CN115346669A - Medical artificial intelligence reasoning method and system based on low-code programming - Google Patents
Medical artificial intelligence reasoning method and system based on low-code programming Download PDFInfo
- Publication number
- CN115346669A CN115346669A CN202210974297.6A CN202210974297A CN115346669A CN 115346669 A CN115346669 A CN 115346669A CN 202210974297 A CN202210974297 A CN 202210974297A CN 115346669 A CN115346669 A CN 115346669A
- Authority
- CN
- China
- Prior art keywords
- medical
- model
- low
- code
- reasoning
- 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
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
- G06F3/0484—Interaction techniques based on graphical user interfaces [GUI] for the control of specific functions or operations, e.g. selecting or manipulating an object, an image or a displayed text element, setting a parameter value or selecting a range
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
- G06F3/0484—Interaction techniques based on graphical user interfaces [GUI] for the control of specific functions or operations, e.g. selecting or manipulating an object, an image or a displayed text element, setting a parameter value or selecting a range
- G06F3/0486—Drag-and-drop
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/20—Software design
- G06F8/24—Object-oriented
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/30—Creation or generation of source code
- G06F8/31—Programming languages or programming paradigms
- G06F8/315—Object-oriented languages
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/30—Creation or generation of source code
- G06F8/38—Creation or generation of source code for implementing user interfaces
Abstract
The invention provides a medical artificial intelligence reasoning method and system based on low code programming, comprising the following steps: acquiring a plurality of medical artificial intelligence models and a plurality of data sets to be inferred; defining a low-code programming template according to the calculation content and the application method of the artificial intelligence model, and converting the medical artificial intelligence model into the low-code model according to the low-code programming template; selecting a model to be executed and a data set to be executed from all low-code models and all data sets according to prompts in a visual interface, dragging and combining the model to be executed and the data set to be executed in the visual interface, and assigning values for parameters of the model to be executed to generate a medical AI reasoning and calculating task; and constructing a cloud computing engine according to the computing environment required by the reasoning task so as to execute the medical AI reasoning computing task on the accessed cloud to obtain an execution result.
Description
Technical Field
The invention relates to the technical field of low-code development and the technical field of artificial intelligence, in particular to a low-code programming technology for medical artificial intelligence reasoning calculation.
Background
Artificial intelligence technology is gaining more attention in medical research, which is exploring to gradually cover various aspects of the medical field. One of the main researches is to combine a special model of artificial intelligence with medical basic data such as medical history, text, sequencing data, image data and the like to implement reasoning and calculation so as to realize auxiliary support for medical work and improve medical work efficiency. For example, in cancer research, doctors and researchers can accurately predict symptoms of cancer, trends of disease conditions, and the like by performing modeling analysis on medical image data; in the research of ophthalmic diseases, doctors and researchers carry out artificial intelligent auxiliary research based on fundus images, which is beneficial to realizing the automation of the aiming of the fundus diseases and improving the accuracy of research and judgment; based on medical record data, an artificial intelligence model of accurate medical treatment is established, which is beneficial to reducing the cost of accurate medical treatment and has positive effects on early intervention of various diseases and the like.
With the development and maturity of these single-modal and multi-modal artificial intelligence models, the inference usage of the artificial intelligence models gradually becomes an important aspect affecting the usability and the usability of the artificial intelligence models. Due to the characteristic of universal iterative updating of the artificial intelligence model, the inference by applying the relevant model in programming is the best mode for realizing flexible application of the artificial intelligence model, and the model is favorably applied to scenes with various requirements as required.
At present, the inference calculation of the artificial intelligence model is mainly performed by a traditional programming method in the field of computers. For medical staff who mainly uses doctors, traditional programming languages such as Python, R, SPSS, and tool platforms are used to some extent in their work, but they are generally difficult to be well mastered by medical staff. The application of the artificial intelligence model has the professional knowledge background, and the threshold of related professional knowledge causes that medical personnel cannot well implement programming of artificial intelligence reasoning calculation. Medical workers face the problems that a programming method is difficult to master, a programming tool threshold is high, support in the medical field is poor and the like when artificial intelligence reasoning and calculation are realized by utilizing the conventional programming technology.
In the aspect of programming methods, doctors often do not have professional programming training experiences or education backgrounds, and particularly under the high competitive output pressure facing scientific research results, the programming process is focused on the implementation of core scientific research services and hopefully realizes scientific research calculation quickly. The traditional programming language has lower abstraction level, and agile development of medical scientific research services requires a programming framework to realize abstract expression of service level. The need to provide "human programmable" code as a service "in medical research services is urgent.
In the aspect of programming tools, the current programming tools have higher requirements on knowledge reserves and implementation capacity required by the implementation of medical scientific research programming. This severely limits the utility of medical research services and the complexity of the computing environment makes the running of programs difficult to implement. Although new IDEs such as Jupytor notewood are derived, the application still needs to master the grammar knowledge of programming languages such as python and understand the program running method. The artificial intelligence reasoning calculation often needs to be performed by means of novel computing environments such as cloud calculation and heterogeneous computing devices such as a GPU, and therefore common medical workers are difficult to develop flexible and efficient artificial intelligence technology programming application.
Disclosure of Invention
Aiming at the problems that a programming method is difficult and tools are difficult to utilize when artificial intelligence reasoning calculation is applied by medical researchers, the invention can shield the bottom level details of professional programming through a medical AI reasoning low code programming template so as to be suitable for mode support high-efficiency programming in the medical field; the invention provides graphical presentation of medical AI reasoning low-code programming by rendering the network web front end facing the medical AI reasoning low-code programming, and supports dynamic real-time display of programming content, programming results, calculation states and the like; the invention programs a cloud computing engine through medical AI reasoning low code, supports the conversion of a low code program and the execution in the cloud, and is butted to various cloud computing environments; on the basis of the previous technologies, the invention constructs a medical AI reasoning low-code programming system and realizes graphical tool support of implementing AI reasoning calculation for medical scientific researchers in a full life cycle.
