Electrical impedance analysis children allergy prediction system

Abstract

The invention relates to the technical field of bioelectrical impedance measurement, in particular to an electrical impedance analysis children allergy prediction system. The system comprises an electrical impedance data acquisition module, an electrical impedance signal processing module, an allergy correlation factor analysis module, an allergy mode depth recognition module, an allergy prediction dynamic adjustment module, an electrical impedance deep learning analysis module, an electrical impedance spectrum comprehensive analysis module and a child allergy risk assessment module. According to the invention, through combination of a multi-spectrum electrical impedance measurement technology and a current source and voltage measurement instrument, measurement of children bioelectrical impedance is realized, Fourier transform is used for signal denoising and smoothing processing, measurement noise and environmental interference are eliminated, and electrical impedance data are analyzed through principal component analysis and a clustering algorithm. The convolutional neural network is applied to feature extraction of the electrophysiological indexes, the accuracy is improved, and the allergy prediction accuracy is improved through time sequence deep learning analysis of the long-short-term memory network on the electrical impedance data and application of wavelet transform on multi-spectrum analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of bioelectrical impedance measurement, in particular to an electrical impedance analysis system for predicting children's allergies. Background Art

[0002] Bioelectrical Impedance Analysis (BIA) is a non-invasive, rapid, and relatively simple method for assessing body composition and other relevant health parameters. This technique sends a weak electrical current through the body and measures the resistance (resistance) and reaction (reactance) of the current as it passes through tissue. Because different tissue types (such as muscle, fat, and bone) offer varying resistance to the current, analyzing this data can assess body water content, fat content, muscle mass, and other parameters. This technology has widespread applications in nutritional assessment, sports science, and medical health monitoring.

[0003] The electrical impedance analysis pediatric allergy prediction system uses bioelectrical impedance measurement technology to predict whether a child is prone to allergies. This system is based on a core assumption: the body's response to allergens is reflected in changes in electrical impedance. Therefore, by measuring and analyzing a child's bioelectrical impedance, allergic tendencies can be detected early, allowing intervention before severe allergic reactions occur. This prediction system aims to identify high-risk groups early, allowing them to receive earlier medical intervention and lifestyle adjustments to reduce the risk of allergies.

[0004] Traditional pediatric allergy prediction systems lack accuracy, sensitivity, and data processing capabilities. Traditional systems lack the ability to continuously measure electrical impedance, resulting in incomplete data and an inability to effectively capture subtle changes in children's bioelectrical impedance. Inaccurate signal processing is susceptible to measurement noise and environmental interference, reducing data quality. Lack of support for advanced algorithms and deep learning technologies often results in inaccurate and in-depth predictions, impacting allergy prediction effectiveness, limiting the reliability of risk assessments, and lacking sufficient data support for allergy management and intervention measures. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose an electrical impedance analysis children's allergy prediction system.

[0006] To achieve the above objectives, the present invention adopts the following technical solutions: the electrical impedance analysis child allergy prediction system includes an electrical impedance data acquisition module, an electrical impedance signal processing module, an allergy-related factor analysis module, an allergy pattern deep recognition module, an allergy prediction dynamic adjustment module, an electrical impedance deep learning analysis module, an electrical impedance spectrum comprehensive analysis module, and a child allergy risk assessment module; The electrical impedance data acquisition module is based on multi-spectral electrical impedance measurement technology, using a current source and a voltage meter to continuously measure the child's bioelectrical impedance, while simultaneously recording the resistance and reactance values ​​at differentiated frequencies to generate a preliminary electrical impedance data set; The electrical impedance signal processing module uses Fourier transform to perform signal denoising and smoothing based on the preliminary electrical impedance data set, eliminates measurement noise and environmental interference, and generates optimized electrical impedance data; The allergy-related factor analysis module uses principal component analysis and clustering algorithms based on the optimized electrical impedance data to analyze patterns and trends in the electrical impedance data, extract electrophysiological parameters associated with allergic reactions, and generate allergy-related electrophysiological indicators; The allergy pattern deep recognition module uses a convolutional neural network for deep learning based on allergy-related electrophysiological indicators, extracts features and performs pattern recognition on the electrophysiological indicators, identifies target patterns associated with allergic reactions, and generates a deep recognition allergy pattern; The allergy prediction dynamic adjustment module uses an adaptive algorithm based on deep recognition of allergy patterns to dynamically adjust the threshold of the allergy prediction model and generate a dynamically adjusted prediction threshold; The electrical impedance deep learning analysis module uses a long short-term memory network to perform time series deep learning analysis on the electrical impedance data based on a dynamically adjusted prediction threshold, identifies long-term and short-term patterns in the data, and generates electrical impedance deep analysis results; The electrical impedance spectrum comprehensive analysis module uses wavelet transform to perform multi-spectrum analysis on the electrical impedance data based on the results of the electrical impedance depth analysis, refines the frequency characteristics of resistance and reactance, improves the precision of allergy prediction, and generates detailed data of the electrical impedance spectrum; The children's allergy risk assessment module is based on the detailed data of the electrical impedance spectrum and uses a support vector machine to establish an allergy risk assessment model, classify and analyze multidimensional data, and generate children's allergy risk assessment results.

[0007] As a further solution of the present invention, the preliminary impedance data set includes resistance values, reactance values ​​and time stamps of differentiated frequencies, the optimized impedance data includes denoised frequency responses and smoothed impedance curves, the allergy-related electrophysiological indicators include key frequency points, key impedance features and clustering groups, the deep identification of allergy patterns includes the impedance patterns of target allergic reactions and features identified by deep learning, the dynamically adjusted prediction threshold includes adaptively adjusted sensitivity parameters and currently set threshold standards, the impedance depth analysis results include long-term and short-term impedance trend analysis, time series data features, the impedance spectrum detail data include spectral features after wavelet transform and refined impedance responses, and the children's allergy risk assessment results include risk level classification and allergy tendency score.

[0008] As a further solution of the present invention, the electrical impedance data acquisition module includes an electrical impedance measurement submodule, a data sampling submodule, and an electrical impedance recording submodule; The electrical impedance measurement submodule is based on multi-spectrum electrical impedance measurement technology. It uses a current source to apply multi-frequency AC current to the child, and a voltage meter captures the corresponding voltage signal. The current source controls and sets the differential frequency, ranging from low frequency to high frequency, covering the electrical impedance characteristics of multiple tissues in the body. The voltage meter includes an AD converter to convert the analog voltage signal into a digital signal. The digital processing unit analyzes the voltage response and captures changes in resistance and reactance values. By adjusting the current frequency, the electrical impedance changes at the differential frequency are captured to generate frequency-adjusted electrical impedance data. The data sampling submodule performs data sampling based on frequency-adjusted electrical impedance data, applies a fast Fourier transform algorithm, sets the sampling rate to 300 times per second, captures the details and changes of bioelectrical impedance, and sets the sampling window size to 5 milliseconds to 1.5 seconds to match the electrical impedance changes of differentiated frequencies. During the data sampling process, the FFT algorithm converts the continuous time signal into a frequency domain signal, extracts the key frequency components, and generates sampled electrical impedance data; The electrical impedance recording submodule records the resistance and reactance values ​​at differentiated frequencies based on the sampled electrical impedance data, applies database storage technology, sets the data storage format to a structured CSV file, classifies and labels the sampled data, applies the ZIP compression algorithm, optimizes storage space and management efficiency, and generates a preliminary electrical impedance data set.

[0009] As a further solution of the present invention, the electrical impedance signal processing module includes an electrical impedance denoising submodule, a data smoothing submodule, and an electrical impedance filtering submodule; The electrical impedance denoising submodule uses fast Fourier transform to perform frequency domain analysis based on the preliminary electrical impedance data set. It converts the data set using the NumPy library's fft function, identifies frequency components, sets a frequency threshold of 5 Hz, filters frequency components below this threshold, and generates denoised electrical impedance data. The data smoothing submodule applies the sliding average method to smooth the data based on the denoised electrical impedance data. It uses the rolling and mean functions of the pandas library, sets the window size to 10, calculates the average value of each window, smoothes the short-term fluctuations in the data, and generates smoothed electrical impedance data. The electrical impedance filtering submodule adopts low-pass filtering based on the smoothed electrical impedance data. Through the signal module of the SciPy library, the cutoff frequency is set to 2 Hz, and the lfilter function is used to filter the data to filter out signals above the cutoff frequency, and extract and generate optimized electrical impedance data.

