CN113066580A - Medical score determining method and device, electronic equipment and storage medium - Google Patents

Medical score determining method and device, electronic equipment and storage medium Download PDF

Info

Publication number
CN113066580A
CN113066580A CN202110285202.5A CN202110285202A CN113066580A CN 113066580 A CN113066580 A CN 113066580A CN 202110285202 A CN202110285202 A CN 202110285202A CN 113066580 A CN113066580 A CN 113066580A
Authority
CN
China
Prior art keywords
target
time
preset
training
characteristic
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.)
Pending
Application number
CN202110285202.5A
Other languages
Chinese (zh)
Inventor
薛佩姣
赖志明
林志哲
薛佳佳
蒋劲峰
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.)
Shanghai Electric Group Corp
Original Assignee
Shanghai Electric Group Corp
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 Shanghai Electric Group Corp filed Critical Shanghai Electric Group Corp
Priority to CN202110285202.5A priority Critical patent/CN113066580A/en
Publication of CN113066580A publication Critical patent/CN113066580A/en
Pending legal-status Critical Current

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
    • 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
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • 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
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • 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
    • G06N3/08—Learning methods
    • 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
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems

Landscapes

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

Abstract

The invention relates to the technical field of medical treatment, and discloses a medical treatment grading determination method, a device, electronic equipment and a storage medium, wherein the method comprises the steps of determining a target characteristic sequence according to a target characteristic value corresponding to a preset characteristic of a target to be graded in a first preset time before the current time; and inputting the target characteristic sequence into a first training LSTM model corresponding to a target area where the target to be evaluated is located to obtain the predicted medical grade of the target to be evaluated at the target moment, wherein the first training LSTM model is obtained by performing migration training on a second training LSTM model according to a first sample in a database corresponding to the target area, and the second training LSTM model is obtained by training an initial LSTM model based on a second sample in a public database. According to the embodiment, the predicted medical score of the target to be scored at the target moment can be accurately obtained.