Aiming at the defects of the prior art, the invention provides a medical artificial intelligence reasoning method based on low-code programming, which comprises the following steps:
step 1, acquiring a plurality of medical artificial intelligence models and a plurality of data sets to be inferred;
step 2, defining a low-code programming template according to the calculation content and the application method of the artificial intelligence model, and converting the medical artificial intelligence model into the low-code model according to the low-code programming template;
step 3, selecting a model to be executed and a data set to be executed from all low-code models and all data sets according to prompts in a visual interface, dragging and combining the model to be executed and the data set to be executed in the visual interface, and assigning values for parameters of the model to be executed to generate a medical AI reasoning calculation task;
and 4, constructing a cloud computing engine according to the computing environment required by the reasoning task so as to execute the medical AI reasoning computing task on the accessed cloud to obtain an execution result.
The medical artificial intelligence reasoning method based on the low-code programming is characterized in that the visual interface comprises a medical computation scene component, a medical AI model pool component, a medical AI data pool component, a medical AI low-code dragging component and a medical AI model template component; the components cooperate to program the graphical web presentation and operation of the low-code programming;
wherein the medical computing scenario component provides access to a medical computing scenario; the medical AI model pool component can display all the low-code models and support the low-code models to be added into a model groove of an operation panel of the medical AI low-code dragging component in a dragging mode; the medical AI data pool component displays all data sets to be inferred, supports the addition of the data sets to a model input slot of an operation panel of the medical AI low-code dragging component in a dragging mode, and binds with the input of a model in the model slot; the medical AI low-code dragging component is composed of a dragging operation panel, supports the dragging of the contents of other components to the operation panel, and realizes the programming operation in an editing mode; the medical AI model template component is used for displaying template information of the currently selected model.
The medical artificial intelligence reasoning method based on low code programming, wherein the step 3 comprises the following steps:
and after the model to be executed is selected, loading a graphical template of the model in the visual interface, dragging and editing parameters appointed by the graphical template, verifying the editing validity, and rendering and modifying the visual effect in real time.
The medical artificial intelligence reasoning method based on low code programming is characterized in that the step 4 comprises the following steps:
interpreting the instantiation operation of the medical AI reasoning and calculating task, executing a calling script of the instantiation operation, and delivering the calculation to a cloud computing engine;
the cloud computing engine comprises a computing task management module and a cloud computing power access module; the computing task management module and the cloud computing power access module are used for realizing the packaging of the computing task and the computing power management through interaction; the calculation task management module performs instantiation management on the medical AI reasoning calculation task, including the state and control of the calculation task; the cloud computing power access module is used for adapting to the unified cloud computing power and distributing the AI inference computation after the instantiation of the medical AI inference computation task to the corresponding computing environment for execution.
The invention also provides a medical artificial intelligence reasoning system based on low code programming, which comprises:
the system comprises an initial module, a data processing module and a data processing module, wherein the initial module is used for acquiring a plurality of medical artificial intelligence models and a plurality of data sets to be inferred;
the conversion module is used for defining a low-code programming template according to the calculation content and the application method of the artificial intelligence model and converting the medical artificial intelligence model into the low-code model according to the low-code programming template;
the visualization module is used for selecting a model to be executed and a data set to be executed from all low-code models and all data sets according to prompts in a visualization interface, dragging and combining the model to be executed and the data set to be executed in the visualization interface, assigning values to parameters of the model to be executed, and generating a medical AI reasoning and calculating task;
and the execution module is used for constructing a cloud computing engine according to the computing environment required by the reasoning task so as to execute the medical AI reasoning computing task on the accessed cloud and obtain an execution result.
The medical artificial intelligence reasoning system based on the low-code programming is characterized in that the visual interface comprises a medical computation scene component, a medical AI model pool component, a medical AI data pool component, a medical AI low-code dragging component and a medical AI model template component; the components cooperate to program the graphical web presentation and operation of the low-code programming;
wherein the medical computing scenario component provides access to a medical computing scenario; the medical AI model pool component can show all the low-code models and support the low-code models to be added into a model slot of an operation panel of the medical AI low-code dragging component in a dragging mode; the medical AI data pool component displays all data sets to be inferred, supports the addition of the data sets to a model input slot of an operation panel of the medical AI low-code dragging component in a dragging mode, and binds with the input of a model in the model slot; the medical AI low-code dragging component is composed of a dragging operation panel, supports the dragging of the contents of other components to the operation panel, and realizes the programming operation in an editing mode; the medical AI model template component is used for displaying template information of the currently selected model.
The medical artificial intelligence reasoning system based on low code programming is characterized in that the visualization module is used for loading a graphical template of the model in the visualization interface after the model to be executed is selected, dragging and editing parameters specified by the graphical template, verifying the editing validity and rendering and modifying the visualization effect in real time.
The medical artificial intelligence reasoning system based on the low code programming is characterized in that the execution module is used for explaining the instantiation operation of the medical AI reasoning calculation task, executing the calling script of the instantiation operation and delivering the calculation to the cloud computing engine;
the cloud computing engine comprises a computing task management module and a cloud computing power access module; the computing task management module and the cloud computing power access module realize the encapsulation of computing tasks and computing power management through interaction; the calculation task management module performs instantiation management on the medical AI reasoning calculation task, including the state and control of the calculation task; the cloud computing power access module is used for adapting to the unified cloud computing power and distributing the AI inference computation after the instantiation of the medical AI inference computation task to the corresponding computing environment for execution.