[0010] As a further solution of the present invention, the allergy-related factor analysis module includes an electrophysiological parameter extraction submodule, an allergy cluster analysis submodule, and an electrical impedance time warping submodule; The electrophysiological parameter extraction submodule uses the Fourier transform function (fft) in the numpy and scipy libraries to extract frequency features based on the optimized electrical impedance data. It then uses the butter function in the scipy.signal library to create a Butterworth low-pass filter with parameters set to a cutoff frequency of 100 Hz and a filter order of 5 to filter the signal. The data is then Z-normalized using the normalization function in numpy to generate an electrophysiological feature dataset. The allergy cluster analysis submodule uses the PCA function in the scikit-learn library to perform principal component analysis based on the electrophysiological feature dataset, reducing the dimensionality to retain the principal components with a cumulative contribution rate of 95%. Then, the KMeans function is applied for clustering. The number of clusters is determined as the optimal number of clusters using the knee_locator function of scikit-learn. Clusters are formed by calculating and minimizing the Euclidean distance from sample points to multiple cluster centers to generate allergy pattern clustering results. The electrical impedance time warping submodule is based on the allergy pattern clustering results and uses the ARMA model in the statsmodels library to perform time series analysis. The p and q parameters of the model are determined by ACF and PACF graphs. p is the number of autoregressive terms and q is the number of moving average terms. The parameter values ​​are determined according to the autocorrelation characteristics of the data. The regularity of electrical impedance data changes over time is analyzed to generate allergy-related electrophysiological indicators.

[0011] As a further solution of the present invention, the allergy pattern deep recognition module includes an electrical impedance feature extraction submodule, a deep pattern learning submodule, and an allergy sequence mining submodule; The electrical impedance feature extraction submodule is based on allergy-related electrophysiological indicators. It uses the TensorFlow and Keras libraries to build a convolutional neural network, uses the Conv2D layer to extract features from the input electrophysiological data, sets 32 filters, each with a kernel size of 3x3, and uses the ReLU activation function to perform feature mapping on the electrophysiological indicators, extract key features, and generate feature extraction results. The deep pattern learning submodule continues to deepen the learning process based on the feature extraction results, adding Conv2D layers and MaxPooling2D layers, setting 64 filters, maintaining the kernel size at 3x3, and using a 2x2 pooling window in the MaxPooling2D layer to extract and identify patterns in the electrophysiological data. A Dropout layer is added with a ratio of 0.25 to avoid overfitting and generate pattern learning results. The allergy sequence mining submodule is based on the pattern learning results, applies a long short-term memory network, sets the number of neurons in the network to 50, uses the Adam optimizer, adjusts the learning rate to 0.01, selects categorical_crossentropy as the loss function, and trains the network through the fit method. During the training process, the parameters of the neural network are cyclically adjusted to capture the time series characteristics in the electrophysiological data, mine sequence patterns associated with allergic reactions, and generate deep recognition allergy patterns.

[0012] As a further solution of the present invention, the allergy prediction dynamic adjustment module includes an allergy reaction adaptive learning submodule, an electrical impedance feedback control submodule, and an allergy prediction optimization submodule; The allergic reaction adaptive learning submodule builds a recurrent neural network based on the TensorFlow library. Specifically, it uses an LSTM layer to process time series data. The parameters include 128 units and a tanh activation function. During the training process, the EarlyStopping callback function is used to monitor the loss value on the validation set. The parameters are set to patience = 10 and min_delta = 0.01 to generate adaptive learning analysis results. The electrical impedance feedback control submodule uses a PID controller for feedback control based on the results of adaptive learning analysis. Specifically, the electrical impedance data is monitored in real time and compared with the preset allergy mode threshold. The PID parameters are dynamically adjusted, including the proportional coefficient P set to 0.1, the integral coefficient I set to 0.01, and the differential coefficient D set to 0.001, to generate feedback control adjustment results. The allergy prediction optimization submodule optimizes the allergy prediction model based on the feedback control adjustment results, using grid search and cross-validation. Specifically, grid search is used to test differentiated LSTM network parameters, including the number of units and activation function, and the optimal parameter settings are determined through 5-fold cross-validation. The allergy prediction threshold is dynamically adjusted to generate a dynamically adjusted prediction threshold.

[0013] As a further solution of the present invention, the electrical impedance deep learning analysis module includes an electrical impedance time series analysis submodule, an electrical impedance feature deepening submodule, and an electrical impedance dynamic modeling submodule; The electrical impedance time series analysis submodule uses TensorFlow and Keras to build a long short-term memory network based on a dynamically adjusted prediction threshold, performs time series analysis on the electrical impedance data, sets 100 neurons, extracts time series features, analyzes the long-term and short-term trends of the electrical impedance data, adjusts the time step and batch size, optimizes the network structure to match the characteristics of the electrical impedance data, and generates preliminary results of the time series analysis; The electrical impedance feature deepening submodule uses a multi-layer LSTM network for deep feature analysis based on the preliminary results of time series analysis. It increases the number of network layers and neurons, deepens the feature extraction of electrical impedance data through a layer-by-layer learning strategy, adjusts the parameters of the LSTM layer, including increasing the number of neurons to 150 and matching the dropout value, mines the electrical impedance data pattern, and generates feature deepening results. The electrical impedance dynamic modeling submodule constructs a dynamic long-short-term memory network model based on the feature deepening results and adjusts the model parameters, including setting the learning rate to 0.001, the number of iterations to 50, using the Adam optimizer for network optimization, setting the loss function to the mean square error, and using the fit method to train the model. The network is optimized to capture the dynamic changes in the electrical impedance data and generate the electrical impedance depth analysis results.

[0014] As a further solution of the present invention, the electrical impedance spectrum comprehensive analysis module includes a multi-spectrum wavelet analysis submodule, an electrical impedance signal decomposition submodule, and an electrical impedance frequency optimization submodule; The multi-spectral wavelet analysis submodule uses PyWavelets to perform wavelet transform based on the results of the electrical impedance depth analysis, selects the db4 wavelet of the Daubechies series, sets the number of decomposition layers to 5, captures the characteristics of the electrical impedance signal at the differentiated frequency level, and generates multi-spectral wavelet analysis results; The electrical impedance signal decomposition submodule uses numpy to perform threshold processing and reconstruction of wavelet coefficients based on the multi-spectral wavelet analysis results, removes noise and enhances the key features of the signal, refines the frequency characteristics of resistance and reactance, and generates electrical impedance signal refinement results; The electrical impedance frequency optimization submodule uses scipy.signal to estimate the signal's spectral density based on the electrical impedance signal refinement result, identifies and optimizes the key frequency components of the electrical impedance signal, and generates electrical impedance spectrum detail data.

