该处理器32可以进一步配置成根据所述病例数据库中每个病种的相关系数向量δ i(X *)确定所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *)。
该处理器32还可以进一步配置成将所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *)代入到概率计算公式中,得到患者患病例数据库中每个病种的概率。
示例性的,上述的概率计算公式为:
Figure PCTCN2018087986-appb-000013
其中,C i表示患者患病例数据库中的病种i的概率,
Figure PCTCN2018087986-appb-000014
中的M表示病例数据库中的M个病种,η为误差矩阵,该h=h 1+h 2+.......+h M+η,其中:h为新来患者的病情特征分布向量,h i为所述患者的病情特征分布向量h中对应各病种i的语义成分(1≤i≤M),而该h i是由病种i下的所有病例的病情特征分布向量的线性组合。
Figure PCTCN2018087986-appb-000015
是L2范式的平方。
示例性的,通过将h i代入公式1可以得到C=[C 1,C 2,...,C M,C η,],,由上述的公式1中可以得知C i满足C 1+C 2+...+C M+C η=1,其中,C η的计算公式如下:
Figure PCTCN2018087986-appb-000016
通过上述的公式1和公式2可以得知,C i反映了患者h属于病种D i 可能性的大小(C η反映了患者h不属于前面任一病种D 1-D M的可能性)。这是因为C i越大,表明构成患者h的病情特征分布向量中包含属于病种D i的病例越多,即患者h位于D i语义子空间的部分越多,则属于病种D i的可能性越大。例如在图2中,假设已知病例空间中一共有三个病种或子空间,分别对应不同的形状:四角星、三角形、六角星,分别对应C的前三个维度,最后一个维度为误差C η。圆形节点表示新来的患者。图2所示的两个圆圈分别表示用来表示新来患者的两种线性组合。第一种仅仅用四角星节点所代表的子空间就可以表示新来患者,第二种需要用所有的三个子空间来表示新来的患者。可以清楚地看到,对于左边的图有C=[1,0,0,0],即患者可能处于四角星所代表的子空间。对于右边的图有C=[0.25,0.375,0.375,0],则很难分析患者处于哪个子空间或属于哪个病种。
示例性的,上述的分诊结果包括可参考的治疗指南、分诊流程以及所涉及的科室等信息。
示例性的,上述的输出器33可以进一步配置成采用以下任一种方式实现:
方式A、输出所有病种中可能性最大的病种作为患者的分诊结果。例如,计算患者患每个病种的可能性大小,且所确定出患者患病种A的可能性最大,则在分诊系统的界面上显示出患者患病种A所对应的分诊结果。
方式B、将所有病种中可能性不为零的病种,按照可能性大小顺序输出作为患者的分诊结果。例如,计算出患者患每个病种的可能性,并将所有病种中可能性不为零的病种按可能性从大到小的顺序排序,则在分诊系统的界面上从大到小依次显示出患者患每个病种的可能性的分诊结果。
在上述实施例提供的分诊装置中,首先,该装置通过获取患者的病情特征数据;然后,根据患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定患者患病例数据库中每个病 种的可能性;最后,根据患者患病例数据库中每个病种的可能性,输出患者的分诊结果,从而实现了对患者的智能分诊,以减少医院的分诊压力。
根据本公开的另一个方面,提供了一种分诊设备400。图4示出了该分诊设备400的结构示意图。如图4所示,该分诊设备400可以包括:一个或多个处理器401;和与处理器401连接的存储器402,其上存储有计算机可执行指令,所述计算机可执行指令被配置为当被所述一个或多个处理器执行时,执行如上所述的任何一种方法的一个或多个步骤。
分诊设备400可以实现为本地计算的计算机产品结构,即在用户侧实现上述实施例所描述的分诊方法;也可以实现为本地和远端交互的计算机产品结构,即在用户侧的终端实现上述实施例所描述的分诊方法的部分步骤,例如病情特征数据的输入、分诊结果的输出;在与用户侧终端连接的网络端实现上述实施例所描述的分诊方法的其它步骤,例如患者患所述病例数据库中每个病种的可能性的计算。
在本公开的一些实施例中,分诊设备的一个或多个处理器可以位于同一计算机产品或不同的计算机产品。例如,部分处理器可以位于用户侧的计算机产品,部分处理器位于远端或云端的服务器的计算机产品,以分别执行分诊方法的部分步骤。
处理器401可以是中央处理单元(CPU)或者现场可编程逻辑阵列(FPGA)或者单片机(MCU)或者数字信号处理器(DSP)或者专用集成电路(ASIC)等具有数据处理能力和/或程序执行能力的逻辑运算器件。
