CN113948185A

CN113948185A - Remote training scheme pushing method and device, computer equipment and storage medium

Info

Publication number: CN113948185A
Application number: CN202111183704.3A
Authority: CN
Inventors: 简俊豪; 王卫苹; 刘啟鸿; 林观泉; 黎韵诗; 周丽冰; 欧阳敏生
Original assignee: Guangdong Deyi Technology Co ltd
Current assignee: Guangdong Deyi Technology Co ltd
Priority date: 2021-10-11
Filing date: 2021-10-11
Publication date: 2022-01-18

Abstract

The application relates to a pushing method and device of a remote training scheme, computer equipment and a storage medium. The method comprises the following steps: acquiring cognitive function evaluation information and medical related information sent by a first terminal; the cognitive function evaluation information is obtained by a first user through remote evaluation on a first terminal; the medical related information comprises medical image information and biological data information; determining abnormal stage information according to the cognitive function evaluation information and the medical related information; determining a training scheme matched with the abnormal stage information according to a plurality of training data in the database, wherein the training scheme comprises training data related to a plurality of cognitive domains; and pushing the training scheme to the first terminal so that the first user can perform training operation according to the training scheme. By adopting the method, a proper evaluation and training scheme can be automatically recommended to the user, so that remote cognitive evaluation and training are realized, and the user can finish home training without special medical equipment.

Description

Remote training scheme pushing method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of intelligent interaction technologies, and in particular, to a method and an apparatus for pushing a remote training scenario, a computer device, and a storage medium.

Background

At present, when a user needs to carry out certain training according to physical conditions, the user needs to go to a professional organization, firstly passes through a series of complex examination processes, and then is guided and accompanied by training by a manual field according to examination results. However, this method requires the user to frequently move to the scene, which is time-consuming and labor-consuming, and is inconvenient.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a pushing method, an apparatus, a computer device and a storage medium for a remote training scheme, which can remotely push a training scheme.

A method of pushing a remote training scenario, the method comprising:

acquiring cognitive function evaluation information and medical related information sent by a first terminal; the cognitive function evaluation information is obtained by a first user through remote evaluation on the first terminal; the medical related information comprises medical image information and biological data information;

determining abnormal stage information according to the cognitive function assessment information and the medical related information;

determining a training scheme matched with the abnormal stage information according to a plurality of training data in a database, wherein the training scheme comprises training data related to a plurality of cognitive domains;

and pushing the training scheme to the first terminal so that the first user can perform training operation according to the training scheme.

In one embodiment, the determining abnormal stage information according to the cognitive function assessment information and the medically related information includes:

carrying out image preprocessing on the medical image information to obtain a preprocessed image;

inputting the preprocessed image into a pre-trained long-time and short-time memory neural network model, and extracting features by the long-time and short-time memory neural network model to obtain hidden representation of image features; the image features comprise a waveband change feature, an edge feature and a gray scale feature;

classifying by utilizing a fully-connected neural network based on the biological data information and the hidden representation of the image characteristics to obtain prediction stage information;

and determining abnormal stage information of the first user according to the cognitive function evaluation information and the prediction stage information.

In one embodiment, the training step of the long-time memory neural network model includes:

acquiring a current sample image, and taking image characteristics in the current sample image as an input value at the current moment;

according to the hidden representation of the image feature at the previous moment, a forgetting gate performs weighted operation on the input value to obtain a first intermediate value, an input gate performs weighted operation on the input value to obtain a second intermediate value and a third intermediate value, and an output gate performs weighted operation on the input value to obtain a fourth intermediate value;

determining the unit state of the image feature at the current moment according to the unit state, the first intermediate value, the second intermediate value and the third intermediate value of the image feature at the previous moment;

obtaining an output value of the current moment according to the unit state of the image characteristic of the current moment and the fourth intermediate value; wherein the output value is a hidden representation of the image feature at the current time;

and obtaining a next sample image, taking the image characteristics in the next sample image as input values of the next moment, returning to the hidden representation according to the image characteristics of the previous moment, performing weighted operation on the input values by a forgetting gate to obtain a first intermediate value, performing weighted operation on the input values by an input gate to obtain a second intermediate value and a third intermediate value, and performing weighted operation on the input values by an output gate to obtain a fourth intermediate value, and continuing to perform the steps until the training termination condition is met, stopping iteration, and obtaining the trained long-time and short-time memory neural network model.

In one embodiment, the training step of the fully-connected neural network comprises:

inputting the image characteristics, the hidden representation of the image characteristics and the unit state of the sample image into a fully-connected neural network, and outputting an abnormal stage prediction result; the abnormal stage prediction result represents probability values in different abnormal stages;

determining image characteristics to be deleted based on the difference between the abnormal stage prediction result and the abnormal stage to which the sample image actually belongs, and determining adjustment parameters based on the image characteristics to be deleted;

determining a reward function based on the abnormal stage prediction result and the adjustment parameter;

and adjusting the adjusting parameters of the fully-connected neural network by taking the maximum reward function as a target, and ending the training until the training termination condition is met to obtain the trained fully-connected neural network.

In one embodiment, the method further comprises:

sending evaluation standard information to a first terminal so that the first terminal can visually display evaluation content to a first user; the evaluation standard information is used for guiding a first user to carry out evaluation test on the first terminal;

the visual display at least comprises one of picture display, character display, video display and voice display.

In one embodiment, the method further comprises:

according to an interactive request initiated by a first terminal or a second terminal, establishing a communication connection between the first terminal and the second terminal, so that the first terminal can receive an evaluation instruction sent by the second terminal and carry out remote evaluation according to the evaluation instruction; wherein the interactive request comprises at least one of a chat request, a voice request, and a video request.

In one embodiment, the method further comprises:

pushing the training scheme to a second terminal so that a second user can modify the training scheme on the second terminal; and

and receiving a modified training scheme sent by a second terminal, and pushing the modified training scheme to the first terminal.

A push device for a remote training regimen, the device comprising:

the acquisition module is used for acquiring the cognitive function evaluation information and the medical related information sent by the first terminal; the cognitive function evaluation information is obtained by remote evaluation of a user on the first terminal; the medical related information comprises medical image information and biological data information;

the determining module is used for determining abnormal stage information according to the cognitive function evaluation information and the medical related information;

the determining module is further configured to determine a training scheme matched with the abnormal stage information according to a plurality of training data in a database, where the training scheme includes training data related to a plurality of cognitive domains;

and the pushing module is used for pushing the training scheme to the first terminal so that a user can carry out training operation according to the training scheme.

