CN116910105A - Medical information query system and method based on pre-training large model - Google Patents
Medical information query system and method based on pre-training large model Download PDFInfo
- Publication number
- CN116910105A CN116910105A CN202311168022.4A CN202311168022A CN116910105A CN 116910105 A CN116910105 A CN 116910105A CN 202311168022 A CN202311168022 A CN 202311168022A CN 116910105 A CN116910105 A CN 116910105A
- Authority
- CN
- China
- Prior art keywords
- medical information
- query
- information
- instruction
- model
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000012549 training Methods 0.000 title claims abstract description 18
- 239000003814 drug Substances 0.000 claims description 11
- 238000001356 surgical procedure Methods 0.000 claims description 11
- 229940079593 drug Drugs 0.000 claims description 9
- 238000009966 trimming Methods 0.000 claims description 7
- 230000009471 action Effects 0.000 claims description 6
- 238000012800 visualization Methods 0.000 claims description 4
- 238000009877 rendering Methods 0.000 claims description 3
- 230000004044 response Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 4
- 238000007689 inspection Methods 0.000 description 4
- 238000003745 diagnosis Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000036772 blood pressure Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000007711 solidification Methods 0.000 description 1
- 230000008023 solidification Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2457—Query processing with adaptation to user needs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/242—Query formulation
- G06F16/2433—Query languages
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2453—Query optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a medical information query system and a method based on a pre-training large model, which belong to the technical field of data processing, and the method comprises the following steps: s1, constructing an ODR database and an instruction fine adjustment data set in a medical data retrieval scene; s2, performing model fine adjustment on the large language model to obtain a medical information retrieval model; s3, receiving medical information inquiry conditions and converting the medical information inquiry conditions into standard retrieval instructions; s4, inquiring and retrieving the required medical information based on the standard retrieval instruction; and S5, visually displaying the retrieved required medical information to realize medical information inquiry. According to the invention, the natural language understanding capability learned by the open source pre-training language big model is utilized, the instruction fine-tuning data set of the medical scene is introduced, the context information is automatically filled through the service system under different service scenes, the data retrieval capability of the natural language conforming to the habit of the doctor is provided for the doctor, the supporting capability of the data retrieval system is expanded, and the use cost of the doctor is reduced.
Description
Technical Field
The invention belongs to the technical field of data processing, and particularly relates to a medical information query system and method based on a pre-training large model.
Background
Current medical information systems are divided into a number of subsystems, such as LIS system (Laboratory Information System), for managing patient examination-related information, including laboratory equipment management, examination applications, and examination results, among others, according to different functions. The PACS system (Picture Archiving and Communication System) is used for managing medical image information of a patient, including storage, transmission and diagnosis of medical images such as CT, MRI, X-ray films and the like. The EMR system (Electronic Medical Record) is an information system for managing the electronic medical records of a patient, including information about the patient's basic information, medical history, diagnostic results, orders, and the like.
The systems are independent of each other, medical staff need to log in different systems for inquiring related information, the inquired conditions and displayed digital fields are relatively solidified, personalized inquiry requirements of doctors cannot be met, for example, the statistics inquiry conditions and the displayed result fields are difficult to exhaust in advance when people treat diabetics and cost are treated in the past 3 years in each department, many information systems solve the long tail requirements by directly opening a custom sql inquiry function, but the requirements on the skills of the medical staff are high through sql inquiry, and the information systems are generally difficult to use.
In summary, the information retrieval function provided by the current medical information system for doctors has the following two problems:
1) The query conditions and result fields are cured in advance, and personalized query requirements of doctors cannot be met.
2) The comprehensive query or custom query interface is very complex, the technical threshold is higher, the study and use cost of doctors is high, and the query efficiency is low.
Disclosure of Invention
Aiming at the defects in the prior art, the medical information query system and method based on the pre-training large model provided by the invention solve the problems of solidification of query conditions, high query technical threshold and low query efficiency of the existing medical information query method.