Description

Medical score determining method and device, electronic equipment and storage medium
Technical Field
The invention relates to the technical field of medical treatment, in particular to a medical treatment score determining method and device, electronic equipment and a storage medium.
Background
With the progress of science and technology, the medical level is rapidly developed. In the medical field, the severity of a disease of a patient can be known according to the medical score of the patient, and the method is very important for diagnosis and treatment of the disease.
In the related art, the relationship between the physical examination information of the sample at a certain time and the medical score at the next time is obtained based on the public database, and by learning the relationship, the medical score at the next time of the patient can be predicted from the physical examination information of the patient at a past certain time.
However, the constitution of patients in different regions is obviously different, and when the above scheme is used for learning all sample data in public data, the relationship between the physical examination information of samples in a region with fewer samples in a period of time and the medical score at the next moment cannot be effectively learned, and the medical score of the patients in the region with fewer samples cannot be accurately determined.
Disclosure of Invention
The invention provides a medical score determining method and device, electronic equipment and a storage medium, which are used for improving the accuracy of determining a medical score.
In a first aspect, an embodiment of the present invention provides a medical score determining method, including:
determining a target characteristic sequence according to a target characteristic value corresponding to a preset characteristic of a target to be evaluated in a first preset time before the current time;
inputting the target feature sequence into a Long Short-Term Memory (LSTM) model of a target to obtain a predicted medical score of the target to be evaluated at a target time, wherein the target time is a second preset time after the current time, and the first training LSTM model is a trained LSTM model corresponding to a target area where the target to be evaluated is located;
obtaining the first trained LSTM model by:
and performing migration training on a second training LSTM model according to the first sample in the database corresponding to the target area to obtain the first training LSTM model, wherein the second training LSTM model is obtained by training the initial LSTM model based on the second sample in the public database.
In the scheme, if the initial LSTM model is trained directly according to the first sample in the database corresponding to the target area, the relevant characteristics of the first sample in the target area can be learned, but the quantity of the first sample may not be enough to train the LSTM model, or the learned LSTM model has poor learning characteristic capability, the second trained LSTM model is obtained by training based on the second sample in the public database, namely the second trained LSTM model with strong learning characteristic capability is trained through enough samples, then the LSTM model is subjected to transfer training according to the first sample, so that the first trained LSTM model has strong characteristic learning capability and can accurately predict the medical score of the patient in the target area, therefore, the target characteristic sequence corresponding to the target to be scored in a past period of time is input into the first trained LSTM model, the predicted medical score of the target to be scored can be accurately obtained.
In a possible implementation manner, performing migration training on a second training LSTM model according to a first sample in a database corresponding to the target area to obtain the first training LSTM model includes:
determining a first sample characteristic sequence according to a first sample characteristic value corresponding to the preset characteristic in a first time interval;
and performing migration training on the second training LSTM model according to the first sample characteristic sequence and a first sample medical score corresponding to the first sample at a first time to obtain the first training LSTM model, wherein the first time is a first preset time period before a first preset time, the first time is a second preset time period after the first preset time, and the first preset time is one or more preset times.
In the above scheme, the first training LSTM model is obtained by training the second training LSTM model according to the first sample feature sequence corresponding to the first sample at a certain duration (first preset duration) before the first preset time and the medical score corresponding to the first sample at a time after the first preset time by another duration (second preset duration), when the first training LSTM model is used, a target feature sequence corresponding to the same time length (first preset time length) before the current time of the target to be scored is input into the first preset time length of the target, the predicted medical score corresponding to the time (target time) after the current time of the target to be scored is output, wherein the time (target time) is the same time length (second preset time length) after the current time, namely, the time intervals corresponding to the medical scores of the characteristic sequences in the training and using processes of the first training LSTM model are the same, so that the accuracy of the medical score prediction is further improved.
In a possible implementation manner, determining a target feature sequence according to a target feature value corresponding to a preset feature of a target to be scored in a first preset time before a current time includes:
and combining target characteristic values corresponding to the preset characteristics of the target to be scored in each time interval into the target characteristic sequence according to the time sequence of each time interval contained in the first preset time interval.
According to the scheme, the LSTM model can learn the corresponding relation between the characteristic values and the medical scores, and also can learn the corresponding relation between the sequence of the characteristic values and the medical scores, namely the same characteristic values, but different sequences can correspond to different medical scores, the first preset duration is divided into a plurality of time intervals, the target characteristic values corresponding to the preset characteristics in the time intervals are combined according to the time sequence of the time intervals, namely the target characteristic sequence is obtained according to the time sequence, and the predicted medical scores can be determined more accurately.
In a possible implementation manner, the determining, before the target feature sequence is determined, the preset features include static features and dynamic features, the first preset time period includes multiple time periods, and according to a target feature value corresponding to a preset feature of the target to be scored in the first preset time period before the current time, the determining further includes:
selecting a target dynamic characteristic value representing the dynamic characteristic of the corresponding time period from all dynamic characteristic values corresponding to the dynamic characteristic of the target to be scored in each time period in the plurality of time periods;
and taking a target static characteristic value corresponding to the static characteristic of the target to be evaluated and a target dynamic characteristic value representing the dynamic characteristic of each time interval as a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the corresponding time interval.
According to the scheme, the target dynamic characteristic value representing the dynamic characteristic of each time interval is selected, the target static characteristic value and the target dynamic characteristic value representing the dynamic characteristic of each time interval are directly used as the target characteristic values corresponding to the preset characteristics of the target to be evaluated in the corresponding time interval, so that a comprehensive and proper target characteristic value can be conveniently obtained, the accuracy of medical evaluation prediction is improved, and the calculation amount of the first training LSTM model is reduced.
In a possible implementation manner, the selecting, from all the dynamic feature values corresponding to the dynamic features of the target to be scored in each of the plurality of time periods, a target dynamic feature value characterizing the dynamic features of the corresponding time period includes:
selecting at least one dynamic characteristic value as a dynamic characteristic value representing the corresponding type characteristic of the corresponding time period from the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value corresponding to each type of characteristic of the multiple types of characteristics of the target to be evaluated in each time period;
and taking the dynamic characteristic values of all types of characteristics representing each time interval as target dynamic characteristic values representing the dynamic characteristics of the corresponding time interval.
According to the scheme, the dynamic characteristics comprise various types of characteristics, the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value corresponding to each type in each time period can reflect the type characteristics of the time period to a certain extent, at least one dynamic characteristic value is selected from the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value to be used as a dynamic characteristic value for representing the corresponding type characteristics of the corresponding time period, the dynamic characteristic values representing all the type characteristics of each time period are used as target dynamic characteristic values representing the dynamic characteristics of the corresponding time period, the target dynamic characteristic values accurately representing the dynamic characteristics of each time period can be obtained, the calculated amount of the first training LSTM model is reduced, the efficiency of determining the predicted medical score is improved, and the method is more suitable for the medical field.
In one possible implementation, the second trained LSTM model is obtained by:
determining a second sample characteristic sequence according to a second sample characteristic value corresponding to the preset characteristic of the second sample in a second time interval;
and training an initial LSTM model according to the second sample feature sequence and a second sample medical score corresponding to the second sample at a second time to obtain a second training LSTM model, wherein the second time period is a first preset time period before a second preset time, the second time is a time which is a second preset time period after the second preset time, and the second preset time is one or more preset times.
In a second aspect, an embodiment of the present invention provides a medical score determining apparatus, including:
the determining module is used for determining a target characteristic sequence according to a target characteristic value corresponding to preset characteristics of a target to be evaluated in a first preset time before the current time;
the score obtaining module is used for inputting the target feature sequence into a first training LSTM model to obtain the predicted medical score of the target to be scored at a target moment, wherein the target moment is a moment with a second preset time length after the current moment, and the first training LSTM model is an LSTM model corresponding to a target area where the target to be scored is located;
