Medical and health field is a hot application field of wireless sensor networks. How to correctly refine and classify telemedicine sensor data is the research focus in related fields. Therefore, a detailed classification mathematical model simulation of telemedicine sensor data based on multi feature fusion is proposed. On the basis of telemedicine sensor data acquisition, it is preprocessed to reduce the computational overhead of detailed classification. The reliability features of the preprocessed telemedicine sensing data are extracted, the extracted features are fused by the principal component analysis method, and the refined classification model of telemedicine sensing data is constructed based on the principle of machine learning. The fused features are input into the model to complete the refined classification of telemedicine sensing data. The experimental results show that the correct refinement classification rate of the proposed method is more than 90%, the refinement classification accuracy is higher than 98.5%, the convergence speed is good, and the refinement classification time is 4 ~ 12 s, which proves that the correct refinement classification rate and accuracy of the proposed method are high, the classification time is short, and has good application performance.
Notes
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
At present, most hospitals begin to adopt information-based management methods [1], and gradually realize the automatic processing of information from outpatient registration to inpatient treatment. With the deepening of information construction, especially the application of wireless sensor network, the amount of telemedicine sensing data will continue to increase, showing the characteristics of big data, These data will be divided into different categories because of different departments and attributes, which requires a lot of costs and resources in the storage process [2]. Facing this situation, how to realize the effective classification of telemedicine sensing data is the research focus.
Jin et al. pointed out that telemedicine provides an attractive perspective for remote monitoring of human health, which provides an opportunity for timely prevention of chronic diseases. A key limitation of promoting telemedicine in clinical application is the lack of noninvasive medical technology and effective monitoring platform, which should be wearable and capable of high-performance remote monitoring of health risks. This method proposes a telemedicine technology based on volatolomics, which continuously and noninvasively evaluates human health by continuously tracking the changes of volatile markers from human breathing or skin. In particular, a flexible electronic product based on nano sensors is specially designed as a powerful platform for implementing the proposed cost-effective health care. Although this method realizes telemedicine health monitoring, it does not process and classify telemedicine sensing data, which needs to be further optimized [3]. Shi et al. shows that precision medicine, as a new medical service model, has developed rapidly in recent years, which summarizes the problems encountered in the development of precision medicine. Combined with the characteristics of telemedicine, it proposes to establish a precision medicine big data service platform based on telemedicine, and studies and analyzes the construction principles, key technologies, technical architecture, network architecture and function realization of the platform Based on clinical data, health data and other kinds of data, the platform can realize applications such as accurate diagnosis, accurate treatment and accurate medication, and has good application prospects. However, this method does not reduce the dimension of multi class data, resulting in low data classification accuracy [4]. Zachrison et al. conducted secondary analysis on the data of the national emergency department inventory survey in 2016 to determine the rural emergency departments that used and did not use telemedicine in 2016. All rural emergency rooms without telemedicine were followed up to ask about the staffing, transfer mode and perceived barriers to the use of telemedicine. The results show that the cost will limit the application scope of telemedicine, but the classification efficiency of this method is low [5].
Advertisement
In order to effectively realize the detailed classification of telemedicine sensing data, a mathematical model simulation of detailed classification of telemedicine sensing data is proposed. The proposed method can refine and classify the telemedicine sensing data accurately, which can be used as a reference for the realization of telemedicine.
2 Telemedicine sensing data preprocessing
2.1 Telemedicine sensing data acquisition
Due to the multi-source correlation in telemedicine sensing data, to collect telemedicine sensing data, it is necessary to create an extraction target containing object and time, and use the target to find the relevant Abstract telemedicine sensing data. The collection process is shown in Fig. 1:
Fig. 1
Telemedicine sensing data acquisition process
×
The specific steps shown in Fig. 1 are as follows:
Step 1: find the acquisition source from the telemedicine sensing data. The definition of telemedicine sensing data collection source is as follows: for all records in telemedicine sensing data, if they have two characteristics of different recording time and different target address, the structure containing time and target address is defined as a telemedicine sensing data collection source;
Step 2: use the telemedicine sensor data collection source to search the relevant log record set in the telemedicine sensor network [6]. Record the data meeting the requirements as:
where α is the fusion degree of telemedicine sensing data, i is a constant, μ is the optimization coefficient, and \(a_{i}\) is the telemedicine sensing data.