A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:

According to the pushing method, the pushing device, the computer equipment and the storage medium of the remote training scheme, the first user carries out remote evaluation on the first terminal, so that cognitive function evaluation information and medical related information sent by the first terminal are obtained, and abnormal stage information is determined according to the cognitive function evaluation information and the medical related information; and determining a training scheme matched with the abnormal stage information according to the determined abnormal stage information, and pushing the training scheme to the first terminal, so that the first user can perform training operation according to the training scheme. Therefore, the system can automatically recommend a proper evaluation and training scheme to the user, remote cognitive evaluation and training are achieved, the user can finish home training without going to the site or with the help of special medical equipment, and the system is very convenient.

Drawings

FIG. 1 is a diagram of an application environment of a push method of a remote training scenario in one embodiment;

FIG. 2 is a schematic flow chart diagram illustrating a push method of the remote training scenario in one embodiment;

FIG. 3 is a schematic diagram of a conceptual framework for determining abnormal stage information according to cognitive function assessment information and medically related information by a server in one embodiment;

FIG. 4 is a flowchart illustrating the steps of the server determining abnormal stage information based on the cognitive function assessment information and the medically relevant information in one embodiment;

FIG. 5 is a schematic flow chart illustrating the training steps of the long-term and short-term memory neural network model in one embodiment;

FIG. 6 is a schematic diagram of a long term memory neural network model according to an embodiment;

FIG. 7 is a schematic flow chart diagram illustrating the training steps for a fully-connected neural network in one embodiment;

FIG. 8 is a block diagram of a push device of the remote training scheme in one embodiment;

FIG. 9 is a block diagram of a pushing device of a remote training scenario in another embodiment;

FIG. 10 is a diagram showing an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The application provides a pushing method and device of a remote training scheme, computer equipment and a storage medium, and a big data analysis and neural network model are combined, so that the system can automatically push a proper evaluation and training scheme, and is more accurate and adaptive compared with a manual training scheme selected through experience, so that a user can realize remote evaluation and training by using a terminal without depending on special medical equipment.

The terms "first" and "second" referred to in the embodiments of the present application are used in the present application to describe different terminals, but the terminals should not be limited by these terms. These terms are only used to distinguish one terminal from another. For example, a first terminal may be referred to as a second terminal, and similarly, a second terminal may be referred to as a first terminal, without departing from the scope of the various described embodiments, but they are not the same terminal unless the context clearly dictates otherwise. A similar situation exists for the first user and the second user.

In this embodiment, the first user refers to a person to be trained, such as a patient or a family member of the patient, which may receive a training scheme by using the first terminal and perform a training operation according to the training scheme. In contrast, the second user refers to a subject person, such as a therapist, doctor, etc., who is able to provide a training opinion, and may view and/or modify, correct, etc. the training program using the second terminal.

The first/second terminals may be, but are not limited to, various electronic devices such as personal computers, notebook computers, smart phones, tablet computers, smart televisions, and portable wearable devices. In some embodiments, the first/second terminal may be loaded with an APP application (e.g., an applet, etc.) or a client with a web/website access capability, and provide a remote evaluation portal, show a training scheme, and show a training scheme, etc. to the first user through the APP application or the client. In some embodiments, the electronic device/processing device includes components such as memory, memory controllers, one or more processing units (CPUs), peripheral interfaces, RF circuits, audio circuits, speakers, microphones, input/output (I/O) subsystems, display screens, other output or control devices, and external ports, which communicate via one or more communication buses or signal lines.

Wherein the training regimen refers to training material that instructs the first user how to perform the training. The training data comprises training data related to cognitive domains such as memory, speech function, attention, executive power, visual space, emotion, sleep, mental behavior, daily life capacity and the like, and the training data comprises one or more of a video course, a text course, voice guidance and the like.

The pushing method of the remote training scheme provided by the application can be applied to the application environment shown in fig. 1. The server 101 is network-connected to a first terminal 102 and a second terminal 103, respectively. The network may be the internet, a mobile network, a Local Area Network (LAN), a wide area network (WLAN), a Storage Area Network (SAN), one or more intranets, etc., or a suitable combination thereof, and the type, protocol, etc. of the communication network between the server and the electronic device on the first/second user side in the embodiments of the present application are not limited in this application.

The server 101 is configured with a data processing system (or platform system, etc.). In some embodiments, the servers may be arranged on one or more physical servers depending on a variety of factors such as function, load, and the like. When distributed in a plurality of entity servers, the server can be composed of servers based on a cloud architecture. For example, a Cloud-based server includes a Public Cloud (Public Cloud) server and a Private Cloud (Private Cloud) server, wherein the Public or Private Cloud server includes Software-as-a-Service (SaaS), Platform-as-a-Service (PaaS), Infrastructure as a Service (IaaS), and Infrastructure as a Service (IaaS). The server may also be comprised of a distributed or centralized cluster of servers. For example, a server cluster is composed of at least one physical server. Each entity server is provided with a plurality of virtual servers, each virtual server runs at least one functional module to realize the pushing method of the remote training scheme, and the virtual servers are communicated through a network.

In one embodiment, as shown in fig. 2, a push method of a remote training scheme is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps:

step S202, acquiring cognitive function evaluation information and medical related information sent by a first terminal; the cognitive function evaluation information is obtained by a first user through remote evaluation on a first terminal; the medically-related information includes medical image information and biographic information.

The cognitive function assessment information includes, but is not limited to, assessment results of cognitive domain related functions such as memory, speech function, attention, executive power, visual space, emotion, sleep, mental behavior, and daily life capability. The cognitive function assessment information is obtained by the first user through remote assessment on the first terminal. For example, the first terminal presents the cognitive function assessment scale to the first user on a display screen; and after the first user completes filling the cognitive function assessment scale, the first terminal acquires the cognitive function assessment information of the first user. The Cognitive function Assessment Scale includes, but is not limited to, one or more of an Assessment of memory impairment (Ascertain destinata 8, AD8), a minimum-Mental State Assessment (MMSE), a Montreal Cognitive Assessment (MOCA), an Activity of Living Scale (ADL), and a memory and execution Scale. And for another example, the first terminal prompts the first user to perform remote evaluation through voice, acquires voice information answered by the first user according to the voice prompt, converts the voice information into cognitive function evaluation information, and sends the cognitive function evaluation information to the server. Of course, the above embodiments are merely examples, and do not limit the scope of the embodiments of the present application; for example, after the first terminal and the second terminal establish a communication connection, voice communication between the first user and the second user is realized, so that the first user completes remote evaluation under the guidance of the second user, and the first terminal sends the cognitive function evaluation information of the first user to the server.

