CN112820371B - Health recommendation system and method based on medical knowledge map - Google Patents

Health recommendation system and method based on medical knowledge map Download PDF

Info

Publication number
CN112820371B
CN112820371B CN202110437769.XA CN202110437769A CN112820371B CN 112820371 B CN112820371 B CN 112820371B CN 202110437769 A CN202110437769 A CN 202110437769A CN 112820371 B CN112820371 B CN 112820371B
Authority
CN
China
Prior art keywords
abnormal
user
health
afx
association
Prior art date
Legal status (The legal status 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 status listed.)
Active
Application number
CN202110437769.XA
Other languages
Chinese (zh)
Other versions
CN112820371A (en
Inventor
裘实
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Health Hope (beijing) Technology Co ltd
Original Assignee
Health Hope (beijing) Technology Co ltd
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 Health Hope (beijing) Technology Co ltd filed Critical Health Hope (beijing) Technology Co ltd
Priority to CN202110437769.XA priority Critical patent/CN112820371B/en
Publication of CN112820371A publication Critical patent/CN112820371A/en
Application granted granted Critical
Publication of CN112820371B publication Critical patent/CN112820371B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment

Landscapes

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

Abstract

The application provides a health recommendation system based on a medical knowledge graph, which comprises an abnormal physical sign library; the abnormal sign library comprises medical abnormal data and time interval abnormal data; the user health portrait can be established through the medical abnormal data and the time interval abnormal data. The present application further provides a health recommendation method, including: obtaining an abnormal category feature label of a user a; acquiring the occurrence time of an i-th class abnormal feature tag AXi, setting the association time period of the i-th class abnormal feature tag AXi as delta ATi, wherein each association time period of the AXi comprises each occurrence time of the AXi; acquiring abnormal category characteristics AF occurring in all associated time periods of the abnormal category characteristics AXi; recording all abnormal category characteristics AF occurring in the association time quantum delta ATi of the abnormal category characteristics AXi as preset association characteristics of the AXi, and constructing a user health portrait according to the associated abnormal category characteristics. The present application provides an optimized health recommendation system.