Step3: calculate the node configuration function. The calculation formula is as follows:
where \(\gamma^{i}\) is the allocated value of telemedicine sensing data channel.
After obtaining the configuration results of telemedicine sensor data nodes, taking the sparse Bayesian DOA parameters of the node interface as the characteristic quantity, the joint characteristic parameter identification model of telemedicine sensor data is established [7], the polarization vector of interference signal is obtained, and the multi measurement vector fusion identification technology is adopted to obtain the optimal node deployment of telemedicine wireless sensor, as shown in Fig. 2.
Fig. 2
Deployment of telemedicine wireless sensor nodes
×
Advertisement
According to the node distribution of telemedicine wireless sensors, the adaptive polarization cancellation control is carried out [8], and the model parameters of node deployment are obtained as follows:
where \(r_{s}\) is the distribution interval of array element nodes of uniform linear array, w is the irrelevant additive characteristic parameter, \(r_{j}\) is the sampling interval of discrete information between error signals j, and c is the block sparse characteristic parameter.
The method of joint subspace class is used for cluster analysis of telemedicine sensing data [9], and the optimal solution of telemedicine sensing data fusion matrix is obtained, i.e.
$$Q = {\text{sgn}} \varphi c^{2} P$$
(4)
where φ is the clustering coefficient of telemedicine sensing data. According to the above analysis, collect telemedicine sensing data.
2.2 Dimension reduction processing of telemedicine sensing data
Telemedicine sensor data has the characteristics of massive and high dimensionality. The feature extraction of telemedicine sensor data refinement classification algorithm needs to have the function of dimensionality reduction. For telemedicine sensing data, the traditional dimensionality reduction method cannot determine the dimensionality of all kinds of data. The amount of data contained in telemedicine sensing data is high, and its workload is also high. Feature extraction methods with high computational efficiency are needed.
The incremental orthogonal component analysis (IOCA) method is selected to reduce the dimension and extract features of telemedicine sensing data, so as to improve the time complexity of detailed classification of telemedicine sensing data. IOCA method does not need to set a fixed target dimension, and the target dimension can be adjusted according to the changes of input data in the learning process [10]. Using this method to preprocess the massive telemedicine sensing data will form a better orthogonal component, avoid the redundancy of the data and compress the dimension well.
IOCA method can use the pre fetched telemedicine sensing data to obtain the orthogonal component space \(\left\{ {d_{1} ,d_{2} , \cdots ,d_{n} } \right\}\) that can automatically determine the dimension, so as to realize the rapid dimensionality reduction of telemedicine sensing data. The dimensionality reduction of telemedicine sensing data needs to be realized through the following two steps:
(1)
Let the existing new data be represented by \(\varphi_{k + 1}\) and the learned orthogonal component space be represented by \(D = \left\{ {d_{1} ,d_{2} , \cdots ,d_{k} } \right\}\). calculate the new potential orthogonal component \(d_{k + 1}\) possibly generated by \(\varphi_{k + 1}\) and the linear independence between them;
(2)
Set the adaptive threshold, and use the set adaptive threshold to judge whether \(d_{k + 1}\) can be used as a new orthogonal component added to D.
The specific calculation process of extracting data features by IOCA method is as follows:
(1)
Initialize the orthogonal component space \(D = \emptyset\) with the initial dimension \(k = \dim \left( D \right) = 0\);
(2)
\(x_{j}\) represents the new input data, and \(j > k\) shall be met;
(3)
Use \(r_{i,k + 1}\) to represent the eigenvector and calculate \(r_{i,k + 1} = x_{j}^{T} d_{i}\);
\(g\) represents the original data dimension. When \(\frac{{r_{k + 1,k + 1} }}{{\left\| {x_{j} } \right\|}} \ge \frac{\dim \left( D \right)}{g}\), it means that \(d_{k + 1}\) belongs to the new orthogonal component. At this time, \(d_{k + 1}\) is added to \(D\);
(8)
Make \(k = k + 1\) and repeat the above process until all data preprocessing is completed.