The medical related information is related medical data obtained after diagnosis by the first user, and comprises one or more of medical image information, biological data information, life and eating habit data and the like. The medical image information includes one or more of Imaging data, Electroencephalogram data, and the like, such as Electroencephalogram (EEG) and Functional Magnetic Resonance Imaging (fMRI). The biological data information includes, but is not limited to, one or more of epidemiological data, baseline data, laboratory data, and Α β 42 and tau protein biomarker data, among others.

Specifically, the first terminal sends cognitive function evaluation information to a server; the server receives and obtains cognitive function assessment information and medically related information associated with the first user for subsequent processing.

And S204, determining abnormal stage information according to the cognitive function evaluation information and the medical related information.

The abnormal stage information is used for representing the information of the stage where the abnormal degree is located. Abnormal stages such as mild, moderate or severe, etc. For example, in the case of alzheimer's disease, the deterioration degree of cognitive ability and physical function can be divided into three stages, the first stage is a mild stage, and symptoms such as hypomnesis, decreased judgment ability, time-oriented disorder, poor visual space ability of complex structures, and few words and vocabularies are presented; the second stage is the middle stage, which is marked by serious impairment of near-far memory, visual space capacity reduction with simple structure, disorientation of time and place, incapability of calculation, various neurological symptoms, visible aphasia, disuse, agnosia and the like; the third stage is the severe stage, manifested by severe memory loss, inability to self-care in daily life, mutism, and stiff limbs.

Specifically, the server inputs the acquired cognitive function assessment information and medical related information of the first user into a stage judgment model for processing, and the stage judgment model predicts corresponding abnormal stage information. Thus, the server obtains the abnormal stage information matched with the first user. The stage judgment model is, for example, a judgment model composed of big data, and outputs a result of an abnormal stage that the first user may be in according to the cognitive function evaluation information and the medical related information of the first user, in combination with tens of thousands of cognitive function evaluation information and medical related information in the big data and corresponding abnormal stage information, thereby obtaining abnormal stage information.

For another example, the stage determination model is, for example, an artificial intelligence model, such as a Long Short-Term Memory (LSTM) neural network model, or an improved model based on the Long-Term Memory neural network model, for example, a neural network model formed by combining the Long-Term Memory neural network model and a fully-connected neural network. Exemplarily, as shown in fig. 3, the server inputs the acquired cognitive function assessment information and medical related information of the first user into the long-term memory neural network model for processing, obtains an intermediate result, inputs the intermediate result into the fully-connected neural network again, and outputs an abnormal stage prediction result by the fully-connected neural network. And the server combines the cognitive function evaluation information according to the abnormal stage prediction result output by the fully-connected neural network to obtain final abnormal stage information. Step S206, according to the training data in the database, determining the training scheme matched with the abnormal stage information, wherein the training scheme comprises the training data related to a plurality of cognitive domains.

Specifically, after determining the abnormal stage information corresponding to the first user, the server determines a training scheme matched with the abnormal stage information, so that the first user can train accordingly. Specifically, the server may pre-establish a database, and store a plurality of training data (including but not limited to one or more or a combination of video training data, text training data, and audio training data) in the database; and after the abnormal stage information is determined, the server searches and matches in the database so as to determine a matched training scheme.

In one embodiment, training schemes matched with different abnormal stage information can be constructed in advance by professionals, uploaded to a server and stored in a database by the server. For example, for stage a, a professional manually combines files such as video and audio corresponding to a plurality of training materials, and sends the combined files to a server as a training scheme corresponding to stage a, and the training scheme is stored by the server. And then, after the server determines that the abnormal stage information corresponding to the first user is stage A, searching in a database so as to directly match the training scheme corresponding to the stage A.

Illustratively, the abnormal stage information is assumed to include stages a to H, wherein the training scheme corresponding to the stage a includes continuous attention training, logical thinking training, reasoning ability training, long-term memory training, advanced daily life training, and video/audio/text training data applying mathematical training. The training scheme corresponding to the stage B comprises video/audio/character training data of retentive attention training, transfer training, spatial memory training, character memory training, calculation training, maze training and poetry understanding training. The training scheme corresponding to the C stage comprises video/audio/text training data of dispersive attention training, work training, speech memory training, short-time memory training and vehicle training. The training scheme corresponding to the D stage comprises video/audio/text training data of transfer attention training, retentive attention training, short-time memory training, drawing training, maze training and poetry understanding training. E training data in the form of video/audio/text for distractive attention training, retentive attention training, controlled training, spatial memory training, character memory training, computational exercise training, puzzle training, and home life ability training. The training scheme corresponding to the F stage comprises video/audio/text training data in the forms of retentivity attention training, work training, control training, short-time memory training, character memory training, place memory training, drawing training, space positioning training, shopping activity training and home life training. The training scheme corresponding to the G stage comprises video/audio/text training data in the forms of retentive attention training, concept formation training, short-time memory training, space positioning training, shopping activity training, home life training and speech memory training. The training scheme corresponding to the H stage comprises video/audio/text training data of the forms of the retentive attention training, the shopping activity training, the home life training, the daily life capacity training, the short-time memory training and the space vision training.

Therefore, the server determines the training scheme which is matched with the abnormal stage information and contains the training data related to the multiple cognitive domains in the database, so that the training scheme is automatically and intelligently adjusted according to different characteristics of the abnormal stage information, and the multiple cognitive domains are conveniently trained in a targeted manner.

Step S208, pushing the training scheme to the first terminal for the first user to perform training operation according to the training scheme.

Specifically, the server sends the determined training scheme to the first terminal, and the first terminal displays the training scheme to the first user in a visual mode so that the first user can perform training operation according to the training scheme. Wherein, the training operation means that the first user trains correspondingly according to the training scheme.