[0015] As a further embodiment of the present invention, the child allergy risk assessment module includes a risk data classification submodule, a model cross-validation submodule, and a risk feature selection submodule; The risk data classification submodule uses the support vector machine algorithm to perform classification analysis on the electrical impedance spectrum detail data. The Scikit-learn library is used to create an SVM, the kernel function is set to the radial basis function, the parameters C are adjusted to 1.0, and gamma is adjusted to 0.1. The electrical impedance data is classified to distinguish the differentiated allergy risk levels and generate classification analysis results. The model cross-validation submodule applies the cross-validation method to test the stability and generalization ability of the SVM model based on the classification analysis results. It uses the cross_val_score function of Scikit-learn and sets a 5-fold cross-validation to generate model validation results by verifying the performance of the differentiated data subsets. The risk feature selection submodule uses feature selection technology to optimize the SVM model based on the model validation results, and uses the SelectKBest function of Scikit-learn to select the top 10 features ranked by allergy risk correlation to generate the children's allergy risk assessment results.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, through the combination of multi-spectrum electrical impedance measurement technology and current source and voltage meter, continuous and accurate measurement of children's bioelectrical impedance is achieved, thereby enhancing the quality and integrity of the data. Fourier transform is used for signal denoising and smoothing, effectively eliminating measurement noise and environmental interference, and improving data accuracy. Principal component analysis and clustering algorithms are used to deeply analyze electrical impedance data, and convolutional neural networks are used for feature extraction of electrophysiological indicators to improve the accuracy of identifying patterns related to allergic reactions. The adaptive algorithm significantly improves the sensitivity and adaptability of the prediction in terms of dynamically adjusting the threshold of the allergy prediction model. The time series deep learning analysis of electrical impedance data by long short-term memory networks and the application of wavelet transform in multi-spectral analysis greatly improve the refinement and accuracy of allergy prediction. The allergy risk assessment model established by support vector machines provides a powerful tool for accurate assessment of children's allergy risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a system flow chart of the present invention; Figure 2 Schematic diagram of the system framework of the present invention; Figure 3 This is a flow chart of the electrical impedance data acquisition module of the present invention; Figure 4 This is a flow chart of the electrical impedance signal processing module of the present invention; Figure 5 This is a flow chart of the allergy-related factor analysis module of the present invention; Figure 6 This is a flow chart of the allergy pattern deep recognition module of the present invention; Figure 7 This is a flow chart of the allergy prediction dynamic adjustment module of the present invention; Figure 8 This is a flow chart of the electrical impedance deep learning analysis module of the present invention; Figure 9 This is a flow chart of the electrical impedance spectrum comprehensive analysis module of the present invention; Figure 10 This is a flow chart of the children's allergy risk assessment module of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0020] Example 1: Please refer to Figures 1 to 2 The electrical impedance analysis children's allergy prediction system includes an electrical impedance data acquisition module, an electrical impedance signal processing module, an allergy-related factor analysis module, an allergy pattern deep recognition module, an allergy prediction dynamic adjustment module, an electrical impedance deep learning analysis module, an electrical impedance spectrum comprehensive analysis module, and a children's allergy risk assessment module; The electrical impedance data acquisition module is based on multi-spectral electrical impedance measurement technology. It uses a current source and a voltage meter to continuously measure the child's bioelectrical impedance, while simultaneously recording the resistance and reactance values ​​at differentiated frequencies to generate a preliminary electrical impedance data set. The electrical impedance signal processing module uses Fourier transform to perform signal denoising and smoothing based on the preliminary electrical impedance data set, eliminating measurement noise and environmental interference, and generating optimized electrical impedance data; The allergy-related factor analysis module uses principal component analysis and clustering algorithms based on the optimized electrical impedance data to analyze patterns and trends in the electrical impedance data, extract electrophysiological parameters associated with allergic reactions, and generate allergy-related electrophysiological indicators; The allergy pattern deep recognition module uses convolutional neural networks for deep learning based on allergy-related electrophysiological indicators to extract features and perform pattern recognition on electrophysiological indicators, identify target patterns associated with allergic reactions, and generate deep recognition allergy patterns. The allergy prediction dynamic adjustment module uses an adaptive algorithm based on deep recognition of allergy patterns to dynamically adjust the threshold of the allergy prediction model and generate a dynamically adjusted prediction threshold; The electrical impedance deep learning analysis module uses a long short-term memory network to perform time series deep learning analysis on electrical impedance data based on a dynamically adjusted prediction threshold, identifying long-term and short-term patterns in the data and generating electrical impedance deep analysis results. The impedance spectrum comprehensive analysis module uses wavelet transform to perform multi-spectral analysis on the impedance data based on the results of the impedance depth analysis, refines the frequency characteristics of resistance and reactance, improves the precision of allergy prediction, and generates detailed impedance spectrum data. The children's allergy risk assessment module is based on the detailed data of the electrical impedance spectrum and uses a support vector machine to establish an allergy risk assessment model, classify and analyze multidimensional data, and generate children's allergy risk assessment results.

[0021] The preliminary impedance data set includes resistance values, reactance values ​​and time stamps of differentiated frequencies. The optimized impedance data includes denoised frequency responses and smoothed impedance curves. Allergy-related electrophysiological indicators include key frequency points, key impedance features and clustering groups. Deep identification of allergy patterns includes impedance patterns of target allergic reactions and features identified by deep learning. The dynamically adjusted prediction threshold includes adaptively adjusted sensitivity parameters and currently set threshold standards. The impedance deep analysis results include long-term and short-term impedance trend analysis and time series data features. The impedance spectrum detail data includes spectrum features after wavelet transform and refined impedance response. The results of children's allergy risk assessment include risk level classification and allergy tendency score.

[0022] In the electrical impedance data acquisition module, the system continuously measures children's bioelectrical impedance using multi-spectral electrical impedance measurement technology, combined with a precise current source and voltage meter. The module transmits small currents at set frequency intervals and uses a voltage meter to record the biological tissue's response to the currents at different frequencies. These responses are recorded as resistance and reactance values, forming a preliminary electrical impedance dataset. This dataset is stored in a multidimensional array format consisting of timestamp, frequency, resistance, and reactance values, providing detailed bioelectrical information for subsequent processing.

[0023] In the electrical impedance signal processing module, the system uses a Fourier transform algorithm to denoise and smooth the raw electrical impedance data. This process involves converting the time-domain electrical impedance signal into a frequency-domain signal, using frequency filtering techniques to remove noise and environmental interference at non-target frequencies. By adjusting filter parameters such as cutoff frequency and bandwidth, signal clarity and accuracy are optimized. The processed electrical impedance data is smoother and clearer, providing a more stable foundation for accurate analysis of allergy-related factors.

[0024] The Allergy-Related Factor Analysis module uses principal component analysis and clustering algorithms to deeply explore hidden patterns and trends within the optimized electrical impedance data. Principal component analysis reduces data dimensionality and highlights the most representative electrophysiological parameters, while clustering algorithms further analyze the correlations between these parameters to identify electrophysiological features closely associated with allergic reactions. This series of operations generates medically significant electrophysiological indicators associated with allergies, providing key clues for further analysis.

[0025] The allergy pattern deep recognition module applies deep learning using convolutional neural networks to extract features and perform pattern recognition on allergy-related electrophysiological indicators. Through its multi-layered structure, convolutional neural networks effectively extract features at different levels and identify electrophysiological patterns closely associated with allergic reactions. During the deep learning process, network parameters such as the convolution kernel size, activation function, and optimizer are adjusted to achieve accurate interpretation of electrophysiological indicators. These operations enable the module to generate specific and reliable allergy pattern recognition results.

[0026] The allergy prediction dynamic adjustment module uses an adaptive algorithm to dynamically adjust the thresholds of the allergy prediction model based on identified allergy patterns. This module automatically adjusts the thresholds based on current data patterns and changes, ensuring prediction accuracy and sensitivity. The adaptive algorithm monitors prediction errors in real time and adjusts prediction parameters, such as weights and biases, to accommodate new data features. This allows the module to dynamically adjust prediction thresholds to reflect the latest data patterns in real time, improving prediction accuracy.

[0027] In the electrical impedance deep learning analysis module, a long short-term memory (LSTM) network is used to perform time series deep learning analysis on electrical impedance data. LSTM is particularly well-suited to processing time series data, capable of identifying and memorizing both long-term and short-term patterns within the data. By adjusting LSTM parameters, such as the number of hidden layer units and the time window size, the module can effectively capture the temporal dependencies and complex patterns in the electrical impedance data. The analysis results generate an electrical impedance deep analysis report, providing crucial time series information for assessing children's allergy risk.

[0028] In the comprehensive electrical impedance spectrum analysis module, wavelet transform technology is used to perform multi-spectral analysis on the electrical impedance data. This step refines the frequency characteristics of resistance and reactance, improving the sophistication and accuracy of allergy predictions. By selecting the appropriate wavelet basis and decomposition level, the module can conduct in-depth analysis of different frequency components. This complex spectrum analysis process generates detailed electrical impedance spectrum data, providing more accurate foundational data for children's allergy risk assessment.

[0029] The Child Allergy Risk Assessment module uses a support vector machine (SVM) to establish an allergy risk assessment model and classify and analyze detailed electrical impedance spectrum data. By adjusting the SVM's kernel function and regularization parameters, the module effectively classifies and identifies distinct allergy risk patterns in a high-dimensional space. This assessment model not only distinguishes between allergic and non-allergic states but also assesses the severity of allergies. These analyses ultimately generate a child allergy risk assessment report, providing physicians with reliable diagnostic evidence and helping to promptly and accurately identify and address childhood allergies.