Description

Health recommendation system and method based on medical knowledge map
Technical Field
The application relates to the field of user health portraits, in particular to a health recommendation system and a health recommendation method based on a medical knowledge graph.
Background
In the current continuous development of economy, the aging of social population is getting worse, the senile diseases are prevalent, and the chronic diseases suffered by young and strong people are gradually propagated under the larger living pressure, so that the statistics, abnormal sign symptoms and tracking of users are facilitated, and a personal health medical record report aiming at each user is gradually formed in the prior art, and the information sharing is realized through the internet and big data; however, most of data in the existing personal health medical record report only relates to physical sign parameters of patient advocacy, and information such as abnormal data, living conditions, work, diet, motion and the like found in daily life of a patient is less acquired and not updated timely, so that a user cannot be subjected to multi-dimensional real-time portrait in all aspects, and therefore the multi-dimensional information of the patient cannot be timely obtained, and treatment timeliness reminding and effectiveness of the user are affected because effective treatment and health care measures cannot be formed; in the prior art, the standard physical characteristics of the same type of patient users are often adopted as the reference for evaluating and predicting the health state of the users, but the standard physical characteristics of the current users cannot be used as the reference, so that the guidance and intervention measures which are most suitable for the patients of the current users cannot be effectively provided for the current users, and the requirements of individuals, enterprises and health supervision units cannot be met.
Disclosure of Invention
In order to solve the above problems, the present application provides a recommendation method for a health recommendation system based on a medical knowledge graph, including:
step S00, providing a health recommendation system based on a medical knowledge graph, which comprises a user health information base, wherein the user health information base comprises an abnormal physical sign base; the abnormal sign library comprises medical abnormal data and time interval abnormal data; the medical abnormal data comprise abnormal conditions and occurrence time of which numerical values are beyond normal indexes in the detection result of the body of the user through the medical instrument, and the time interval abnormal data comprise the abnormal conditions and the occurrence time obtained when the user checks himself in life; a user health portrait can be established through medical abnormal data and time interval abnormal data, wherein the user health portrait comprises user personal information and an abnormal category characteristic label of an associated abnormal condition; setting a correlation time period of the abnormal category feature tags according to the time of the abnormal condition, wherein the abnormal category feature tags with correlation occur in the same correlation time period; the user health portrait further comprises life behavior information and travel information of the user, wherein the life behavior information comprises life behaviors and behavior execution time; the travel information comprises travel time, travel location and travel environment of the user; the user health portrait further comprises life behavior information and travel information in a time period when the abnormal category characteristic label of the correlated abnormal condition occurs; the system can also classify users according to occupation, establish occupation category health images according to personal information and occupation characteristics, age characteristics or regional characteristics, and judge the health risk of the result of the abnormal category characteristic label of the abnormal condition through comparison;
step S10, setting the total number of the category feature labels of the abnormal conditions in the abnormal physical sign library as n, wherein the category feature labels comprise X1，X2 ,…,Xn ；
Step S20, acquiring all data of user a in the abnormal physical sign library, and acquiring the abnormal category feature label a = { AX ] of user a1，AX2 ,…,AXmM is less than or equal to n, wherein AXiThe class i abnormal category feature label of the user a is obtained; obtaining an i-th abnormal class characteristic label AXiTime of occurrence ATi={At1，At2 ,…,AtkK is an abnormal class feature label AXiNumber of occurrences, AtkFeature tag AX for anomaly classiTime of occurrence of the t-th time;
step S30, setting the i-th abnormal class characteristic label AXiAssociated time period of Δ ATi，AXiEach associated time period of (a) includes AXiEach occurrence time of (A), AXiAssociated time period Δ ATi={△at1,△at2,△at3…△atkWhere atj=[Atj-△T1，Atj+△T2](ii) a Wherein, AtjIs AXiJ-th occurrence time,. DELTA.atjIs AXiIs associated with the jth time period, Δ T1、△T2A time greater than 0;
step S40, obtaining an abnormal class characteristic label AXiAbnormal class features AF = { AFX) occurring in all associated time periods1，AFX2 ,…,AFXcWherein c is more than or equal to 1 and less than or equal to k, AXi∉AF，AFXcIs AXiClass c exception category characteristics occurring in the associated time period of (a); labeling abnormal class characteristics AXiAssociated time period Δ ATiAll anomaly class features AF occurring within are recorded as AXiThe preset correlation characteristics;
step S50, when the abnormal class feature label AXiWhen the association time period has the preset association characteristic, the preset association is carried outAdding 1 to the correlation count of the feature, and recording the correlation counts of other abnormal class features as 0 to finally obtain AXiPreset associated feature count AL = { NUM) at all associated time periods1 ，NUM2 ，…，NUM cWherein, NUMjClass i exception class feature label AX for user aiJ th preset correlation characteristic AFXjAnd AXiThe total number of associations;
step S60, setting the association threshold value as W, and setting the preset association characteristic
Figure 817513DEST_PATH_IMAGE001
Has a weight of
Figure 997958DEST_PATH_IMAGE002
According to a predetermined association characteristic
Figure 535250DEST_PATH_IMAGE001
Weights and associated counts
Figure 142949DEST_PATH_IMAGE003
Obtaining a weight association count
Figure 195219DEST_PATH_IMAGE004
Taking the preset association feature with the weight association count larger than the association threshold value W as AXiThe associated characteristics of (1);
and step S70, constructing the user health portrait according to the related abnormal category characteristic labels, and displaying the abnormal category characteristic labels of other related abnormal conditions and corresponding health item recommendations to the client.
Wherein, in step S30, Δ at is preferably setjIncluding an anomaly class signature AXiAt most, the shortest time period, wherein1And Δ T2May be equal times or may be unequal times.
Preferably, in step S60, the abnormal class feature label AXiJ th preset correlation characteristic AFXjWeight of (2)
Figure 179355DEST_PATH_IMAGE005
The calculation method comprises the following steps:
Figure 367891DEST_PATH_IMAGE006
wherein the content of the first and second substances,
Figure 146491DEST_PATH_IMAGE007
is AFXjAnd AXiThe closeness of the association;
Figure 686057DEST_PATH_IMAGE008
is AFXjWith AX in a first association periodiThe number of times of association of (a),
Figure 208305DEST_PATH_IMAGE009
is AFXjWith AX in j-th association periodiThe number of times of association of (a),
Figure 782506DEST_PATH_IMAGE010