and the model training module is used for carrying out migration training on a second training LSTM model according to a first sample in the database corresponding to the target area to obtain the first training LSTM model, wherein the second training LSTM model is obtained by training an initial LSTM model based on a second sample in a public database.
In a possible implementation manner, the model training module is specifically configured to:
determining a first sample characteristic sequence according to a first sample characteristic value corresponding to the preset characteristic in a first time interval;
and performing migration training on the second training LSTM model according to the first sample characteristic sequence and a first sample medical score corresponding to the first sample at a first time to obtain the first training LSTM model, wherein the first time is a first preset time period before a first preset time, the first time is a second preset time period after the first preset time, and the first preset time is one or more preset times.
In a possible implementation manner, the determining module determines the target feature sequence according to a target feature value corresponding to a preset feature of the target to be scored in a first preset time before the current time, where the determining module includes:
and combining target characteristic values corresponding to the preset characteristics of the target to be scored in each time interval into the target characteristic sequence according to the time sequence of each time interval contained in the first preset time interval.
In a possible implementation manner, the preset features include a static feature and a dynamic feature, the first preset duration includes a plurality of time periods, and the determining module is further configured to:
before determining a target characteristic sequence according to a target characteristic value corresponding to a preset characteristic of a target to be evaluated in a first preset time before the current time,
selecting a target dynamic characteristic value representing the dynamic characteristic of the corresponding time period from all dynamic characteristic values corresponding to the dynamic characteristic of the target to be scored in each time period in the plurality of time periods;
and taking a target static characteristic value corresponding to the static characteristic of the target to be evaluated and a target dynamic characteristic value representing the dynamic characteristic of each time interval as a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the corresponding time interval.
In a possible implementation manner, the dynamic feature includes multiple types of features, and the determining module selects a target dynamic feature value characterizing the dynamic feature of a corresponding time period from all dynamic feature values corresponding to the dynamic feature of the target to be scored in each time period of the multiple time periods, including:
selecting at least one dynamic characteristic value as a dynamic characteristic value representing the corresponding type characteristic of the corresponding time period from the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value corresponding to each type of characteristic of the multiple types of characteristics of the target to be evaluated in each time period;
and taking the dynamic characteristic values of all types of characteristics representing each time interval as target dynamic characteristic values representing the dynamic characteristics of the corresponding time interval.
In one possible implementation, the model training module is further configured to:
determining a second sample characteristic sequence according to a second sample characteristic value corresponding to the preset characteristic of the second sample in a second time interval;
and training an initial LSTM model according to the second sample feature sequence and a second sample medical score corresponding to the second sample at a second time to obtain a second training LSTM model, wherein the second time period is a first preset time period before a second preset time, the second time is a time which is a second preset time period after the second preset time, and the second preset time is one or more preset times.
In a third aspect, an embodiment of the present invention provides an electronic device, including: a processor and a memory;
wherein the memory stores program code which, when executed by the processor, causes the processor to perform a medical scoring method as described in the first aspect above.
In a fourth aspect, embodiments of the present invention provide a computer-readable storage medium on which is stored a computer program which, when executed by a processor, implements a medical scoring method as described in the first aspect above.
In addition, for technical effects brought by any one implementation manner of the second aspect to the fourth aspect, reference may be made to technical effects brought by different implementation manners of the first aspect, and details are not described here.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
Fig. 1 is a schematic flow chart of a first medical score determination method according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a second medical score determination method provided by an embodiment of the invention;
FIG. 3 is a schematic flow chart of a first method for obtaining a first training LSTM model according to the embodiment of the present invention;
FIG. 4 is a schematic flow chart of a second method for obtaining a first trained LSTM model according to the embodiment of the present invention;
fig. 5 is a schematic structural diagram of a medical score determining apparatus according to an embodiment of the present invention;
fig. 6 is a schematic block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail with reference to the accompanying drawings, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. 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 invention.
The term "and/or" in the embodiments of the present invention describes an association relationship of associated objects, and indicates that three relationships may exist, for example, a and/or B may indicate: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.
The terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present application, "a plurality" means two or more unless otherwise specified.
In the description of the present application, it is to be noted that, unless otherwise explicitly stated or limited, the term "connected" is to be understood broadly, and may for example be directly connected, indirectly connected through an intermediate medium, or be a communication between two devices. The specific meaning of the above terms in the present application can be understood in a specific case by those of ordinary skill in the art.
In the medical field, the severity of a disease of a patient can be known according to the medical score of the patient, and the method is very important for diagnosis and treatment of the disease. In the related art, the relationship between the physical examination information of the sample at a certain time and the medical score at the next time is obtained based on the public database, and by learning the relationship, the medical score at the next time of the patient can be predicted and obtained according to the physical examination information of the patient at a past certain time.
However, the constitution of patients in different regions is obviously different, and when all sample data in the public data are learned, the above scheme cannot effectively learn the relationship between the physical examination information of samples in the regions with fewer samples in a period of time and the medical score at the next moment, so that the medical score of the patients in the regions with fewer samples cannot be accurately determined.
The embodiment of the invention provides a method, a device, electronic equipment and a storage medium for determining a medical score, wherein a second training LSTM model is obtained by training based on a second sample in a public database, namely, the second training LSTM model with stronger learning characteristic capacity is trained by enough samples, and then the second training LSTM model is subjected to transfer training according to the first sample, so that the obtained first training LSTM model has stronger characteristic learning capacity and can accurately predict the medical score of a patient in a target area, and therefore, a target characteristic sequence corresponding to a target to be scored in the target area within a past period of time is input into the first training LSTM model, and the predicted medical score of the target to be scored at a target moment can be accurately obtained.
The following describes the technical solutions of the present application and how to solve the above technical problems with specific embodiments. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present application will be described below with reference to the accompanying drawings.
Fig. 1 is a schematic flow chart of a first medical score determining method provided in an embodiment of the present application, and as shown in fig. 1, the method may include:
step 101: and determining a target characteristic sequence according to a target characteristic value corresponding to a preset characteristic of the target to be evaluated in a first preset time before the current time.
In this embodiment, the future medical grade of the target to be evaluated needs to be predicted according to the related historical data of the target to be evaluated, and based on this, the historical data of the target to be evaluated needs to be obtained first and processed correspondingly.
The target characteristic value corresponding to the preset characteristic of the target to be evaluated in the first preset time before the current time is the parameter corresponding to the different body indexes of the target to be evaluated in the time, and the determined target characteristic sequence reflects the change of the parameter corresponding to the different body indexes of the target to be evaluated in the time along with the time. In this embodiment, the physical index may be various physical indexes, such as body weight, urine volume, PF ratio, mechanical ventilation, antibiotics, age, lymphocyte count, blood glucose level, prothrombin time, and the like.
In this embodiment, a first preset duration may be set according to an actual application scenario, where the longer the first preset duration is, the more accurate the prediction result is, but the calculation amount may also increase. Take the first preset time period as 5 hours as an example:
the first preset time period is 5 hours, and the current time is 12 pm, so that the first preset time period before the current time is 7 pm-12 pm earlier.
The above parameters are merely exemplary, and the present embodiment does not specifically limit the above parameters.
Step 102: and inputting the target characteristic sequence into a first training LSTM model to obtain the predicted medical score of the target to be scored at the target moment.
The target time is a time which is a second preset time after the current time, and the first training LSTM model is a trained LSTM model corresponding to a target area where the target to be evaluated is located.
In this embodiment, a future medical score of the target to be scored needs to be predicted according to the relevant historical data of the target to be scored, the first training LSTM model corresponding to the target area where the target to be scored is located learns the relation between the relevant historical data of the first sample in the database corresponding to the target area and the future medical score, and the medical score at the time of a second preset duration after the current time can be obtained by inputting the target feature sequence of the target to be scored in the historical time into the first training LSTM model.
In this embodiment, the second preset duration may be set according to an actual application scenario, which is not limited in this embodiment.