[0030] See also Figure 3 ,The electrical impedance data acquisition module includes an electrical impedance measurement submodule, a data sampling submodule, and an electrical impedance recording submodule; The electrical impedance measurement submodule is based on multi-spectrum electrical impedance measurement technology. It uses a current source to apply multi-frequency AC current to the child, and a voltage meter captures the resulting voltage signal. The current source controls the setting of differential frequencies, ranging from low to high frequencies, covering the electrical impedance characteristics of multiple tissues in the body. The voltage meter includes an A / D converter that converts the analog voltage signal into a digital signal. The digital processing unit analyzes the voltage response and captures changes in resistance and reactance values. By adjusting the current frequency, the changes in electrical impedance at differential frequencies are captured, generating frequency-adjusted electrical impedance data. The data sampling submodule adjusts the electrical impedance data based on frequency and performs data sampling. It applies the fast Fourier transform algorithm and sets the sampling rate to 300 times per second to capture the details and changes of the bioelectrical impedance. The sampling window size is set to 5 milliseconds to 1.5 seconds to match the electrical impedance changes of the differentiated frequency. During the data sampling process, the FFT algorithm converts the continuous time signal into a frequency domain signal, extracts the key frequency components, and generates the sampled electrical impedance data. The electrical impedance recording submodule records the resistance and reactance values ​​at differentiated frequencies based on the sampled electrical impedance data. It applies database storage technology, sets the data storage format to a structured CSV file, classifies and labels the sampled data, applies the ZIP compression algorithm, optimizes storage space and management efficiency, and generates a preliminary electrical impedance dataset.

[0031] In the electrical impedance measurement submodule, the system utilizes multi-spectrum electrical impedance measurement technology. A current source applies multi-frequency AC current to the child, while a voltage meter captures the resulting voltage signal. The core of this process lies in the control of the current source, which is set to varying frequencies, from low to high, to cover the electrical impedance characteristics of multiple tissues. The current source precisely adjusts the current frequency to ensure frequency accuracy as the current passes through the tissue, thereby capturing the tissue's response to currents of varying frequencies. The analog-to-digital converter (AD converter) embedded in the voltage meter converts the captured analog voltage signal into a digital signal for analysis by the digital processing unit. The digital processing unit analyzes the voltage response at each frequency point, accurately capturing changes in resistance and reactance. The current frequency is adjusted dynamically, spanning a spectrum from low to high, ensuring that the captured impedance data reflects the impedance characteristics of the tissue at different frequencies. The output of this submodule is frequency-adjusted impedance data, which serves as the basis for the subsequent data sampling submodule.

[0032] In the data sampling submodule, frequency-adjusted electrical impedance data is processed using the Fast Fourier Transform (FFT) algorithm. This submodule sets the sampling rate to 300 times per second, effectively capturing the details and variations of bioelectrical impedance. The sampling window size can be set from 5 milliseconds to 1.5 seconds to match the impedance variation characteristics at different frequencies. The FFT algorithm is used to convert continuous time signals into frequency domain signals, with the focus of this conversion process on extracting key frequency components. The FFT algorithm decomposes the time signal and extracts the main frequency components for subsequent analysis and processing. The key to this step lies in the FFT algorithm parameter settings, such as the window size and sampling rate. Accurately setting these parameters is crucial for capturing the true characteristics of the electrical impedance signal. The result of this processing is sampled electrical impedance data, which is more suitable for subsequent recording and analysis.

[0033] The impedance recording submodule performs data recording and storage based on the sampled impedance data. This submodule utilizes database storage technology and selects the structured CSV file format as the data storage format. The CSV format is suitable for storing large amounts of impedance data due to its structured and highly compatible nature, while also facilitating subsequent data processing and analysis. The sampled data is classified and labeled before storage, a step that facilitates subsequent data retrieval and analysis. The ZIP compression algorithm is applied to optimize storage space and management efficiency, particularly when processing large amounts of impedance data, as the compression algorithm can significantly reduce the required storage space. This series of operations ultimately generates a preliminary impedance dataset, which will serve as input data for subsequent modules of the impedance analysis children's allergy prediction system.

[0034] Imagine a scenario where the system captures a series of electrical impedance data, including resistance and reactance values ​​at different frequencies. For example, at frequencies of 10Hz, 50Hz, 100Hz, and 500Hz, the system captures resistance values ​​of 30Ω, 28Ω, 26Ω, and 24Ω, and reactance values ​​of 15Ω, 13Ω, 11Ω, and 9Ω, respectively. This data is processed by the data sampling submodule, with a sampling rate of 300 times per second and a sampling window size of 1 second. An FFT algorithm is applied to this data to extract the primary frequency components and generate sampled electrical impedance data. Finally, the electrical impedance recording submodule saves this data as a CSV file and compresses it using a ZIP compression algorithm to form a preliminary electrical impedance dataset. This dataset will provide critical foundational data for allergy prediction.

[0035] See also Figure 4 ,The electrical impedance signal processing module includes an electrical impedance denoising submodule, a data smoothing submodule, and an electrical impedance filtering submodule; The electrical impedance denoising submodule uses fast Fourier transform to perform frequency domain analysis based on the preliminary electrical impedance dataset. The NumPy library's fft function is used to transform the dataset, identify frequency components, set the frequency threshold to 5 Hz, filter frequency components below this threshold, and generate denoised electrical impedance data. The data smoothing submodule applies the sliding average method to the denoised electrical impedance data. It uses the rolling and mean functions of the pandas library, sets the window size to 10, calculates the average value of each window, smoothes out short-term fluctuations in the data, and generates smoothed electrical impedance data. The electrical impedance filter submodule uses low-pass filtering based on the smoothed electrical impedance data. Through the signal module of the SciPy library, the cutoff frequency is set to 2 Hz, and the lfilter function is used to filter the data to filter out signals above the cutoff frequency and extract the optimized electrical impedance data.

[0036] In the electrical impedance denoising submodule, the system first performs frequency domain analysis on the preliminary electrical impedance dataset, primarily through the fast Fourier transform (FFT). The FFT is performed using the NumPy library's fft function, which converts the time-domain electrical impedance signal into a frequency-domain signal, revealing the characteristics of different frequency components. After performing the FFT, the system filters the frequency components by setting a frequency threshold (5 Hz in this example). The goal is to remove frequency components below this threshold, which are considered noise or irrelevant signals. This processing method effectively reduces the noise level in the data and improves signal quality. This process generates denoised electrical impedance data, which is more accurate and reliable for subsequent analysis.

[0037] The data smoothing submodule then processes the denoised electrical impedance data to further improve data quality. This submodule uses a sliding average method, implemented using the rolling and mean functions in the pandas library. Specifically, a fixed-size window (in this example, 10) is set and moved across the dataset, calculating the average value of the data within each window. This sliding average method effectively smooths short-term fluctuations in the data and reduces the impact of random fluctuations on the analysis results. This step generates smoothed electrical impedance data that are more continuous and consistent, providing a solid foundation for subsequent, more complex analyses.

[0038] The electrical impedance filtering submodule operates on the smoothed electrical impedance data to further optimize data quality. This submodule employs low-pass filtering, primarily implemented through the signal module in the SciPy library. The key to filtering is setting an appropriate cutoff frequency, 2 Hz in this case. The data is filtered using the lfilter function in the signal module to remove signals above the cutoff frequency that fall outside the target range for analysis. Low-pass filtering effectively extracts low-frequency components from the signal, which contain the primary information about bioelectrical impedance. This series of processing generates optimized electrical impedance data that are more focused and accurate in the frequency domain, providing crucial data support for identifying and analyzing specific electrophysiological patterns.

[0039] Imagine the following scenario: The system receives a set of raw electrical impedance data containing resistance and reactance values ​​at multiple frequency points. For example, the data set contains resistance values ​​of 25Ω, 23Ω, and 21Ω at 10Hz, 50Hz, and 100Hz, respectively. The impedance denoising submodule analyzes this data using FFT, removing frequency components below 5Hz to produce denoised data. Subsequently, the data smoothing submodule applies a sliding average method to smooth the denoised data and reduce data volatility. Finally, the impedance filtering submodule performs low-pass filtering to retain only frequency components below 2Hz, generating optimized impedance data. This data is used for further analysis, such as identifying and classifying allergy patterns, providing more accurate predictions for childhood allergies.