According to the pushing method of the remote training scheme, the first user carries out remote evaluation on the first terminal, so that cognitive function evaluation information and medical related information sent by the first terminal are obtained, and abnormal stage information is determined according to the cognitive function evaluation information and the medical related information; and determining a training scheme matched with the abnormal stage information according to the determined abnormal stage information, and pushing the training scheme to the first terminal, so that the first user can perform training operation according to the training scheme. Therefore, the system can automatically recommend a proper evaluation and training scheme to the user, remote cognitive evaluation and training are achieved, and the user can finish home training without special medical equipment.

In a specific implementation manner, the embodiment of the application builds a MySQL (relational database management system) environment on an airy cloud server by using a computer, App and applet multilinear multidata intercommunication technology, and builds a complete software system by using a MySQL database visualization management tool and an information system, a scale system, an accessory system, a video system, a multimedia system, a cognitive training system, a remote data system and a data encryption transmission system of a first user. The first user can conduct remote cognitive assessment on a tablet personal computer and a mobile phone, push a proper training scheme through big data analysis and artificial intelligence technology of a server, and adjust the training scheme in a self-adaptive mode according to cognitive assessment characteristics of the first user.

When a first/second user operates at a corresponding first/second terminal, client software installed on the terminal monitors operation events of the user and makes a response, when data needs to be acquired for analysis, the client software sends a request to a server interface, the server performs corresponding operations of adding, deleting, checking and modifying on the MySQL database according to the request, packages the data into a Json style and returns the Json style to the client software. The request is transferred through the Windows Communication development platform (WCF) service, and the client is not directly connected with the database, so that the pressure of the MySQL link pool can be relieved to the maximum extent, and the data security of the database is ensured to a certain extent.

In order to further improve the accuracy and efficiency of identifying the abnormal stage information, in some embodiments, as shown in fig. 4, the server determines the abnormal stage information according to the cognitive function assessment information and the medically related information, and the step of determining the abnormal stage information includes:

step S402, image preprocessing is carried out on the medical image information to obtain a preprocessed image.

The medical image information includes an electroencephalogram image and a functional magnetic resonance image. Specifically, the server first needs to perform image preprocessing on the electroencephalogram image and the functional magnetic resonance image, so that the neural network can analyze and extract the change characteristics of the alpha wave band in the electroencephalogram image and the edge characteristics and gray characteristics hidden in the functional magnetic resonance image, and meanwhile, the accuracy of the result can be improved, and the influence of noise can be eliminated.

Illustratively, for electroencephalography images, the server first filters the data using a low-pass filter to preserve the more useful low-frequency structure while reducing high-frequency variations of the data. Secondly, in order to ensure the time sparsity of the data, the image needs to be sampled secondarily, so that the correlation between the time dimension and the data is reduced. For functional magnetic resonance images, the server first performs a corrective registration, minimizing the mismatch between the images. Secondly, the functional magnetic resonance image is normalized and smoothed, so that the influence of interference signals is eliminated. Thereby, the server obtains the preprocessed medical image information, i.e. the preprocessed image.

And S404, inputting the preprocessed image into a pre-trained long-time and short-time memory neural network model, and extracting the features of the long-time and short-time memory neural network model to obtain the hidden representation of the image features.

The image features include a band variation feature, an edge feature, and a grayscale feature. The wave band change characteristic is mainly an alpha wave band change characteristic, the alpha wave band frequency distribution is 8-13HZ, and the wave band change characteristic mainly comprises two wave bands: μ 1(8-10Hz) and μ 2(10-13Hz), with amplitudes of about 20-100 μ V.

Specifically, the server inputs the preprocessed image into a long-time and short-time memory neural network model trained in advance, performs feature extraction through the long-time and short-time memory neural network model, performs processing through an input gate, an output gate, a forgetting gate and the like in the long-time and short-time memory neural network model, and outputs the hidden representation of the image feature.

Step S406, based on the biological data information and the hidden representation of the image feature, classifying by using a full-connection neural network to obtain the prediction stage information.

Specifically, the server takes the output obtained by processing the long-time memory neural network model, namely the hidden representation of the image characteristics, as the input of the fully-connected neural network, and the fully-connected neural network classifies the output to obtain the information of the prediction stage.

And step S408, determining abnormal stage information of the first user according to the cognitive function evaluation information and the prediction stage information.

Specifically, the server combines the prediction stage information obtained by classifying the full-connection neural network with the cognitive function evaluation information of the first user, and comprehensively judges the abnormal stage information of the first user. For example, if the server determines that the predicted stage information obtained by classifying the fully-connected neural network matches the abnormal stage information reflected by the cognitive function assessment information of the first user, the server takes the predicted stage information as the abnormal stage information of the first user. For another example, if the server determines that the predicted stage information obtained by classifying the fully-connected neural network is the E stage in the foregoing embodiment, and determines that the stage information of the first user may be located in the C stage in the foregoing embodiment according to the cognitive function assessment information, the server may determine that the abnormal stage information of the first user is the E stage or the D stage.

In the embodiment, the medical image information is analyzed by combining the long-time memory neural network model and the full-connection neural network, so that the identification accuracy of the abnormal stage information is improved, the influence of artificial subjective factors is avoided compared with a traditional manual on-site determination mode, and the efficiency is higher.

In some embodiments, as shown in fig. 5, the training step of the long-term and short-term memory neural network model includes:

step S502, acquiring a current sample image, and taking the image characteristics in the current sample image as the input value of the current moment.

Step S504, according to the hidden representation of the image feature at the previous moment, a forgetting gate performs weighted operation on the input value to obtain a first intermediate value, an input gate performs weighted operation on the input value to obtain a second intermediate value and a third intermediate value, and an output gate performs weighted operation on the input value to obtain a fourth intermediate value.

Step S506, determining the unit state of the image feature at the current time according to the unit state, the first intermediate value, the second intermediate value, and the third intermediate value of the image feature at the previous time.

Step S508, obtaining an output value of the current moment according to the unit state of the image characteristic of the current moment and the fourth intermediate value; wherein the output value is a hidden representation of the image feature at the current time.

And step S510, acquiring a next sample image, taking the image characteristics in the next sample image as the input value of the next moment, returning to the hidden representation according to the image characteristics of the previous moment, performing weighted operation on the input value by a forgetting gate to obtain a first intermediate value, performing weighted operation on the input value by an input gate to obtain a second intermediate value and a third intermediate value, and performing weighted operation on the input value by an output gate to obtain a fourth intermediate value, continuing to execute the steps until a training termination condition is met, and stopping iteration to obtain a trained long-time and short-time memory neural network model.