The invention also provides a storage medium for storing a program for executing the medical artificial intelligence reasoning method based on the low-code programming.
The invention also provides a client used for the optional medical artificial intelligence reasoning system based on the low-code programming.
According to the scheme, the invention has the advantages that: the method can support medical scientific research personnel to quickly construct a required reasoning calculation program in a dragging mode according to the low-code programming module, and support the corresponding program to execute calculation under a required calculation environment, and meanwhile, the calculation state can be fed back to the scientific research personnel in real time in the calculation process; the invention can also support the provider of the model to package the model quickly according to the low code programming module, and the model is merged into the platform to provide service. FIG. 1 illustrates an interactive interface of the present invention, which can be seen to provide a draggable, graphical and concise way of medical AI inference programming.
Drawings
FIG. 1 is an exemplary diagram of an AI inference low code programming page in accordance with the present invention;
FIG. 2 is a hierarchical diagram of a medical AI inference low code programming system;
FIG. 3 is a block diagram of a method embodying the present invention;
FIG. 4 is a flowchart of AI model inference task construction according to the present invention.
Detailed Description
Aiming at solving the problem that a non-computer professional in the medical field is difficult to develop inference programming based on an artificial intelligence model, the invention provides a medical AI inference low-code programming template by combining the characteristics of the medical artificial intelligence model and the low-code programming technology aiming at the problem that a professional programming method is highly complex, and can support the medical staff to realize high-level general AI inference programming on the premise of not needing to understand the programming knowledge in the computer field; the invention provides a medical AI reasoning low-code programming system aiming at the problem that a professional programming tool is complex and combining a medical AI reasoning low-code programming template, the system adopts a web system framework, a web front-end rendering facing the medical AI reasoning low-code programming is designed in the construction of a web front end, the visual presentation of the AI reasoning low-code programming template and the programming can be realized, a medical AI reasoning low-code programming cloud computing engine technology is provided in the construction of a back-end system, the technical processing of an AI reasoning program at a cloud end can be supported, the system can support medical researchers to develop simple and graphical programming, computing, task management, multiplexing and access management of a model pool and a data pool resource and the like, and efficient graphical programming is realized.
Specifically, the present application includes the following key points:
key point 1: medical AI inference low code programming template
The medical AI inference low-code programming template is a high-level abstract definition of an artificial intelligence model capable of realizing low-code programming, when a medical AI inference programming platform is constructed, a maintainer of the platform or a provider of the artificial intelligence model in the platform can efficiently convert an operable artificial intelligence model into a programmable operator in the platform according to the definition mode of the template, the programmer (such as a doctor) supporting the operator constructs an inference program in a low-code or no-code mode without paying attention to the implementation mechanism, calling mechanism and source of the operator, the configuration of operator input can be realized according to the prompt of operator information, and the output result of the operator is obtained.
The medical AI inference low code model comprises a group of metadata information, such as the fields in the following table 1, and can be packaged into a standard operator, so that the AI model can be used in programming in the form of operator units; when the method is used, a programmer configures input information of an operator instantiation, and after the operator instantiation is executed, the operator instantiation output can be obtained. The definition of the template is shown in table 1, where different fields are defined by people in different roles.
TABLE 1 medical AI inference low code programming template description
The technical effects are as follows: the artificial intelligence model is defined and multiplexed in an operator form, and low-code efficient cooperation of a model provider and an inference user is realized.
Key point 2: medical AI-inference oriented low code programming-oriented web front-end rendering
The web front-end rendering for the medical AI inference low-code programming comprises a plurality of front-end components supporting the medical scene programming, including a medical computation scene component, a medical AI model pool component, a medical AI data pool component, a medical AI low-code dragging component, a medical AI model template component, and the like. The components cooperate to realize graphical web display and operation of the medical AI inference low-code programming.
The components cooperate to support the implementation of medical AI drag-and-drop programming. The medical computation scene component provides access to a medical computation scene, corresponds to a model scene label in the AI low-code programming template, provides scene switching, and can cooperatively switch contents along with scene switching of other components; the medical AI model pool component can display available AI models in the platform and support the components in the model pool to be added into a model groove of an operation panel of the medical AI low-code dragging component in a dragging mode; the medical AI data pool component can display the available data set in the platform, and supports the components in the data pool to be added to the model input slot of the operation panel of the medical AI low-code dragging component in a dragging mode and to be bound with the model input in the panel; the medical AI low-code dragging component mainly comprises a dragging operation panel, and can support the dragging of the contents of other components to the operation panel to realize programming operation in an editing mode; the medical AI model template component can display template information of a currently selected model.
Wherein, the medical AI drag-type programming front end can support the following specific operations: after a specific AI model is dragged and selected, a graphical template of the model is loaded on the panel, dragging and editing of the model designated by the template and corresponding input parameters and the like are supported, after editing action occurs, the front end checks operation validity, and renders and modifies a visual effect in real time. On the basis, the edited model can be instantiated into a computing task, the computing task is submitted from the front end to the back end to be run, the running result can be visually displayed in the panel of the dragging component in real time, and a visual rendering flow closed loop from programming to result viewing is formed.
The technical effects are as follows: the programming configuration, the task generation and the result viewing of AI inference are realized in a low-code or no-code dragging mode, and the low-threshold, simple and visual programming operation is supported.
