CN112100286A

CN112100286A - Computer-aided decision-making method, device and system based on multi-dimensional data and server

Info

Publication number: CN112100286A
Application number: CN202010822614.3A
Authority: CN
Inventors: 吴凯; 雷炳业; 韩俊南
Original assignee: Guangzhou Shuangyou Biotechnology Co ltd; South China University of Technology SCUT
Current assignee: Guangzhou Shuangyou Biotechnology Co ltd; South China University of Technology SCUT
Priority date: 2020-08-14
Filing date: 2020-08-14
Publication date: 2020-12-18

Abstract

The invention discloses a computer-aided decision-making method, a device, a system and a server based on multi-dimensional data, wherein the method comprises the steps of data acquisition, data processing, data analysis and decision making; the device comprises an acquisition and input unit, a processing and analysis unit, a storage unit and a decision and display unit; the system comprises acquisition and input equipment, a distributed server and user side equipment; the distributed server is used for arranging distributed queues, databases and containers; the system sends the acquired information data to a distributed queue, information formatting, preprocessing and feature extraction are carried out by containerization service, then the information data are stored in a distributed database, model training and decision making are carried out by a data analysis service container, and an auxiliary decision result is sent to a user side. The invention has the advantages of cross-platform performance, decentralization removal, privacy safety, utilization of big data and artificial intelligence technology, and suitability for assisting doctors in assisting decision scenes by utilizing objective biological criteria.

Description

Computer-aided decision-making method, device and system based on multi-dimensional data and server

Technical Field

The invention relates to a computer-aided decision method, a device, a system and a server based on multi-dimensional data, belonging to the field of computer-aided decision.

Background

With the development of computer-aided decision-making technology, more and more disease early warning models are used to assist doctors in disease diagnosis and prognosis. However, because the diagnosis of neuropsychiatric diseases temporarily lacks objective biological criteria and depends on subjective judgment of doctors, the research of potential factors of the neuropsychiatric diseases and the establishment of a disease early warning model become urgent. Clinically, there are several methods for determining neuropsychiatric diseases and their severity: collecting medical history of a testee, performing mental examination and physical and laboratory examination, wherein the medical history comprises past history, personal history and family history; mental exams are evaluated primarily by means of clinical scales; the body and laboratory examination mainly comprises: routine examination, neuroelectrophysiological and imaging examination.

At present, the popular neuropsychiatric disease assistant decision-making research in academia mainly adopts methods such as statistics or machine learning to analyze the multidimensional medical data of a testee, search potential predisposing factors and establish an effective early warning model. However, since the medical history, the scale data and the medical examination data all contain a large amount of private information, the data maintenance and management are difficult, and an effective unified database is lacked. Meanwhile, the existing models are designed aiming at small data sets, so that the requirements of large data analysis and model modeling on the information collected by the unified database cannot be met, the research is difficult, even wrong results are obtained sometimes, and correct and appropriate assistance cannot be provided for doctors.

Disclosure of Invention

The invention aims to solve the defects of the prior art, and provides a computer-aided decision-making method, a device, a system and a server based on multi-dimensional data, which are based on multi-site distributed data storage and analysis, are suitable for computer-aided decision-making scenes requiring privacy security and multi-dimensional big data analysis, can provide reliable aided decision-making results while ensuring data security, and users of the system are doctors and testees, and have the advantages of complete functions, simplicity in operation, privacy protection, stability and reliability.

The invention aims to provide a computer-aided decision method based on multi-dimensional data.

A second object of the present invention is to provide a multidimensional data processing apparatus for computer-aided decision-making.

A third object of the present invention is to provide a data management and analysis system for computer-aided decision-making.

It is a fourth object of the present invention to provide a multi-centric distributed server for computer-aided decision-making.