3 Build a detailed classification model of telemedicine sensing data
Using the corresponding orthogonal space to extract the features of telemedicine sensing data, obtain multiple groups of feature vectors with low dimensions [11], and build a detailed classification model of telemedicine sensing data based on machine learning to realize the detailed classification of telemedicine sensing data.
3.1 Extracting features of telemedicine sensing data
Because the reliability of the extracted telemedicine sensing data is inconsistent, which is mainly determined by the reputation value, the clustering analysis algorithm [12, 13] is used to cluster the telemedicine sensing data to comprehensively analyze the extracted information. Then the weighted data is clustered. Because the sensor node [14] collects data through the monitoring object, the result of data clustering mainly includes a large data set, while other data not included in the large data set by the clustering algorithm are unreliable data.
According to the above definition, cluster the weighted data received by the nodes in the telemedicine sensing data features [15], and identify the telemedicine sensing data features with the reliability of telemedicine sensing data. Assuming that a set of observations in the telemedicine sensing data is \(\left( {x_{1} ,x_{2} , \ldots ,x_{n} } \right)\), and each observation is an actual vector, calculate whether the markers of the observations are the same by using the following formula:
$$J = \frac{z}{{q\left| {x/o} \right|}}$$
(5)
Among them, J is the minimum set parameter of the observation value, q represents the function of the distance from each point in the set to the cluster center, \(x/o\) represents the clustering parameter of telemedicine sensing data, and z is the vector parameter of the observation value.
The larger the total number of clusters, the higher the accuracy of feature recognition of telemedicine sensing data. Given cluster \(c^{\prime} \in C_{i}\) [16], the entropy of \(c^{\prime}\) is:
where y represents the ratio of the labeled target in \(c^{\prime}\), \(H\left( {c^{\prime}} \right)\) represents the sum of all entropy, and m represents the average entropy of the clustering result.
According to the above process, the telemedicine sensing data clustering is completed, and the data reliability is defined here to identify the characteristics of telemedicine sensing data. The calculation formula is as follows:
where \(G\left| H \right|\) represents the data reliability in the telemedicine sensing data cluster, κ represents the data reliability judgment parameter, \(V_{i}\) represents the unreliable data in the data set, and f is the calculation constant.
In this way, the characteristic reliability of telemedicine sensing data is calculated, the reliability of telemedicine sensing data is measured by using the above formula [17, 18], and the characteristics of telemedicine sensing data are identified according to the calculation results.
3.2 Telemedicine sensing data feature fusion
The detailed classification model construction method of telemedicine sensing data based on machine learning [19, 20] adopts the principal component analysis method to fuse the above extracted data features.
Feature p is extracted from n telemedicine sensing data to obtain the original data feature matrix, which is expressed as follows:
The original variable \(X_{1} ,X_{2} ,X_{3} , \cdots ,X_{p}\) can be used to linearly represent the comprehensive variables obtained after principal component analysis:
$$\left\{ \begin{gathered} Y_{1} = U_{1}^{T} X \hfill \\ Y_{2} = U_{2}^{T} X \hfill \\ \vdots \hfill \\ Y_{p} = U_{p}^{T} X \hfill \\ \end{gathered} \right.$$
(9)
Let \(b_{ij}\) represent the covariance between the i-th feature and the j-th feature [21, 22], and its calculation formula is as follows:
The eigenvalues are sorted from large to small to obtain each principal component. The eigenvalue of telemedicine sensing data is the variance corresponding to each principal component, and \(\lambda_{1} ,\lambda_{2} , \cdots ,\lambda_{p}\) is used to describe the non-zero eigenvalue corresponding to eigenvector \(U_{1} ,U_{2} , \cdots ,U_{p}\).
Let \(\lambda_{k} /\sum\limits_{i = 1}^{p} {\lambda_{i} }\) represent the contribution rate corresponding to the first principal component \(Y_{k}\), describe the share of the information extracted by the k principal component in the total information, and obtain the cumulative contribution rate \(\sum\limits_{i = 1}^{m} {\lambda_{i} } (\sum\limits_{i = 1}^{p} {\lambda_{i} } )^{ - 1}\) on this basis. Determine the number of eigenvectors and principal components of the transformation, and obtain the transformation matrix. The transformation matrix is calculated through principal component analysis and sample original features to complete feature fusion. The specific process is shown in Fig. 3.