Key point 3: medical AI inference low code programming cloud computing engine
The medical AI reasoning low-code programming cloud computing engine is one of components of medical AI reasoning low-code programming back-end services, and can support the formed medical AI reasoning computing tasks to be executed on an accessed cloud, and the computing environments comprise physical machines, virtual machines, containers, server-free computing and the like. The engine comprises a calculation task management module and a cloud computing power access module. And the computing task management module and the cloud computing power access module realize the packaging of the computing task and the computing power management through interaction. And the calculation task management module performs instantiation management on the medical AI reasoning calculation task submitted by the front end at the rear end, including the state, control and the like of the calculation task. The cloud computing power access module supports unified adaptation of various different types of cloud computing power, and according to the definition of the AI inference low code programming template, the instantiated AI inference computation can be executed by distributing the computation to the corresponding computing environment through the computing power access module.
The technical effects are as follows: the method supports the program formed by the medical AI reasoning low-code template to realize the tasked management and is executed on various cloud computing engines.
Key point 4: medical AI reasoning low code programming system
In the face of the service requirement and the complexity of a computing environment of AI model reasoning and the efficient abstract expression of each complex factor in a programming model, the programmability and friendly interaction of a user are improved, and the medical AI reasoning low-code programming system can support medical scientific research personnel to realize efficient multi-scene low-code visual programming based on the AI model without concerning programming work except for services and code operation management work.
The system is constructed in a web form, web front-end rendering oriented to medical AI reasoning low-code programming is applied in a web front end, a medical AI reasoning low-code programming cloud computing engine is applied in a back end, and the system comprises a necessary business management module for medical scientific research.
The system supports medical personnel to carry out inference of an AI model by using an existing data set, specifically drags the AI model, the data set and parameters required by a user to a specified position to generate an inference task in a dragging mode according to prompts in a drawing board frame, sends an operation request to a back-end server, returns result data to render and present a front-end through processing and high-speed calculation of the server corresponding to the AI model, and responds to preview and downloading of a result document.
(1) Establishing an AI model pool, wherein the AI model Chi Zhongbao comprises a plurality of mature AI models or a user newly-built model;
(2) Establishing a data pool, wherein the data pool comprises a plurality of data sets to be inferred;
(3) Providing a visual dragging type interface layer, selecting a model and a data set selected by an inference task from a constructed AI model pool and a data pool, responding to display detailed information by the model or the data set, and dragging a required AI model, the data set and parameters to a specified position in the visual interface layer according to prompts in a drawing board frame to generate the inference task;
(4) After the inference task is generated, processing a data set through a back-end operation model, and finally returning a result to a visual interface layer for rendering and displaying, and supporting the preview and downloading of a result document.
The technical effects are as follows: the complexity of the user for reasoning tasks by using the AI model is reduced, the programmability of the user is improved, the interaction is simple and convenient, and meanwhile, the page is attractive and is friendly to operate.
The medical artificial intelligence low-code programming system is shown in fig. 2 and forms a 3-layer architecture comprising a web display layer, a back-end service layer and a computing environment. The construction implementation of the system can be developed from 5 aspects, and according to the flow shown in fig. 3, the construction implementation comprises the following steps:
(1) And constructing an AI inference low code programming template. An AI reasoning low-code programming template is established for the medical artificial intelligence low-code programming system, and the application method of different roles to the template according to the contents of low-code calculation realized by the template is defined.
(2) The medical AI inference low code programming web front end constructs. And constructing a web front-end interface for low-code AI inference.
(3) The medical AI reasoning low code programming back-end service constructs. And constructing a back-end related service and a service module aiming at the medical AI reasoning calculation.
(4) And accessing a cloud computing environment. And accessing computing environments including a local physical machine, a cloud virtual machine, a container, server-free computing and the like in the system according to needs.
(5) And accessing a medical AI model pool and a data pool. And accessing an available AI model pool, a data pool and the like outside the system in the system.
The specific implementation method comprises the following steps:
(1) AI inference low code programming template construction
Medical AI inference low code programming templates are high-level abstract definitions of artificial intelligence models that can implement low code programming. When a medical AI inference programming platform is constructed, a maintainer of the platform or a provider of an artificial intelligence model in the platform can efficiently convert an operable artificial intelligence model into a programmable operator in the platform according to a definition mode of a template, the operator supports a user (such as a doctor) of the operator to construct an inference program in a low-code or code-free mode without concerning an implementation mechanism, a calling mechanism and a source of the operator, and can realize the configuration of operator input and acquire an output result of the operator according to an operator information prompt.
The medical AI inference low code model comprises a group of metadata information and can be packaged into a standard operator, so that the AI model can be used in programming in the form of operator units; when the method is used, a programmer configures input information of an operator instantiation, and after the operator instantiation is executed, the operator instantiation output can be obtained. The definition of the template is shown in table 1, where different fields are defined by people of different roles.
According to the characteristics of medical AI calculation development, an AI inference model is often constructed by professionals and provided for medical researchers to use. By utilizing the characteristic, the invention realizes the user conversion and multiplexing of the medical AI inference model by using the AI inference low code programming template. The construction of the AI inference low code programming template comprises operations of digital definition, template generation definition, template programming, template instantiation operation and the like.
Step 1: the digital definition is based on the basic data format of the medical AI inference low code programming template. The medical AI inference low code programming template shown in table 1 can be defined by using an information mark language as a medium to form a definition file with a format of XML, JSON, YAML and the like for organizing information. The file has good readability, and can be efficiently analyzed by a computing system.
And 2, step: and constructing a generation tool based on the medical AI inference low-code programming template. The contributors of the medical AI model may utilize the tool or mode to transform the medical AI model they contributed into a template that is reusable by others. In the template generation definition, fields such as a model name, a model scene tag, model information, a model sample, a model input configuration template, a model definition parameter template, a model output configuration template, a model calling script and the like in the template need to be declared according to a template basic data format, wherein the fields such as the model name, the model input configuration template, the model calling script and the like are basic fields which need to be set, and other fields can be gradually improved in future use according to a use mode of the model.