The first purpose of the invention can be achieved by adopting the following technical scheme:

a computer-aided decision-making method based on multidimensional data, the method comprising:

multi-dimensional data acquisition, input of information of a testee and sending of the information to a distributed queue by front-end equipment;

performing information formatting, preprocessing and feature extraction on the data in the queue, and storing the processed data in a distributed database;

performing targeted analysis on each dimension data extracted from the distributed database, training to obtain a corresponding auxiliary decision model, and storing the model into the database;

and deducing and deciding by using a specific auxiliary decision model according to the input data of each dimension, outputting the predicted data and the similar associated type data, and sending the predicted data and the similar associated type data to user end equipment for a doctor and a testee to check.

Further, the multidimensional data comprises neuroimaging data, scale evaluation data and biochemical information data, and the information of the testee comprises demographic information, clinical information and disease history. Further, the air conditioner is provided with a fan,

the assistant decision model is specific to specific data, including a machine learning model, a deep neural network model, and a graph neural network model, wherein,

the machine learning model takes structured data as input, comprises calculated neural image and biochemical information characteristic data, and quantitative scale evaluation, demographic information, clinical information and patient history, takes a loss function as an evaluation criterion, adopts an optimization algorithm to solve the machine learning model parameter with the minimum loss, outputs characteristic weight and decision result, and solves the following parameter formula:

θ＝argmin(L(y,f(x)))

wherein θ is a machine learning model parameter, L is a loss function, f is a machine learning model function, y is a label, and x is an input;

the deep neural network model takes image data or structured data which are the same as those of a machine learning model as input, a loss function as an evaluation criterion and back propagation as an optimization algorithm, and uses a multilayer neuron to fit an input and output mapping function to output a characteristic contribution degree and a decision result, wherein a neural network layer operator in the deep neural network model is as follows:

Z＝a(W·A)

wherein Z is the layer output, a is the activation function, W is the weight matrix, and A is the layer input;

the neural network model of the graph takes a brain connection network node feature vector and an adjacency matrix which are calculated and constructed by neural image data as input, and according to the characteristics of a topological structure, the high-level features of fusion node features and network information are supervised learned by aggregating network nodes and connection edge features, a significance graph aiming at a decision target is excavated, and a decision result is output, wherein a graph neural network layer operator in the neural network model of the graph is as follows:

Z＝a(L·H·W)

where Z is the layer output, a is the activation function, L is the Laplace matrix, H is the layer input, and W is the weight matrix.

The second purpose of the invention can be achieved by adopting the following technical scheme:

a multidimensional data processing apparatus for computer-aided decision making, the apparatus comprising:

the acquisition and input unit is used for acquiring multi-dimensional data by equipment and manually inputting the information of the testee;

the processing and analyzing unit is used for formatting information, preprocessing data, extracting characteristics and analyzing data in the database to obtain an auxiliary decision-making model;

the storage unit is used for storing the data queue to be processed, the data to be analyzed, the analysis model and the analysis result data;

and the decision and display unit is used for making a decision on the input data by using the model, outputting the prediction data and the similar associated type data, integrating and displaying result information and assisting a doctor in diagnosing diseases.

The third purpose of the invention can be achieved by adopting the following technical scheme:

a data management and analysis system for computer-aided decision-making, the system comprising:

the acquisition and input device is used for acquiring multidimensional data comprising neuroimaging data, scale evaluation data and biochemical information data and inputting information of a testee comprising demographic information, clinical information and a patient history;

the distributed server is used for storing a message queue to be processed, managing and scheduling the data processing and analyzing container, instantiating the container service to perform data processing and analysis, storing the processed and analyzed data, and sending an auxiliary decision result to the user side;

and the user side equipment is used for receiving, integrating and displaying the auxiliary decision result, wherein the auxiliary decision result comprises a model performance index, a contribution map of the model input characteristics, a decision result and confidence of the model on input data, clinical information of similar cases and a diagnosis result.

Further, the distributed server comprises a distributed queue, a distributed database and a distributed container arrangement;

the distributed queue is used for receiving and temporarily storing the data to be processed of the multiple sites and sending the data to the available server for processing and analysis;

the distributed database is distributed in each data source and used for storing multi-site data, ensuring data privacy and safety and meeting distributed data analysis;

the distributed container arrangement is used for deploying containerization services to perform data processing and analysis, and the main node server performs unified container arrangement to ensure high availability of the system;

further, the distributed database is used for storing structured and unstructured data by taking a tested person as a unit, wherein the structured and unstructured data comprise neuroimaging data, scale evaluation data, biochemical information data, demographic information, clinical information, disease history, intermediate results of data analysis, data models and aid decision making results.