is a preset correlation characteristic AFXjFor AX in all associated time periodsiThe sum of the number of associations;
Figure 732007DEST_PATH_IMAGE011
wherein the content of the first and second substances,
Figure 758869DEST_PATH_IMAGE012
is AFXjAt AXiThe degree of importance in the anomaly category characteristics over all associated time periods;
Figure 819229DEST_PATH_IMAGE013
is the current preset correlation characteristic AFXjThe total number of occurrences on the user,
Figure 982357DEST_PATH_IMAGE014
is other anomalies in all associated time periodsThe total number of times the category feature occurs on the user,
Figure 365409DEST_PATH_IMAGE015
is in contact with AFXjThe total number of anomaly category features that have an association,
Figure 613988DEST_PATH_IMAGE016
is in other abnormal class characteristics and AFXjThe total number of times of association;
the preset correlation characteristic AFXjWeight of (2)
Figure 681301DEST_PATH_IMAGE017
；
Weight association count Lj=NUMj×dw(AFXj)。
Wherein preferably, in step S70, the system recommends a physical examination item, a diet or a reminder to purchase a drug in advance according to the correlated abnormality type characteristic tags.
Preferably, in step S70, obtaining the abnormal category characteristics in the affected time period by obtaining the history information of the previous disease of the user in the abnormal sign library, and establishing the neural network model by using the abnormal category characteristics in the affected time period and the corresponding disease condition as input and output data samples (x, y), the specific method is as follows: receiving D input vectors x = [ x ]1; x2 ; …; xD]
Figure 964515DEST_PATH_IMAGE018
z represents a weighted sum of inputs, where w = [ w ]1; w2;…; w D]Is the weight vector of the multidimensional input, b ∈ R is the bias.
The activation function used may be a Logistic function or a ReLU function, and when a ReLU function is used:
Figure 255819DEST_PATH_IMAGE019
in a multi-layer feedforward neural network, order
Figure 991694DEST_PATH_IMAGE020
The feed-forward neural network continuously iterates the following formula to carry out information propagation layer by layer, wherein the formula is as follows:
Figure 659436DEST_PATH_IMAGE021
；
the composite function is:
Figure 797156DEST_PATH_IMAGE022
，
wherein the content of the first and second substances,
Figure 259361DEST_PATH_IMAGE023
and
Figure 482532DEST_PATH_IMAGE024
representing the connection weights and offsets for all layers in the network,
Figure 688386DEST_PATH_IMAGE025
is the number of layers of the neural network,
Figure 680613DEST_PATH_IMAGE026
is as follows
Figure 579298DEST_PATH_IMAGE027
The number of layer neurons;
Figure 289765DEST_PATH_IMAGE028
is as follows
Figure 33731DEST_PATH_IMAGE029
Layer to layer
Figure 408693DEST_PATH_IMAGE030
A weight matrix of the layer;
Figure 212700DEST_PATH_IMAGE031
is as follows
Figure 410464DEST_PATH_IMAGE029
Layer to layer
Figure 958120DEST_PATH_IMAGE032
Biasing of the layers;
Figure 924939DEST_PATH_IMAGE033
is as follows
Figure 165427DEST_PATH_IMAGE034
Output of layer neurons;
using a cross-entropy loss function, for sample (x, y) the loss function is:
Figure 788169DEST_PATH_IMAGE035
wherein the content of the first and second substances,
Figure 936254DEST_PATH_IMAGE036
representing by a one-hot vector corresponding to y;
given a training set of
Figure 757580DEST_PATH_IMAGE037
Each sample is sampled
Figure 168969DEST_PATH_IMAGE038
Input to a feedforward neural network to obtain a network output of
Figure 279008DEST_PATH_IMAGE039
The risk function on the data set is:
Figure 168466DEST_PATH_IMAGE040
wherein W and b represent all weight matrices and bias vectors in the network, respectively;
Figure 109878DEST_PATH_IMAGE041
is regularizationTerms to prevent overfitting; lambda [ alpha ]>0 is a long parameter, λ>W is closer to 0 as 0 is larger:
Figure 957748DEST_PATH_IMAGE042
in each iteration of the gradient descent method, the update mode of the parameters W and b is as follows:
Figure 351820DEST_PATH_IMAGE043
wherein α is a learning rate;
calculating the gradient of the l-th layer weight and bias, δ(l)Error term for layer i:
Figure 245302DEST_PATH_IMAGE044
Figure 775641DEST_PATH_IMAGE045
obtaining an iterative formula:
Figure 528833DEST_PATH_IMAGE046
Figure 410201DEST_PATH_IMAGE047
。
the invention also provides a health recommendation system for operating the recommendation method.
The beneficial effect that this application realized is as follows:
according to the method and the system, under the era background of artificial intelligence and big data, information such as life, social contact, diet and movement of a user is acquired timely, the user is subjected to multidimensional all-around real-time portrait by acquiring abnormal type characteristics with relevance, effective reminding and health care measures are formed, and guidance and intervention measures which are most suitable for the current user are effectively provided for the current user by collecting abnormal conditions with relevance.
Meanwhile, the method and the system can help the individual or enterprise user to analyze the health dynamic state and provide the recommendation of related health services; then recommending suggestions for improving or maintaining the health state for the enterprise according to the health change dynamics of the enterprise population; and finally, providing trend analysis and recommending relevant decisions for local health supervision departments according to the health change characteristics of regional people.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings can be obtained by those skilled in the art according to the drawings.
FIG. 1 is a flow chart of the method of the present application.
Detailed Description
The technical solutions in the embodiments of the present application are clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
As shown in fig. 1, the present application provides a health recommendation system based on a medical knowledge-graph;
the health recommendation system obtains personal basic information of a user through a user information collection page, the user can directly fill in the user information collection page, and also can perform associated collection on the personal basic information in terminal equipment in a manner of associating other applications, so that the user information of the user can be obtained, wherein the user information comprises name, gender, age, identification number, occupation category, duty, working age, geographical area and the like;
after user information is acquired, a user health information base is correspondingly established, wherein the health information base comprises an abnormal physical sign base, and the abnormal physical sign base comprises a previous disease, a family genetic disease history, medical abnormal data and time interval abnormal data;
the medical abnormal data comprises all abnormal conditions except normal indexes and the time when the abnormal conditions occur in the results of the detection (including medical examination and annual physical examination) of the body by the medical instrument performed by the user over the years; the medical abnormal condition may include electrocardiogram abnormality, liver function five-item numerical abnormality, lung X-ray abnormality, B-ultrasonic image abnormality, etc.
The time interval abnormal data includes abnormal conditions obtained when the user performs self-check in life and the time when the abnormal conditions occur, and the types of the time interval abnormal data are blood pressure abnormality, blood sugar abnormality, body temperature abnormality, heartbeat abnormality, sudden pain and the like.
