Medical intelligent system application in schizophrenia treatment satisfaction based on neural networks

Aiming at the problems of poor convergence and uneven energy consumption of heed routing protocol, this paper proposes an opfh routing protocol based on options clustering algorithm for the first time. Firstly, the protocol divides the network into several top-level groups by using options clustering algorithm, then elects each top-level group at the same time, and designs and develops the medical intelligent system through wireless sensor. The medical intelligent system developed in this paper not only facilitates the completion of rehabilitation work, but also facilitates the management of hospital rehabilitation, greatly reducing manpower and material resources. It plays an important role in the treatment of schizophrenia. However, patients with schizophrenia have high inner sensitivity, often accompanied by anxiety and depression. Nurses need to take the initiative to communicate with patients, understand and meet the needs of patients, and use the system to manage the negative emotions of patients, so as to improve the confidence and compliance of patients. In order to improve the treatment satisfaction of patients, the large-scale data set without marked target task was used for domain transfer learning of small-scale data set. In this paper, the research of wireless sensor is applied to medical intelligent system, so as to promote the vigorous development of medical intelligent system.


Introduction
Wireless sensor network technology is one of the most in uential technologies currently, widely used in military, medical, environmental monitoring and other elds.The wireless sensor network is composed of many small sensor nodes randomly distributed in a certain area to collect data.It has certain storage capacity, communication capacity and computing capacity, but the resources are limited.Traditional network protocols are not directly applicable to wireless sensor networks [1].Therefore, it is of great signi cance to study energy-e cient routing protocols for wireless sensor networks.It is one of the current research hotspots, and this article studies how to build a schizophrenia medical intelligent system [2].The main steps include: knowledge extraction, fusion and storage.Among them, knowledge extraction adopts different extraction methods for different data sources, mainly using web crawler technology, Bi-LSTM-CRF-based named entity discovery model and other related technologies to achieve entity extraction, which promotes the emergence of two-way LSTM [3].Use the attention mechanism to extract the relationship, use the method of aligning the similar attributes of the entity relationship to do knowledge fusion work, and use the Neo4j graph database to store the data of the extracted entity relationship and other items [4].To study how to complete the question and answer in the eld of medical intelligence based on the medical intelligence system, there are mainly two parts: semantic search and intelligent question and answer [5].The SVM-based classi er is used to classify the sentence intent declared by the user, and the cosine similarity matching can calculate the similarity between entities and return the most similar result.The Stanford CoreNLP tool performs dependency analysis on the question set to determine the user's intention to ask questions [6].It uses the Cypher query language that comes with the Neo4j knowledge graph database for keyword search and matching, and nally uses the WeChat o cial account platform and interactive user interface.Build an intelligent question-and-answer service platform integrating medical and health consultation services and treatment satisfaction surveys [7].
Implementation process of rehabilitation training: According to the characteristics of patients with schizophrenia and the development stage of the patient's condition, the nursing staff is responsible for formulating and implementing individual rehabilitation plans to promote the gradual recovery of the patients' physical and social functions [8].The test of the medical intelligent system shows that the system has basically met the expected demand [9].From the operating results, it can be concluded that the system can provide perfect services to patients and doctors [10].The system provides interactive medical follow-up services for patients and doctors in a cost-effective manner, which not only facilitates patient feedback, but also improves the e ciency of the hospital follow-up process [11].