Further, in the distributed container arrangement, each node server stores all container images including a data processing service container, a data analysis service container and a user side service container, and the master node performs container creation and service scheduling uniformly.

The fourth purpose of the invention can be achieved by adopting the following technical scheme:

a single server comprises a central processing unit, a graphic processing unit and a storage unit, wherein when the central processing unit executes a program stored in the storage unit to run containerization service, the computer-aided decision method based on multi-dimensional data is realized.

Compared with the prior art, the invention has the following beneficial effects:

1. the computer-aided decision making system adopts a browser/server architecture, can be used by multiple users, is cross-platform and is convenient and quick; the background adopts a distributed server, can be respectively deployed at a plurality of sites, supports decentralized data management, and is safe and effective; the method has the advantages that the learning and using cost of a user side is low, the method is user-friendly, the customization is supported, particularly for doctors needing objective criteria to assist in decision making, a model decision result can be quickly generated according to data information input by doctors and collected by equipment, and a detailed and effective result report is provided for the doctors by combining disease prediction and similar cases to assist the doctors in decision making.

2. According to the invention, a distributed big data management and analysis technology is adopted to safely and reliably manage and analyze data distributed in different hospitals and other sites, and a large amount of multidimensional data can be analyzed by combining the big data and an artificial intelligence technology while the privacy and the safety of the data of each site are protected, and the data is used for constructing a decision model to obtain a reliable and effective assistant decision result.

3. The visualized content displayed by the user side through the front-end webpage page comprises a disease prediction report and a similar case report, wherein the disease prediction report comprises model performance indexes, a contribution map of model input characteristics, a decision result and confidence of the model on input data, and helps a doctor to obtain a direct disease prediction result while judging the reliability of the model; the similar case report comprises case information with high similarity degree, which is obtained by the background model through similarity retrieval in the distributed database by using the input data, and the case information comprises clinical information and diagnosis results, so that doctors are helped to further evaluate the illness state of the testee.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the structures shown in the drawings without creative efforts.

FIG. 1 is a flowchart of a computer-aided decision-making method according to embodiment 1 of the present invention;

FIG. 2 is a block diagram of a computer-aided decision device according to embodiment 2 of the present invention;

fig. 3 is a general structural diagram of a computer-aided decision making system according to embodiment 3 of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer and more complete, the technical solutions in the embodiments of the present invention will be described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments, and all other embodiments obtained by a person of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the protection scope of the present invention.

Example 1:

as shown in fig. 1, the present embodiment provides a computer-aided decision method based on multidimensional data, which is implemented by a distributed server and includes the following steps:

s101, multi-dimensional data acquisition, information input of a testee, and sending to a distributed queue by front-end equipment.

The multidimensional data and the testee information collected and input by the equipment are directly sent to a distributed queue, and further data processing is carried out by a containerization service deployed on a distributed server.

In this embodiment, the multidimensional data includes neuroimage data, scale evaluation data, and biochemical information data, and the specific descriptions of the neuroimage data, the scale evaluation data, and the biochemical information data are as follows:

1. neuro-image data: including Structural and Functional Magnetic Resonance Imaging (s/fMRI for short), Diffusion Tensor Imaging (DTI for short), Electroencephalogram (EEG for short);

2. scale assessment data: including self-rating And other rating scales commonly used in neuropsychiatric disease diagnosis And research, Positive And Negative Symptom Scales (PANSS), Hamilton depression, Hamilton anxiety, Young mania rating, UKU side effect Scale, self-awareness And Treatment Attitude Questionnaire (ITAQ), gross function rating, Clinical Global Impression Scale (CGI), PHQ-9 depression screening Scale, health profile (SF-12), symptom self-rating Scale (SCLC-report inventory), General Well-Being Scale (GWS B).