Meanwhile, the life behavior information and the travel information of the user can be acquired through a network tool or other special application programs by combining with the user information of the user, wherein the life behavior information comprises life behaviors and time for executing the behaviors, and the types of the life behaviors comprise medicine taking behaviors (comprising medicine names and medicine taking time), eating behaviors (recipes, food volumes and eating time), work and rest time, exercise intensity, duration and the like;
the category of the travel information includes a vehicle (walking, riding or car, airplane) used by the user, an area traveled, an environmental condition of the traveled area, a time of traveling through different areas, and the like.
In specific implementation, for example, the abnormal situation category labels may include labels for recording senses such as joint pain, limb pain, headache, muscle pain, neuropathic pain, and the like, and in other embodiments, the abnormal situation category labels may also be provided as labels for daily physical examinations including hypertension, high heart rate, high blood sugar, high blood fat, and the like; in other embodiments, the abnormal situation category label can be set as a label with visual expression of fever, cough, dyspnea, body weakness, local red swelling and the like;
in the specific implementation process, preferably, the value range of n is [10,200], more preferably [20,50], the value of the category total n determines the size of the range for dividing the abnormal condition, when n is larger, the number of the abnormal condition tags is more, the attribute of the abnormal condition is more finely divided, for example, the joint pain can be further refined into wrist joint pain, hip joint pain and ankle joint pain, and the specialty and the pertinence of the user can be further facilitated.
In addition, in order to numerically represent the category label of the abnormal situation, the total number of the category feature labels of the abnormal situation in the abnormal physical sign library is set as n, and the category feature label comprises X1，X2 ,…,XnN is more than or equal to 1; meanwhile, as will be appreciated by those skilled in the art, the above-mentioned expression of the attribute tag is only an exemplary description, and is not a sole example for limiting the protection scope of the present invention.
The health recommendation system of the medical knowledge graph disclosed by the invention can be arranged on terminal equipment, and also can be arranged on a host and a system, and when the system starts to work, the following recommendation method of the health recommendation system based on the medical knowledge graph can be realized:
setting the total number of category feature labels of abnormal conditions in the abnormal physical sign library as n, wherein the category feature labels comprise X1，X2 ,…,Xn ；
Acquiring all data of the user a in an abnormal physical sign library, and acquiring an abnormal class feature label A = { AX of the user a1，AX2 ,…,AXmM is less than or equal to n, wherein AXiA label of the i-th abnormal category characteristic of the user a; obtaining an i-th abnormal class characteristic label AXiTime of occurrence ATi=[At1，At2 ,…,Atk ] TWherein k is an abnormal class feature label AXiThe number of occurrences;
setting an i-th class exception class feature label AXiAssociated time period of Δ ATi，AXiEach associated time period of (a) includes AXiEach occurrence time of (A), AXiAssociated time period Δ ATi={△at1,△at2,△at3…△atkWhere atj=[Atj-△T1，Atj+△T2]（△T1＞0，△T2> 0) wherein, AtjIs AXiJ-th occurrence time,. DELTA.atjIs AXiIs associated with the jth time period, Δ T1、△T2For times greater than 0, Δ T1、△T2May or may not be equal;
obtaining an anomaly class feature AXiAbnormal class features AF = { AFX) occurring in all associated time periods1，AFX2,…,AFXcWherein c is more than or equal to 1 and less than or equal to k, AXi∉ AF; characterizing the anomaly class AXiAssociated time period Δ ATiAll anomaly class features AF occurring within are recorded as AXiThe preset correlation characteristics;
when abnormal class feature label AXiWhen the preset associated feature exists in the associated time period, adding 1 to the associated count of the preset associated feature, recording the associated counts of other abnormal class features as 0, and finally obtaining the AXiPreset associated feature count AL = { NUM) at all associated time periods1 ，NUM2 ，…，NUM cWherein, NUMjClass i exception class feature label AX for user aiJ th preset correlation characteristic AFXjAnd AXiTotal number of associations.
For example, let user a's first class anomaly signature AX1Occurring 3 times in total, namely 7/2/2001, 7/12/2001, and 7/15/2001, 4 associated abnormal features are respectively AX in abnormal time periods including 7/2/20014、AX6、AX7, AX11There are 3 associated abnormalities characterized in the abnormal time period including 7, 12 and 2001: are each AX4、AX5、AX8(ii) a There are 2 associated abnormal features in the abnormal time period including 7, 12 and 2001, each of which is AX4、AX8(ii) a The first class anomaly class feature label AX can be obtained1Is preset associated with the anomaly characteristic AX4、AX5、AX6、AX7、AX8、AX11(ii) a And the association count is {3,1,1,1,2,1}
Wherein the anomaly class feature label AXIIs associated with the anomaly characteristic AFXjWeight of (2)dw(AFXj) The calculation method comprises the following steps:
Figure 907042DEST_PATH_IMAGE048
wherein the content of the first and second substances,
Figure 557466DEST_PATH_IMAGE049
is AFXjAnd AXiThe degree of association; wherein the content of the first and second substances,
Figure 215980DEST_PATH_IMAGE050
is a preset correlation characteristic AFXjWith AX during a jth association periodiThe number of times of association of (a),
Figure 584645DEST_PATH_IMAGE051
is AFXjWith AX in a first association periodiThe number of times of association of (a),
Figure 619597DEST_PATH_IMAGE052
is a correlation feature AFXjFor AX in all associated time periodsiThe sum of the number of associations;
Figure 390107DEST_PATH_IMAGE053
wherein the content of the first and second substances,
Figure 485102DEST_PATH_IMAGE054
is AFXjAt AXiThe degree of importance in the anomaly category characteristics over all associated time periods;
Figure 341062DEST_PATH_IMAGE055
is the current preset correlation characteristic AFXjThe total number of occurrences on the user,
Figure 179705DEST_PATH_IMAGE056
is the total number of times other abnormality category features have occurred on the user for all associated time periods,
Figure 539143DEST_PATH_IMAGE015
is in contact with AFXjThe total number of anomaly category features that have an association,
Figure 805039DEST_PATH_IMAGE016
is in other abnormal class characteristics and AFXjTotal number of times of association.
The preset correlation characteristic AFXjWeight of (2)
Figure 879787DEST_PATH_IMAGE057
；
Weight association count: l isj=NUMj×dw(AFXj)。
Setting the association threshold value as W, and taking the abnormal class characteristic with the weight association count larger than the association threshold value as AX according to the weightiThe associated features of (1).
Through the scheme, the system can construct the abnormal physical sign portrait of the user according to the medical abnormal data and the abnormal time period data of the user, and when the abnormal type characteristics of the client occur, the system prompts other related abnormal conditions which may occur simultaneously to the client, and asks the client to pay attention to the related detection, diet attention, advance medicine purchase and the like.
In addition, the system can also acquire the disease time by acquiring the historical information of the previous diseases of the users in the abnormal physical sign libraryAnd (3) establishing a neural network model by taking the abnormal class characteristics in the affected time period and the corresponding disease conditions as input and output data samples (x, y) according to the abnormal class characteristics in the period, wherein the specific method comprises the following steps: receiving D input vectors x = [ x ]1; x 2 ; …; x D ]