Related Work
The literature introduces the construction of knowledge graphs.It mainly starts with the steps of knowledge graph construction [12].First of all, the basic part of knowledge acquisition is knowledge graph construction.Here we introduce the relevant steps of each knowledge source in combination with data sources.Secondly, it introduces the knowledge fusion part of knowledge graph construction and the main algorithms used in knowledge fusion.Finally, enter the knowledge storage part after the creation of the knowledge graph [13].The literature introduces the design and implementation of an intelligent question answering system based on knowledge graphs.First introduce the requirements analysis and system development environment [14].From the perspective of the system, from the knowledge map intelligent answering medical and health eld, the experimental effect of the system is demonstrated.The literature introduces LEACH's improved WSN routing algorithm.Based on the LEACH protocol, in view of its shortcomings, improvements were proposed from various aspects such as threshold function, cluster head selection, transmission path, etc., and a simulation model was established through Matlab to test the performance of LEACH EPM algorithm [15].The literature introduces the WSN mobile convergence routing algorithm of MobileSink.Using the clustering method of LEACH-C protocol, combined with the clustering algorithm of K-Means, a new clustering strategy is proposed.For solutions with different realtime data requirements, the path of the mobile receiver will be planned separately to compensate for the power consumption of the network [16].And through Matlab simulation to verify the performance of the proposed algorithm.The literature introduces the main process of the retrieval question and answer system, and details the construction and annotation method of the recovery question and answer system data set [17].Next, we introduce the semantic matching model and attention mechanism based on dynamic routing proposed in this paper, and combine the model it is compared with some existing matching models [18].At the same time, an ablation comparison experiment was performed on each model architecture, and the results show that the model proposed in this paper can better model semantic information and sentence matching interactive information.

Wireless sensor
The rst step of traditional routing protocols is to select an optimal transmission path, and then make full use of bandwidth to transmit data on the optimal path found.Traditional routing protocols usually provide high-quality services and effective bandwidth usage, while WSN resources are limited.More e cient collaboration is needed to complete different tasks, and different protocols need to be designed to manage the WSN routing protocol according to different application scenarios.The following points should also be noted: The routing protocol is a popular protocol widely used at present, which is widely used in high-density wireless sensor networks with hundreds of sensor nodes.In this case, the routing protocol must have a good resource management strategy, data fusion, load balancing and high robustness.The routing protocol consists of two stages: pool and data transmission.The rst task in the clustering stage is to select cluster heads.The choice of leaders is divided into central choice and distributed choice.The centralized selection of cluster heads is mainly completed by the base station, and the distributed process is mainly completed between sensor nodes.It is produced through mutual competition, and the conditions of competition can be probability or node energy.After selecting the last cluster head, each cluster head establishes a transmission path with the base station in a single-hop or multi-hop manner.
When sending data to the terminal, it is responsible for collecting and merging data from the nodes in the group, and then sending the data to the base station through a pre-con gured transmission path.

Key technologies of medical intelligent system
Csuanhu uses a contradictory learning framework to switch learning between word segmentation tasks and named entity recognition tasks.The main purpose of named entity detection is to determine the boundaries of entities and divide them into prede ned categories.The Chinese named entity recognition task has very little data, and the Chinese named entity recognition task and the Chinese word segmentation task have many identical words, and each task has its own uniqueness.Therefore, they proposed a new adversarial transfer learning framework, which makes full use of words and named entities to recognize shared information of the task.As shown in Fig. 1, the word segmentation task and the named entity recognition task have similarities.
Common text presentation levels and a task-speci c text presentation level.In this model, two-way LSTM is used to render text, and then the shared layer and private layer of the task are used as the last task for subsequent tasks.The opponent's loss is introduced in the public presentation layer so that these two tasks can be handled.By conducting experiments on two named entity recognition data sets, they proved the effectiveness of the contradiction learning algorithm framework for this task.