3. Biochemical information data: the method comprises the following steps of (1) including intestinal flora data and blood index data, wherein the intestinal flora data specifically comprises the results of alpha diversity, beta diversity, flora composition analysis, flora difference analysis and KEGG (Kyoto Encyclopedia of Genes and genomics, Kyoto Encyclopedia of Genes and Genomes) function analysis of intestinal microorganisms of a testee; the blood index data specifically include blood cell number, immune factor, and oxidative stress.

In this embodiment, the information of the subject includes demographic information, clinical information, and disease history, and the specific descriptions of the demographic information, clinical information, and disease history are as follows:

1. demographic information: including basic information such as the age and the sex of the testee;

2. clinical information: including physical disease, clinical symptoms, and clinical assessment by a doctor of a subject;

3. the history of the disease: including personal medical history and family medical history.

And S102, performing information formatting, preprocessing and feature extraction on the data in the queue, and storing the processed data in a distributed database.

In the implementation, the information formatting comprises the quantification and the structuralization processing of the information of the testee input by the doctor, so that the data consistency is ensured.

In this embodiment, the preprocessing and feature extraction include preprocessing and feature extraction of neuroimage data, and the extracted features are specifically described as follows:

1. structural magnetic resonance imaging: gray matter volumes, white matter volumes, and structural connectivity networks;

2. functional magnetic resonance imaging: the system comprises a regional consistency, a low-frequency oscillation amplitude, a degree centrality and a functional connection network;

3. and (3) diffusion tensor imaging: anisotropy fraction, mean dispersion, radial dispersion, axial dispersion and tensor network;

4. electroencephalogram: event-related potentials, complexity, and brain function network topology attributes.

S103, analyzing the data extracted from the distributed database, training an auxiliary decision model, and storing the auxiliary decision model in the database.

In this embodiment, the assistant decision model includes a graph neural network model, a machine learning model, and a deep neural network model, and the graph neural network model, the machine learning model, and the deep neural network model are specifically described as follows:

1. graph neural network model: according to the topological structure characteristics of network data, by aggregating the node and edge characteristics in the network graph data, the high-level characteristics including the node characteristics and the network information are supervised and learned, and the node and graph classification and prediction tasks are completed;

2. a machine learning model: selecting a proper model from a model space aiming at vectorization characteristics of the structured data, and fitting a decision function by using an optimization algorithm to complete data discrimination and prediction tasks;

3. deep neural network model: and aiming at vectorization characteristics of the structured data, updating neuron variables in the deep neural network by using an optimization algorithm based on forward deduction and back propagation so as to fit an original model function, and completing data discrimination and prediction tasks.

And S104, making a decision by using the appointed auxiliary decision model based on the input data, outputting the predicted data and the similar associated type data, and sending the predicted data and the similar associated type data to user end equipment for a doctor and a testee to check.

In this embodiment, the content output to the client device includes the disease prediction and the similar case data, and the specific description of the disease prediction and the similar case data is as follows:

1. disease prediction: the method comprises model performance indexes, a contribution map of model input features, and decision results and confidence degrees of the model on input data, and helps a doctor to obtain a direct disease prediction result while judging the reliability of the model;

2. similar cases: the method comprises the steps that a background model uses input data to carry out similarity retrieval in a distributed database to obtain case information with high similarity degree, including clinical information and diagnosis results, and helps doctors to further evaluate the illness state of a testee.

The data, model and results in steps S102-S104 described above are persisted to a distributed database.

Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above embodiments may be implemented by a program to instruct related hardware, and the corresponding program may be stored in a computer-readable storage medium.

It should be noted that although the method operations of the above-described embodiments are depicted in the drawings in a particular order, this does not require or imply that these operations must be performed in this particular order, or that all of the illustrated operations must be performed, to achieve desirable results. Rather, the depicted steps may change the order of execution. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions.