And step 3: programming operations of the AI template are provided. The medical AI inference template can be modeled as an operator with uncertain number of parameters, which are determined according to the generation definition of the template in step 2, and programmed to operate as a process of applying the operator. The medical science scientific research personnel can complete the programming operation by assigning values to each parameter in the model input parameters and the model definition parameters in the template, and the system can interpret the parameters as a computable operator.
And 4, step 4: and (5) instantiating and running the AI template. And 3, according to a calculable operator formed by programming the AI model in the step 3, the system can start the instantiation operation of the template, check the legality of the template operator and start the interpretation execution after the correctness is checked.
(2) Medical AI inference low code programming web front-end construction
The medical AI inference low code programmed web front end includes related web front end interfaces and components, etc. The web front-end rendering oriented to the medical AI inference low-code programming comprises a plurality of front-end components supporting the medical scene programming, including a medical computation scene component, a medical AI model pool component, a medical AI data pool component, a medical AI low-code dragging component, a medical AI model template component and the like. The components cooperate to realize graphical web display and operation of the medical AI inference low-code programming.
The components cooperatively support the implementation of medical AI drag-and-drop programming. The medical computation scene component provides access to a medical computation scene, corresponds to a model scene label in the AI low-code programming template, provides scene switching, and can cooperatively switch contents along with scene switching of other components; the medical AI model pool component can display available AI models in the platform and support the components in the model pool to be added into a model groove of an operation panel of the medical AI low-code dragging component in a dragging mode; the medical AI data pool component can display the available data sets in the platform, and supports the components in the data pool to be added to a model input slot of an operation panel of the medical AI low-code dragging component in a dragging mode and bound with the model input in the panel; the medical AI low-code dragging component mainly comprises a dragging operation panel, and can support the dragging of the contents of other components to the operation panel to realize programming operation in an editing mode; the medical AI model template component may present template information for the currently selected model.
Aiming at the implementation of the medical AI reasoning low-code programming web front end, the method specifically comprises the following steps:
step 1: and configuring a front-end server. According to the requirement of running management of a web front-end interface, a front-end server, such as Nginx, is configured in the system, so that an external user can access a static page through a browser. Here, cross-domain access of the medical AI inference low code programming web page can be realized through the vue proxy + Nginx reverse proxy method.
Step 2: front-end rendering pages and core component development.
The front-end interface can be realized by adopting a standardized front-end research and development technology, such as a VUE, CSS, javaScript and other technologies. If the page is constructed based on the VUE, page codes can be written through < template > </template >, and element-ui related components are introduced into the page codes to realize development of the function page; writing page interaction logic in < script > </script > by adopting js language; the page frame is styled in the < style > </style > using the CSS language. The core front-end component of the low-code programming for medical AI inference can be defined based on a similar approach.
Particularly, for a medical AI model pool component, a medical AI data pool component and a medical AI low code dragging component which relate to dragging operation, data transfer between the components needs to be realized in an important way so as to support interactive cooperation of the components. The specific operation of the medical AI low-code programming front end can support: after a specific AI model is dragged and selected, a graphical template of the model is loaded on the panel, dragging and editing of the model designated by the template, corresponding input parameters and the like are supported, after editing action occurs, the front end verifies operation legality, and the visual effect is rendered and modified in real time. On the basis, the edited model can be instantiated into a computing task, the computing task is submitted from the front end to the back end to be run, the running result can be visually displayed in the panel of the dragging component in real time, and a visual rendering flow closed loop from programming to result viewing is formed.
And step 3: a routing rule is added. By setting the page routing router, different page contents, namely front-end routing, are accessed according to different user events, and a front-end access routing address is defined, so that a user can access the system through the routing address.
And 4, step 4: and defining an interface path and finishing interface calling. Js, and uploading the page frame, the page source code and the style design file to a corresponding server according to different user URL requests, acquiring different return contents, and generating a web front-end interactive page.
As shown in FIG. 4, a medical scientist may implement AI inference low code programming that implements:
1) The user can switch and display the AI model and the scene to which the data belongs, and can click to obtain the corresponding classification;
2) The mouse floats above the AI model and the data and is highlighted, and corresponding detailed information can be checked by clicking;
3) Selecting an AI model, a data set and parameters, dragging to the center position of a frame of the drawing board corresponding to the prompt information, and storing;
4) After dragging the configuration parameters, responding to the click addition parameters to obtain parameter popup configuration parameters;
5) Clicking the 'clear' drawing board, and dragging the drawing board again to construct a task;
6) Clicking the 'generating task', and checking a 'task list' after inputting task information. The task information may include task names and/or task description information;
7) After clicking 'run', clicking 'pause' and 'end' to control the task state;
8) And in the task list, checking the running state of the historical task, and checking the result of the completed task.