The first trained LSTM model may be obtained by:
and performing migration training on a second training LSTM model according to the first sample in the database corresponding to the target area to obtain the first training LSTM model, wherein the second training LSTM model is obtained by training the initial LSTM model based on the second sample in the public database.
If the initial LSTM model is trained directly according to the first sample in the database corresponding to the target region, the relevant features of the first sample in the target region can be learned, but generally, the number of samples in the database corresponding to one region is small, so the number of the first samples may not be enough to train the LSTM model, or the trained LSTM model has poor feature learning ability; if the initial LSTM model is trained directly from the second sample in the public database, the relevant features of the first sample in the target region may not be well learned.
In this embodiment, a second training LSTM model is obtained by training based on a second sample in a public database, that is, a second training LSTM model with strong learning characteristic capability is trained through enough samples, and then the LSTM model is subjected to migration training according to the first sample, so that the obtained first training LSTM model has strong characteristic learning capability and can accurately predict the medical score of a patient in the target region, and therefore, a target characteristic sequence corresponding to a target to be evaluated in a past period of time is input into the first training LSTM model, and the predicted medical score of the target to be evaluated can be accurately obtained.
In this embodiment, the database corresponding to the target area and the public database are not limited, and the public database may include a database corresponding to the target area, that is, the public database includes all the first samples; the public database may also partially include a database corresponding to the target region, that is, the public database includes a part of the first sample; the common database may also not comprise the database corresponding to the target area at all, i.e. the common database does not contain any first samples.
The public database may be any public clinical database containing a number of samples greater than a predetermined number threshold, such as a Medical Information Market for Intensive Care (MIMIC) database.
The present embodiment does not limit the type of the medical score, for example, the medical score is a Sequential Organ Failure Assessment (SOFA) score.
In some specific embodiments, the medical score is a SOFA score, and after a predicted SOFA score of the target to be scored at the target time is obtained, information carrying the predicted SOFA score is sent in a preset notification manner, so that after the information is received by a target person, symptom analysis is performed on the target to be scored based on the predicted SOFA score, treatment measures are taken in advance, and the like.
According to the scheme, the second training LSTM model is obtained through training based on the second sample in the public database, namely the second training LSTM model with strong learning characteristic capacity is trained through enough samples, and then the LSTM model is subjected to transfer training according to the first sample, so that the obtained first training LSTM model has strong characteristic learning capacity and can accurately predict the medical score of the patient in the target area, therefore, the target characteristic sequence corresponding to the target to be evaluated in the target area within a past period of time is input into the first training LSTM model, and the predicted medical score of the target to be evaluated can be accurately obtained.
Based on that the LSTM model learns not only the correspondence between the feature values and the medical scores but also the correspondence between the order of the feature values and the medical scores, i.e. the same feature values, but different orders may correspond to different medical scores, in the medical score determining method of some embodiments, the first preset duration may be divided into a plurality of time intervals to determine the predicted medical scores more accurately, as shown in fig. 2, and the method may include:
step 201: and selecting a target dynamic characteristic value representing the dynamic characteristic of the corresponding time period from all the dynamic characteristic values corresponding to the dynamic characteristic of the target to be scored in each time period in the plurality of time periods.
In this embodiment, the feature values of some features for different preset features are not changed over time or in a short time (a first preset time period), for example: body weight, urine volume, PF ratio, mechanical ventilation, antibiotics, age, etc.;
and the characteristic values of some characteristics can be changed in real time, such as lymphocyte count, blood sugar content, prothrombin time and the like.
Based on this, the preset features are classified into two types: static features that do not change over time or for short periods of time, and dynamic features that can change in real time.
In this embodiment, the feature values of the target to be scored corresponding to the static features in the multiple time periods are not changed, and the feature values corresponding to the static features in each time period do not need to be determined; the feature values of the target to be evaluated corresponding to the dynamic features in the plurality of time periods may change, and the feature values corresponding to the dynamic features in each time period need to be determined.
In this embodiment, if there are many types of dynamic features and feature values are collected more frequently, the dynamic features in each time period correspond to a large number of dynamic feature values, and the dynamic feature values are directly used as target dynamic feature values representing the dynamic features in the corresponding time period, so that the calculation amount of the first training LSTM model is large due to the large number of target dynamic feature values, the efficiency of determining the predicted medical score is low, and the timeliness requirement of the medical field for determining the predicted medical score is high.
Based on this, in some embodiments, the selecting of the target dynamic feature value representing the dynamic feature of the corresponding time period from all the dynamic feature values corresponding to the dynamic feature of the target to be scored in each time period of the multiple time periods may be implemented by, but is not limited to:
selecting at least one dynamic characteristic value as a dynamic characteristic value representing the corresponding type characteristic of the corresponding time period from the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value corresponding to each type of characteristic of the multiple types of characteristics of the target to be evaluated in each time period;
and taking the dynamic characteristic values of all types of characteristics representing each time interval as target dynamic characteristic values representing the dynamic characteristics of the corresponding time interval.
The first preset time is 5 hours, the current time is 12 pm, and the first preset time before the current time is 7 am-12 pm, as an example, the following description is made:
each hour is divided into 5 periods, namely 7 am-12 am: 7 am-8 am, noted as epoch 1, 8 am-9 am, noted as epoch 2, 9 am-10 am, noted as epoch 3, 10 am-11 am, noted as epoch 4, 11 am-12 am, noted as epoch 5. Each time interval has different types of dynamic characteristics, which are marked as characteristic 1, characteristic 2, characteristic 3, … … and characteristic n, and the characteristic value of the characteristic 1 of the object to be evaluated in the time interval 1 has A11、A12、……A1kWherein A is11Is the maximum of these characteristic values, A13Is the minimum of these characteristic values, A1kThe median of these characteristic values is designated A1A can be substituted11、A13、A1kAnd A1As a dynamic characteristic value of characteristic 1 characterizing period 1; the characteristic value of the characteristic 1 of the object to be evaluated in the time interval 2 has B11、B12、……B1kWherein B is15Is the maximum of these characteristic values, B13Is the minimum of these characteristic values, B12The median of these characteristic values is designated B1B can be substituted by15、B13、B12B1As the dynamic feature value of the feature 1 in the characterization time period 2, the dynamic feature values of the feature 1 in the characterization time periods 3, 4, and 5 and the dynamic feature values of the various types of features in the other time periods are determined in the same manner, which is not described herein again.
The above embodiment has been described by taking the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value, and the average dynamic characteristic value as the dynamic characteristic values representing the corresponding type characteristics of the corresponding time periods, but the embodiment is not limited thereto, and for example, any one or more of the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value, and the average dynamic characteristic value may be selected as the dynamic characteristic values representing the corresponding type characteristics of the corresponding time periods. It will be appreciated that the dynamic feature values characterizing different types of features for different time periods need to be selected in the same manner.
By selecting at least one dynamic characteristic value as a dynamic characteristic value for representing the corresponding type characteristic of the corresponding time period from the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value, and using the dynamic characteristic values representing all the type characteristics of each time period as target dynamic characteristic values representing the dynamic characteristics of the corresponding time period, the target dynamic characteristic values accurately representing the dynamic characteristics of each time period can be obtained, the calculated amount of the first training LSTM model is reduced, the efficiency of determining the predicted medical score is improved, and the method is more suitable for the medical field.
Step 202: and taking a target static characteristic value corresponding to the static characteristic of the target to be evaluated and a target dynamic characteristic value representing the dynamic characteristic of each time interval as a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the corresponding time interval.
As described above, the preset features include static features that do not change with time or in a short time, and dynamic features that can change in real time. And obtaining a target dynamic characteristic value representing the dynamic characteristic of each time interval through the steps, and taking the target dynamic characteristic value and a target static characteristic value corresponding to the static characteristic as target characteristic values corresponding to preset characteristics in the time interval.
The first preset time is 5 hours, the current time is 12 pm, the first preset time before the current time is 7 am-12 pm, and each hour is taken as an example to be explained:
a in the characteristic value of the characteristic 1 of the time interval 111(maximum value), A13(minimum value), A1k(median) and A1(average) as the dynamic eigenvalue of the characteristic 1 characterizing the period 1; a in the feature value of feature 2 of time interval 124(maximum value), A22(minimum value), A23(median) and A2(average value) as the dynamic feature value characterizing feature 2 of time period 1, … …, A is the feature value of feature n of time period 1n3(maximum value), An6(minimum value), An1(median) and An(mean) as the dynamics of the feature n characterizing the period 1And (4) characteristic value. The target static characteristic value corresponding to the static characteristic and A corresponding to the characteristic 111、A13、A1kAnd A1Feature 2 corresponds to A24、A22、A23And A2… …, A corresponding to feature nn3、An6、An1And AnAnd target characteristic values corresponding to preset characteristics in the time interval 1 are taken as targets to be evaluated. Target characteristic values corresponding to preset characteristics of the target to be scored in the time periods 2, 3, 4 and 5 are respectively determined in the same manner, and details are not repeated here.