In order to achieve the aim of the invention, the invention adopts the following technical scheme: a medical information query system based on a pre-trained large model, comprising:
model training module: the method comprises the steps of fine-tuning a language big model according to an instruction fine-tuning data set in a medical data retrieval scene to obtain a medical information retrieval model;
ODR database: the information query database is used for storing a medical information data table based on the sql query statement and used as a medical information retrieval model;
front-end business workstation: the medical information query condition is used for receiving medical information query conditions of a user;
an information retrieval module: the medical information search module is used for perfecting medical information search conditions and converting the medical information search conditions into standard search instructions to be input into a medical information search model;
and the information query module: the method comprises the steps of generating an sql query statement according to a standard retrieval instruction input into a medical information retrieval model, and then querying and retrieving required medical information in an ODR database;
an information visualization component: the medical information query system is embedded in the front-end business workstation and is used for visually displaying the needed medical information for the user to realize medical information query.
Further, the fine tuning instruction content in the instruction fine tuning data set comprises a data table field description, current context information and query description text, and the corresponding instruction action is an sql query statement;
the data table field is described as a data table prompt word of medical information, the current context information is a data table field value related to the data table prompt word, and the query description text is a target query medical information.
Further, the medical information data table stored in the ODR database includes a patient information table, a hospitalization record table, an order record table, a check record table, a surgery schedule table, a surgery process record table, a vital sign record table, a surgery consumable consumption table, and a surgery medicine consumption table;
each data table is provided with a plurality of corresponding medical information details, and any medical information detail is used as a data table prompt word of the corresponding data table.
Further, the medical information retrieval model comprises a text-based medical information retrieval model and a voice-based medical information retrieval model;
wherein the text-based medical information retrieval model corresponds to the instruction fine-tuning dataset in text format;
the voice-based medical information retrieval model fine-tunes the data set in response to instructions after converting the voice to text.
A medical information query method, comprising the steps of:
s1, constructing an ODR database and an instruction fine adjustment data set in a medical data retrieval scene;
s2, performing model fine adjustment on the large language model by using the instruction fine adjustment data set to obtain a medical information retrieval model;
s3, receiving medical information query conditions of the user and converting the medical information query conditions into standard retrieval instructions;
s4, inquiring and retrieving the required medical information based on the standard retrieval instruction;
and S5, visually displaying the retrieved required medical information to realize medical information inquiry.
Further, in the step S1, the instruction trimming dataset is used to generate sql query statements for implementing medical information query;
the sql query statement generated by the instruction trimming dataset comprises a single-table query and a multi-table association query;
the fine tuning instruction content in the instruction fine tuning data set comprises a data table field description, current context information and a query description text, and the corresponding instruction action is an sql query statement; the data table field is described as a data table prompt word of medical information, the current context information is a data table field value related to the data table prompt word, and the query description text is a target query medical information.
Further, the step S3 specifically includes:
s31, a received medical information query condition;
the medical information inquiry conditions comprise voice information conditions and text information conditions, and when the medical information inquiry conditions are voice information conditions, the medical information inquiry conditions are converted into corresponding text information conditions;
s32, automatically filling and perfecting the context information of the text information condition according to the current medical scene, and converting the context information into a standard retrieval instruction;
wherein, the standard retrieval instruction is a data table field description+current context information+query description text.
Further, the step S4 specifically includes:
s41, outputting a corresponding sql query statement by using a medical information retrieval model according to a standard retrieval instruction;
s42, calling the needed medical information in the ODR database according to the sql query statement.
Further, the step S5 specifically includes:
s51, performing data rendering according to the data format of the medical information required to be called;
and S52, displaying the medical information rendered by the data in each business scene of each front-end business workstation in a general page, so as to realize medical information inquiry.
The beneficial effects of the invention are as follows:
according to the invention, the natural language understanding capability learned by the open source pre-training language big model is utilized, the instruction fine-tuning data set of the medical scene is introduced on the basis of the semantic relation among the words, and the context information is automatically filled through the business system under different business scenes, so that the data retrieval capability of the natural language conforming to the habit of a doctor can be provided for the doctor, the supporting capability of the data retrieval system is greatly expanded, and meanwhile, the use cost of the doctor is greatly reduced.