In order to make the aforementioned features and advantages of the present invention more comprehensible, embodiments are described in detail below. The tumor recurrence risk assessment model is mainly based on collaborative mining of relevant information such as clinical indexes, survival indexes and omics indexes of tumor patients, extracts key indexes to construct a risk assessment model, and performs data analysis by using the model, so that the tumor recurrence risk of the patients is predicted and assessed, doctors are better assisted in assessing the tumor recurrence risk of the patients, and the disease conditions of the patients are further reminded and prevented. The specific implementation steps of the tumor recurrence risk assessment task constructed by the model and the like are as follows:
1) A user switches scenes left and right in scene selection, selects the scenes when switching to the scenes of tumor diseases, displays a model of tumor recurrence risk assessment corresponding to a model pool, and displays clinical table data, omics data, survival data and sensor data corresponding to a data pool;
2) When a mouse floats above a data set such as a tumor recurrence risk assessment model or clinical table data, a corresponding icon is highlighted (the model is red, the data is blue), and detailed information of the corresponding model or data set, including information such as a sequence number, a name, a scene, description and release time, can be checked by clicking;
3) Dragging a model (tumor recurrence risk assessment) from a model pool, and clicking a left key to place the model at the center of a drawing board; selecting three data sets (clinical table data, omics data, survival data) from a data pool, respectively dragging and placing the data sets into drawing board slots (storing the data sets according to corresponding prompt information), and finally dragging the data sets (tumor model parameters) from parameter configuration into corresponding parameter slots;
4) After dragging the configuration parameters, clicking the addition parameters to obtain parameter popup configuration parameters, filling parameter information and output path information, wherein the model can not be filled without additional parameters;
5) If the model and the parameters need to be dragged again, the model can be directly dragged to cover or click the 'clear' drawing board, and after the contents of the drawing board are cleared, the drawing can be carried out again according to the steps to generate a task;
6) Clicking the "generate task" button enters the task name (e.g.: after the tumor recurrence risk assessment task 1) and the task description information are clicked, the generated task can be checked in a task list;
7) After clicking 'operation', a task operation timing prompt box appears, and a user can click 'pause' or 'end' to control the task operation state; after the operation is successful, a page pops up a tumor recurrence risk assessment task result, and the result comprises 1 risk assessment line graph and four groups of data (csv format). The doctor can carry out the association analysis of the recurrence risk on the cancer population according to the result, finds the number of markers (including clinical factors such as sex, age and the like, risk genes, risk sites and the like) corresponding to the indicated risk, and utilizes the markers to construct a risk assessment model, which can remarkably divide the population into two subgroups of low risk and high risk, thereby better assessing the tumor recurrence risk of the patient.
8) After the task training, a doctor can check the running condition of the tumor recurrence risk assessment task in a task list, and can click the result check result to review the historical task.
(3) Medical AI inference low code programming backend service construction
Step 1: the medical AI reasoning low code programming back-end service constructs. The medical AI reasoning low-code programming back-end service comprises modules of model management, data management, task management, public resource management, authority management, system monitoring, cloud computing management, data persistence and the like. The whole framework for the back-end service can be constructed by self-development or open-source framework implementation, such as flash and the like.
The model management explains and executes instantiation operation of a medical AI inference low-code programming template, and the key process of the model management is to execute a calling script in the AI template, deliver calculation to a cloud computing engine for specific execution, and transmit a generated calculation task state to a task management module; in the data management module, managing the calculated input parameters (such as data sets), accessing data according to data access addresses or data assignments, and submitting the data to the model management module; in the task management module, services convert instantiated computation into tasks, and monitor the computation state and store the results; the configuration of scenes, parameters and the like can be provided in the common resource management module; in the authority management, the configuration management can be carried out on the authority of the system user; in the system monitoring management, the system can interact with a cloud computing management module to monitor the computing execution state in real time
Step 2: and constructing data persistence management. The system relates to persistent storage of data such as medical AI inference programming calculation tasks and the like, and can provide support by adopting various types of data, such as real-time storage of task states by adopting a memory database Redis and structured storage of data by adopting mysql. For the adopted database, the related storage structure and the like are required to be set. In addition, the rapid deployment of the database and the like can be realized by means of a computing engine such as docker and the like.
And step 3: and setting a cloud computing engine. And constructing an extensible cloud computing engine in a back-end service according to a medical AI reasoning computing environment needing to be supported. The medical AI reasoning low-code programming cloud computing engine is one of components of medical AI reasoning low-code programming back-end services, and can support the formed medical AI reasoning computing tasks to be executed on an accessed cloud, and the computing environments comprise physical machines, virtual machines, containers, server-free computing and the like. The engine comprises a calculation task management module and a cloud computing power access module. And the computing task management module and the cloud computing power access module realize the encapsulation of computing tasks and computing power management through interaction. And the calculation task management module performs instantiation management on the medical AI reasoning calculation task submitted by the front end at the rear end, including the state, control and the like of the calculation task. The cloud computing power access module supports unified adaptation to various different types of cloud computing power, and according to the definition of the AI inference low code programming template, the instantiated AI inference computation can be distributed to corresponding computing environments for execution through the computing power access module.
(4) Cloud computing environment access
According to a medical AI reasoning computing environment which needs to be supported, a cloud computing environment which needs to be relied on, such as a computing environment managed by a user, computing environments provided by other public cloud manufacturers and private cloud manufacturers and the like, is accessed in the system in a configuration mode.
(5) Medical AI model pool and data pool access
And in the back-end service, the system platform is supported to access an external AI model pool and a data pool through configuration.
The following are system examples corresponding to the above method examples, and this embodiment can be implemented in cooperation with the above embodiments. The related technical details mentioned in the above embodiments are still valid in this embodiment, and are not described herein again in order to reduce repetition. Accordingly, the related-art details mentioned in the present embodiment can also be applied to the above-described embodiments.
The invention also provides a medical artificial intelligence reasoning system based on low code programming, which comprises:
the system comprises an initial module, a data processing module and a data processing module, wherein the initial module is used for acquiring a plurality of medical artificial intelligence models and a plurality of data sets to be inferred;
the conversion module is used for defining a low-code programming template according to the calculation content and the application method of the artificial intelligence model and converting the medical artificial intelligence model into the low-code model according to the low-code programming template;
the visualization module is used for selecting a model to be executed and a data set to be executed from all low-code models and all data sets according to prompts in a visualization interface, dragging and combining the model to be executed and the data set to be executed in the visualization interface, assigning values to parameters of the model to be executed, and generating a medical AI reasoning and calculating task;
and the execution module is used for constructing a cloud computing engine according to the computing environment required by the reasoning task so as to execute the medical AI reasoning computing task on the accessed cloud and obtain an execution result.