The training termination condition includes, but is not limited to, that the number of times of training reaches a preset number of times, the training duration reaches a preset duration, or the accuracy reaches a preset threshold, and the like.

Specifically, the server acquires a plurality of medical images as training samples and inputs the medical images into a long-time and short-time memory neural network model for training. For a certain training process, the server firstly obtains a sample image of the current training as a current sample image, and performs feature extraction through a long-time and short-time memory neural network model to obtain the image features of the current sample image. Since the long-term memory neural network is a recurrent neural network, the current input needs to be combined with the last output. Therefore, the server performs weighted operation on the input value by the forgetting gate to obtain a first intermediate value, performs weighted operation on the input value by the input gate to obtain a second intermediate value and a third intermediate value, and performs weighted operation on the input value by the output gate to obtain a fourth intermediate value according to the hidden representation of the image feature at the previous moment, and determines the unit state of the image feature at the current moment according to the unit state of the image feature at the previous moment, the first intermediate value, the second intermediate value and the third intermediate value. Then, the server obtains an output value at the current time, namely the hidden representation of the image feature at the current time, by combining the unit state of the image feature at the current time and the fourth intermediate value. Meanwhile, the server stores the hidden representation of the image feature at the current moment for being used as an input value in the next training, so that an iterative loop is performed.

And when the next training is carried out, the server acquires the next sample image, takes the image characteristics in the next sample image as the input value of the next moment, returns to the step S504 to continue executing until the training termination condition is met, and stops iteration to obtain the trained long-time and short-time memory neural network model.

Exemplarily, as shown in fig. 6, the server inputs the sample image into the long-term and short-term memory neural network model, and performs feature extraction through the long-term and short-term memory neural network model to obtain an image feature x of the current sample image_t(including all image features of the sample image) as an input value at the current time (i.e., time t). Then, the hidden representation h of the image characteristic at the last moment (namely the t-1 moment) obtained by the last training is represented_t-1Input value x from forgetting gate_tCarrying out weighting operation to obtain a first intermediate value f_t. Wherein, forgetting to input value x_tCarrying out weighting operation to obtain a first intermediate value f_tThe calculation formula of (c) can be represented by the following formula:

f_t＝σ(W_f·[h_t-1,x_t]+b_f)

where σ represents a sigmoid function, W_fWeight matrix for forgetting gate, b_fIs a bias parameter.

In the long-time memory neural network model, a forgetting gate is used for processing, namely, the output of the sigmoid function is multiplied by the unit state by the sigmoid function through the input value at the current moment and the output value at the last moment. If the sigmoid function outputs 0, the part of information needs to be forgotten, otherwise, the part of information continues to be passed on in the state of the unit. At the same time, processing is performed by the input gate, updating the old cell state. The previous forget gate determines which information to forget or add, is implemented by the input gate, and is processed by the output gate to determine the final output value.

Illustratively, the long-term memory neural network model inputs a value x from an input gate_tPerforming weighting operation to obtain a second intermediate value i_tAnd a third intermediate value

Wherein the input gate inputs the value x_tRespectively obtaining second intermediate values i by weighting operation_tAnd a third intermediate value

The calculation formula of (c) can be represented by the following formula:

i_t＝σ(W_i·[h_t-1,x_t]+b_i)

wherein, W_iAnd W_CWeight matrix of input gates, b_iAnd b_CIn order to input the offset parameter of the gate,

is a vector of candidate values used to decide which information can be added to the cell state.

Illustratively, the long-term memory neural network model is implemented by an output gate for an input value x_tPerforming weighting operation to obtain a fourth intermediate value o_t. Wherein the output gate pair inputs the value x_tPerforming weighting operation to obtain fourth intermediate values o_tThe calculation formula of (c) can be represented by the following formula:

o_t＝σ(W_o·[h_t-1,x_t]+b_o)

wherein, W_oAs a weight matrix of output gates, b_oIs the bias parameter of the output gate.

After the first intermediate value, the second intermediate value and the third intermediate value are obtained, the server obtains the unit state C of the image characteristic at the last moment_t-1Determining the cell state C of the image feature at the current time_t. I.e. the old state C is used first_t-1Multiplication of corresponding points by f_tI.e. C_t-1·f_tDiscarding the information that has been determined to be forgotten, and adding it

The final constituent cell state C_t. It can be formulated as:

then, the server combines the unit state C of the image feature at the current moment_tAnd a fourth intermediate value o_tObtaining the output value of the current time, namely the hidden representation h of the image characteristic of the current time_t. Meanwhile, the server hides the image characteristics at the current moment to represent h_tAnd storing the data for being used as an input value in the next training, thereby performing an iterative loop.

In the embodiment, the medical image information is analyzed by memorizing the neural network model at long time, so that the identification accuracy of the abnormal stage information can be improved, compared with the traditional remote mode in which the diagnosis is carried out manually, the influence of artificial subjective factors is avoided, and the accuracy and the efficiency are higher.

In some embodiments, as shown in fig. 7, the step of training the fully-connected neural network comprises:

step S702, inputting the image characteristics, the hidden representation of the image characteristics and the unit state of the sample image into a fully-connected neural network, and outputting an abnormal stage prediction result; the abnormal stage prediction result represents the probability values of different abnormal stages.

Step S704, determining an image feature to be deleted based on a difference between the abnormal stage prediction result and the abnormal stage to which the sample image actually belongs, and determining an adjustment parameter based on the image feature to be deleted.

Step S706, based on the abnormal stage prediction result and the adjustment parameter, determining a reward function.

Step S708, aiming at maximizing the reward function, adjusting the adjusting parameters of the fully-connected neural network until the training termination condition is met, and ending the training to obtain the trained fully-connected neural network.

Specifically, the server inputs image features of the sample image extracted by the long-and-short-term memory neural network model, hidden representations of the image features output by the long-and-short-term memory neural network model processing, and unit states in the long-and-short-term memory neural network model into the fully-connected neural network, classifies the fully-connected neural network, and outputs an abnormal stage prediction result. The abnormal stage prediction result represents the probability values of different abnormal stages. Typically, the output abnormal stage predictors are probability distributions, for example, for the three abnormal stages c1, c2, c3, the output abnormal stage predictor P (y | X) is a probability distribution for the diseased stage prediction, where y ∈ { c1, c2, c3 }. For each sample image, the server is pre-marked with the abnormal stage to which the sample image actually belongs. Therefore, the server can compare the abnormal stage prediction result output by the fully-connected neural network with the abnormal stage to which the sample image actually belongs, and determine whether the image characteristics of the sample image are retained or deleted according to the difference between the two abnormal stage prediction results. Based on the image features that should be deleted, the server may determine adjustment parameters and construct a reward function in conjunction with the abnormal stage prediction results.