The above description is only an example to illustrate how to obtain the target characteristic value corresponding to the preset characteristic of the target to be scored in each time period, and the embodiment does not limit the selection of the target dynamic characteristic value and the selection of other parameters.
By selecting the target dynamic characteristic value representing the dynamic characteristic of each time interval, and directly taking the target static characteristic value and the target dynamic characteristic value representing the dynamic characteristic of each time interval as the target characteristic value corresponding to the preset characteristic of the target to be evaluated in the corresponding time interval, the comprehensive and proper target characteristic value can be conveniently obtained, the accuracy of medical evaluation prediction is improved, and the calculated amount of the first training LSTM model is reduced.
Step 203: and combining target characteristic values corresponding to the preset characteristics of the target to be scored in each time interval into the target characteristic sequence according to the time sequence of each time interval contained in the first preset time interval.
As described above, the LSTM model not only learns the correspondence between the feature values and the medical scores, but also learns the correspondence between the order of the feature values and the medical scores, that is, the same feature values, but different orders may correspond to different medical scores, and after the first preset duration is divided into a plurality of time periods, the target feature values corresponding to the preset features in each time period need to be combined according to the time order of each time period, that is, the target feature sequence is obtained according to the time sequence.
The first preset time is 5 hours, the current time is 12 pm, the first preset time before the current time is 7 am-12 pm, and each hour is taken as an example to be explained:
7 am-8 am, noted as epoch 1, 8 am-9 am, noted as epoch 2, 9 am-10 am, noted as epoch 3, 10 am-11 am, noted as epoch 4, 11 am-12 am, noted as epoch 5. According to the time sequence of the time periods 1-5, combining a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the time period 1, a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the time period 2, a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the time period 3, a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the time period 4 and a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the time period 5 into a target characteristic sequence.
The above is only an example of how to obtain the target feature sequence, and the embodiment is not limited thereto.
The first preset duration is divided into a plurality of time intervals, and the target characteristic values corresponding to the preset characteristics in the time intervals are combined according to the time sequence of the time intervals, namely the target characteristic sequence is obtained according to the time sequence, so that the predicted medical grade can be determined more accurately.
Step 204: and inputting the target characteristic sequence into a first training LSTM model to obtain the predicted medical score of the target to be scored at the target moment.
Step 204 is implemented in the same manner as step 102, and is not described herein again.
In some specific embodiments, after obtaining the actual medical score of the target to be scored at the target time, the first training LSTM model is optimized based on the target feature sequence and the actual medical score.
Fig. 3 is a schematic flowchart of a first method for obtaining a first trained LSTM model according to an embodiment of the present application, and as shown in fig. 3, the method may include:
s301: and determining a first sample characteristic sequence according to a first sample characteristic value corresponding to the preset characteristic in a first time period of a first sample in the database corresponding to the target area.
The first time interval is a first preset time interval before a first preset time, and the first preset time is preset one or more times.
In this embodiment, the determining the first sample feature sequence is similar to the determining the target feature sequence, and the difference is only that the current time is replaced by one or more preset times, which can be referred to the above embodiments and is not described herein again.
S302: and performing migration training on the second training LSTM model according to the first sample characteristic sequence and the first sample medical score corresponding to the first sample at the first moment to obtain the first training LSTM model.
And the first moment is a moment which is a second preset duration after the first preset moment, and the second training LSTM model is obtained by training an initial LSTM model based on a second sample in a public database.
In some embodiments, the first trained LSTM model described above may be trained by:
preprocessing data corresponding to the first sample (such as removing redundant data, filling and smoothing incomplete data, and the like); dividing the first sample into a first verification sample, a first training sample and a first test sample based on a preset proportion; determining basic information (such as the number of hidden layer features, the number of full-connection layer features, the learning rate of training rounds and the like) according to the first verification sample; training the second training LSTM model by using the first training sample according to the determined basic information to obtain a trained LSTM model; and testing the trained LSTM model by using the first test sample, and if the target accuracy is reached, taking the trained LSTM model as a first trained LSTM model.
In some specific embodiments, after obtaining the actual medical score of the target to be scored at the target time, the first training LSTM model is optimized based on the target feature sequence and the actual medical score.
In the above scheme, the first training LSTM model is obtained by training the second training LSTM model according to the first sample feature sequence corresponding to the first sample at a certain duration (first preset duration) before the first preset time and the medical score corresponding to the first sample at a time after the first preset time by another duration (second preset duration), when the first training LSTM model is used, a target feature sequence corresponding to the same time length (first preset time length) before the current time of the target to be scored is input into the first preset time length of the target, the predicted medical score corresponding to the time (target time) after the current time of the target to be scored is output, wherein the time (target time) is the same time length (second preset time length) after the current time, namely, the time intervals corresponding to the medical scores of the characteristic sequences in the training and using processes of the first training LSTM model are the same, so that the accuracy of the medical score prediction is further improved.
Fig. 4 is a schematic flowchart of a second method for obtaining a first trained LSTM model according to an embodiment of the present application, and as shown in fig. 4, the method may include:
s401: and determining a second sample characteristic sequence according to a second sample characteristic value corresponding to the preset characteristic in a second time period of a second sample in the public database.
The second time interval is a time interval of the first preset duration before a second preset time, and the second preset time is preset one or more times.
In this embodiment, the determining the second sample feature sequence is similar to the determining the target feature sequence, and the difference is only that the current time is replaced by one or more preset times, which can be referred to in the foregoing embodiments and is not described herein again.
S402: and training the initial LSTM model according to the second sample characteristic sequence and the second sample medical score corresponding to the second sample at the second time to obtain a second training LSTM model.
And the second moment is a moment which is a second preset time after the second preset moment.
In some embodiments, the first trained LSTM model described above may be trained by:
preprocessing the data corresponding to the second sample (e.g., removing redundant data, padding and smoothing incomplete data, etc.); dividing the second sample into a second verification sample, a second training sample and a second test sample based on a preset proportion; determining basic information (such as the number of hidden layer features, the number of full-connection layer features, the learning rate of training rounds and the like) according to the second verification sample; training the initial LSTM model by using a second training sample according to the determined basic information to obtain a trained LSTM model; and testing the trained LSTM model by using a second test sample, and if the target accuracy is reached, taking the trained LSTM model as a second trained LSTM model.
S403: and determining a first sample characteristic sequence according to a first sample characteristic value corresponding to the preset characteristic in a first time period of a first sample in the database corresponding to the target area.
S404: and performing migration training on the second training LSTM model according to the first sample characteristic sequence and the first sample medical score corresponding to the first sample at the first moment to obtain the first training LSTM model.
The first time interval is a time interval of the first preset time before the first preset time, the first time is a time interval of the second preset time after the first preset time, and the first preset time is preset one or more times.
The steps 403 and 404 are implemented in the same manner as the steps 201 and 202, and are not described herein again.
As shown in fig. 5, based on the same inventive concept, an embodiment of the present invention provides a medical score determining apparatus 500, including: a determination module 501, a score obtaining module 502 and a model training module 503.
The determining module 501 is configured to determine a target feature sequence according to a target feature value corresponding to a preset feature of a target to be scored within a first preset time before a current time;
a score obtaining module 502, configured to input the target feature sequence into a first training LSTM model, so as to obtain a predicted medical score of the target to be scored at a target time, where the target time is a time after a current time by a second preset time, and the first training LSTM model is an LSTM model corresponding to a target area where the target to be scored is located;
the model training module 503 is configured to perform migration training on a second training LSTM model according to a first sample in the database corresponding to the target area to obtain the first training LSTM model, where the second training LSTM model is obtained by training an initial LSTM model based on a second sample in a public database.
In a possible implementation manner, the model training module 503 is specifically configured to:
determining a first sample characteristic sequence according to a first sample characteristic value corresponding to the preset characteristic in a first time interval;