[0040] See also Figure 5 ,The allergy-related factor analysis module includes the electrophysiological ,parameter extraction submodule, the allergy cluster analysis submodule, and the ,electrical impedance time warping submodule; The electrophysiological parameter extraction submodule uses the Fourier transform function (fft) in the numpy and scipy libraries to extract frequency features based on the optimized electrical impedance data. It then uses the butter function in the scipy.signal library to create a Butterworth low-pass filter with a cutoff frequency of 100 Hz and a filter order of 5 to filter the signal. The data is then Z-normalized using the normalization function in numpy to generate an electrophysiological feature dataset. The allergy cluster analysis submodule uses the PCA function in the scikit-learn library to perform principal component analysis on the electrophysiological feature dataset, reducing the dimensionality to retain the principal components with a cumulative contribution rate of 95%. The KMeans function is then used for clustering. The optimal number of clusters is determined using the knee_locator function in scikit-learn. Clusters are formed by calculating and minimizing the Euclidean distance from sample points to multiple cluster centers, generating allergy pattern clustering results. The electrical impedance time warping submodule is based on the allergy pattern clustering results and uses the ARMA model in the statsmodels library to perform time series analysis. The p and q parameters of the model are determined by ACF and PACF plots. p is the number of autoregressive terms, and q is the number of moving average terms. The parameter values ​​are determined according to the autocorrelation characteristics of the data. The regularity of electrical impedance data changes over time is analyzed to generate allergy-related electrophysiological indicators.

[0041] In the electrophysiological parameter extraction submodule, the system extracts and analyzes frequency features based on the optimized electrical impedance data. This process primarily utilizes the Fourier transform (FFT) function in the NumPy and SciPy libraries, which converts time series data into frequency domain data. Specifically, the system first applies the FFT function to the optimized electrical impedance data to analyze the presence and intensity of different frequency components. After completing the frequency domain analysis, to further purify the signal, the system creates a Butterworth low-pass filter using the butter function in the scipy.signal library. Key parameters of this filter include a cutoff frequency of 100 Hz and a filter order of 5. These settings aim to preserve signal components below 100 Hz while effectively suppressing higher-frequency noise. Following filtering, the system applies Z-score normalization to the data using the normalize function in NumPy. This step eliminates absolute differences between measurements, making the data more suitable for subsequent analysis and comparison. These processing steps generate an electrophysiological feature dataset containing important frequency features and electrophysiological information, providing a foundation for identifying specific physiological states and patterns.

[0042] The allergy clustering analysis submodule performs further data analysis and pattern recognition based on the electrophysiological feature dataset. First, principal component analysis (PCA) is performed using the scikit-learn library's PCA function. The goal is to reduce the data dimensionality while preserving as much important information as possible from the original data. During PCA, the system targets the principal components with a cumulative contribution rate of 95%. This setting aims to reduce data complexity while preserving the most effective information. After dimensionality reduction, the system applies the KMeans function to cluster the data. The number of clusters is determined using the scikit-learn knee_locator tool to ensure the optimal number of clusters. The clustering process essentially calculates the Euclidean distance from a sample point to multiple cluster centers and minimizes this distance to form distinct clusters. This process generates clustering results for allergy patterns, which reveal similarities and differences between different allergy patterns and are important for understanding the mechanisms and characteristics of allergic reactions.

[0043] The impedance time warping submodule performs time series analysis based on the allergy pattern clustering results. This submodule uses the ARMA model from the statsmodels library. The ARMA model is a commonly used time series analysis tool suitable for analyzing and predicting patterns and trends in time series data. When using the ARMA model, the system first determines the model parameters: the number of autoregressive terms (AR) p and the number of moving average terms (MA) q. These parameters are determined based on an analysis of the data's autocorrelation function (ACF) and partial autocorrelation function (PACF). By analyzing the ACF and PACF plots, the system effectively identifies time-dependent characteristics in the data, providing a basis for setting the ARMA model parameters. After the parameters are determined, the system applies the ARMA model to analyze the temporal patterns of the impedance data, ultimately generating allergy-related electrophysiological indices. These indices reveal the temporal patterns of the impedance data and are valuable for understanding and predicting the development of allergic reactions.

[0044] Imagine the following scenario: The system processes a set of electrical impedance data, consisting of resistance and reactance values ​​at different time points. For example, a dataset might include resistance values ​​measured every second for several consecutive minutes. The electrophysiological parameter extraction submodule analyzes this data using FFT to identify frequency components, filters it using a Butterworth low-pass filter, and then performs Z-score normalization to generate an electrophysiological feature dataset. The allergy clustering analysis submodule then performs PCA dimensionality reduction and K-means clustering to reveal the characteristics of different allergy patterns. Finally, the electrical impedance time warping submodule uses the ARMA model to analyze the temporal changes in the electrical impedance data and generate allergy-related electrophysiological indicators. These indicators are used for further analysis and prediction, providing important evidence for the diagnosis and treatment of childhood allergies.

[0045] See also Figure 6 ,The allergy pattern deep recognition module includes the ,electrical impedance feature extraction submodule, the deep pattern learning submodule, and the ,allergy sequence mining submodule; The electrical impedance feature extraction submodule uses the TensorFlow and Keras libraries to build a convolutional neural network based on allergy-related electrophysiological indicators. It uses the Conv2D layer to extract features from the input electrophysiological data. 32 filters are set, each with a kernel size of 3x3. The ReLU activation function is used to perform feature mapping on the electrophysiological indicators, extract key features, and generate feature extraction results. Based on the feature extraction results, the deep pattern learning submodule continues to deepen the learning process by adding Conv2D and MaxPooling2D layers, setting 64 filters and maintaining the kernel size at 3x3. The MaxPooling2D layer uses a 2x2 pooling window to extract and identify patterns in the electrophysiological data. A Dropout layer is added with a ratio of 0.25 to avoid overfitting and generate pattern learning results. The allergy sequence mining submodule is based on the pattern learning results, applies the long short-term memory network, sets the number of neurons in the network to 50, uses the Adam optimizer, adjusts the learning rate to 0.01, selects the categorical_crossentropy loss function, and trains the network through the fit method. During the training process, the parameters of the neural network are cyclically adjusted to capture the time series characteristics in the electrophysiological data, mine the sequence patterns associated with allergic reactions, and generate deep recognition allergy patterns.

[0046] In the electrical impedance feature extraction submodule, the system uses the TensorFlow and Keras libraries to build a convolutional neural network (CNN) to perform deep feature extraction on the input electrophysiological data. This process begins with one or more convolutional layers (Conv2D), each with 32 filters and a 3x3 kernel size. These filters scan and extract local features from the input data. The 3x3 kernel size is used to balance detailed feature extraction with computational efficiency. Following the convolutional layer, the Rectified Linear Unit (ReLU) activation function is used because it increases nonlinearity while avoiding the vanishing gradient problem, effectively improving the network's learning ability. Through this series of operations, the system extracts key features from the electrophysiological indicators and maps these features into a new feature space, generating feature extraction results. This result contains the core information of the electrophysiological data, laying the foundation for subsequent in-depth analysis.

[0047] In the deep pattern learning submodule, the learning process is further deepened based on the feature extraction results. During this stage, the system adds additional convolutional layers (Conv2D) and pooling layers (MaxPooling2D) to enhance the model's learning and generalization capabilities. The newly added convolutional layer is configured with 64 filters, maintaining a kernel size of 3x3, to extract more complex features. The subsequent MaxPooling2D layer uses a 2x2 pooling window to reduce the spatial dimension of the feature map while retaining the most important feature information. Furthermore, to prevent model overfitting, the system introduces a Dropout layer with a ratio of 0.25. The Dropout layer randomly drops a portion of neurons during training, forcing the network to learn more robust feature representations. Through these steps, the system generates pattern learning results that incorporate deep patterns and associations in the electrophysiological data, providing important information for identifying specific allergic reactions.

[0048] In the allergy sequence mining submodule, based on the pattern learning results, a long short-term memory (LSTM) network is applied for deep learning. LSTM is particularly well-suited for processing time series data because it can capture long-term dependencies in the data. In this submodule, the system sets the number of neurons in the LSTM network to 50, selects Adam as the optimizer, and adjusts the learning rate to 0.01. The loss function uses categorical_crossentropy, which is suitable for multi-classification problems. The network is trained using the fit method. This training process involves repeatedly adjusting the network parameters to better capture the time series characteristics of the electrophysiological data. After training, the system is able to mine sequential patterns associated with allergic reactions and generate results that deeply identify allergy patterns. This result provides important time series information for predicting childhood allergies, enhancing the accuracy and reliability of allergy predictions.