Fig. 3
Feature fusion process
×
The center vector \(U_{k} = [u_{k1} ,u_{k2} , \cdots ,u_{kn} ]\) is obtained through average calculation, the distance \(l_{k}^{2}\) between the data element \(X = [x_{1} ,x_{2} , \cdots ,x_{n} ]\) to be identified and the center vector is calculated on the basis of machine learning principle, the category of data element \(X = [x_{1} ,x_{2} , \cdots ,x_{n} ]\) is determined, and the detailed classification model of telemedicine sensing data is constructed:
According to the above process, the detailed classification of telemedicine sensing data is realized.
4 Experiment and analysis
In order to verify the overall effectiveness of the mathematical model for the refinement and classification of telemedicine sensing data, the telemedicine sensing data refinement and classification model (method 1), the precision medical big data service platform based on telemedicine (method 2) and the obstacle data analysis method for telemedicine implementation in rural emergency departments (method 3) were used for testing. The experimental environment for this test is: the CPU is Pentium (R) dual core E5200, the main frequency is 2.5 GHz, the memory is 2G, and the experimental platform is Matlab 2019b.
4.1 Experimental setup
Before the experiment, simulate the workload of the telemedicine sensor data storage system. In order to be close to the real use situation as much as possible, use the partial access log of the hospital system to simulate the workload, set the number of concurrent clients to 20, simulate the workload generator under different circumstances, and set the request rate between 1000 ~ 3000req / s. After obtaining the access log, remove the noise data in the log to avoid some useless data occupying the contents of the log. Select 8% of the most visited pages in the remaining records, randomly select the processed 5% and convert it into read–write operations, and draw the file size distribution in the test file for subsequent analysis. The file size distribution is shown in Fig. 4.
Fig. 4
Schematic diagram of experimental file size distribution
×
In the face of the mathematical model of detailed classification of telemedicine sensing data, a large data cluster of detailed classification was built with four servers in the experiment to support the operation of each method and analyze the compression resistance of the method when the file size distribution is known.
4.2 Analysis of experimental results
4.2.1 Correctly refine classification test
Let ACC represent the proportion of correctly refined classification in all samples, and its calculation formula is as follows:
$$ACC = \frac{TP + TN}{{TP + TN + FP + FN}}$$
(13)
Among them, FN represents false negative; TP stands for true positive; FP represents false positive; TN stands for true negative.
Test the correct refinement classification rate of different methods, and the test results are shown in Fig. 5.
Fig. 5
Test results of correctly refining classification rate
×
By analyzing the data in Fig. 5, it can be seen that when using method 2 and method 3 to refine and classify the telemedicine sensing data, the correct refinement classification rate is between 80 and 89%, and when using method 1 to refine and classify the telemedicine sensing data, the correct refinement classification rate obtained in multiple iterations is more than 90%, It shows that method 1 can accurately complete the detailed classification of telemedicine sensing data, because method 1 integrates the characteristics of telemedicine sensing data, and classifies the telemedicine sensing data according to the fused characteristics, which improves the correct detailed classification rate of the method.
4.2.2 Detailed classification accuracy test
Considering that there are a large number of interference data in wireless sensor networks, the anti-interference performance of refinement classification algorithm is very important. The refinement classification accuracy of telemedicine sensor data with different sizes of white noise is counted by using three algorithms. The statistical results are shown in Fig. 6.
Fig. 6
Comparison of refined classification accuracy under interference
×
As can be seen from the experimental results in Fig. 6, when the noise interference of 5 ~ 40 dB is added, when the method 2 is used for thinning classification, the thinning classification accuracy is between 95.5 ~ 98.7; When method 3 is used for detailed classification, the precision of detailed classification is between 95.8 ~ 98.5; When method 1 is used for detailed classification, the precision of detailed classification is between 99 ~ 99.8; It shows that the refinement classification accuracy of method 1 under different white noise is significantly higher than that of the other comparison methods, which verifies that method 1 has high anti-interference performance. This algorithm has high precision, fast speed and high anti-interference performance. It can effectively resist many noise interferences in wireless sensor networks, and can be applied to the practical refinement classification of telemedicine sensor data.