存储器402可以是各种由任何类型的易失性或非易失性存储设备或者它们的组合实现,如静态随机存取存储器(SRAM),电可擦除可编程只读存储器(EEPROM),可擦除可编程只读存储器(EPROM),可编程只读存储器(PROM),只读存储器(ROM),磁存储器,快闪存储器,磁盘或光盘。
计算机指令包括了一个或多个由对应于处理器的指令集架构定义的处理器操作,这些计算机指令可以被一个或多个计算机程序在逻辑上包含和表示。
该分诊设备400还可以连接各种输入设备403(例如用户界面、键 盘等)、各种输出设备404(例如扬声器等)、以及显示设备405等实现计算机产品与其它产品或用户的交互,本文在此不再赘述。
连接可以是通过网络模块406连接,例如无线网络、有线网络、和/或无线网络和有线网络的任意组合。网络可以包括局域网、互联网、电信网、基于互联网和/或电信网的物联网(Internet of Things)、和/或以上网络的任意组合等。有线网络例如可以采用双绞线、同轴电缆或光纤传输等方式进行通信,无线网络例如可以采用3G/4G/5G移动通信网络、蓝牙、Zigbee或者Wi-Fi等通信方式。
根据本公开的另一个方面,还提供了一种计算机可读存储介质,其上包含有计算机可执行指令,所述指令在由一个或多个处理器执行时,使所述一个或多个处理器执行如上所述的任何一种方法的一个或多个步骤。
上述分诊设备和计算机可读存储介质的具体实现方式可以参照前面针对分诊方法的描述,在此不再赘述。本领域普通技术人员能够理解,上述分诊设备和计算机可读存储介质同样能够实现对患者的智能分诊,可以减少医院的分诊压力。
通过以上的实施方式的描述,所属领域的技术人员可以清楚地了解到,为描述的方便和简洁,仅以上述各功能器的划分进行举例说明,实际应用中,可以根据需要而将上述功能分配由不同的功能器完成,即将装置的内部结构划分成不同的功能器,以完成以上描述的全部或者部分功能。上述描述的系统,装置和单元的具体工作过程,可以参考前述方法实施例中的对应过程,在此不再赘述。
在本申请所提供的几个实施例中,应该理解到,所揭露的分诊装置,可以通过其它的方式实现。例如,以上所描述的装置的实施例仅仅是示意性的,例如,所述模块或单元的划分,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,例如多个单元或组件可以结合或者可以集成到另一个系统,或一些特征可以忽略,或不执行。另一点,所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口,装置或单元的间接耦合或通信连接,可以是电性,机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的,作为单元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单元来实现本实施例方案的目的。
另外,在本公开各个实施例中的各功能单元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,也可以采用软件功能单元的形式实现。
所述集成的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,可以存储在一个计算机可读取存储介质中。基于这样的理解,本公开的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)或处理器(processor)执行本公开各个实施例所述方法的全部或部分步骤。而前述的存储介质包括:U盘、移动硬盘、只读存储器(ROM,Read-Only Memory)、随机存取存储器(RAM,Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
可以理解的是,以上所述仅为本公开的示例性实施方式,但本公开的保护范围并不局限于此。应当指出的是,在不脱离本公开的精神和原理的前提下,本领域的普通技术人员可轻易想到各种变化或替换,这些变化或替换都应涵盖在本公开的保护范围之内。因此,本公开的保护范围应以所附权利要求的保护范围为准。
在权利要求书中,任何置于括号中的附图标记都不应当解释为限制权利要求。术语“包括”并不排除除了权利要求中所列出的元件或步骤之外的元件或步骤的存在。元件前的词语“一”或“一个”并不排除存在多个这样的元件。本公开可以借助于包括若干分离元件的硬件来实现,也可以通过适当编程的软件或固件来实现,或者通过它们的任意组合来实现。
在列举了若干装置的设备或系统权利要求中,这些装置中的一个或多个能够在同一个硬件项目中体现。仅仅某些措施记载在相互不同的从属权利要求中这个事实并不表明这些措施的组合不能被有利地使用。