[0049] Imagine the system receives a set of electrophysiological parameter data, including impedance and reactance values ​​at different time points. For example, a dataset might contain resistance values ​​measured every second for several minutes. The impedance feature extraction submodule extracts key features using a convolutional neural network. The deep pattern learning submodule further deepens the learning of these features and uses a dropout layer to prevent overfitting. Finally, the allergy sequence mining submodule uses a long short-term memory network to deeply analyze the time series characteristics of this data and discover patterns associated with allergic reactions. Through this process, the system is able to generate a deep analysis model for predicting childhood allergies, providing accurate prediction results.

[0050] See also Figure 7 ,The allergy prediction dynamic adjustment module includes the allergy reaction adaptive learning submodule, the ,electrical impedance feedback control submodule, and the allergy prediction optimization submodule; The allergic reaction adaptive learning submodule builds a recurrent neural network based on the TensorFlow library. Specifically, it uses an LSTM layer to process time series data. Parameters include 128 units and a tanh activation function. During training, the EarlyStopping callback function is used to monitor the loss value on the validation set. The parameters are set to patience = 10 and min_delta = 0.01 to generate adaptive learning analysis results. The electrical impedance feedback control submodule uses a PID controller for feedback control based on the results of adaptive learning analysis. Specifically, the electrical impedance data is monitored in real time and compared with the preset allergy mode threshold. The PID parameters are dynamically adjusted, including the proportional coefficient P set to 0.1, the integral coefficient I set to 0.01, and the differential coefficient D set to 0.001, to generate the feedback control adjustment results. The allergy prediction optimization submodule optimizes the allergy prediction model based on the feedback control adjustment results, using grid search and cross-validation. Specifically, grid search is used to test differentiated LSTM network parameters, including the number of units and activation function, and the optimal parameter settings are determined through 5-fold cross-validation. The allergy prediction threshold is dynamically adjusted to generate a dynamically adjusted prediction threshold.

[0051] In the allergic reaction adaptive learning submodule, the system builds a recurrent neural network (RNN) based on the TensorFlow library, specifically employing a long short-term memory (LSTM) layer to process time series data. LSTM is an improvement on RNN, better able to handle long-term dependencies in time series data. The LSTM layer parameters are set to 128 units and a tanh activation function. This setting of 128 units aims to provide sufficient model complexity to capture complex patterns in electrophysiological data. The tanh activation function adds nonlinear features to the network, helping the model learn more complex data structures. During training, the system uses an EarlyStopping callback function to monitor the loss on the validation set to avoid overfitting. The EarlyStopping parameters are set to patience=10 and min_delta=0.01, meaning that training is terminated if the loss does not decrease by more than 0.01 within 10 epochs. This setting improves learning efficiency while maintaining the model's generalization ability. The final product of adaptive learning is the analysis result, which contains the learned time series characteristics of allergic reactions and provides a basis for subsequent prediction and analysis.

[0052] In the electrical impedance feedback control submodule, a PID controller is used for feedback control based on the results of adaptive learning analysis. The PID controller is a classic control system that consists of three parts: proportional (P), integral (I), and differential (D), and can adjust the output based on the system's real-time feedback. In this system, the PID controller is used to monitor the electrical impedance data in real time and compare it with the preset allergy mode threshold. Dynamic adjustment of the PID parameters is key, with the proportional coefficient P set to 0.1, the integral coefficient I set to 0.01, and the differential coefficient D set to 0.001. This parameter setting is designed to achieve a precise response to changes in electrical impedance while avoiding instability caused by over-adjustment. The result of the feedback control adjustment is that the system can more sensitively and accurately monitor and respond to changes in electrical impedance, thereby providing higher precision and efficiency in monitoring allergic reactions.

[0053] In the allergy prediction optimization submodule, grid search and cross-validation are used to optimize the allergy prediction model based on the feedback control adjustment results. Grid search is a method that systematically iterates over multiple parameter combinations to optimize model performance. During this process, the system tests different LSTM network parameters, including different numbers of neurons and different types of activation functions. This operation aims to find the network structure and parameter settings that best suit the current dataset. 5-fold cross-validation is a method for evaluating the generalization ability of a model. It divides the dataset into five parts, rotating one of them as a test set and the others as a training set, to evaluate the model's performance on different data subsets. This method can determine the optimal parameter settings and dynamically adjust the allergy prediction threshold. The effect of dynamically adjusting the prediction threshold is to improve the accuracy and adaptability of the allergy prediction model, enabling the model to better adapt to data changes under different individuals and conditions, thereby improving the accuracy of predicting allergic reactions.

[0054] Imagine a system processing a set of electrophysiological data, such as electrical impedance values ​​monitored over several consecutive hours. The allergy response adaptive learning submodule uses an LSTM layer to analyze this time series data and optimizes the learning process through early stopping. The impedance feedback control submodule applies a PID controller to adjust the model output in real time to accommodate changes in the impedance data. Finally, the allergy prediction optimization submodule uses grid search and cross-validation to find the optimal LSTM parameters and optimize the allergy prediction model. This system generates an allergy prediction model that accurately adapts to changes in impedance, providing important technical support for the early identification and intervention of childhood allergies.

[0055] See also Figure 8 ,The electrical impedance deep learning analysis module includes the ,electrical impedance time series analysis sub-module, the electrical impedance feature deepening ,sub-module, and the electrical impedance dynamic modeling sub-module; The electrical impedance time series analysis submodule uses TensorFlow and Keras to build a long short-term memory network based on a dynamically adjusted prediction threshold. It performs time series analysis on the electrical impedance data, sets 100 neurons, extracts time series features, analyzes the long-term and short-term trends of the electrical impedance data, adjusts the time step and batch size, optimizes the network structure to match the characteristics of the electrical impedance data, and generates preliminary results of the time series analysis. Based on the preliminary results of time series analysis, the electrical impedance feature deepening submodule applies a multi-layer LSTM network for deep feature analysis. This increases the number of network layers and neurons, deepens the feature extraction of the electrical impedance data through a layer-by-layer learning strategy, adjusts the parameters of the LSTM layer, including increasing the number of neurons to 150 and matching the dropout value, and mines the electrical impedance data patterns to generate feature deepening results. The impedance dynamic modeling submodule builds a dynamic long-short-term memory network model based on the feature deepening results and adjusts the model parameters, including setting the learning rate to 0.001, the number of iterations to 50, using the Adam optimizer for network optimization, setting the loss function to the mean square error, and using the fit method to train the model. The network is optimized to capture the dynamic changes in the impedance data and generate the impedance depth analysis results.

[0056] In the electrical impedance time series analysis submodule, the system performs time series analysis on the electrical impedance data using a long short-term memory (LSTM) network built using the TensorFlow and Keras libraries. The LSTM network is designed to capture and analyze the time series characteristics of the electrical impedance data, particularly long- and short-term trends. To achieve this, the system sets the number of neurons in the LSTM layer to 100. This number ensures sufficient network complexity to learn and identify complex patterns in the electrical impedance data. Simultaneously, the system adjusts the time step and batch size to optimize the network structure to match the characteristics of the electrical impedance data. Adjusting the time step affects the time window within which the network learns data, while the batch size affects the amount of data processed during each training step. These adjustments enable the network to more effectively learn the time series characteristics of the electrical impedance data and generate preliminary results for the time series analysis. These preliminary results capture the time series characteristics of the electrical impedance data and provide a foundation for further analysis.

[0057] In the electrical impedance feature deepening submodule, based on the preliminary results of time series analysis, a multi-layer LSTM network is applied for deeper feature analysis. During this process, the system increases the number of network layers and neurons, specifically increasing the number of LSTM neurons in each layer to 150. This design aims to deepen the feature extraction of the electrical impedance data through a layer-by-layer learning strategy, enabling the network to learn more complex and deep data patterns. To prevent overfitting, the system uses appropriate dropout values ​​in the LSTM layers, which helps improve the model's generalization ability. Through these steps, the system can more deeply explore patterns in the electrical impedance data and generate feature deepening results. This result provides in-depth feature information of the electrical impedance data, laying the foundation for building more accurate prediction models.