Mathematical model
The task of semantic matching is to evaluate the semantic consistency of two given texts.The rst task is to capture and express the semantic information in the text.In natural language processing tasks, text is generally composed of fully connected neural networks, convolutional neural networks, or recurrent neural networks.
The formal expression of speci c attention is shown in formulas 1 to 3: 1 2 3 Among them, (,) is the calculation function of the similarity between the query and any key.There are many different implementations here, and   is the normalized weight corresponding to different values, and nally get the query for all < key, value > pair of weighted attention vector.In this model, the attention mechanism is used to interact with two sentence vectors, and the vector dot product is introduced as the similarity calculation function, and query, key, and value are the vector representations of the two texts respectively.
For the input text, this article uses two different forms of vectorized representation, the xed word vector representation based on Word2Vec and the dynamic word vector based on BERT and the vectorized representation of the two input texts is shown in formula (4): 4 For the input vector, it will rst pass through a highway layer to make a preliminary representation of the input vector.The speci c form of the highway layer is as follows: In order to obtain the interactive attention matching information of two texts, the attention calculation of dot multiplication is performed on the input two texts.Calculate the attention weight of each word in the text  1 with each word in the text  2 to get   , which represents the degree of similarity between the word at the i-th position in  1 and the word at the j-th position in  2 .The speci c calculation formula is:

Simulation analysis
When extracting knowledge from unstructured information data, named entity detection method is used for entity extraction.The entity recognition uses the Bi-LSTM-CRF algorithm model, and the experiment proves that it is indeed better than the LSTM and Bi-LSTM models.The main reason is that the bidirectional LSTM can concatenate the hidden layer vectors before and after the merging, and fully consider the context information.Compared with the LSTM, the entity labeling will be more accurate.The Bi-LSTM-CRF model adds a CRF layer on the basis of Bi-LSTM.CRF can be considered as the decoding layer of the model.It has broadcast properties to test whether the predicted response is reasonable; therefore, the effect of the Bi-LSTM-CRF model is better.Figure 2  feature word in each sentence has a different contribution to the sentence, the attention mechanism can automatically identify which features of the sentence are important keywords, which undoubtedly plays a key role in the role ranking.The purpose of adding the attention mechanism to the original model is to assign different weights to each feature, so as to achieve a good classi cation effect.Figure 3 below is the result of the model evaluation experiment.

System overview
Schizophrenia is a mental illness.Lack of mental, emotional and thinking coordination is the most important symptom.The morbidity rate is very high, and it continues to increase.This is because elderly patients have poor self-care ability and are prone to negative emotions such as memory disorders, sleep disorders, anxiety and depression, which leads to poor compliance of many patients to treatment, which affects the treatment effect.Without timely and effective care, adverse events such as suffocation, falling, and getting out of bed are prone to occur, which is not conducive to the recovery of the patient's health.The concept of high-quality care is to put patients rst and provide adequate nursing interventions to help them recover.
Good patient satisfaction is a prerequisite for a harmonious doctor-patient relationship, and good doctorpatient communication through a friendly system is also an important means to improve patient satisfaction.The system is designed for this purpose, and it is a medical monitoring system that can perform various types of monitoring.
Generally speaking, there are mainly the following three ways.1. Careful follow-up to provide reminders for patients, such as: follow-up examination, medication, life, health education, etc. 2. Management monitoring: satisfaction survey, medical ethics survey, complaint and suggestion feedback, management level improvement.3. Monitoring scienti c research: recording patients, statistical analysis, and improving medical standards.
The medical follow-up system is a service system that integrates the functions of appointment registration, post-diagnosis follow-up, late registration, hospital information, follow-up management and other functions.In the meantime, the appointment le and hospital consultation are combined with the existing system and data of the hospital, so I won't repeat them in this article.The post-diagnosis monitoring, follow-up noti cation and other functions developed by this system are explained in detail in the requirements and design stages.The purpose of the system is to provide patients with comfortable medical services and high-quality medical expertise, save doctors' time and money, and simplify the entire follow-up process.The system includes a front-end small program module for lling out the followup form, and the construction of a public WeChat platform equipped with small programs (including doctor-patient message templates), voice smartphones, and consultation platform for the back-end of data collection and processing Information management function.