The medical artificial intelligence reasoning system based on the low-code programming is characterized in that the visual interface comprises a medical computation scene component, a medical AI model pool component, a medical AI data pool component, a medical AI low-code dragging component and a medical AI model template component; the components cooperate to program the graphical web presentation and operation of the low-code programming;
wherein the medical computing scenario component provides access to a medical computing scenario; the medical AI model pool component can display all the low-code models and support the low-code models to be added into a model groove of an operation panel of the medical AI low-code dragging component in a dragging mode; the medical AI data pool component displays all data sets to be inferred, supports the addition of the data sets to a model input slot of an operation panel of the medical AI low-code dragging component in a dragging mode, and binds with the input of a model in the model slot; the medical AI low-code dragging component is composed of a dragging operation panel, supports the dragging of the contents of other components to the operation panel, and realizes the programming operation in an editing mode; the medical AI model template component is used for displaying template information of the currently selected model.
The medical artificial intelligence reasoning system based on low code programming is characterized in that the visualization module is used for loading a graphical template of the model in the visualization interface after the model to be executed is selected, dragging and editing parameters specified by the graphical template, verifying the editing validity and rendering and modifying the visualization effect in real time.
The medical artificial intelligence reasoning system based on the low code programming is characterized in that the execution module is used for explaining the instantiation operation of the medical AI reasoning calculation task, executing the calling script of the instantiation operation and delivering the calculation to the cloud computing engine;
the cloud computing engine comprises a computing task management module and a cloud computing power access module; the computing task management module and the cloud computing power access module realize the encapsulation of computing tasks and computing power management through interaction; the calculation task management module performs instantiation management on the medical AI reasoning calculation task, including the state and control of the calculation task; the cloud computing power access module is used for adapting to the unified cloud computing power and distributing the AI inference computation after the instantiation of the medical AI inference computation task to the corresponding computing environment for execution.
The invention also provides a storage medium for storing a program for executing the medical artificial intelligence reasoning method based on the low-code programming.
The invention also provides a client used for the medical artificial intelligence reasoning system based on the low-code programming.
Claims (10)
1. A medical artificial intelligence reasoning method based on low code programming is characterized by comprising the following steps:
step 1, acquiring a plurality of medical artificial intelligence models and a plurality of data sets to be inferred;
step 2, defining a low-code programming template according to the calculation content and the application method of the artificial intelligence model, and converting the medical artificial intelligence model into the low-code model according to the low-code programming template;
step 3, selecting a model to be executed and a data set to be executed from all low-code models and all data sets according to prompts in a visual interface, dragging and combining the model to be executed and the data set to be executed in the visual interface, and assigning values for parameters of the model to be executed to generate a medical AI reasoning calculation task;
and 4, constructing a cloud computing engine according to the computing environment required by the reasoning task so as to execute the medical AI reasoning computing task on the accessed cloud to obtain an execution result.
2. The low-code programming based medical artificial intelligence reasoning method of claim 1, wherein the visual interface comprises a medical computation scenario component, a medical AI model pool component, a medical AI data pool component, a medical AI low-code dragging component, a medical AI model template component; the components cooperate to program the graphical web presentation and operation of the low-code programming;
wherein the medical computing scenario component provides access to a medical computing scenario; the medical AI model pool component can show all the low-code models and support the low-code models to be added into a model slot of an operation panel of the medical AI low-code dragging component in a dragging mode; the medical AI data pool component displays all data sets to be inferred, supports the addition of the data sets to a model input slot of an operation panel of the medical AI low-code dragging component in a dragging mode, and binds with the input of a model in the model slot; the medical AI low-code dragging component is composed of a dragging operation panel, supports the dragging of the contents of other components to the operation panel, and realizes the programming operation in an editing mode; the medical AI model template component is used for displaying template information of the currently selected model.
3. The low-code programming based medical artificial intelligence reasoning method of claim 1, wherein step 3 comprises:
and after the model to be executed is selected, loading a graphical template of the model in the visual interface, dragging and editing the specified parameters of the graphical template, verifying the editing validity, and rendering and modifying the visual effect in real time.
4. A medical artificial intelligence reasoning method based on low code programming according to claim 1, wherein the step 4 comprises:
interpreting instantiation operation of the medical AI reasoning calculation task, executing a calling script of the instantiation operation, and delivering calculation to a cloud computing engine;
the cloud computing engine comprises a computing task management module and a cloud computing power access module; the computing task management module and the cloud computing power access module realize the encapsulation of computing tasks and computing power management through interaction; the calculation task management module performs instantiation management on the medical AI reasoning calculation task, including the state and control of the calculation task; the cloud computing power access module is used for adapting to the unified cloud computing power and distributing the AI inference computation after the instantiation of the medical AI inference computation task to the corresponding computing environment for execution.
5. A medical artificial intelligence reasoning system based on low code programming, comprising:
the system comprises an initial module, a data processing module and a data processing module, wherein the initial module is used for acquiring a plurality of medical artificial intelligence models and a plurality of data sets to be inferred;
the conversion module is used for defining a low-code programming template according to the calculation content and the application method of the artificial intelligence model and converting the medical artificial intelligence model into the low-code model according to the low-code programming template;
the visualization module is used for selecting a model to be executed and a data set to be executed from all low-code models and all data sets according to prompts in a visualization interface, dragging and combining the model to be executed and the data set to be executed in the visualization interface, assigning values to parameters of the model to be executed and generating a medical AI reasoning and calculating task;
and the execution module is used for constructing a cloud computing engine according to the computing environment required by the reasoning task so as to execute the medical AI reasoning computing task on the accessed cloud and obtain an execution result.