Therefore, the server can train based on the reward value returned by the constructed reward function, the reward function is maximized, the adjustment parameters of the fully-connected neural network are adjusted, iteration is continuously and circularly performed until the training termination condition is met, and the trained fully-connected neural network is obtained. The training termination condition includes, but is not limited to, that the number of times of training reaches a preset number of times, the training duration reaches a preset duration, or the accuracy reaches a preset threshold, and the like.

Illustratively, the reward function may be calculated by the following formula:

wherein R is_LFor the bonus value, P (c | X) is the prediction result in the abnormal stage obtained by inputting the sample image feature. L' refers to the number (times) of features deleted from a sample, which is used to delete a feature when the effect of the feature on the final result is small. Accordingly, L is the number of complete features in the sample image. Gamma is a hyperparameter, and the larger gamma is, the more the deletion is prone toCharacteristic; conversely, smaller γ tends to retain the feature. b is also a super parameter, and the reward value is made positive or negative through manual setting.

In the embodiment, self-learning of the neural network is completed through deep reinforcement learning, the long-time memory neural network and the short-time memory neural network are combined with the fully-connected neural network, the stage judgment model is built, the prediction result and the actual result of the analysis model are compared, and the identification accuracy of the abnormal stage is improved.

In some embodiments, the above method further comprises: sending evaluation standard information to the first terminal so that the first terminal can visually display evaluation content to the first user; the evaluation criterion information is used for guiding the first user to perform an evaluation test on the first terminal, and includes but is not limited to a cognitive function evaluation scale, an evaluation guidance video, or other materials capable of guiding the first user to perform evaluation. The visual display includes, but is not limited to, one or more of a picture display, a text display, a video display, a voice display, and the like, or a combination of more.

Specifically, the server may send evaluation criterion information to the first terminal based on a remote evaluation request of the first terminal, so that the first terminal can visually display evaluation content to the first user. For example, a first user performs a touch operation or the like on a first terminal, and the first terminal sends a remote evaluation request to the server accordingly. After receiving the remote evaluation request sent by the first terminal, the server searches corresponding evaluation standard information (such as various measuring tables) in the database and sends the evaluation standard information to the first terminal. After receiving the evaluation standard information sent by the server, the first terminal visually displays the scale to the first user through a display screen, for example, displays the scale to the first user through a page, or displays the scale to the first user in combination with an audio prompt, or plays a video for guiding the first user to perform remote evaluation to the first user, and the like.

In the embodiment, the evaluation standard information is sent to the first terminal, so that the user can finish remote evaluation without going to a hospital, and the evaluation is convenient and flexible.

In some embodiments, the above method further comprises: and establishing communication connection between the first terminal and the second terminal according to an interactive request initiated by the first terminal or the second terminal so that the first terminal can receive an evaluation instruction sent by the second terminal and carry out remote evaluation according to the evaluation instruction. The interactive request includes, but is not limited to, one or more of a chat request, a voice request, a video request, and the like. The evaluation instruction is used for guiding the first user to perform remote evaluation, such as a voice instruction, a text tutorial, or a video tutorial sent by the second user.

Specifically, the server may further receive an interaction request initiated by the first terminal, and request a communication connection from the second terminal. After the second user agrees to the communication connection on the second terminal, the server can establish the communication connection between the first terminal and the second terminal. Therefore, the first user can finish the evaluation according to the evaluation instruction sent by the second user through the second terminal. Or the server receives an interaction request initiated by the second terminal and requests the first terminal for communication connection. After the first user agrees to the communication connection on the first terminal, the server can establish the communication connection between the first terminal and the second terminal. Thus, the second user may send an evaluation instruction to the first user, and the first user may perform the evaluation according to the evaluation instruction.

In the above embodiment, the communication connection between the first user and the second user is established through an interaction request initiated by one of the first user and the second user, so that remote evaluation can be actively and timely performed, for example, when the second user visits back regularly or finds an abnormality, the situation of the first user is actively known; for another example, the first user actively performs remote evaluation when needed, so that the interactivity is strong and the user experience is good. Meanwhile, the second user can conduct remote guidance cognitive function assessment according to professional opinions, the remote assessment result of the first user is more accurate, and further the subsequent training scheme is more effective and more targeted.

In some embodiments, the above method further comprises: pushing the training scheme to a second terminal so that a second user can modify the training scheme on the second terminal; and receiving the modified training scheme sent by the second terminal, and pushing the modified training scheme to the first terminal.

Specifically, after determining the corresponding training scheme, the server may also push the training scheme to the second terminal at the same time, so that the second user may view the training scheme on the second terminal or modify the training scheme. And after the second user modifies the training scheme, the second terminal sends the modified training scheme to the server, and the server receives the modified training scheme sent by the second terminal and pushes the modified training scheme to the first terminal, so that the first user can perform training operation according to the modified training scheme.

In the above embodiment, the training scheme obtained by automatic matching is sent to the second user, so that the second user can further perform summary analysis according to the cognitive function evaluation information stored in the database by the server, and perform professional analysis by combining with the image data, so that professional modification or adjustment is performed on the training scheme, and the training scheme is more accurate and more targeted.

It should be understood that, although the steps in the flowcharts of fig. 2, 4, 6-7 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2, 4, 6-7 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or in alternation with other steps or at least some of the other steps.

In one embodiment, as shown in fig. 8, there is provided a push device 800 of a remote training scenario, comprising: an obtaining module 801, a determining module 802, and a pushing module 803, wherein:

an obtaining module 801, configured to obtain cognitive function assessment information and medical related information sent by a first terminal; the cognitive function evaluation information is obtained by remote evaluation of a user on a first terminal; the medically-related information includes medical image information and biographic information.

The determining module 802 is configured to determine abnormal stage information according to the cognitive function assessment information and the medical related information.

The determining module 802 is further configured to determine a training scheme matching the abnormal stage information according to a plurality of training materials in the database, where the training scheme includes training materials related to a plurality of cognitive domains.