System requirement analysis
The combination of computer technology and traditional medicine saves manpower and material resources when providing services to patients, saves hospital resources, makes patients' medical conditions more e cient and humanized, and promotes the progress of the medical industry.Medical tracking systems are also used for this purpose.Let's analyze the main functional requirements of the system.This system mainly solves the following problems: A. Make an appointment for registration through the system The system is linked with the reservation record function of the hospital o cial website, and can query the reservation record of patients in the system.

B. Hospital information
Hospital information includes doctor introduction, service introduction, medical disclosure, interruption notice, treatment information, hospital location and other related information.Patients can query the relevant content through the system.
C. The patients were followed up Before using the doctor's monitoring system, the hospital also monitored the patients after diagnosis.The main form of follow-up is paper-based follow-up; Follow up examination is required manually by the hospital.The monitoring effect of this monitoring method is not good, and it costs the hospital human and material resources.
The current follow-up is basically satisfactory follow-up and follow-up after diagnosis.Different types of follow-ups need to be created to encourage patients to participate in the follow-up.The patient's satisfaction is recorded only once to prevent some people from maliciously viewing the evaluation.Satisfaction monitoring is an evaluation of the hospital's medical environment and medical quality.In the follow-up examination after the diagnosis, the patient's condition is asked.If the patient has positive feedback after the consultation, the corresponding doctor can receive feedback on the template message.In addition, follow-up scienti c examinations can also be performed on some patients according to the needs of the hospital.

D. Appointment reminder
Intelligent voice phone is the key function of tracking service.If the patient is not involved in the follow-up diagnosis after the diagnosis is associated with the o cial account of the WeChat public, the system will call patients based on the matching telephone equipment of the electronic questionnaire.Then collect the tracking data to complete all the tracking.

E. Statistical analysis management
A large amount of data generated by the monitoring system is a reference for hospitals to make diagnosis and treatment plans, analyze the work of departments, and support management decisions.It connects to the database, reads the table structure from the database and extracts the related data elds.The cleaned data is displayed in the statistical management module of information management system in the form of graph or table.Doctors or managers can view and export the corresponding reports through the corresponding follow-up time, follow-up department and the required nal statistical report type (such as general practice statistical report and problem summary report), and make corresponding work adjustment.

System design
The medical monitoring system is divided into two parts.The mobile device based on the o cial WeChat account is used for interaction between patients and the system, and the network terminal used by doctors and system administrators to interact with the system.The web terminal in this article refers to the browser page used by doctors and system administrators for system interaction.The server is also connected to the o cial WeChat platform to provide background services for the o cial WeChat account.The data in the database is shared through the server, WeChat public platform and the Internet.The o cial WeChat platform is the back end of the o cial WeChat account.The public platform and WeChat server are designed with C/S architecture, and tasks are shared between the client and the server, reducing the communication volume of the system.At the same time, the client can cache general content, reducing the load on the server.The server mainly provides data management, data exchange, system and data maintenance, and concurrency control services, and the client program is mainly responsible for speci c user activities.The web and server are designed with B/S architecture.The web pulls corresponding data from the server and presents the results in the browser.This mode concentrates the execution core of the system function on the server, which simpli es the development, maintenance and use of the system.The physical architecture diagram of the system is shown in Fig. 4.
Terminals refer to devices such as mobile phones used by patients and PC terminals operated by doctors.
The display layer includes the patient's o cial WeChat public account, the WeChat applet interface for participating in follow-up care, and the web page for doctors to manage follow-up information.The presentation layer contains all the header pages and the jump logic between the pages.This layer is used to display information to users, and at the same time receive operating instructions from users, and then send requests to the server.The business layer includes the controller layer and the model layer.The controller layer implements the business logic of each module.These modules can handle all system functions together by calling each other.The model layer provides data access interfaces and de ned data models for the entire business layer to perform data operations.The database entity layer includes a MySQL database and a Redis cache database con gured on the Tencent Cloud server, which is used to store the entity and relationship tables involved in the system, as well as a large amount of data generated by the system's post-provisioning.