Drawings
Fig. 1 is a flowchart of a medical information query method based on a pre-training large model provided by the invention.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and all the inventions which make use of the inventive concept are protected by the spirit and scope of the present invention as defined and defined in the appended claims to those skilled in the art.
Example 1:
the embodiment of the invention provides a medical information query system based on a pre-training large model, which comprises the following steps:
model training module: the method comprises the steps of fine-tuning a language big model according to an instruction fine-tuning data set in a medical data retrieval scene to obtain a medical information retrieval model;
ODR database: the information query database is used for storing a medical information data table based on the sql query statement and used as a medical information retrieval model;
front-end business workstation: the medical information query condition is used for receiving medical information query conditions of a user;
an information retrieval module: the medical information search module is used for perfecting medical information search conditions and converting the medical information search conditions into standard search instructions to be input into a medical information search model;
and the information query module: the method comprises the steps of generating an sql query statement according to a standard retrieval instruction input into a medical information retrieval model, and then querying and retrieving required medical information in an ODR database;
an information visualization component: the medical information query system is embedded in the front-end business workstation and is used for visually displaying the needed medical information for the user to realize medical information query.
In the embodiment of the invention, a plurality of open-source language large models (such as ziya13B and chatGLM 6B) are available at present, the language large models perform unsupervised learning on a plurality of different data sets, including wiki, gitsub, text2sql and the like, the semantic relation of language context is learned, and the knowledge is deposited in the network parameters of the large models, so that the input text can be understood and perceived to a certain extent.
In the embodiment of the invention, a command fine adjustment data set is added aiming at a medical data information query scene based on such a large model, and the fine adjustment command content in the command fine adjustment data set in the embodiment of the invention comprises a data table field description, current context information and a query description text, wherein the corresponding command action is an sql query statement;
the data table field is described as a data table prompt word of medical information, the current context information is a data table field value related to the data table prompt word, and the query description text is a target query medical information; specifically, the main function of the context information is to automatically complement the query limit conditions, so that medical staff is saved from repeatedly describing the query limit conditions by natural language.
In the embodiment of the invention, the current open-source pre-training large model can generate correct sql (such as form query) for some simple natural languages, but the open-source pre-training large model can not generate correct sql for some complex scenes (such as multi-table association query), at the moment, a fine adjustment sample needs to be added, fine adjustment in instruction fine adjustment data in the embodiment can comb out two tables and multi-table association query samples according to actual application scenes, and even if all scene samples are not exhausted, the language large model has certain generalization capability; meanwhile, in the actual running process, the support can be rapidly upgraded by adding a sample mode, and compared with the traditional mode of upgrading by modifying codes, the method is greatly improved.
In the embodiment, the context information is added in the fine adjustment sample, so that the convenience of user inquiry is improved, and the medical information retrieval model after fine adjustment automatically acquires inquiry conditions from the context information.
Examples of instruction trimming data sets for medical data scenes provided by the embodiment of the invention are shown in table 1:
table 1: instruction fine-tuning dataset examples
The medical information data table stored in the ODR database in the embodiment of the invention comprises a patient information table, a hospitalization record table, a doctor's advice record table, an inspection record table, a surgery schedule table, a surgery process record table, a vital sign record table, a surgery consumable consumption table and a surgery medicine consumption table;
each data table is provided with a plurality of corresponding medical information details, and any medical information detail is used as a data table prompt word of the corresponding data table.
Specifically, in the present embodiment, the details of the medical information corresponding to each medical information data table are as follows:
1. patient information table (t_trans_patient): patient_no, patient name, gender, and update_time.
2. Hospitalization record table (t_event_in_record): patient no, hospital no, admissions date, chip completions, admissions diagnosis, diagnostic physics and update time.
3. Order record table (t_trans_media_ad device): patient no, hospital no, type (type: 1-test, 2-check, 3-treatment, 4-western, 5-consumable, 6-transfusion, 7-care, 8-meal), content, status (status: nurse received, executing), sign_sector (doctor down), sign_time (time down), and update_time (time up).