The pushing module 803 is configured to push the training scheme to the first terminal, so that the user performs a training operation according to the training scheme.

In one embodiment, the determining module is further configured to perform image preprocessing on the medical image information to obtain a preprocessed image; inputting the preprocessed image into a pre-trained long-time and short-time memory neural network model, and extracting features by using the long-time and short-time memory neural network model to obtain hidden representation of image features; the image features comprise a waveband change feature, an edge feature and a gray feature; classifying by utilizing a fully-connected neural network based on the biological data information and the hidden representation of the image characteristics to obtain information of a prediction stage; and determining abnormal stage information of the first user according to the cognitive function evaluation information and the prediction stage information.

In one embodiment, as shown in fig. 9, the apparatus further includes a training module 804, where the training module is configured to obtain a current sample image, and use an image feature in the current sample image as an input value at a current time; according to the hidden representation of the image feature at the previous moment, a forgetting gate performs weighted operation on an input value to obtain a first intermediate value, an input gate performs weighted operation on the input value to obtain a second intermediate value and a third intermediate value, and an output gate performs weighted operation on the input value to obtain a fourth intermediate value; determining the unit state of the image feature at the current moment according to the unit state, the first intermediate value, the second intermediate value and the third intermediate value of the image feature at the previous moment; obtaining an output value of the current moment according to the unit state of the image characteristic of the current moment and the fourth intermediate value; wherein the output value is a hidden representation of the image feature at the current moment; and obtaining a next sample image, taking the image characteristics in the next sample image as input values of the next moment, returning to the hidden representation according to the image characteristics of the previous moment, performing weighted operation on the input values by a forgetting gate to obtain a first intermediate value, performing weighted operation on the input values by an input gate to obtain a second intermediate value and a third intermediate value, and performing weighted operation on the input values by an output gate to obtain a fourth intermediate value, and continuing to perform the steps until a training termination condition is met, stopping iteration, and obtaining the trained long-time and short-time memory neural network model.

In one embodiment, the training module is further configured to input the image features, the hidden representations of the image features, and the unit states of the sample images into a fully-connected neural network, and output an abnormal stage prediction result; the abnormal stage prediction result represents probability values in different abnormal stages; determining image characteristics to be deleted based on the difference between the abnormal stage prediction result and the abnormal stage to which the sample image actually belongs, and determining adjustment parameters based on the image characteristics to be deleted; determining a reward function based on the abnormal stage prediction result and the adjustment parameter; and adjusting the adjusting parameters of the fully-connected neural network by taking the maximum reward function as a target, and ending the training until the training termination condition is met to obtain the trained fully-connected neural network.

In one embodiment, the apparatus further includes a sending module, configured to send evaluation criterion information to the first terminal, so that the first terminal visually displays evaluation content to the first user; the evaluation standard information is used for guiding the first user to carry out evaluation test on the first terminal; the visual display at least comprises one of picture display, character display, video display and voice display.

In one embodiment, the apparatus further includes an interaction module, configured to establish a communication connection between the first terminal and the second terminal according to an interaction request initiated by the first terminal or the second terminal, so that the first terminal receives an evaluation instruction sent by the second terminal, and performs remote evaluation according to the evaluation instruction; wherein the interactive request at least comprises one of a chat request, a voice request and a video request.

In one embodiment, the pushing module is further configured to push the training scheme to the second terminal, so that the second user can modify the training scheme on the second terminal; and receiving the modified training scheme sent by the second terminal, and pushing the modified training scheme to the first terminal.

For specific definition of the pushing device of the remote training scheme, reference may be made to the above definition of the pushing method of the remote training scheme, which is not described herein again. The modules in the pushing device of the remote training scheme described above may be implemented in whole or in part by software, hardware, and combinations thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, and the computer device may be the server in the foregoing embodiments, and its internal structure diagram may be as shown in fig. 10. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing data such as cognitive function assessment information, medical related information, training data and the like. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a push method of a remote training scenario.

Those skilled in the art will appreciate that the architecture shown in fig. 10 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having a computer program stored therein, the processor implementing the following steps when executing the computer program: acquiring cognitive function evaluation information and medical related information sent by a first terminal; the cognitive function evaluation information is obtained by a first user through remote evaluation on a first terminal; the medical related information comprises medical image information and biological data information; determining abnormal stage information according to the cognitive function evaluation information and the medical related information; determining a training scheme matched with the abnormal stage information according to a plurality of training data in the database, wherein the training scheme comprises training data related to a plurality of cognitive domains; and pushing the training scheme to the first terminal so that the first user can perform training operation according to the training scheme.

In one embodiment, the processor, when executing the computer program, further performs the steps of: carrying out image preprocessing on the medical image information to obtain a preprocessed image; inputting the preprocessed image into a pre-trained long-time and short-time memory neural network model, and extracting features by using the long-time and short-time memory neural network model to obtain hidden representation of image features; the image features comprise a waveband change feature, an edge feature and a gray feature; classifying by utilizing a fully-connected neural network based on the biological data information and the hidden representation of the image characteristics to obtain information of a prediction stage; and determining abnormal stage information of the first user according to the cognitive function evaluation information and the prediction stage information.

In one embodiment, the processor, when executing the computer program, further performs the steps of: acquiring a current sample image, and taking the image characteristics in the current sample image as an input value at the current moment; according to the hidden representation of the image feature at the previous moment, a forgetting gate performs weighted operation on an input value to obtain a first intermediate value, an input gate performs weighted operation on the input value to obtain a second intermediate value and a third intermediate value, and an output gate performs weighted operation on the input value to obtain a fourth intermediate value; determining the unit state of the image feature at the current moment according to the unit state, the first intermediate value, the second intermediate value and the third intermediate value of the image feature at the previous moment; obtaining an output value of the current moment according to the unit state of the image characteristic of the current moment and the fourth intermediate value; wherein the output value is a hidden representation of the image feature at the current moment; and obtaining a next sample image, taking the image characteristics in the next sample image as input values of the next moment, returning to the hidden representation according to the image characteristics of the previous moment, performing weighted operation on the input values by a forgetting gate to obtain a first intermediate value, performing weighted operation on the input values by an input gate to obtain a second intermediate value and a third intermediate value, and performing weighted operation on the input values by an output gate to obtain a fourth intermediate value, and continuing to perform the steps until a training termination condition is met, stopping iteration, and obtaining the trained long-time and short-time memory neural network model.