Figure 522120DEST_PATH_IMAGE058
z represents a weighted sum of inputs, where w = [ w ]1; w2;…; w D]Is the weight vector of the multidimensional input, b ∈ R is the bias.
The activation function used may be a Logistic function or a ReLU function, and when a ReLU function is used:
Figure 1643DEST_PATH_IMAGE019
in a multi-layer feedforward neural network, order
Figure 438441DEST_PATH_IMAGE059
The feed-forward neural network continuously iterates the following formula to carry out information propagation layer by layer, wherein the formula is as follows:
Figure 3414DEST_PATH_IMAGE060
the composite function is:
Figure 387122DEST_PATH_IMAGE061
where W and b represent the connection weights and offsets for all layers in the network.lIs the number of layers of the neural network,M l is as followslThe number of layer neurons;
Figure 721152DEST_PATH_IMAGE062
is as follows
Figure 328851DEST_PATH_IMAGE063
Layer to layerlA weight matrix of the layer;
Figure 381120DEST_PATH_IMAGE064
is as followsl1 layer to the secondlBiasing of the layers;
Figure 99678DEST_PATH_IMAGE065
is as followslOutput of layer neurons;
using a cross-entropy loss function, for sample (x, y) the loss function is:
Figure 553793DEST_PATH_IMAGE066
wherein the content of the first and second substances,
Figure 332393DEST_PATH_IMAGE067
representing by a one-hot vector corresponding to y;
given a training set of
Figure 871959DEST_PATH_IMAGE068
Each sample is sampled
Figure 394207DEST_PATH_IMAGE069
Input to a feedforward neural network to obtain a network output of
Figure 702828DEST_PATH_IMAGE070
The risk function on the data set is:
Figure 938417DEST_PATH_IMAGE071
wherein W and b represent all weight matrices and bias vectors in the network, respectively;
Figure 699699DEST_PATH_IMAGE072
is a regularization term to prevent overfitting; lambda [ alpha ]>0 is a long parameter, λ>W is closer to 0 as 0 is larger:
Figure 25639DEST_PATH_IMAGE042
the way in which the parameters W and b are updated in each iteration of the gradient descent method
Figure 923187DEST_PATH_IMAGE073
Figure 43590DEST_PATH_IMAGE074
Figure 823327DEST_PATH_IMAGE075
Figure 687378DEST_PATH_IMAGE076
Figure 705013DEST_PATH_IMAGE077
By establishing the model, the state of illness or possible treatment directions can be judged according to the abnormal category characteristics with relevance, and a recommended rehabilitation scheme such as exercise and diet can be given to help the abnormal situations encountered by the client.
In addition, by acquiring the life behavior information and the travel information of the user in the associated time period and judging whether abnormal life behavior information and abnormal travel information exist in the current associated time period, a behavior-abnormality model of the user can be generated, and the specific method comprises the following steps: acquiring user behavior information, wherein the behavior information comprises user living behavior information and travel information, and the living behavior information comprises living behaviors and behavior execution time; the travel information comprises travel time, travel place and travel environment of the user, and behavior information data of the user a is set to be AR = { AR = }1 ,AR2 ,…,ARdThe time when the action occurs is ART = { ARt = }1 ,ARt2 ,…,ARtdIn which are ARtiAs behavior information ARiThe occurrence time of the user is obtained, and the behavior characteristic association time period delta AT of the i-th class abnormal class characteristic AXi of the user is obtainedi＇={△at1＇,△at2＇,△at3＇…△atk' where Δ atj' is AXiIs associated with the time period, Δ at, of the jth behavioral characteristic ofj＇=[Atj-△T1-△T3，Atj+△T2-△T4]（△T3＞0，△T4> 0), known as Δ atj=[Atj-△T1，Atj+△T2]Is AXiIs said AX, isiIs associated with the j-th behavior feature of (1) by the time period Δ atjOccurring in a feature association period Δ atjBefore, or including a feature association time period Δ atjFor a period of time. Delta T3、△T4For times greater than 0, Δ T3、△T4May or may not be equal.
Obtaining Δ AT in a behavioral association time periodiOf the behavioral characteristics that occurred, abnormal behavioral characteristics AQ = { AQX = [, ]1，AQX2 ,…,AQXdD is more than or equal to 1 and less than or equal to k; characterizing the anomaly class AXiTime interval Δ AT associated with behavior characteristicsiAll abnormal behaviour characteristics occurring in' AQ are recorded as AXiPresetting associated behavior characteristics;
when abnormal class feature AXiWhen the behavior associated time period has abnormal behavior characteristics, the association count of the preset associated behavior characteristics is added with 1, the association counts of other behavior characteristics are marked as 0, and finally the AX is obtainediPreset associated feature count AL' = { NUM) at all behavior associated time periods1＇ ，NUM2＇，…，NUM c', wherein, NUMj' is the i-th type exception category feature AX of user aiJ < th > preset associated behavior feature AQXjAnd AXiTotal number of associations.
For example, in the time period of 6-7 months in 2014, the abnormal type feature [ topical red rash ] of the user occurs 3 times, namely 12 days in 6-6 months in 2014, 28 days in 6-28 months in 2014 and 19 days in 7-19 months in 2014; the abnormal type characteristics with relevance are [ fever ], [ running nose ], abnormal information in the life behavior information, such as [ business trip ], and [ business trip place ] [ environmental condition ] in the business trip information are checked in the life behavior information and the trip information of the user in the time period, whether people with the relevant abnormal type characteristics exist in the business trip place or not is checked in the period of 6 months to 7 months in the abnormal physical sign library, after a certain amount of same type abnormal condition data is found, the trip abnormal information is recorded, when the customer triggers the trip abnormal information again, namely the business trip arrives at the place, the relevant abnormal type characteristics are prompted, and suitable clothes, external medicines, food contraindications and the like are recommended.
In addition, the system can classify users according to occupation, establish an occupation category comparison model according to personal information and occupation characteristics, age characteristics or regional characteristics, compare abnormal physical sign images of the person undergoing physical examination with the occupation category comparison model in the system, judge whether the result of the abnormal characteristics has a health risk or not, if so, send the health risk to the person through a recommendation system, and provide a proposal for improving the risk.
Setting an abnormal value threshold according to personal conditions such as age, work and race, physical conditions and the like, and alarming under a high risk condition when the slope of an abnormal value curve is greater than a safety slope threshold; when the maximum value of the abnormal numerical value is larger than the safety threshold value, alarming under the high risk condition; and when the person has the behavior which can cause the abnormal data to continue to deteriorate, the person is warned;
if the trend of the numerical value of the abnormal data at this time is mild compared with the numerical value of the historical abnormal data, recommending the personnel to continue using the recommendation scheme;
in addition, an enterprise physical examination information portrait can be further constructed through abnormal physical sign pictures, and an enterprise models from multiple dimensions by acquiring the abnormal physical sign picture information of all current employees of the enterprise and considering the attribute information of the enterprise, the environmental data of the region where the enterprise is located and the like, so as to construct a comprehensive abnormal type characteristic model.