4. Check record table (t_test_rec): the method includes the steps of event_no (patient number), hospital_no (hospitalization number), test_application_dept_no (examination application department number), test_application_dept_name (examination application department name), application_dot_no (application doctor number), test_report_dept_no (examination report department number), test_report_dept_name (examination report department name), report_dot_no (report_dot_no), test_type (examination category), test_index_name (examination index name), test_index_result (examination index result), test_res_type_name (examination result name), and update_time (update time).
5. Surgical schedule (t_trans_operation_schedule): the patient_no, the hospital_no, the schedule_id, the octor_name, the operation_status, the room_name, the sequence_time, the schedule_time (operation date, format: 2020-01-01), the operation_name, and the update_time (update time).
6. Surgical procedure record table (t_trans_operation_process): patient no, hospital no, schedule id, operation node name, operator time, and update time.
7. Vital sign record table (t_trans_operation_vital sign): patient no, hospital no, schedule id, heart rate, blood pressure, saO2, measurement time, and update time.
8. Surgical consumable consumption table (t_trans_operation_cst): patient no, hospital no, schedule id, cst num, update time.
9. Surgical drug consumption table (t_trans_operation_drug): patient no, hospital no, schedule id, drug num, drug unit, and update time.
Based on the medical information data table, an example of a complete fine tuning instruction is as follows:
Input:
there are 9 medical information tables: 1. patient information table (t_trans_patient): patient_no, patient name, gender, and update_time. 2. Hospitalization record table (t_event_in_record): patient no, hospital no, admissions date, chip completions, admissions diagnosis, diagnostic physics and update time. 3. Order record table (t_trans_media_ad device): patient no, hospital no, type (type: 1-test, 2-check, 3-treatment, 4-western, 5-consumable, 6-transfusion, 7-care, 8-meal), content, status (status: nurse received, executing), sign_sector (doctor down), sign_time (time down), and update_time (time up). 4. Check record table (t_test_rec): the method includes the steps of event_no (patient number), hospital_no (hospitalization number), test_application_dept_no (examination application department number), test_application_dept_name (examination application department name), application_dot_no (application doctor number), test_report_dept_no (examination report department number), test_report_dept_name (examination report department name), report_dot_no (report_dot_no), test_type (examination category), test_index_name (examination index name), test_index_result (examination index result), test_res_type_name (examination result name), and update_time (update time). 5. Surgical schedule (t_trans_operation_schedule): the patient_no, the hospital_no, the schedule_id, the octor_name, the operation_status, the room_name, the sequence_time, the schedule_time (operation date, format: 2020-01-01), the operation_name, and the update_time (update time). 6. Surgical procedure record table (t_trans_operation_process): patient no, hospital no, schedule id, operation node name, operator time, and update time. 7. Vital sign record table (t_trans_operation_vital sign): patient no, hospital no, schedule id, heart rate, blood pressure, saO2, measurement time, and update time. 8. Surgical consumable consumption table (t_trans_operation_cst): patient no, hospital no, schedule id, cst num, update time. 9. Surgical drug consumption table (t_trans_operation_drug): patient no, hospital no, schedule id, drug num, drug unit, and update time.
Current context information: (room_name=28, sequence=2, parent_no=123, schedule_id=12345).
Generating the correct sql statement: inquiring the next operation preparation condition.
Output:
SELECT*
FROMt_trans_operation_schedule
WHEREroom_name='28'
ANDsequence>2orderbysequencelimit1
The model subjected to fine adjustment based on the fine adjustment instruction has certain generalization capability, and corresponding sql sentences can be generated according to the prompt instruction.
The medical information retrieval model in the embodiment of the invention comprises a text-based medical information retrieval model and a voice-based medical information retrieval model;
wherein the text-based medical information retrieval model corresponds to the instruction fine-tuning dataset in text format;
the voice-based medical information retrieval model fine-tunes the data set in response to instructions after converting the voice to text.