Example 2:

as shown in fig. 2, the present embodiment provides a multidimensional data processing apparatus for computer-aided decision, which is applied to implement the above method, and includes an acquisition and entry unit 501, a processing and analysis unit 502, a storage unit 503, and a decision and display unit 504, where specific functions of each unit are as follows:

the acquisition and input unit 501 is used for acquiring multi-dimensional data by equipment and manually inputting information of a testee;

the processing and analyzing unit 502 is configured to format information, pre-process data, extract features, and analyze data in a database to obtain a model;

the storage unit 503 is configured to store a to-be-processed data queue, to-be-analyzed data, an analysis model, and analysis result data;

the decision and display unit 504 is configured to use the model to make a decision on input data, output a disease prediction and a similar case, integrate and display result information, and assist a doctor in disease diagnosis.

It should be noted that the apparatus provided in the foregoing embodiment is only illustrated by the division of the functional units, and in practical applications, the above function allocation may be performed by different functional units according to needs, that is, the internal structure is divided into different functional units to perform all or part of the functions described above.

Example 3:

as shown in fig. 3, the present embodiment provides a data management and analysis system for computer-aided decision-making, which includes a collection and entry device, a distributed server and a user end device.

The acquisition and input equipment acquires multidimensional data including neuroimaging data, scale evaluation data and biochemical information data, and inputs the information of a testee including demographic information, clinical information and patient history. And transmitting the acquired and recorded data to a distributed queue stored on a distributed server through an http protocol. And in addition, the distributed server manages and schedules the data processing and analyzing container, instantiates the container service to perform data processing and analysis, stores the processed and analyzed data, and sends an auxiliary decision result to the user side. And the user end equipment receives, integrates and displays the auxiliary decision result, including the model performance index, the contribution map of the model input characteristics, the decision result and confidence of the model on the input data, the clinical information of the similar cases and the diagnosis result.

In this embodiment, the acquisition and entry device integrates a data transmission module, which includes a processor, a memory, an input/output unit, and other components, where the processor runs a program stored in the memory, and transmits the acquired data to a designated network address through the input/output unit according to a network transmission protocol.

In this implementation, the distributed server is deployed with a distributed queue, a distributed database, and a distributed container arrangement, and the specific description of the distributed queue, the distributed database, and the distributed container arrangement is as follows:

1. the distributed queue is used for receiving and temporarily storing the data to be processed of the multiple sites and sending the data to the available server for processing and analysis;

2. the distributed database is distributed in each data source, and stores structured and unstructured data by taking a tested person as a unit, wherein the structured and unstructured data comprise neuroimaging data, scale evaluation data, biochemical information data, demographic information, clinical information, disease history, data analysis intermediate results, data models and auxiliary decision results; the decentralized storage ensures the data privacy and safety and meets the requirement of distributed data analysis;

3. and the distributed container arrangement is used for deploying containerized services to perform data processing and analysis, the master node server performs unified container arrangement to ensure high availability of the system, each slave node server stores all container images including a data processing service container, a data analysis service container and a user side service container, and the master node performs unified container creation and service scheduling.

In this embodiment, the container mirror image includes a data processing service container, a data analysis service container, and a user side service container, and the specific descriptions of the data processing service container, the data analysis service container, and the user side service container are as follows:

1. data processing service container: including the system environment, software dependencies, and service interface programs needed for the data processing services. The service interface program integrates the preprocessing and feature extraction programs in the embodiment 1, can process corresponding data in a targeted manner, is connected with a distributed database, and stores the processing result in the distributed database;

2. data analysis service container: including the system environment, software dependencies, and service interface programs needed for the data analysis service. The service interface program integrates the auxiliary decision model training and deduction program in the embodiment 1, can train and deduct specific data by using a specific model interface, is connected with a distributed database, and stores the obtained auxiliary decision model and result data in the distributed database;

3. a user side service container: including the system environment, software dependencies, static resources and service interface procedures required by the client services. The service interface program can return a specific response and a static interface according to the user request, is connected with the distributed database, integrates the data of the user request, returns the data to the user side and is displayed in the static interface.