Claims (20)

  1. 一种分诊方法,所述方法包括:
    获取患者的病情特征数据;
    根据所述患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定所述患者患所述病例数据库中每个病种的可能性;
    根据所述患者患所述病例数据库中每个病种的可能性,输出所述患者的分诊结果。
  2. 根据权利要求1所述的方法,其中,所述病情特征数据包括病情症状信息和体征检测参数的至少一者。
  3. 根据权利要求1或2所述的方法,其中,所述获取患者的病情特征数据的步骤进一步包括:
    根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将输入的患者的病情特征转换成患者的病情特征分布向量,所述患者的病情特征分布向量中的每个元素指示所述输入的患者的病情特征中是否出现该元素所在位置对应的病情特征。
  4. 根据权利要求3所述的方法,其中,所述根据所述患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定所述患者患所述病例数据库中每个病种的可能性的步骤进一步包括:
    基于患者的病情特征分布向量,通过对关系模型h=DX求解X来确定满足预定条件的相关系数向量X *,其中,h为所述患者的病情特征分布向量,h和X均为列向量,D为由所述病例数据库中每个病种下各病例的病情特征分布向量组成的矩阵,D=[D 1,D 2,......,D i,......,D M],其中:D i=[D i,1,D i,2,…,D i,j,…,Di ,K],D i,j为所述病例数据库中病种i的第j个病例的病例病情特征分布向量,所述K表示所述病例数据库中病种i包括K个病例,所述M表示病例数据库中包括M种疾病;
    根据所述确定的相关系数向量X *,确定所述病例数据库中每个病种的可能性。
  5. 根据权利要求4所述的方法,其中,所述方法还包括:
    根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将每个病种下各病例的病情特征转换成病例病情特征分布向量;所述病例的病情特征分布向量中的每个元素指示所述病例的病情特征中是否出现该元素所在位置对应的病情特征。
  6. 根据权利要求4所述的方法,其中,所述预定条件包括第一预定条件||DX-h|| 2≤ε以及第二预定条件X *=arg min||X|| 1,其中:||·|| 1是L1范式,||·|| 2是L2范式,所述ε为预设参数。
  7. 根据权利要求4所述的方法,其中,根据所述确定的相关系数向量X *确定所述病例数据库中每个病种的可能性的步骤进一步包括:
    从所述确定的相关系数向量X *中确定出所述病例数据库中每个病种的相关系数向量δ i(X *),δ i(X *)表示通过将系数向量X *中属于病种Di的维度乘以1,其余维度乘以0所得到的一个向量;
    根据所述病例数据库中每个病种的相关系数向量δ i(X *)确定所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *);
    将所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *)代入到概率计算公式中,得到所述患者患所述病例数据库中每个病种的概率;
    其中,所述概率计算公式为:
    Figure PCTCN2018087986-appb-100001
    C i表示所述患者患所述病例数据库中的病种i的概率,
    Figure PCTCN2018087986-appb-100002
    中的M表示所述病例数据库中的M个病种,所述η为误差向量,所述
    Figure PCTCN2018087986-appb-100003
    是L2范式的平方。
  8. 根据权利要求1所述的方法,其中,所述根据所述病例数据库 中每个病种的可能性,输出所述患者的分诊结果的步骤进一步包括:将所有病种中可能性不为零的病种,按照可能性大小顺序输出作为所述患者的分诊结果。
  9. 根据权利要求8所述的方法,其中,输出所有病种中可能性最大的病种作为所述患者的分诊结果。
  10. 一种分诊装置,包括:
    获取器,配置成获取患者的病情特征数据;
    处理器,配置成根据所述患者的病情特征数据、以及病例数据库中所有病种下各病例的病情特征数据,确定所述患者患所述病例数据库中每个病种的可能性;
    输出器,配置成根据所述患者患所述病例数据库中每个病种的可能性,输出所述患者的分诊结果。