6. The low-code programming based medical artificial intelligence reasoning system of claim 5, wherein the visualization interface comprises a medical computation scenario component, a medical AI model pool component, a medical AI data pool component, a medical AI low-code drag component, a medical AI model template component; the components cooperate to perform graphical web presentation and operation of low-code programming;
wherein the medical computing scenario component provides access to a medical computing scenario; the medical AI model pool component can display all the low-code models and support the low-code models to be added into a model groove of an operation panel of the medical AI low-code dragging component in a dragging mode; the medical AI data pool component displays all data sets to be inferred, supports the addition of the data sets to a model input slot of an operation panel of the medical AI low-code dragging component in a dragging mode, and binds with the input of a model in the model slot; the medical AI low-code dragging component is composed of a dragging operation panel, supports the dragging of the contents of other components to the operation panel, and realizes the programming operation in an editing mode; the medical AI model template component is used for displaying template information of the currently selected model.
7. The low-code programming-based medical artificial intelligence reasoning system of claim 5, wherein the visualization module is configured to load a graphical template of the model in the visualization interface after the model to be executed is selected, drag and edit parameters specified by the graphical template, verify editing validity, and render and modify the visualization effect in real time.
8. The low-code programming-based medical artificial intelligence reasoning system of claim 5, wherein the execution module is used for interpreting and executing the instantiated calculation of the medical AI reasoning calculation task, and delivering the calculation to the cloud computing engine;
the cloud computing engine comprises a computing task management module and a cloud computing power access module; the computing task management module and the cloud computing power access module realize the encapsulation of computing tasks and computing power management through interaction; the calculation task management module performs instantiation management on the medical AI reasoning calculation task, including the state and control of the calculation task; the cloud computing power access module is used for adapting to the unified cloud computing power and distributing the AI reasoning computation after the medical AI reasoning computation task is instantiated to the corresponding computing environment for execution.
9. A storage medium storing a program for executing the medical artificial intelligence reasoning method based on low code programming according to any one of claims 1 to 4.
10. A client for use in the low code programming based medical artificial intelligence reasoning system of any one of claims 5 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210974297.6A CN115346669A (en) | 2022-08-15 | 2022-08-15 | Medical artificial intelligence reasoning method and system based on low-code programming |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210974297.6A CN115346669A (en) | 2022-08-15 | 2022-08-15 | Medical artificial intelligence reasoning method and system based on low-code programming |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115346669A true CN115346669A (en) | 2022-11-15 |
Family
ID=83951586
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210974297.6A Pending CN115346669A (en) | 2022-08-15 | 2022-08-15 | Medical artificial intelligence reasoning method and system based on low-code programming |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115346669A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115810137A (en) * | 2023-02-09 | 2023-03-17 | 数据堂（北京）科技股份有限公司 | Method for constructing interactive artificial intelligence technical evaluation scheme |
-
2022
- 2022-08-15 CN CN202210974297.6A patent/CN115346669A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115810137A (en) * | 2023-02-09 | 2023-03-17 | 数据堂（北京）科技股份有限公司 | Method for constructing interactive artificial intelligence technical evaluation scheme |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lanza et al. | Polymetric views-a lightweight visual approach to reverse engineering | |
| Rasure et al. | The khoros application development environment | |
| US9116672B2 (en) | Method and system for displaying selectable autocompletion suggestions and annotations in mapping tool | |
| CN105593844B (en) | Infrastructure is customized when operation | |
| US20200319856A1 (en) | Visual virtual programming machine for real-time interactive creation, playback, execution, inspection and manipulation of programming elements | |
| Van Mierlo et al. | Debugging parallel DEVS | |
| CN106484394A (en) | A kind of fast software development system with double engines | |
| Françoise et al. | Marcelle: composing interactive machine learning workflows and interfaces | |
| EP3163436A1 (en) | Visual software modeling method based on software meta-view for constructing software view | |
| CN107943691A (en) | A kind of method and device for the functional test page for automatically generating intelligent contract | |
| JP2022177793A (en) | Operator registration method and apparatus for deep learning framework, device and storage medium | |
| Ayer et al. | Why scientists should learn to program in Python | |
| Oberhauser | VR-UML: The unified modeling language in virtual reality–an immersive modeling experience | |
| Wanderley et al. | A framework to diminish the gap between the business specialist and the software designer | |
| CN115346669A (en) | Medical artificial intelligence reasoning method and system based on low-code programming | |
| CN114270313A (en) | Computer-implemented method, computer program and physical computing environment | |
| Pierro | Smart-graph: Graphical representations for smart contract on the ethereum blockchain | |
| Bolte et al. | Vis-a-vis: visual exploration of visualization source code evolution | |
| Cunningham et al. | Object-oriented implementation of a graphical-programming system | |
| Schramm et al. | Rapid UI development for enterprise applications: Combining manual and model-driven techniques | |
| St-Aubin et al. | Analytics and visualization of spatial models as a service | |
| Nunez | An extended spreadsheet paradigm for data visualisation systems, and its implementation | |
| Hoffer et al. | Visual reflection library: a framework for declarative GUI programming on the Java platform | |
| Design et al. | MIT Architecture | |
| Lee et al. | Geosheet: A distributed visualization tool for geometric algorithms |
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 |