In this embodiment, by setting a voice-based medical information retrieval model, for a voice recognition service provided by a medical staff in a scene (such as an operating room) where it is inconvenient to input characters by hand, the voice model may adopt an open-source deep neural network pre-training model (such as a wenet), and a medical scene fine-tuning data set is added on the basis of the pre-training model. The data set can automatically generate corresponding voice data through a text2speech tool by using the natural language text of the instruction fine tuning data set in the last step as input to form a voice-text fine tuning data set under a medical retrieval scene, and then training a voice recognition model of the medical retrieval scene based on the data set.
Example 2:
the embodiment of the invention provides a medical information query method based on the medical information query system in the embodiment 1, as shown in fig. 1, comprising the following steps:
s1, constructing an ODR database and an instruction fine adjustment data set in a medical data retrieval scene;
s2, performing model fine adjustment on the large language model by using the instruction fine adjustment data set to obtain a medical information retrieval model;
s3, receiving medical information query conditions of the user and converting the medical information query conditions into standard retrieval instructions;
s4, inquiring and retrieving the required medical information based on the standard retrieval instruction;
and S5, visually displaying the retrieved required medical information to realize medical information inquiry.
In step S1 of the embodiment of the present invention, the instruction trimming dataset is used to generate sql query statements for implementing medical information query;
the sql query statement generated by the instruction trimming dataset comprises a single-table query and a multi-table association query;
the fine tuning instruction content in the instruction fine tuning data set comprises a data table field description, current context information and a query description text, and the corresponding instruction action is an sql query statement; the data table field is described as a data table prompt word of medical information, the current context information is a data table field value related to the data table prompt word, and the query description text is a target query medical information.
The step S3 of the embodiment of the invention specifically comprises the following steps:
s31, a received medical information query condition;
the medical information inquiry conditions comprise voice information conditions and text information conditions, and when the medical information inquiry conditions are voice information conditions, the medical information inquiry conditions are converted into corresponding text information conditions;
s32, automatically filling and perfecting the context information of the text information condition according to the current medical scene, and converting the context information into a standard retrieval instruction;
wherein, the standard retrieval instruction is a data table field description+current context information+query description text.
In this embodiment, the context information, i.e. the environment information in which the user is located, is automatically acquired by the front-end service system in which the user is located, i.e. the inter-operative (room_name) in the operating room scene, the operating table time (sequence), the patient number (patient), the schedule_id (schedule_id), and the so-called auto-filling is that the information is assembled into a trimming instruction string, such as "current context information: (room_name=28, sequence=2, parent_no=123, schedule_id=12345) ".
Specifically, in an operating room scenario, the context information includes an operating room name (room_name=28), a current session (sequence=2), a patient number (patient_no=123), and an operation number (schedule_id=12345). The business system delivers the parameters current context information + query description text to invoke the natural language retrieval service. The service system automatically fills the context information, so that the integrity requirement of the input information of a user can be greatly reduced, the user experience is improved, for example, in an operating room scene, the inspection report of the current patient is required to be inquired, the complete information is not required to be input, the complete information is in the form of ' the inspection information of 28 patients with the name of 28 operating rooms ' is required to be inquired ', the ' the inspection information of the patient is required to be input ', and the search service can automatically acquire the inquiry condition according to the context information transmitted by the service system and convert the inquiry condition into a standard search instruction.
The step S4 of the embodiment of the invention specifically comprises the following steps:
s41, outputting a corresponding sql query statement by using a medical information retrieval model according to a standard retrieval instruction;
s42, calling the needed medical information in the ODR database according to the sql query statement.
Specifically, when a query request of a service system is received, a complete prompt instruction is filled in, the query request is formatted into a data table field description, current context information and query description text, then the complete prompt instruction is used as input of a medical information retrieval model service, a large model which is finely tuned based on a medical retrieval scene instruction set is called to output corresponding sql statements, a database is queried according to the sql statements returned by the medical information retrieval model, and corresponding needed medical information data is acquired.
In the embodiment of the present invention, step S5 specifically includes:
s51, performing data rendering according to the data format of the medical information required to be called;
and S52, displaying the medical information rendered by the data in each business scene of each front-end business workstation in a general page, so as to realize medical information inquiry.