Database design
The patient information table stores the personal information lled in by the patient on the homepage, as well as related information and part of the tracking content of the patient's o cial WeChat public account link, as shown in Table 1.openID is the number used by WeChat to identify patients.Everyone who uses WeChat has openID.

Conclusion
This article examines the HEED routing protocol, introduces the Head of Group Choice method, competition method, grouping method and routing mechanism of the HEED algorithm in detail, then discusses the shortcomings of the algorithm and proposes an improved method for these shortcomings.In OPTIK's clustering algorithm, fuzzy logic control is used to set the optimal clustering radius, which is mainly to improve the convergence of the wireless sensor network, alleviate the "hot spot" problem, and balance the network load; nally combined with LEACH to improve -C and PSO-C The advantages of the algorithm, a non-uniform packet routing protocol based on simulation algorithm is proposed.Based on the needs of the project and the needs of the hospital for patient care, this paper mainly designs and implements a doctor-centered follow-up system that allows patients to make appointments, follow-up after diagnosis, hospital information, aftercare management and other service systems to provide medical services to patients.At the same time, the hospital can also receive follow-up reports from patients and record the patient's condition in time; on the other hand, the hospital can also meet the country's requirements for patient follow-up care.

Declarations
Compliance with Ethical Standards

Con ict of interest
The authors declare that they have no con ict of interests

Ethical approval
This article does not contain any studies with human participants performed by any of the authors.

S 1 = 7 It
(s 11 , s 12 , … , s 1n ) , S 2 = (s 21 , s 22 , … , s 2n )ŝlj = tanh (w f s ij + b f )6 is composed of the text coding layer and the attention mechanism presentation layer.This model uses the bidirectional long and short-term memory network BiLSTM as the text encoder.LSTM is a common variant of the cyclic neural network.It introduces 3 gate structures.The forgetting gate controls the proportion of time historical information before the control, and the output gate controls the proportion of output information.,The speci c form is shown in formulas 8 to 13:

Page 7/ 17 15
According to the calculated weight   , the attention vector of position i in the text  1 can be obtained: 16 In order to obtain the fusion of the text encoding representation and the attention representation, this model splices the text encoding representation and the attention representation at each position of the text to obtain the nal output representation c 1i of the text representation layer: 17 Input the output of the text presentation layer into a layer of BiLSTM again to mix the information to obtain the mixed sequence representation: 18 below is the experimental results of each model.By extracting the relationship, we found that the extracting relationship can be regarded as a multiclassi cation problem in nature.The most important thing for the classi cation problem is the task selection feature and model.In order to better handle more complex sentences and improve the effect of different semantics and expression skills of different sentence lengths, this paper chooses to focus on the classi cation model of the Bi-LSTM top-level attention mechanism for relation extraction.For each M (c 11 , c 12 , … , c 1n ) , S c 2 = BiLST M (c 21 , c 22 , … , c 2n )

Figures
Figures

Figure 2 Entity recognition experiment result graph Figure 3
Figure 2

Figure 4 Physical architecture diagram
Figure 4

Table 1
Doctor pro le is the content of doctors that can be viewed on the WeChat public platform, as shown in Table2Doctor pro le table.The primary key in the doctor information table is DoctorID.DoctorPicture saves the location of the doctor's picture in the cos memory object.upreport is generated every time the patient completes the follow-up.Each tracking report corresponds to an ID tracking order number.FollowupID can be used to nd the patient's follow-up records.followupID is the primary key of the patient follow-up record table.A patient can correspond to multiple follow-up records.The patient follow-up record table is shown in Table3.The doctor can manage the follow-up template.The follow-up template table is shown in Table4.The master saves the template information created by the medical user.followupType represents the followup type, which is used to distinguish different follow-up types.For example: the tracking template table is suitable for clinicians, mainly clinicians who manage tracking templates.QuestionNumbers is the number of questions in the follow-up template questionnaire that the doctor will create.

Table 4
Follow-up template information table