  11. 根据权利要求10所述的装置,其中,所述病情特征数据包括病情症状信息和体征检测参数的至少一者。
  12. 根据权利要求10或11所述的装置,其中,所述获取器进一步配置成:
    根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将输入的患者的病情特征转换成患者的病情特征分布向量,所述患者的病情特征分布向量中的每个元素指示所述输入的患者的病情特征中是否出现该元素所在位置对应的病情特征。
  13. 根据权利要求12所述的装置,其中,所述处理器进一步配置成:
    基于患者的病情特征分布向量,通过对关系模型h=DX求解X来确定满足预定条件的相关系数向量X *,其中,h为所述患者的病情特征分布向量,h和X均为列向量,D为由所述病例数据库中每个病种下各病例的病情特征分布向量组成的矩阵,D=[D 1,D 2,......,D i,......,D M],其中:D i=[D i,1,D i,2,…,D i,j,…,D i,K],D i,j为所述病例数据库中病种i的第j个病例的病例病情特征分布向量,所述K表示所述病例数据库中病种i包括K个病例,所述M表示病例数据库中包括M种疾病;
    根据所述确定的相关系数向量X *,确定所述病例数据库中每个病种的可能性。
  14. 根据权利要求13所述的装置,还包括转换器,该转换器配置成根据预设的向量元素的位置与向量元素所表示的病情特征的对应关系,将每个病种下各病例的病情特征转换成病例病情特征分布向量,所述病例病情特征分布向量中的每个元素指示病例的病情特征中是否出现该元素所在位置对应的病情特征。
  15. 根据权利要求13所述的装置,其中,所述预定条件包括第一预定条件||DX-h|| 2≤ε以及第二预定条件X *=arg min||X|| 1,其中:||·|| 1是L1范式,||·|| 2是L2范式,所述ε为预设参数。
  16. 根据权利要求13所述的装置,其中,所述处理器在根据所述确定的相关系数向量X *,确定所述病例数据库中每个病种的可能性时,进一步配置成:
    从所述确定的相关系数向量X *中确定出所述病例数据库中每个病种的相关系数向量δ i(X *),δ i(X *)表示通过将系数向量X *中属于病种Di的维度乘以1,其余维度乘以0所得到的一个向量;
    根据所述病例数据库中每个病种的相关系数向量δ i(X *)确定所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *);
    将所述患者的病情特征分布向量h中对应每个病种的语义成分h i=D*δ i(X *)代入到概率计算公式中,得到所述患者患所述病例数据库中每个病种的概率;
    其中,所述概率计算公式为:
    Figure PCTCN2018087986-appb-100004
    C i表示所述患者患所述病例数据库中的病种i的概率,
    Figure PCTCN2018087986-appb-100005
    中的M表示所述病例数据库中的M个病种,所述η为误差向量,所述
    Figure PCTCN2018087986-appb-100006
    是L2范式的平方。
  17. 根据权利要求10所述的装置,其中,所述输出器进一步配置成:
    将所有病种中可能性不为零的病种,按照可能性大小顺序输出作为所述患者的分诊结果。
  18. 根据权利要求17所述的装置,其中,所述输出器进一步配置成:输出所有病种中可能性最大的病种作为所述患者的分诊结果。
  19. 一种分诊设备,包括:
    一个或多个处理器;和
    存储器,其上存储有计算机可执行指令,所述计算机可执行指令配置成当被所述一个或多个处理器执行时,执行如权利要求1-9中任何一项所述的方法的一个或多个步骤。
  20. 一种计算机可读存储介质,其上包含有计算机可执行指令,所述指令在由一个或多个处理器执行时,使所述一个或多个处理器执行如权利要求1-9中任何一项所述的方法的一个或多个步骤。
PCT/CN2018/087986 2017-06-13 2018-05-23 分诊方法、装置和设备以及计算机可读存储介质 WO2018228151A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710444355.3 2017-06-13
CN201710444355.3A CN107153775B (zh) 2017-06-13 2017-06-13 一种智能分诊方法及装置