Specifically, the front page of the information visualization component can be rendered into different styles according to the returned data format, for example, text-value two-column data can be displayed in a form and histogram mode, for example, time-value two-column data can be returned, and can be displayed in a line diagram mode, and can be displayed in a two-dimensional form by default.
The principles and embodiments of the present invention have been described in detail with reference to specific examples, which are provided to facilitate understanding of the method and core ideas of the present invention; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Those of ordinary skill in the art will recognize that the embodiments described herein are for the purpose of aiding the reader in understanding the principles of the present invention and should be understood that the scope of the invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific modifications and combinations from the teachings of the present disclosure without departing from the spirit thereof, and such modifications and combinations remain within the scope of the present disclosure.
Claims (9)
1. A medical information query system based on a pre-trained large model, comprising:
model training module: the method comprises the steps of fine-tuning a language big model according to an instruction fine-tuning data set in a medical data retrieval scene to obtain a medical information retrieval model;
ODR database: the information query database is used for storing a medical information data table based on the sql query statement and used as a medical information retrieval model;
front-end business workstation: the medical information query condition is used for receiving medical information query conditions of a user;
an information retrieval module: the medical information search module is used for perfecting medical information search conditions and converting the medical information search conditions into standard search instructions to be input into a medical information search model;
and the information query module: the method comprises the steps of generating an sql query statement according to a standard retrieval instruction input into a medical information retrieval model, and then querying and retrieving required medical information in an ODR database;
an information visualization component: the medical information query system is embedded in the front-end business workstation and is used for visually displaying the needed medical information for the user to realize medical information query.
2. The medical information query system based on a pre-trained large model according to claim 1, wherein the fine tuning instruction content in the instruction fine tuning dataset comprises a data table field description, current context information and query description text, the corresponding instruction action of which is an sql query statement;
the data table field is described as a data table prompt word of medical information, the current context information is a data table field value related to the data table prompt word, and the query description text is a target query medical information.
3. The pre-trained large model based medical information query system of claim 1, wherein the medical information data tables stored in the ODR database include a patient information table, a hospitalization record table, an order record table, a test record table, a surgical shift table, a surgical procedure record table, a vital sign record table, a surgical consumable consumption table, and a surgical drug consumption table;
each data table is provided with a plurality of corresponding medical information details, and any medical information detail is used as a data table prompt word of the corresponding data table.
4. The pre-trained large model based medical information query system of claim 1, wherein the medical information retrieval model comprises a text-based medical information retrieval model and a speech-based medical information retrieval model;
wherein the text-based medical information retrieval model corresponds to the instruction fine-tuning dataset in text format;
the voice-based medical information retrieval model fine-tunes the data set in response to instructions after converting the voice to text.
5. The medical information query method of the medical information query system based on the pre-training large model according to any one of claims 1 to 4, comprising the steps of:
s1, constructing an ODR database and an instruction fine adjustment data set in a medical data retrieval scene;
s2, performing model fine adjustment on the large language model by using the instruction fine adjustment data set to obtain a medical information retrieval model;
s3, receiving medical information query conditions of the user and converting the medical information query conditions into standard retrieval instructions;
s4, inquiring and retrieving the required medical information based on the standard retrieval instruction;
and S5, visually displaying the retrieved required medical information to realize medical information inquiry.
6. The medical information query method according to claim 5, wherein in the step S1, the instruction trim dataset is used to generate sql query statements for implementing medical information queries;
the sql query statement generated by the instruction trimming dataset comprises a single-table query and a multi-table association query;
the fine tuning instruction content in the instruction fine tuning data set comprises a data table field description, current context information and a query description text, and the corresponding instruction action is an sql query statement; the data table field is described as a data table prompt word of medical information, the current context information is a data table field value related to the data table prompt word, and the query description text is a target query medical information.
7. The medical information query method according to claim 5, wherein the step S3 specifically includes:
s31, a received medical information query condition;
the medical information inquiry conditions comprise voice information conditions and text information conditions, and when the medical information inquiry conditions are voice information conditions, the medical information inquiry conditions are converted into corresponding text information conditions;
s32, automatically filling and perfecting the context information of the text information condition according to the current medical scene, and converting the context information into a standard retrieval instruction;
wherein, the standard retrieval instruction is a data table field description+current context information+query description text.