4.2.3 Convergence rate test
RDV (relative difference value) is selected as an important evaluation index to measure the convergence speed of the model. The lower the RDV value of the model, the better the convergence speed of the model. The calculation formula of RDV value is as follows:
where \(P^{A} \left( {w^{ * } } \right)\) and \(P^{A} \left( w \right)\) respectively represent the current value of the optimal solution and the operation process of obtaining the optimal solution.
The RDV values of telemedicine sensing data refined and classified by different models are shown in Fig. 7.
Fig. 7
RDV value comparison
×
As can be seen from the experimental results in Fig. 7, under different operation times, the highest RDV value of method 2 is 0.05, the highest RDV value of method 3 is 0.04, and the highest RDV value of method 1 is 0.035, which is the lowest of the three methods. It is proved that the convergence speed of this method is the best, indicating that method 1 can still maintain a high convergence speed while maintaining high refinement classification accuracy, Thirdly, it is verified that the refinement classification of telemedicine sensor data has high refinement classification effectiveness and convergence speed.
4.2.4 Classification time test
Methods 1, 2 and 3 are used to refine and classify telemedicine sensing data. The time spent in refining and classifying different methods is compared. The test results are shown in Fig. 8.
Fig. 8
Detailed classification time test results
×
According to the analysis of the data in Fig. 8, with the increase of the amount of telemedicine sensing data, the refinement classification time used in method 1, method 2 and method 3 continues to increase. However, when the amount of telemedicine sensing data is the same, the refinement classification time used in method 2 is 7 ~ 15 s, the refinement classification time used in method 3 is 8 ~ 16S, and the refinement classification time used in method 1 is 4 ~ 12 s, The refinement classification time used in method 1 is much lower than that used in methods 2 and 3. Because method 1 uses the principal component analysis method to integrate the extracted telemedicine sensing data features, and carries out the dimensionality reduction processing on the features. According to the dimensionality reduction processing features, it constructs the telemedicine sensing data refinement classification model, and uses the model to complete the refinement classification of telemedicine sensing data, which shortens the time of refinement classification and improves the refinement classification efficiency of method 1.
5 Conclusion
More and more genetic diseases have an impact on people's lives. With the application of wireless sensor networks, the amount of relevant telemedicine sensing data has increased sharply. Therefore, it is of great significance to classify telemedicine sensing data. The rapid development of computer technology and medical imaging technology promotes the application of computer-aided diagnosis and treatment in clinical experiment and medical research. In order to improve the correct classification rate and low classification efficiency of telemedicine sensing data, the mathematical model simulation of detailed classification of telemedicine sensing data is proposed, the features of telemedicine sensing data are extracted and fused, and the detailed classification model of telemedicine sensing data is established to realize the detailed classification of telemedicine sensing data. It is verified that the proposed method is correct, the thinning classification rate is more than 90%, the thinning classification accuracy is higher than 98.5%, the convergence speed is good, the classification time is short, and has good application performance. It can provide technical support for telemedicine sensing data sharing, facilitate remote monitoring of human health, accurately control all kinds of monitoring data, and promote the smooth operation of the medical data monitoring platform.
Acknowledgements
This work was supported by the Scientific research project of Hubei Vocational and technical education society (No.ZJGB2021061). Authors would like to acknowledge contribution to this research from the Rector of the Silesian University of Technology, Gliwice, Poland under pro-quality grant no. 09/010/RGJ22/0068.
Declarations
Ethics
The authors have no relevant financial or non-financial interests to disclose. Haiying Chen provided the algorithm and experimental results, wrote the manuscript, Marcin Woźniak revised the paper, supervised and analyzed the experiment. We also declare that data availability and ethics approval is not applicable in this paper.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Die Fachzeitschrift ATZelektronik bietet für Entwickler und Entscheider in der Automobil- und Zulieferindustrie qualitativ hochwertige und fundierte Informationen aus dem gesamten Spektrum der Pkw- und Nutzfahrzeug-Elektronik.
Lassen Sie sich jetzt unverbindlich 2 kostenlose Ausgabe zusenden.
ATZlectronics worldwide is up-to-speed on new trends and developments in automotive electronics on a scientific level with a high depth of information.
Order your 30-days-trial for free and without any commitment.