For the description of the multidimensional data and the user side display content in this embodiment, refer to embodiment 1 above, and are not described herein again.

Example 4:

the embodiment provides a multi-center distributed server for computer-aided decision-making, which comprises a central processing unit, a graphic processing unit and a storage unit, and is characterized in that the distributed server respectively realizes distributed queue, distributed database and distributed container arrangement. When the server operates the containerization service, the computer-aided decision method based on the multi-dimensional data of the embodiment 1 is implemented as follows:

storing the multidimensional data and the testee information sent by the front-end equipment into a distributed queue;

analyzing data extracted from a distributed database, training an auxiliary decision model, and storing the auxiliary decision model in the database;

and (4) making a decision by using a specified auxiliary decision model based on the input data, outputting the predicted data and the similar associated type data, and sending the predicted data and the similar associated type data to the user end equipment for the doctor and the testee to view.

In conclusion, the system can realize computer-aided decision-making, provides an aided decision-making system with simple operation and complete functions for doctors or testees, and comprises a multidimensional data analysis model based on a graph neural network, machine learning and a deep neural network, and distributed queues, databases and container arrangement, thereby protecting the privacy of the medical information of the testees, providing a way for big data analysis, improving the reliability of the aided decision-making model and providing an aided decision-making result.

The above description is only for the preferred embodiments of the present invention, but the protection scope of the present invention is not limited thereto, and any person skilled in the art can substitute or change the technical solution and the inventive concept of the present invention within the scope of the present invention.

Claims

1. A computer-aided decision method based on multidimensional data, the decision method comprising:

performing targeted analysis on each dimension data extracted from the distributed database, training to obtain a corresponding assistant decision model, and storing the assistant decision model into the database;

and performing inference and decision by using a specified auxiliary decision model according to the input data of each dimension, and outputting the predicted data and the similar association type data.

2. The multidimensional data processing and analyzing method of claim 1, wherein the multidimensional data comprises neuroimaging data, scale assessment data, biochemical information data, and the subject information comprises demographic information, clinical information, and medical history.

3. The multidimensional data processing and analysis method of claim 1, wherein the aided decision models comprise a machine learning model, a deep neural network model, and a graph neural network model, wherein,

θ＝argmin(L(y,f(x)))

Z＝a(W·A)

the neural network model takes a brain connection network node feature vector and an adjacent matrix which are calculated and constructed by neural image data as input, and outputs a significant feature map and a decision result, and a neural network layer operator in the neural network model is as follows:

Z＝a(L·H·W)

4. A multidimensional data processing apparatus for computer-aided decision making, the apparatus comprising:

and the decision and display unit is used for making a decision on the input data by using the auxiliary decision model and outputting the prediction data and the similar association type data.

5. A data management and analysis system for computer-aided decision-making, the system comprising:

and the user side equipment is used for receiving, integrating and displaying an auxiliary decision result, wherein the auxiliary decision result comprises a model performance index, a contribution map of the model input characteristics, a decision result and confidence of the model on input data, and similar case clinical information.

6. The data management and analysis system of claim 5, wherein the distributed server comprises a distributed queue, a distributed database, and a distributed container arrangement;

the distributed container arrangement is used for deploying containerization services to perform data processing and analysis, and the main node server performs unified container arrangement to ensure high availability of the system.

7. The data management and analysis system of claim 6, wherein the distributed database is configured to store structured and unstructured data in units of subjects, including neuroimaging data, scale assessment data, biochemical information data, demographic information, clinical information, medical history, intermediate results of data analysis, data models, and decision-making assistance results.

8. The data management and analysis system of claim 6, wherein the distributed container arrangement comprises a plurality of node servers, each node server stores all container images including a data processing service container, a data analysis service container and a user side service container, and the master node performs unified container creation and service scheduling.

9. A multi-center distributed server for computer-aided decision-making, wherein a single server comprises a central processing unit, a graphic processing unit and a storage unit, and the central processing unit implements the computer-aided decision-making method based on multi-dimensional data according to claims 1-3 when executing a containerization service executed by a program stored in the storage unit.