8. The medical information query method according to claim 7, wherein the step S4 specifically includes:
s41, outputting a corresponding sql query statement by using a medical information retrieval model according to a standard retrieval instruction;
s42, calling the needed medical information in the ODR database according to the sql query statement.
9. The medical information query method according to claim 8, wherein the step S5 specifically includes:
s51, performing data rendering according to the data format of the medical information required to be called;
and S52, displaying the medical information rendered by the data in each business scene of each front-end business workstation in a general page, so as to realize medical information inquiry.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311168022.4A CN116910105A (en) | 2023-09-12 | 2023-09-12 | Medical information query system and method based on pre-training large model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311168022.4A CN116910105A (en) | 2023-09-12 | 2023-09-12 | Medical information query system and method based on pre-training large model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116910105A true CN116910105A (en) | 2023-10-20 |
Family
ID=88360609
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311168022.4A Pending CN116910105A (en) | 2023-09-12 | 2023-09-12 | Medical information query system and method based on pre-training large model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116910105A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117709441A (en) * | 2024-02-06 | 2024-03-15 | 云南联合视觉科技有限公司 | Method for training professional medical large model through gradual migration field |
| CN117709441B (en) * | 2024-02-06 | 2024-05-03 | 云南联合视觉科技有限公司 | Method for training professional medical large model through gradual migration field |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109036585A (en) * | 2018-07-24 | 2018-12-18 | 合肥康之恒机械科技有限公司 | A kind of medical platform construction method of servicing and system online |
| CN111813802A (en) * | 2020-09-11 | 2020-10-23 | 杭州量之智能科技有限公司 | Method for generating structured query statement based on natural language |
| CN111897829A (en) * | 2020-05-26 | 2020-11-06 | 华瑞新智科技（北京）有限公司 | Natural language query method and equipment for medical software |
| CN113495900A (en) * | 2021-08-12 | 2021-10-12 | 国家电网有限公司大数据中心 | Method and device for acquiring structured query language sentences based on natural language |
| CN113918589A (en) * | 2020-07-10 | 2022-01-11 | 阿里巴巴集团控股有限公司 | Query statement generation method, correlation method and device |
| CN114168720A (en) * | 2021-10-22 | 2022-03-11 | 福建亿榕信息技术有限公司 | Natural language data query method and storage device based on deep learning |
| CN114528312A (en) * | 2022-02-16 | 2022-05-24 | 京东科技信息技术有限公司 | Method and device for generating structured query language statement |
| CN114547072A (en) * | 2022-02-10 | 2022-05-27 | 招商银行股份有限公司 | Method, system, equipment and storage medium for converting natural language query into SQL |
| CN114897163A (en) * | 2022-05-23 | 2022-08-12 | 阿里巴巴（中国）有限公司 | Pre-training model data processing method, electronic device and computer storage medium |
| CN115292352A (en) * | 2022-08-22 | 2022-11-04 | 中国工商银行股份有限公司 | Question query method, device, equipment, medium and program product |
| CN116340352A (en) * | 2023-03-20 | 2023-06-27 | 中国工商银行股份有限公司 | Data query method and device, storage medium and electronic equipment |
| CN116340584A (en) * | 2023-05-24 | 2023-06-27 | 杭州悦数科技有限公司 | Implementation method for automatically generating complex graph database query statement service |
| CN116501758A (en) * | 2023-06-28 | 2023-07-28 | 四创科技有限公司 | NetCDF data query method and terminal |
-
2023
- 2023-09-12 CN CN202311168022.4A patent/CN116910105A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109036585A (en) * | 2018-07-24 | 2018-12-18 | 合肥康之恒机械科技有限公司 | A kind of medical platform construction method of servicing and system online |
| CN111897829A (en) * | 2020-05-26 | 2020-11-06 | 华瑞新智科技（北京）有限公司 | Natural language query method and equipment for medical software |
| CN113918589A (en) * | 2020-07-10 | 2022-01-11 | 阿里巴巴集团控股有限公司 | Query statement generation method, correlation method and device |