According to the enterprise physical examination information picture, the physical examination results with the same abnormal data value and the persons with similar personal information and behavior information can be independently established as an abnormal type characteristic model;
for example, a 15 male employee at a position 35-38 years old may have cardiac function index data that exceeds a first health risk value; reminding to push employees who fit this type of feature, such as employees who are "35-38 years old," male "," management position "," work time per month over 260 hours ", the current behavioral features may result in health risks and prompt from which behaviors should be improved.
When the physical examination is carried out again, when the personal information of the staff comprises the characteristic types, the physical examination items are preferentially added to the numerical value of the heart function direction of the staff;
meanwhile, other behavior information of the staff is added, such as daily movement, walking step number, rest interval time, sleeping time, diet and the like, and is added into the characteristic types according to the occurrence times and the weight;
in pushing, consider the situation that some diseases are improved by staff, and other values are out of limits, and [0034] consider whether the improvement scheme affects the values.
And acquiring an overproof data curve graph of the personnel in the historical physical examination data, comparing each overproof data with the current physical examination data, watching the development condition of the overproof data in a historical physical examination data center, and if the standard is better, improving the recommendation score of [0035] in the improvement scheme of the personnel, and optimizing and adjusting the improvement scheme.
If the situation is not improved or continues to deteriorate, the recommended score in the improvement scheme is reduced, and the improvement scheme of the overproof numerical value of the person is adjusted.
Similarly, a medical unit physical examination image can be established by the abnormal physical sign image, and the medical unit physical examination image expresses all the health condition change rules of the medical unit physical examination person, namely the medical unit physical examination person reflects the health condition of a region. The composition of the medical unit physical examination image is more complex than that of the enterprise physical examination information image. When constructing the portrait, the physical examination data of workers engaged in different health industries needs to be comprehensively considered, and the accurate physical examination information portrait of the medical unit needs to be generated by combining the data of local environment and the like.
In addition, according to different classifications of physical examination information images, personalized health service recommendations with different directions are also provided, and after the physical examination information figure image is constructed, the related figure image needs to be analyzed and applied. When generating health service recommendations for individual users, the explicit needs and implicit needs of the individual users need to be analyzed in depth. Where explicit needs refer to things that a user subjectively thinks about, and implicit needs are things that a user does not realize but is important to. The following application scenarios are given: for an individual user, each abnormal physical sign information such as physical examination data is collected, even if physical examinations are all in a normal value range, certain single values fluctuate, the data fluctuation conditions can be deeply mined, the living state of the current user is presumed from the fluctuation conditions, and relevant health recommendation services such as diet work and rest suggestions are generated for the user, so that the user can be guaranteed to successfully pass through the physical examinations next time.
The management personnel of the enterprise need to stand at the height of the enterprise to make relevant decisions according to the health data of the employees. The following application scenarios are given: for the enterprise manager, the health fluctuation condition of the current employee is fed back to the enterprise manager through the related abnormal condition data of the employee, for example, although all the employees are qualified in the health certificate physical examination project, the employees present the same change rules on the data of certain indexes, and the change rules may be related to the working system of the employees, such as night shift and the like. In order to ensure that the staff can pass the examination in the next health examination, it is necessary to recommend relevant analysis conditions and health services to the enterprise manager. And for enterprises with similar structure and function, health safety accidents of the enterprises with similar structure and function can be recommended to play a role in warning.
The health unit considers the whole health condition of a region and makes a relevant health policy according to the local health data change condition. The following application scenarios are given: from the perspective of health regulatory agencies, they need to comprehensively consider the health changes of people in each area and give changes to the environment or other relevant policies. For example, there may be a significant difference in physical examination data between two persons in adjacent management areas, and the reason for the difference may be the environment. Therefore, by analyzing the health data of the whole administrative district, relevant result analysis and recommendation can be generated for health supervision units.
In the process of establishing the abnormal characteristic model, the data acquisition, the data cleaning and the data analysis are included.
In the data acquisition process, abnormal physical sign data can be directly acquired from relevant medical knowledge map big data, and other data such as weather data can be directly captured from the Internet by using a crawler tool.
In the process of data cleaning, the obtained data is preprocessed, including processing of residual values, abnormal values and the like, so that subsequent analysis is facilitated.
In the data analysis process, the cleaned data is subjected to correlation analysis by using a correlation technology of data mining, such as clustering, PCA and the like, and the correlation among the data is found.
After the portrait is constructed, personalized recommendations need to be generated for users of different levels according to the portrait, a suitable recommendation algorithm can be adopted to model the problem, and a recommendation model is optimized by combining with an actual application scene, such as time for generating a recommendation list, so as to achieve the goal of online use.
After the research of the recommendation model is completed, a system needs to be designed for the user according to the real application scenario and corresponding services need to be provided. The development framework of the system can use the main development framework SSM and the like to ensure the normal use of users.
After the system is developed, all functional modules of the system need to be tested, and a black box testing method can be adopted to ensure the normal operation of all functions of the system. After the test is finished, the system can be deployed on line for the actual use of the user, and then more detailed adjustment can be carried out according to the feedback used by the user.
While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application. It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.