and performing migration training on the second training LSTM model according to the first sample characteristic sequence and a first sample medical score corresponding to the first sample at a first time to obtain the first training LSTM model, wherein the first time is a first preset time period before a first preset time, the first time is a second preset time period after the first preset time, and the first preset time is one or more preset times.
In a possible implementation manner, the determining module 501 determines the target feature sequence according to a target feature value corresponding to a preset feature of the target to be scored in a first preset time before the current time, where the determining includes:
and combining target characteristic values corresponding to the preset characteristics of the target to be scored in each time interval into the target characteristic sequence according to the time sequence of each time interval contained in the first preset time interval.
In a possible implementation manner, the preset features include a static feature and a dynamic feature, the first preset duration includes a plurality of time periods, and the determining module 501 is further configured to:
before determining a target characteristic sequence according to a target characteristic value corresponding to a preset characteristic of a target to be evaluated in a first preset time before the current time,
selecting a target dynamic characteristic value representing the dynamic characteristic of the corresponding time period from all dynamic characteristic values corresponding to the dynamic characteristic of the target to be scored in each time period in the plurality of time periods;
and taking a target static characteristic value corresponding to the static characteristic of the target to be evaluated and a target dynamic characteristic value representing the dynamic characteristic of each time interval as a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the corresponding time interval.
In a possible implementation manner, the dynamic features include multiple types of features, and the determining module 501 selects a target dynamic feature value characterizing the dynamic features of a corresponding time period from all dynamic feature values corresponding to the dynamic features of the target to be scored in each time period of the multiple time periods, including:
selecting at least one dynamic characteristic value as a dynamic characteristic value representing the corresponding type characteristic of the corresponding time period from the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value corresponding to each type of characteristic of the multiple types of characteristics of the target to be evaluated in each time period;
and taking the dynamic characteristic values of all types of characteristics representing each time interval as target dynamic characteristic values representing the dynamic characteristics of the corresponding time interval.
In a possible implementation manner, the model training module 503 is further configured to:
determining a second sample characteristic sequence according to a second sample characteristic value corresponding to the preset characteristic of the second sample in a second time interval;
and training an initial LSTM model according to the second sample feature sequence and a second sample medical score corresponding to the second sample at a second time to obtain a second training LSTM model, wherein the second time period is a first preset time period before a second preset time, the second time is a time which is a second preset time period after the second preset time, and the second preset time is one or more preset times.
Since the apparatus is the apparatus in the method in the embodiment of the present invention, and the principle of the apparatus for solving the problem is similar to that of the method, the implementation of the apparatus may refer to the implementation of the method, and repeated details are not repeated.
As shown in fig. 6, based on the same inventive concept, an embodiment of the present invention provides an electronic device 600 including: a processor 601 and a memory 602;
a memory 602 for storing computer programs executed by the processor 601. The memory 602 may be a volatile memory (volatile memory), such as a random-access memory (RAM); the memory 602 may also be a non-volatile memory (non-volatile memory) such as, but not limited to, a read-only memory (rom), a flash memory (flash memory), a Hard Disk Drive (HDD) or a solid-state drive (SSD), or any other medium which can be used to carry or store desired program code in the form of instructions or data structures and which can be accessed by a computer. The memory 602 may be a combination of the above.
The processor 601 may include one or more Central Processing Units (CPUs), Graphics Processing Units (GPUs), or digital Processing units (dsps), etc.
The embodiment of the present invention does not limit the specific connection medium between the memory 602 and the processor 601. In fig. 6, the memory 602 and the processor 601 are connected by a bus 603, the bus 603 is represented by a thick line in fig. 6, and the connection manner between other components is merely illustrative and not limited. The bus 603 may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown in FIG. 6, but this is not intended to represent only one bus or type of bus.
Wherein the memory 602 stores program code which, when executed by the processor 601, causes the processor 601 to perform the following:
determining a target characteristic sequence according to a target characteristic value corresponding to a preset characteristic of a target to be evaluated in a first preset time before the current time;
inputting the target characteristic sequence into a first training LSTM model to obtain the predicted medical score of the target to be scored at a target moment, wherein the target moment is a moment with a second preset duration after the current moment, and the first training LSTM model is a trained LSTM model corresponding to a target area where the target to be scored is located;
obtaining the first trained LSTM model by:
and performing migration training on a second training LSTM model according to the first sample in the database corresponding to the target area to obtain the first training LSTM model, wherein the second training LSTM model is obtained by training the initial LSTM model based on the second sample in the public database.
In one possible implementation, the processor 601 is specifically configured to:
determining a first sample characteristic sequence according to a first sample characteristic value corresponding to the preset characteristic in a first time interval;
and performing migration training on the second training LSTM model according to the first sample characteristic sequence and a first sample medical score corresponding to the first sample at a first time to obtain the first training LSTM model, wherein the first time is a first preset time period before a first preset time, the first time is a second preset time period after the first preset time, and the first preset time is one or more preset times.
In one possible implementation, the processor 601 is specifically configured to:
and combining target characteristic values corresponding to the preset characteristics of the target to be scored in each time interval into the target characteristic sequence according to the time sequence of each time interval contained in the first preset time interval.
In a possible implementation manner, the preset features include a static feature and a dynamic feature, the first preset duration includes a plurality of time periods, and the processor 601 is further configured to:
before determining a target characteristic sequence according to a target characteristic value corresponding to a preset characteristic of a target to be evaluated in a first preset time before the current time,
selecting a target dynamic characteristic value representing the dynamic characteristic of the corresponding time period from all dynamic characteristic values corresponding to the dynamic characteristic of the target to be scored in each time period in the plurality of time periods;
and taking a target static characteristic value corresponding to the static characteristic of the target to be evaluated and a target dynamic characteristic value representing the dynamic characteristic of each time interval as a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the corresponding time interval.
In one possible implementation, the processor 601 is specifically configured to:
selecting at least one dynamic characteristic value as a dynamic characteristic value representing the corresponding type characteristic of the corresponding time period from the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value corresponding to each type of characteristic of the multiple types of characteristics of the target to be evaluated in each time period;
and taking the dynamic characteristic values of all types of characteristics representing each time interval as target dynamic characteristic values representing the dynamic characteristics of the corresponding time interval.
In one possible implementation, the processor 601 is specifically configured to:
determining a second sample characteristic sequence according to a second sample characteristic value corresponding to the preset characteristic of the second sample in a second time interval;
and training an initial LSTM model according to the second sample feature sequence and a second sample medical score corresponding to the second sample at a second time to obtain a second training LSTM model, wherein the second time period is a first preset time period before a second preset time, the second time is a time which is a second preset time period after the second preset time, and the second preset time is one or more preset times.
In the embodiment of the present invention, the electronic device may be a device with certain computing capability, such as a personal computer, a mobile phone, a tablet computer, a notebook, an e-book reader, and the like.
Since the electronic device is the electronic device that executes the method in the embodiment of the present invention, and the principle of the electronic device to solve the problem is similar to that of the method, the implementation of the electronic device may refer to the implementation of the method, and repeated details are not described again.
Embodiments of the present invention provide a computer-readable medium having stored thereon a computer program that, when executed by a processor, performs the steps of the medical score determination method described above. The storable medium may be, among other things, a non-volatile storable medium.
The present application is described above with reference to block diagrams and/or flowchart illustrations of methods, apparatus (systems) and/or computer program products according to embodiments of the application. It will be understood that one block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, and/or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer and/or other programmable data processing apparatus, create means for implementing the functions/acts specified in the block diagrams and/or flowchart block or blocks.
Accordingly, the subject application may also be embodied in hardware and/or in software (including firmware, resident software, micro-code, etc.). Furthermore, the present application may take the form of a computer program product on a computer-usable or computer-readable storage medium having computer-usable or computer-readable program code embodied in the medium for use by or in connection with an instruction execution system. In the context of this application, a computer-usable or computer-readable medium may be any medium that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
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 invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (10)