| CN111813802A (en) * | 2020-09-11 | 2020-10-23 | 杭州量之智能科技有限公司 | Method for generating structured query statement based on natural language |
| CN113495900A (en) * | 2021-08-12 | 2021-10-12 | 国家电网有限公司大数据中心 | Method and device for acquiring structured query language sentences based on natural language |
| CN114168720A (en) * | 2021-10-22 | 2022-03-11 | 福建亿榕信息技术有限公司 | Natural language data query method and storage device based on deep learning |
| CN114547072A (en) * | 2022-02-10 | 2022-05-27 | 招商银行股份有限公司 | Method, system, equipment and storage medium for converting natural language query into SQL |
| CN114528312A (en) * | 2022-02-16 | 2022-05-24 | 京东科技信息技术有限公司 | Method and device for generating structured query language statement |
| CN114897163A (en) * | 2022-05-23 | 2022-08-12 | 阿里巴巴（中国）有限公司 | Pre-training model data processing method, electronic device and computer storage medium |
| CN115292352A (en) * | 2022-08-22 | 2022-11-04 | 中国工商银行股份有限公司 | Question query method, device, equipment, medium and program product |
| CN116340352A (en) * | 2023-03-20 | 2023-06-27 | 中国工商银行股份有限公司 | Data query method and device, storage medium and electronic equipment |
| CN116340584A (en) * | 2023-05-24 | 2023-06-27 | 杭州悦数科技有限公司 | Implementation method for automatically generating complex graph database query statement service |
| CN116501758A (en) * | 2023-06-28 | 2023-07-28 | 四创科技有限公司 | NetCDF data query method and terminal |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117709441A (en) * | 2024-02-06 | 2024-03-15 | 云南联合视觉科技有限公司 | Method for training professional medical large model through gradual migration field |
| CN117709441B (en) * | 2024-02-06 | 2024-05-03 | 云南联合视觉科技有限公司 | Method for training professional medical large model through gradual migration field |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11586808B2 (en) | Insertion of standard text in transcription | |
| CN109584975B (en) | Medical data standardization processing method and device | |
| US20220044812A1 (en) | Automated generation of structured patient data record | |
| JP4649405B2 (en) | Structured document creation method and apparatus | |
| US20130138457A1 (en) | Electronic health record system and method for patient encounter transcription and documentation | |
| US20140343957A1 (en) | Clinical content analytics engine | |
| US20070198250A1 (en) | Information retrieval and reporting method system | |
| US20020069056A1 (en) | Methods and systems for generating documents from voice interactions | |
| CN110782973B (en) | Medical image information grading storage method and device, computer equipment and medium | |
| EP4170670A1 (en) | Medical data processing method and system | |
| DE102005021860A1 (en) | System for managing stored medical audio information | |
| KR20100129016A (en) | Searching system and method of medical information | |
| US20180330820A1 (en) | Content-driven problem list ranking in electronic medical records | |
| US20240020342A1 (en) | System and method for prioritization and presentation of heterogeneous medical data | |
| CN110706774A (en) | Medical record generation method, terminal device and computer readable storage medium | |
| CN112837772A (en) | Pre-inquiry case history generation method and device | |
| US20140278553A1 (en) | Dynamic Superbill Coding Workflow | |
| Liao et al. | Medical data inquiry using a question answering model | |
| CN114912887A (en) | Clinical data entry method and device based on electronic medical record | |
| JP6647668B1 (en) | Data structure, learning method and information providing system for machine learning | |
| JP2020205070A (en) | Medial report creation device and control method therefor, medical report creation system, and program | |
| CN110047569B (en) | Method, device and medium for generating question-answer data set based on chest radiography report | |
| CN116737879A (en) | Knowledge base query method and device, electronic equipment and storage medium | |
| CN116910105A (en) | Medical information query system and method based on pre-training large model | |
| CN116738998A (en) | Medical dialogue multi-granularity semantic annotation system and method based on Web |
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 |