Claims (6)

1. A recommendation method of a health recommendation system based on a medical knowledge graph comprises the following steps:
step S00, providing a health recommendation system based on a medical knowledge graph, which comprises a user health information base, wherein the user health information base comprises an abnormal physical sign base; the abnormal sign library comprises medical abnormal data and time interval abnormal data; the medical abnormal data comprise abnormal conditions and occurrence time of which numerical values are beyond normal indexes in the detection result of the body of the user through the medical instrument, and the time interval abnormal data comprise the abnormal conditions and the occurrence time obtained when the user checks himself in life; a user health portrait can be established through medical abnormal data and time interval abnormal data, wherein the user health portrait comprises user personal information and an abnormal category characteristic label of an associated abnormal condition; setting a correlation time period of the abnormal category feature tags according to the time of the abnormal condition, wherein the abnormal category feature tags with correlation occur in the same correlation time period; the user health portrait further comprises life behavior information and travel information of the user, wherein the life behavior information comprises life behaviors and behavior execution time; the travel information comprises travel time, travel location and travel environment of the user; the user health portrait further comprises life behavior information and travel information in a time period when the abnormal category characteristic label of the correlated abnormal condition occurs; the system can also classify users according to occupation, establish occupation category health images according to personal information and occupation characteristics, age characteristics or regional characteristics, and judge the health risk of the result of the abnormal category characteristic label of the abnormal condition through comparison;
step S10, setting the total number of the category feature labels of the abnormal conditions in the abnormal physical sign library as n, wherein the category feature labels comprise X1，X2 ,…,Xn ；
Step S20, acquiring all data of user a in the abnormal physical sign library, and acquiring the abnormal category feature label a = { AX ] of user a1，AX2 ,…,AXmM is less than or equal to n, wherein AXiThe class i abnormal category feature label of the user a is obtained; obtaining an i-th abnormal class characteristic label AXiTime of occurrence ATi={At1，At2 ,…,AtkK is an abnormal class feature label AXiNumber of occurrences, AtkFeature tag AX for anomaly classiTime of occurrence of the t-th time;
step S30, setting the i-th abnormal class characteristic label AXiAssociated time period of Δ ATi，AXiEach associated time period of (a) includes AXiEach occurrence time of (A), AXiAssociated time period Δ ATi={△at1,△at2,△at3…△atkWhere atj=[Atj-△T1，Atj+△T2](ii) a Wherein, AtjIs AXiJ-th occurrence time,. DELTA.atjIs AXiIs associated with the jth time period, Δ T1、△T2A time greater than 0;
step S40, obtaining an abnormal class characteristic label AXiAbnormal class features AF = { AFX) occurring in all associated time periods1，AFX2 ,…,AFXcIn which 1 is≤c≤k，AXi∉AF，AFXcIs AXiClass c exception category characteristics occurring in the associated time period of (a); labeling abnormal class characteristics AXiAssociated time period Δ ATiAll anomaly class features AF occurring within are recorded as AXiThe preset correlation characteristics;
step S50, when the abnormal class feature label AXiWhen the preset associated feature exists in the associated time period, adding 1 to the associated count of the preset associated feature, recording the associated counts of other abnormal class features as 0, and finally obtaining the AXiPreset associated feature count AL = { NUM) at all associated time periods1 ，NUM2 ，…，NUM cWherein, NUMjClass i exception class feature label AX for user aiJ th preset correlation characteristic AFXjAnd AXiThe total number of associations;
step S60, setting the association threshold value as W, and setting the preset association characteristic
Figure 874529DEST_PATH_IMAGE001
Has a weight ofdw(AFXj) According to a predetermined association characteristic
Figure 856391DEST_PATH_IMAGE001
Weights and associated counts
Figure 599219DEST_PATH_IMAGE002
Obtaining a weight association count
Figure 183784DEST_PATH_IMAGE003
Taking the preset association feature with the weight association count larger than the association threshold value W as AXiThe associated characteristics of (1);
and step S70, constructing the user health portrait according to the related abnormal category characteristic labels, and displaying the abnormal category characteristic labels of other related abnormal conditions and corresponding health item recommendations to the client.
2. The method of claim 1The method for recommending a health recommendation system based on a medical knowledge graph, wherein in step S30, Δ atjIncluding an anomaly class signature AXiAt most, the shortest time period, wherein1And Δ T2May be equal times or may be unequal times.
3. The recommendation method of the health recommendation system based on medical knowledge-graph as claimed in claim 1, wherein in step S60, the abnormality category feature label AXiJ th preset correlation characteristic AFXjWeight of (2)
Figure 695668DEST_PATH_IMAGE004
The calculation method comprises the following steps:
Figure 582853DEST_PATH_IMAGE005
；
wherein the content of the first and second substances,
Figure 937611DEST_PATH_IMAGE006
is AFXjAnd AXiThe closeness of the association;
Figure 201233DEST_PATH_IMAGE007
is AFXjWith AX in a first association periodiThe number of times of association of (a),
Figure 302044DEST_PATH_IMAGE008
is AFXjWith AX in j-th association periodiThe number of times of association of (a),
Figure 15922DEST_PATH_IMAGE009
is a preset correlation characteristic AFXjFor AX in all associated time periodsiThe sum of the number of associations;
Figure 467763DEST_PATH_IMAGE010
wherein the content of the first and second substances,
Figure 535076DEST_PATH_IMAGE011
is AFXjAt AXiThe degree of importance in the anomaly category characteristics over all associated time periods;
Figure 21553DEST_PATH_IMAGE012
is the current preset correlation characteristic AFXjThe total number of occurrences on the user,
Figure 375173DEST_PATH_IMAGE013
is the total number of times other abnormality category features have occurred on the user for all associated time periods,
Figure 314311DEST_PATH_IMAGE014
is in contact with AFXjThe total number of anomaly category features that have an association,
Figure 44369DEST_PATH_IMAGE015
is in other abnormal class characteristics and AFXjThe total number of times of association;
the preset correlation characteristic AFXjWeight of (2)
Figure 385352DEST_PATH_IMAGE016
；
Weight association count Lj=NUMj×dw(AFXj)。
4. The recommendation method of the medical knowledge-graph-based health recommendation system according to claim 1, wherein in step S70, the system recommends a physical examination item, a diet or a reminder to purchase a drug in advance according to the correlated abnormality type characteristic tags.
5. The recommendation method of the medical knowledge-graph-based health recommendation system of claim 1, wherein in the step of recommending the health, the medical knowledge-graph-based health recommendation system is used for recommending the healthIn step S70, obtaining the abnormal category characteristics in the affected time period by obtaining the historical information of the previous disease in the abnormal sign library, taking the abnormal category characteristics in the affected time period and the corresponding disease state as input and output data samples (x, y), and establishing a neural network model, the specific method is as follows: receiving D input vectors x = [ x ]1; x2 ; …; xD]
Figure 785240DEST_PATH_IMAGE017
z represents a weighted sum of inputs, where w = [ w ]1; w2;…; w D]Is the weight vector of the multidimensional input, b is the bias for R;
the activation function used may be a Logistic function or a ReLU function, and when a ReLU function is used:
Figure 336307DEST_PATH_IMAGE018
in a multi-layer feedforward neural network, order
Figure 745423DEST_PATH_IMAGE019
The feed-forward neural network continuously iterates the following formula to carry out information propagation layer by layer, wherein the formula is as follows:
Figure 940912DEST_PATH_IMAGE020
；
the composite function is:
Figure 42860DEST_PATH_IMAGE021
，
wherein
Figure 815644DEST_PATH_IMAGE022
And
Figure 762872DEST_PATH_IMAGE023
representing the connection weights and offsets for all layers in the network,
Figure 78447DEST_PATH_IMAGE024
is the number of layers of the neural network,
Figure 944771DEST_PATH_IMAGE025
is as follows
Figure 345797DEST_PATH_IMAGE026
The number of layer neurons;
Figure 96715DEST_PATH_IMAGE027
is as follows
Figure 125851DEST_PATH_IMAGE028
Layer to layer
Figure 569602DEST_PATH_IMAGE029
A weight matrix of the layer;
Figure 51399DEST_PATH_IMAGE030
is as follows
Figure 606008DEST_PATH_IMAGE031
Layer to layer
Figure 630596DEST_PATH_IMAGE032
Biasing of the layers;
Figure 104303DEST_PATH_IMAGE033
is as follows
Figure 214341DEST_PATH_IMAGE034
Output of layer neurons;
using a cross-entropy loss function, for sample (x, y) the loss function is:
Figure 53201DEST_PATH_IMAGE035
wherein the content of the first and second substances,
Figure 322509DEST_PATH_IMAGE036
representing by a one-hot vector corresponding to y;
given a training set of
Figure 108062DEST_PATH_IMAGE037
Each sample is sampled
Figure 705397DEST_PATH_IMAGE038
Input to a feedforward neural network to obtain a network output of
Figure 460863DEST_PATH_IMAGE039
The risk function on the data set is:
Figure 194464DEST_PATH_IMAGE040
wherein W and b represent all weight matrices and bias vectors in the network, respectively;
Figure 150919DEST_PATH_IMAGE041
is a regularization term to prevent overfitting; lambda [ alpha ]>0 is a long parameter, λ>W is closer to 0 as 0 is larger:
Figure 94604DEST_PATH_IMAGE042
;
in each iteration of the gradient descent method, the update mode of the parameters W and b is as follows:
Figure 794707DEST_PATH_IMAGE043
wherein α is a learning rate;
calculating the gradient of the l-th layer weight and bias, δ(l)As the first layerError term:
Figure 648393DEST_PATH_IMAGE044
Figure 369224DEST_PATH_IMAGE045
obtaining an iterative formula:
Figure 941151DEST_PATH_IMAGE046
Figure 179366DEST_PATH_IMAGE047
。
6. health recommendation system for running the recommendation method according to any of claims 1-5.
CN202110437769.XA 2021-04-22 2021-04-22 Health recommendation system and method based on medical knowledge map Active CN112820371B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110437769.XA CN112820371B (en) 2021-04-22 2021-04-22 Health recommendation system and method based on medical knowledge map