In one embodiment, the processor, when executing the computer program, further performs the steps of: inputting the image characteristics, the hidden representation of the image characteristics and the unit state of the sample image into a fully-connected neural network, and outputting an abnormal stage prediction result; the abnormal stage prediction result represents probability values in different abnormal stages; determining image characteristics to be deleted based on the difference between the abnormal stage prediction result and the abnormal stage to which the sample image actually belongs, and determining adjustment parameters based on the image characteristics to be deleted; determining a reward function based on the abnormal stage prediction result and the adjustment parameter; and adjusting the adjusting parameters of the fully-connected neural network by taking the maximum reward function as a target, and ending the training until the training termination condition is met to obtain the trained fully-connected neural network.

In one embodiment, the processor, when executing the computer program, further performs the steps of: sending evaluation standard information to the first terminal so that the first terminal can visually display evaluation content to the first user; the evaluation standard information is used for guiding the first user to carry out evaluation test on the first terminal; the visual display at least comprises one of picture display, character display, video display and voice display.

In one embodiment, the processor, when executing the computer program, further performs the steps of: establishing communication connection between the first terminal and the second terminal according to an interactive request initiated by the first terminal or the second terminal, so that the first terminal receives an evaluation instruction sent by the second terminal and carries out remote evaluation according to the evaluation instruction; wherein the interactive request at least comprises one of a chat request, a voice request and a video request.

In one embodiment, the processor, when executing the computer program, further performs the steps of: pushing the training scheme to a second terminal so that a second user can modify the training scheme on the second terminal; and receiving the modified training scheme sent by the second terminal, and pushing the modified training scheme to the first terminal.

The computer equipment realizes remote cognitive assessment and training, so that a user can finish home training without special medical equipment.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of: acquiring cognitive function evaluation information and medical related information sent by a first terminal; the cognitive function evaluation information is obtained by a first user through remote evaluation on a first terminal; the medical related information comprises medical image information and biological data information; determining abnormal stage information according to the cognitive function evaluation information and the medical related information; determining a training scheme matched with the abnormal stage information according to a plurality of training data in the database, wherein the training scheme comprises training data related to a plurality of cognitive domains; and pushing the training scheme to the first terminal so that the first user can perform training operation according to the training scheme.

In one embodiment, the computer program when executed by the processor further performs the steps of: : carrying out image preprocessing on the medical image information to obtain a preprocessed image; inputting the preprocessed image into a pre-trained long-time and short-time memory neural network model, and extracting features by using the long-time and short-time memory neural network model to obtain hidden representation of image features; the image features comprise a waveband change feature, an edge feature and a gray feature; classifying by utilizing a fully-connected neural network based on the biological data information and the hidden representation of the image characteristics to obtain information of a prediction stage; and determining abnormal stage information of the first user according to the cognitive function evaluation information and the prediction stage information.

In one embodiment, the computer program when executed by the processor further performs the steps of: : acquiring a current sample image, and taking the image characteristics in the current sample image as an input value at the current moment; according to the hidden representation of the image feature at the previous moment, a forgetting gate performs weighted operation on an input value to obtain a first intermediate value, an input gate performs weighted operation on the input value to obtain a second intermediate value and a third intermediate value, and an output gate performs weighted operation on the input value to obtain a fourth intermediate value; determining the unit state of the image feature at the current moment according to the unit state, the first intermediate value, the second intermediate value and the third intermediate value of the image feature at the previous moment; obtaining an output value of the current moment according to the unit state of the image characteristic of the current moment and the fourth intermediate value; wherein the output value is a hidden representation of the image feature at the current moment; and obtaining a next sample image, taking the image characteristics in the next sample image as input values of the next moment, returning to the hidden representation according to the image characteristics of the previous moment, performing weighted operation on the input values by a forgetting gate to obtain a first intermediate value, performing weighted operation on the input values by an input gate to obtain a second intermediate value and a third intermediate value, and performing weighted operation on the input values by an output gate to obtain a fourth intermediate value, and continuing to perform the steps until a training termination condition is met, stopping iteration, and obtaining the trained long-time and short-time memory neural network model.

In one embodiment, the computer program when executed by the processor further performs the steps of: : inputting the image characteristics, the hidden representation of the image characteristics and the unit state of the sample image into a fully-connected neural network, and outputting an abnormal stage prediction result; the abnormal stage prediction result represents probability values in different abnormal stages; determining image characteristics to be deleted based on the difference between the abnormal stage prediction result and the abnormal stage to which the sample image actually belongs, and determining adjustment parameters based on the image characteristics to be deleted; determining a reward function based on the abnormal stage prediction result and the adjustment parameter; and adjusting the adjusting parameters of the fully-connected neural network by taking the maximum reward function as a target, and ending the training until the training termination condition is met to obtain the trained fully-connected neural network.

In one embodiment, the computer program when executed by the processor further performs the steps of: : sending evaluation standard information to the first terminal so that the first terminal can visually display evaluation content to the first user; the evaluation standard information is used for guiding the first user to carry out evaluation test on the first terminal; the visual display at least comprises one of picture display, character display, video display and voice display.

In one embodiment, the computer program when executed by the processor further performs the steps of: : establishing communication connection between the first terminal and the second terminal according to an interactive request initiated by the first terminal or the second terminal, so that the first terminal receives an evaluation instruction sent by the second terminal and carries out remote evaluation according to the evaluation instruction; wherein the interactive request at least comprises one of a chat request, a voice request and a video request.

In one embodiment, the computer program when executed by the processor further performs the steps of: : pushing the training scheme to a second terminal so that a second user can modify the training scheme on the second terminal; and receiving the modified training scheme sent by the second terminal, and pushing the modified training scheme to the first terminal.

The computer readable storage medium realizes remote cognitive assessment and training, so that a user can finish home training without special medical equipment.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. A method for pushing a remote training scenario, the method comprising:

2. The method of claim 1, wherein determining abnormal stage information from the cognitive function assessment information and medically relevant information comprises:

3. The method of claim 2, wherein the training step of the long-and-short-term memory neural network model comprises:

4. The method of claim 2, wherein the step of training the fully-connected neural network comprises:

5. The method of claim 1, further comprising:

6. The method of claim 1, further comprising:

7. The method of claim 1, further comprising:

8. A push device for a remote training scenario, the device comprising:

9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 7.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.