1. A method for determining a medical score, comprising:
determining a target characteristic sequence according to a target characteristic value corresponding to a preset characteristic of a target to be evaluated in a first preset time before the current time;
inputting the target characteristic sequence into a first training long-short term memory network (LSTM) model to obtain a predicted medical score of the target to be evaluated at a target moment, wherein the target moment is a moment with a second preset duration after the current moment, and the first training LSTM model is a trained LSTM model corresponding to a target area where the target to be evaluated is located;
obtaining the first trained LSTM model by:
and performing migration training on a second training LSTM model according to the first sample in the database corresponding to the target area to obtain the first training LSTM model, wherein the second training LSTM model is obtained by training the initial LSTM model based on the second sample in the public database.
2. The method of claim 1, wherein performing migration training on a second training LSTM model according to a first sample in a database corresponding to the target region to obtain the first training LSTM model, comprises:
determining a first sample characteristic sequence according to a first sample characteristic value corresponding to the preset characteristic in a first time interval;
and performing migration training on the second training LSTM model according to the first sample characteristic sequence and a first sample medical score corresponding to the first sample at a first time to obtain the first training LSTM model, wherein the first time is a first preset time period before a first preset time, the first time is a second preset time period after the first preset time, and the first preset time is one or more preset times.
3. The method according to claim 1, wherein determining the target feature sequence according to a target feature value corresponding to a preset feature of the target to be scored within a first preset time period before the current time comprises:
and combining target characteristic values corresponding to the preset characteristics of the target to be scored in each time interval into the target characteristic sequence according to the time sequence of each time interval contained in the first preset time interval.
4. The method according to claim 1, wherein the preset features include static features and dynamic features, the first preset time period includes a plurality of time periods, and before determining the target feature sequence according to a target feature value corresponding to the preset features of the target to be scored in the first preset time period before the current time, the method further includes:
selecting a target dynamic characteristic value representing the dynamic characteristic of the corresponding time period from all dynamic characteristic values corresponding to the dynamic characteristic of the target to be scored in each time period in the plurality of time periods;
and taking a target static characteristic value corresponding to the static characteristic of the target to be evaluated and a target dynamic characteristic value representing the dynamic characteristic of each time interval as a target characteristic value corresponding to the preset characteristic of the target to be evaluated in the corresponding time interval.
5. The method according to claim 4, wherein the dynamic features comprise a plurality of types of features, and selecting a target dynamic feature value characterizing the dynamic features of a corresponding time period from all dynamic feature values corresponding to the dynamic features of the target to be scored in each time period of the plurality of time periods comprises:
selecting at least one dynamic characteristic value as a dynamic characteristic value representing the corresponding type characteristic of the corresponding time period from the minimum dynamic characteristic value, the maximum dynamic characteristic value, the median dynamic characteristic value and the average dynamic characteristic value corresponding to each type of characteristic of the multiple types of characteristics of the target to be evaluated in each time period;
and taking the dynamic characteristic values of all types of characteristics representing each time interval as target dynamic characteristic values representing the dynamic characteristics of the corresponding time interval.
6. The method of any of claims 1 to 5, wherein the second trained LSTM model is obtained by:
determining a second sample characteristic sequence according to a second sample characteristic value corresponding to the preset characteristic of the second sample in a second time interval;
and training an initial LSTM model according to the second sample feature sequence and a second sample medical score corresponding to the second sample at a second time to obtain a second training LSTM model, wherein the second time period is a first preset time period before a second preset time, the second time is a time which is a second preset time period after the second preset time, and the second preset time is one or more preset times.
7. A medical score determination device, comprising:
the determining module is used for determining a target characteristic sequence according to a target characteristic value corresponding to preset characteristics of a target to be evaluated in a first preset time before the current time;
the score obtaining module is used for inputting the target feature sequence into a first training LSTM model to obtain the predicted medical score of the target to be scored at a target moment, wherein the target moment is a moment with a second preset time length after the current moment, and the first training LSTM model is an LSTM model corresponding to a target area where the target to be scored is located;
and the model training module is used for carrying out migration training on a second training LSTM model according to a first sample in the database corresponding to the target area to obtain the first training LSTM model, wherein the second training LSTM model is obtained by training an initial LSTM model based on a second sample in a public database.
8. The apparatus of claim 7, wherein the model training module is specifically configured to:
determining a first sample characteristic sequence according to a first sample characteristic value corresponding to the preset characteristic in a first time interval;
and performing migration training on the second training LSTM model according to the first sample characteristic sequence and a first sample medical score corresponding to the first sample at a first time to obtain the first training LSTM model, wherein the first time is a first preset time period before a first preset time, the first time is a second preset time period after the first preset time, and the first preset time is one or more preset times.
9. An electronic device, comprising: a processor and a memory;
wherein the memory stores program code which, when executed by the processor, causes the processor to perform the steps of the medical score determination method of any of claims 1-6.
10. A storage medium having stored therein a computer program which, when executed by a processor, implements a medical score determination method according to any one of claims 1 to 6.
CN202110285202.5A 2021-03-17 2021-03-17 Medical score determining method and device, electronic equipment and storage medium Pending CN113066580A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110285202.5A CN113066580A (en) 2021-03-17 2021-03-17 Medical score determining method and device, electronic equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110285202.5A CN113066580A (en) 2021-03-17 2021-03-17 Medical score determining method and device, electronic equipment and storage medium