[0058] In the electrical impedance dynamic modeling submodule, a dynamic long short-term memory network model is constructed based on the results of feature deepening. In this submodule, the system adjusts the key parameters of the model, including setting the learning rate to 0.001 and the number of iterations to 50 times. The setting of the learning rate affects the speed of parameter updates during model training, while the number of iterations determines the total number of rounds of model training. The system uses the Adam optimizer for network optimization because the Adam optimizer combines the multiple advantages of gradient descent and is suitable for processing large data sets. The loss function is chosen to be the mean square error, which is suitable for regression problems and helps to quantify the difference between model predictions and actual data. The model is trained using the fit method to optimize the network to capture the dynamic changes in the electrical impedance data. After training, the generated electrical impedance deep analysis results can reveal the complex dynamic patterns in the electrical impedance data, providing important support for the accurate prediction of allergic reactions.

[0059] Imagine a system processing a time series of electrical impedance data, for example, impedance values ​​recorded every minute for several hours. The Impedance Time Series Analysis submodule analyzes this data using an LSTM network to capture both long-term and short-term trends. The Impedance Feature Deepening submodule further increases the depth and number of neurons in the LSTM layers to deeply learn the complex patterns in the data. Finally, the Impedance Dynamic Modeling submodule uses an optimized LSTM network model to comprehensively analyze the impedance data and generate in-depth analytical results. This result will provide a precise analytical foundation for predicting childhood allergies, improving the accuracy and reliability of the prediction model.

[0060] See also Figure 9 ,The electrical impedance spectrum comprehensive analysis module includes a multi-spectrum wavelet analysis sub-module, an electrical impedance signal decomposition sub-module, and an electrical impedance frequency optimization sub-module; The multi-spectral wavelet analysis submodule uses PyWavelets for wavelet transform based on the results of the electrical impedance depth analysis. It selects the db4 wavelet of the Daubechies series and sets the number of decomposition layers to 5 to capture the characteristics of the electrical impedance signal at differentiated frequency levels and generate multi-spectral wavelet analysis results. The electrical impedance signal decomposition submodule uses numpy to perform threshold processing and reconstruction of wavelet coefficients based on the results of multi-spectral wavelet analysis, removes noise and enhances the key features of the signal, refines the frequency characteristics of resistance and reactance, and generates electrical impedance signal refinement results; The electrical impedance frequency optimization submodule uses scipy.signal to estimate the signal's spectral density based on the electrical impedance signal refinement results, identify and optimize the key frequency components of the electrical impedance signal, and generate electrical impedance spectrum detail data.

[0061] In the multi-spectral wavelet analysis submodule, the system utilizes the PyWavelets library to perform wavelet transforms, specifically analyzing the spectral features of the electrical impedance deep analysis results. Wavelet transforms are an effective signal processing tool, particularly suitable for revealing the local characteristics of non-stationary signals. The system selects the db4 wavelet from the Daubechies family for analysis due to its excellent smoothness and compactness, making it suitable for analyzing bioelectrical signals such as electrical impedance signals. During the wavelet transform process, the system sets the decomposition level to five, a depth designed to capture the details and characteristics of the electrical impedance signal at different frequency levels. By performing multi-level decomposition of the electrical impedance signal, the system can reveal subtle changes and characteristics of the signal at each spectral level, generating multi-spectral wavelet analysis results. This result provides detailed spectral information for subsequent signal decomposition and optimization, enabling more in-depth and precise analysis of the electrical impedance signal.

[0062] In the electrical impedance signal decomposition submodule, the numpy library is used to perform threshold processing and reconstruction on the wavelet coefficients based on the results of multi-spectral wavelet analysis. The purpose of threshold processing is to remove noise components from the signal and enhance its key features. During this process, the system sets appropriate thresholds to filter the wavelet coefficients, retaining only the important coefficients and removing the small coefficients that represent noise. The subsequent reconstruction process reconstructs the signal based on these filtered wavelet coefficients to retain more useful information. This processing not only reduces the impact of noise but also improves the clarity of the frequency characteristics of resistance and reactance. Through these refinement operations, the system generates an electrical impedance signal refinement result, which shows more refined and clear frequency characteristics in the electrical impedance signal, providing a foundation for further frequency optimization.

[0063] In the electrical impedance frequency optimization submodule, the scipy.signal library is used to estimate the signal's spectral density based on the impedance signal refinement results. The purpose of spectral density estimation is to identify and optimize the key frequency components in the impedance signal. During this process, the system analyzes the signal's spectral distribution and identifies the frequency components that are most critical to the impedance change. Through this analysis, the system can more accurately understand the frequency characteristics of the impedance signal and optimize the signal processing and analysis strategies. Ultimately, the system generates detailed impedance spectrum data that reveals the most important frequency components and characteristics of the impedance signal, providing critical frequency information for subsequent allergy prediction and analysis.

[0064] Assume that the system processes a set of electrical impedance deep analysis results, including resistance and reactance values ​​at different time points. The multi-spectral wavelet analysis submodule uses the DB4 wavelet function to perform hierarchical analysis on this data, revealing the characteristics of the electrical impedance signal at different spectral levels. The electrical impedance signal decomposition submodule further refines these spectral features, using thresholding and reconstruction to clarify the frequency characteristics of resistance and reactance. Finally, the electrical impedance frequency optimization submodule uses these refined results to perform spectral density estimation, identify key frequency components, and generate detailed electrical impedance spectrum data. This data provides the electrical impedance analysis-based pediatric allergy prediction system with more in-depth and precise spectral information, improving the accuracy and efficiency of allergy prediction.

[0065] See also Figure 10 ,The pediatric allergy risk assessment module includes a risk data classification submodule, a model cross-validation submodule, and a risk feature selection submodule; The risk data classification submodule uses the support vector machine algorithm to perform classification analysis on the electrical impedance spectrum detail data. The Scikit-learn library is used to create an SVM, setting the kernel function to the radial basis function, adjusting the parameters C to 1.0 and gamma to 0.1, classifying the electrical impedance data, distinguishing the differentiated allergy risk levels, and generating classification analysis results. The model cross-validation submodule applies the cross-validation method to test the stability and generalization ability of the SVM model based on the classification analysis results. It uses the cross_val_score function of Scikit-learn and sets a 5-fold cross-validation to generate model validation results by verifying the performance of differentiated data subsets. The risk feature selection submodule uses feature selection technology to optimize the SVM model based on the model validation results. The SelectKBest function of Scikit-learn is used to select the top 10 features with the highest correlation with allergy risk and generate the children's allergy risk assessment results.

[0066] In the risk data classification submodule, the system uses a support vector machine (SVM) algorithm to classify and analyze the electrical impedance spectrum detail data. This process is implemented using the Scikit-learn library. First, an SVM classifier is created. The SVM kernel is configured with a radial basis function (RBF) because it excels at handling nonlinear relationships and is well-suited to the complex characteristics of electrical impedance data. Parameters C and gamma are set to 1.0 and 0.1, respectively. The C parameter controls the model's tolerance for classification errors. Larger C values ​​make the model less tolerant of classification errors and tend to overfit; smaller C values ​​increase the model's tolerance and tend to underfit. The gamma parameter influences the distribution of the data after mapping it to the new feature space. Larger gamma values ​​result in more complex decision boundaries, also leading to overfitting. By adjusting these parameters, the SVM model can better adapt to the characteristics of the electrical impedance data and effectively classify differentiated allergy risk levels. The classification analysis results clearly distinguish different allergy risk levels, providing an important basis for subsequent risk assessment and management.

[0067] In the model cross-validation submodule, the stability and generalization ability of the SVM model are tested based on the results of risk data classification. Cross-validation is a method for evaluating model performance, especially for evaluating the generalization ability of a model on a limited data set. The system uses Scikit-learn's cross_val_score function to perform 5-fold cross-validation. This means that the original data is divided into five parts, one of which is used as the test set in turn, and the rest as the training set. By training and testing on these five different data subsets, the system can evaluate the performance of the SVM model on different data samples and ensure the stability and generalization ability of the model. The results of model cross-validation are a series of performance indicators, such as accuracy and recall, which provide important information for further model tuning and verification.