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110437769.XA CN112820371B (en) 2021-04-22 2021-04-22 Health recommendation system and method based on medical knowledge map

Publications (2)

Publication Number Publication Date
CN112820371A CN112820371A (en) 2021-05-18
CN112820371B true CN112820371B (en) 2021-08-03

Family

ID=75862558

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110437769.XA Active CN112820371B (en) 2021-04-22 2021-04-22 Health recommendation system and method based on medical knowledge map

Country Status (1)

Country Link
CN (1) CN112820371B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113221012A (en) * 2021-05-31 2021-08-06 北京海纳数聚科技有限公司 Community fitness activity recommendation method based on group portrait
CN114974508B (en) * 2022-05-20 2024-05-24 西南大学 Generating device of sports prescription and electronic equipment
CN115036034B (en) * 2022-08-11 2022-11-08 之江实验室 Similar patient identification method and system based on patient characterization map
CN117727456B (en) * 2023-12-28 2024-05-24 江苏智慧智能软件科技有限公司 Obstetrical psychological assessment model modeling system and method based on artificial intelligence

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018045228A3 (en) * 2016-08-31 2018-04-12 Intermountain Invention Management, Llc Interpreting genomic results and providing targeted treatment options in cancer patients
CN108922623A (en) * 2018-07-12 2018-11-30 中国铁道科学研究院集团有限公司 A kind of health risk assessment and Disease Warning Mechanism information system
WO2019212833A1 (en) * 2018-04-30 2019-11-07 The Board Of Trustees Of The Leland Stanford Junior University System and method to maintain health using personal digital phenotypes
CN111354469A (en) * 2020-03-31 2020-06-30 浙江禾连网络科技有限公司 User health condition comprehensive evaluation method and system

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104462858A (en) * 2014-12-30 2015-03-25 天津迈沃医药技术有限公司 Health warning method based on multi-order hidden Markov model
CN112037934A (en) * 2019-06-03 2020-12-04 合肥市行思拓教育科技有限公司 Intelligent health examination system based on mobile device and learning method
CN112002429B (en) * 2020-10-27 2021-02-19 北京梦天门科技股份有限公司 Physical examination portrait and service recommendation system and method for public health enterprise personnel

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018045228A3 (en) * 2016-08-31 2018-04-12 Intermountain Invention Management, Llc Interpreting genomic results and providing targeted treatment options in cancer patients
WO2019212833A1 (en) * 2018-04-30 2019-11-07 The Board Of Trustees Of The Leland Stanford Junior University System and method to maintain health using personal digital phenotypes
CN112639990A (en) * 2018-04-30 2021-04-09 小利兰·斯坦福大学托管委员会 System and method for maintaining health using personal digital phenotype
CN108922623A (en) * 2018-07-12 2018-11-30 中国铁道科学研究院集团有限公司 A kind of health risk assessment and Disease Warning Mechanism information system
CN111354469A (en) * 2020-03-31 2020-06-30 浙江禾连网络科技有限公司 User health condition comprehensive evaluation method and system

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Effects of lifelong testosterone exposure on health and disease using Mendelian randomization;Mohammadi-Shemirani;《ELIFE》;20201016;第9卷;全文 *
丁伟;全民健康管理系统软件在慢病预警与干预管理中的开发应用;《科技成果》;20160401;全文 *

Also Published As

Publication number Publication date
CN112820371A (en) 2021-05-18

Similar Documents

Publication Publication Date Title
CN112820371B (en) Health recommendation system and method based on medical knowledge map
Zarandi et al. Designing a general type-2 fuzzy expert system for diagnosis of depression
Kamalraj et al. Interpretable filter based convolutional neural network (IF-CNN) for glucose prediction and classification using PD-SS algorithm
JP6463572B1 (en) Support system to estimate the internal state of the target system
JP6881687B2 (en) Methods and systems for modeling the user&#39;s mental / emotional state
US10617351B2 (en) Cognitive biometric systems to monitor emotions and stress
Kraft et al. Data mining in healthcare information systems: case study of a veterans' administration spinal cord injury population
US20090240529A1 (en) Method for selecting a high risk patient for participation in a care management program for patients having poor prognoses
CN112002429B (en) Physical examination portrait and service recommendation system and method for public health enterprise personnel
Tsang et al. Application of machine learning to support self-management of asthma with mHealth
Dooshima et al. A predictive model for the risk of mental illness in Nigeria using data mining
Elbayoudi et al. The human behaviour indicator: A measure of behavioural evolution
Leevy et al. Investigating the relationship between time and predictive model maintenance
Mohammed et al. A Novel Predictive Analysis to Identify the Weather Impacts for Congenital Heart Disease Using Reinforcement Learning
Zolnoori et al. Fuzzy rule-based expert system for evaluating level of asthma control
Shamseddine et al. On the feasibility of federated learning for neurodevelopmental disorders: Asd detection use-case
Anderson et al. Physician utilization of computers in medical practice: policy implications based on a structural model
CN116434957A (en) Psychological health state identification and assessment method
Yu et al. Air traffic controllers' mental fatigue recognition: A multi-sensor information fusion-based deep learning approach
Hilton et al. The roles of antisociality and neurodevelopmental problems in criminal violence and clinical outcomes among male forensic inpatients
Shaposhnyk et al. Inferring cognitive load level from physiological and personality traits
US11250723B1 (en) Visuospatial disorders detection in dementia using a computer-generated environment based on voting approach of machine learning algorithms
Beaulah et al. AI in Quality Measurement
Jindal et al. Predicting Susceptibility to Covid Stress Using Data Mining
Anakal et al. Clinical Decision Support System for Diagnosis and Treatment of COPD Using Ensemble Methods

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
PE01 Entry into force of the registration of the contract for pledge of patent right

Denomination of invention: A health recommendation system and method based on medical knowledge map

Effective date of registration: 20211203

Granted publication date: 20210803

Pledgee: Beijing ustron Tongsheng financing Company limited by guarantee

Pledgor: HEALTH HOPE (BEIJING) TECHNOLOGY CO.,LTD.

Registration number: Y2021990001151

PE01 Entry into force of the registration of the contract for pledge of patent right