Publications (1)

Publication Number Publication Date
CN113066580A true CN113066580A (en) 2021-07-02

Family

ID=76561146

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110285202.5A Pending CN113066580A (en) 2021-03-17 2021-03-17 Medical score determining method and device, electronic equipment and storage medium

Country Status (1)

Country Link
CN (1) CN113066580A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114297454A (en) * 2021-12-30 2022-04-08 医渡云（北京）技术有限公司 Method and device for discretizing features, electronic equipment and computer readable medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111275691A (en) * 2020-01-22 2020-06-12 北京邮电大学 Small sample tumor necrosis rate classification prediction device based on deep learning
CN111798980A (en) * 2020-07-10 2020-10-20 哈尔滨工业大学（深圳） Complex medical biological signal processing method and device based on deep learning network
CN112365978A (en) * 2020-11-10 2021-02-12 北京航空航天大学 Method and device for establishing early risk assessment model of tachycardia event
CN112382394A (en) * 2020-11-05 2021-02-19 苏州麦迪斯顿医疗科技股份有限公司 Event processing method and device, electronic equipment and storage medium

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111275691A (en) * 2020-01-22 2020-06-12 北京邮电大学 Small sample tumor necrosis rate classification prediction device based on deep learning
CN111798980A (en) * 2020-07-10 2020-10-20 哈尔滨工业大学（深圳） Complex medical biological signal processing method and device based on deep learning network
CN112382394A (en) * 2020-11-05 2021-02-19 苏州麦迪斯顿医疗科技股份有限公司 Event processing method and device, electronic equipment and storage medium
CN112365978A (en) * 2020-11-10 2021-02-12 北京航空航天大学 Method and device for establishing early risk assessment model of tachycardia event

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114297454A (en) * 2021-12-30 2022-04-08 医渡云（北京）技术有限公司 Method and device for discretizing features, electronic equipment and computer readable medium

Similar Documents

Publication Publication Date Title
US9861308B2 (en) Method and system for monitoring stress conditions
US20120270199A1 (en) Methods and systems for assessing latent traits using probabilistic scoring
US20170092145A1 (en) System, method and non-transitory computer readable storage medium for truly reflecting ability of testee through online test
CN111651677B (en) Course content recommendation method, apparatus, computer device and storage medium
US20230080350A1 (en) Methods and Apparatus for Diagnosis of Progressive Kidney Function Decline Using a Machine Learning Model
CN107610009B (en) Trinity enrollment probability prediction method based on neural network
EP4027348A2 (en) Affinity prediction method and apparatus, method and apparatus for training affinity prediction model, device and medium
EP3618080A1 (en) Control method and reinforcement learning for medical system
CN112635053A (en) Big data-based resident health early warning method, device, equipment and system
CN112447270A (en) Medication recommendation method, device, equipment and storage medium
CN107506606A (en) Common disease Risk Forecast Method and system
CN113066580A (en) Medical score determining method and device, electronic equipment and storage medium
CN117612712A (en) Method and system for detecting and improving cognition evaluation diagnosis precision
CN112927788B (en) Physical examination item recommendation method, device, equipment and storage medium
CN116994751A (en) Method and device for constructing pre-eclampsia early-stage risk prediction model
CN109615204B (en) Quality evaluation method, device and equipment of medical data and readable storage medium
CN109192306A (en) A kind of judgment means of diabetes, equipment and computer readable storage medium
CN112669973B (en) Disease collaborative progressive prediction method based on big data deep learning and robot
CN114090733A (en) Learning resource recommendation method and device, storage medium and electronic equipment
CN114330859A (en) Optimization method, system and equipment for real-time quality control
Schena et al. Understanding patient needs and predicting outcomes in IgA nephropathy using data analytics and artificial intelligence: a narrative review
Ruan et al. Discrimination, calibration, and point estimate accuracy of GRU-D-Weibull architecture for real-time individualized endpoint prediction
CN116525123B (en) Medical examination ground element feedback system and method based on analysis model
Sørensen et al. Medical databases
US20220223284A1 (en) Systems and methods for modelling a human subject

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