Publications (1)

Publication Number Publication Date
WO2018228151A1 true WO2018228151A1 (zh) 2018-12-20

Family

ID=59795730

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/087986 WO2018228151A1 (zh) 2017-06-13 2018-05-23 分诊方法、装置和设备以及计算机可读存储介质

Country Status (2)

Country Link
CN (1) CN107153775B (zh)
WO (1) WO2018228151A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3572959A4 (en) * 2017-01-19 2021-01-06 BOE Technology Group Co., Ltd. DATA ANALYSIS METHOD AND DEVICE

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107153775B (zh) * 2017-06-13 2020-03-10 京东方科技集团股份有限公司 一种智能分诊方法及装置
CN107910073A (zh) * 2017-12-21 2018-04-13 苏州麦迪斯顿医疗科技股份有限公司 一种急诊预检分诊方法及装置
CN108399945A (zh) * 2018-02-10 2018-08-14 武汉大学中南医院 一种门急诊智能分诊方法和系统
CN108831544A (zh) * 2018-06-27 2018-11-16 中国人民解放军陆军军医大学第三附属医院(野战外科研究所) 一种门诊自动分诊方法及系统
CN110619959A (zh) * 2019-08-09 2019-12-27 贵阳叁玖互联网医疗有限公司 一种智能分诊的方法及系统
CN116631597B (zh) * 2023-07-24 2024-01-16 深圳捷工智能电气股份有限公司 一种移动端、医生端、护士端就近身份信息比对确认方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184315A (zh) * 2011-04-02 2011-09-14 中国医学科学院医学信息研究所 基于诊断要素分析的科室分诊系统
CN104376409A (zh) * 2014-11-07 2015-02-25 深圳市前海安测信息技术有限公司 一种基于网络医院的分诊数据处理方法及系统
CN105023073A (zh) * 2014-04-16 2015-11-04 北京百和园科技有限公司 一种基于人工神经网络的医院智能评估分诊系统
CN107153775A (zh) * 2017-06-13 2017-09-12 京东方科技集团股份有限公司 一种智能分诊方法及装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102156812A (zh) * 2011-04-02 2011-08-17 中国医学科学院医学信息研究所 基于症状相似度分析的就医辅助决策方法
CN103855707B (zh) * 2014-02-20 2017-04-19 深圳供电局有限公司 一种含分布式电源配电网的供电可靠性评估方法
CN105653859A (zh) * 2015-12-31 2016-06-08 遵义医学院 一种基于医疗大数据的疾病自动辅助诊断系统及方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102184315A (zh) * 2011-04-02 2011-09-14 中国医学科学院医学信息研究所 基于诊断要素分析的科室分诊系统
CN105023073A (zh) * 2014-04-16 2015-11-04 北京百和园科技有限公司 一种基于人工神经网络的医院智能评估分诊系统
CN104376409A (zh) * 2014-11-07 2015-02-25 深圳市前海安测信息技术有限公司 一种基于网络医院的分诊数据处理方法及系统
CN107153775A (zh) * 2017-06-13 2017-09-12 京东方科技集团股份有限公司 一种智能分诊方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3572959A4 (en) * 2017-01-19 2021-01-06 BOE Technology Group Co., Ltd. DATA ANALYSIS METHOD AND DEVICE
US11195114B2 (en) 2017-01-19 2021-12-07 Boe Technology Group Co., Ltd. Medical data analysis method and device as well as computer-readable storage medium

Also Published As

Publication number Publication date
CN107153775A (zh) 2017-09-12
CN107153775B (zh) 2020-03-10

Similar Documents

Publication Publication Date Title
WO2018228151A1 (zh) 分诊方法、装置和设备以及计算机可读存储介质
US11488694B2 (en) Method and system for predicting patient outcomes using multi-modal input with missing data modalities
CN111414393B (zh) 一种基于医学知识图谱的语义相似病例检索方法及设备
CN111292853B (zh) 基于多参数的心血管疾病风险预测网络模型及其构建方法
US7809660B2 (en) System and method to optimize control cohorts using clustering algorithms
US20170147777A1 (en) Method and apparatus for predicting health data value through generation of health data pattern
US11195114B2 (en) Medical data analysis method and device as well as computer-readable storage medium
US20170147753A1 (en) Method for searching for similar case of multi-dimensional health data and apparatus for the same
US11875277B2 (en) Learning and applying contextual similiarities between entities
CN107908819B (zh) 预测用户状态变化的方法和装置
JP7357614B2 (ja) 機械支援対話システム、ならびに病状問診装置およびその方法
WO2021139116A1 (zh) 相似患者智能分群方法、装置、设备和存储介质
US20160147960A1 (en) Apparatus and method for providing customized personal health service
US20190034810A1 (en) Data processing method, medical term processing system and medical diagnostic system
WO2022227203A1 (zh) 基于对话表征的分诊方法、装置、设备及存储介质
Fati et al. A continuous cuffless blood pressure estimation using tree-based pipeline optimization tool
CN113343696A (zh) 电子病历命名实体的识别方法、装置、远程终端及系统
Guangming et al. Nonparametric statistical inference for p (x< y< z)
CN113488157B (zh) 智能导诊处理方法、装置、电子设备及存储介质
WO2020132918A1 (zh) 药品预测方法、装置、计算机设备及存储介质
US11238988B2 (en) Large scale identification and analysis of population health risks
CN111986815B (zh) 基于共现关系的项目组合挖掘方法及相关设备
Islam et al. Distinguishing l and h phenotypes of Covid-19 using a single X-ray image
CN112435757A (zh) 一种急性肝炎的预测装置及系统
Hao et al. Optimal minimax variable selection for large-scale matrix linear regression model

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: 18817371

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 07/04/2020)

122 Ep: pct application non-entry in european phase

Ref document number: 18817371

Country of ref document: EP

Kind code of ref document: A1