[0068] In the risk feature selection submodule, the SVM model is optimized based on the model validation results, and the model effect is improved by adopting feature selection technology. Feature selection is a commonly used technique in machine learning, which aims to select the most useful features from a large number of features to improve the performance and interpretability of the model. The system uses Scikit-learn's SelectKBest function to implement feature selection. This function scores the features based on the correlation between the features and the target variable and selects the top 10 features with the highest scores. This selection is based on the assumption that the features with the highest correlation with the target variable also contribute the most to the model's prediction results. In this way, the system can extract the key features that are most relevant to children's allergy risk and generate a children's allergy risk assessment result. This assessment result not only improves the accuracy of allergy risk prediction, but also makes the prediction results easier for doctors and researchers to understand and apply.

[0069] Suppose the system processes a set of detailed electrical impedance spectrum data, which contains resistance and reactance values ​​at multiple frequency points. The risk data classification submodule uses SVM to classify this data and distinguish different allergy risk levels. The model cross-validation submodule evaluates the stability and generalization ability of the SVM model through 5-fold cross-validation. Finally, the risk feature selection submodule optimizes the model based on the cross-validation results, selects the features most relevant to allergy risk, and generates the final child allergy risk assessment results. This result provides important decision-making support for the prevention and intervention of childhood allergies.

[0070] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. Electrical impedance analysis children's allergy prediction system, characterized by: The system comprises: The electrical impedance data acquisition module is based on multi-spectral electrical impedance measurement technology. It uses a current source and a voltage meter to continuously measure the child's bioelectrical impedance, while simultaneously recording the resistance and reactance values ​​at differentiated frequencies to generate a preliminary electrical impedance data set. The electrical impedance signal processing module uses Fourier transform to perform signal denoising and smoothing based on the preliminary electrical impedance data set, eliminating measurement noise and environmental interference, and generating optimized electrical impedance data; The allergy-related factor analysis module uses principal component analysis and clustering algorithms based on the optimized electrical impedance data to analyze patterns and trends in the electrical impedance data, extract electrophysiological parameters associated with allergic reactions, and generate allergy-related electrophysiological indicators; The allergy pattern deep recognition module uses convolutional neural networks for deep learning based on allergy-related electrophysiological indicators to extract features and perform pattern recognition on electrophysiological indicators, identify target patterns associated with allergic reactions, and generate deep recognition allergy patterns. The allergy prediction dynamic adjustment module is based on deep recognition of allergy patterns and adopts an adaptive algorithm to dynamically adjust the threshold of the allergy prediction model and generate a dynamically adjusted prediction threshold.

2. The electrical impedance analysis children's allergy prediction system according to claim 1, characterized in that: The system further comprises: The electrical impedance deep learning analysis module uses a long short-term memory network to perform time series deep learning analysis on electrical impedance data based on a dynamically adjusted prediction threshold, identifying long-term and short-term patterns in the data and generating electrical impedance deep analysis results. The impedance spectrum comprehensive analysis module uses wavelet transform to perform multi-spectral analysis on the impedance data based on the results of the impedance depth analysis, refines the frequency characteristics of resistance and reactance, improves the precision of allergy prediction, and generates detailed impedance spectrum data. The children's allergy risk assessment module uses support vector machines to establish an allergy risk assessment model based on detailed data from the electrical impedance spectrum, classifies and analyzes multidimensional data, and generates children's allergy risk assessment results; The preliminary impedance data set includes resistance values, reactance values ​​and time stamps of differentiated frequencies, the optimized impedance data includes denoised frequency responses and smoothed impedance curves, the allergy-related electrophysiological indicators include key frequency points, key impedance features and clustering groups, the deep recognition of allergy patterns includes impedance patterns of target allergic reactions and features identified by deep learning, the dynamically adjusted prediction threshold includes adaptively adjusted sensitivity parameters and currently set threshold standards, the impedance depth analysis results include long-term and short-term impedance trend analysis, time series data features, the impedance spectrum detail data includes spectrum features after wavelet transformation and refined impedance responses, and the children's allergy risk assessment results include risk level classification and allergy tendency score.

3. The electrical impedance analysis children's allergy prediction system according to claim 2, characterized in that: The electrical impedance data acquisition module includes an electrical impedance measurement submodule, a data sampling submodule, and an electrical impedance recording submodule; The electrical impedance measurement submodule is based on multi-spectrum electrical impedance measurement technology. It uses a current source to apply multi-frequency alternating current to the child, and a voltage meter captures the corresponding voltage signal. The current source controls and sets the differential frequency, ranging from low frequency to high frequency, covering the electrical impedance characteristics of multiple tissues in the body. The voltage meter includes an analog-to-digital conversion module to convert the analog voltage signal into a digital signal. The digital processing unit analyzes the voltage response and captures changes in resistance and reactance values. By adjusting the current frequency, the electrical impedance changes at the differential frequency are captured to generate frequency-adjusted electrical impedance data. The data sampling submodule performs data sampling based on frequency-adjusted electrical impedance data, applies a fast Fourier transform algorithm, sets the sampling rate to 300 times per second, captures the details and changes of bioelectrical impedance, and sets the sampling window size to 5 milliseconds to 1.5 seconds to match the electrical impedance changes of differentiated frequencies. During the data sampling process, the fast Fourier transform algorithm converts the continuous time signal into a frequency domain signal, extracts the key frequency components, and generates sampled electrical impedance data; The electrical impedance recording submodule records the resistance and reactance values ​​at differentiated frequencies based on the sampled electrical impedance data, applies structured data storage technology, sets the data storage format to a structured CSV file, classifies and labels the sampled data, applies a data compression algorithm, optimizes storage space and management efficiency, and generates a preliminary electrical impedance data set.

4. The electrical impedance analysis children's allergy prediction system according to claim 3, characterized in that: The electrical impedance signal processing module includes an electrical impedance denoising submodule, a data smoothing submodule, and an electrical impedance filtering submodule; The electrical impedance measurement submodule is based on multi-spectrum electrical impedance measurement technology. It uses a current source to apply multi-frequency alternating current to the child, and a voltage meter captures the corresponding voltage signal. The current source controls and sets the differential frequency, ranging from low frequency to high frequency, covering the electrical impedance characteristics of multiple tissues in the body. The voltage meter includes an analog-to-digital conversion device to convert the analog voltage signal into a digital signal. The digital processing unit analyzes the voltage response and captures changes in resistance and reactance values. By adjusting the current frequency, the electrical impedance changes at the differential frequencies are captured to generate frequency-adjusted electrical impedance data. The data sampling submodule performs data sampling based on frequency-adjusted electrical impedance data, applies a fast Fourier transform algorithm, sets the sampling rate to 300 times per second, captures the details and changes of bioelectrical impedance, and sets the sampling window size to 5 milliseconds to 1.5 seconds to match the electrical impedance changes of differentiated frequencies. During the data sampling process, the fast Fourier transform algorithm converts the continuous time signal into a frequency domain signal, extracts the key frequency components, and generates sampled electrical impedance data; The electrical impedance recording submodule records the resistance and reactance values ​​at differentiated frequencies based on the sampled electrical impedance data, applies relational database storage technology, sets the data storage format to a structured CSV file, classifies and labels the sampled data, applies a data compression algorithm, optimizes storage space and management efficiency, and generates a preliminary electrical impedance data set.

5. The electrical impedance analysis children's allergy prediction system according to claim 4, characterized in that: The allergy-related factor analysis module includes an electrophysiological parameter extraction submodule, an allergy cluster analysis submodule, and an electrical impedance time warping submodule; The electrophysiological parameter extraction submodule uses the fast Fourier transform algorithm to extract frequency features based on the optimized electrical impedance data, and then uses the Butterworth low-pass filter algorithm to filter the signal with the parameters set to a cutoff frequency of 100 Hz and a filter order of 5. The data is then normalized using the Z-score normalization method to generate an electrophysiological feature dataset; The allergy cluster analysis submodule uses the principal component analysis algorithm to reduce the dimension of the electrophysiological feature dataset, retaining the principal components with a cumulative contribution rate of 95%. Then, the K-means clustering algorithm is applied to cluster the data. The optimal number of clusters is determined by the elbow rule. Clusters are formed by calculating and minimizing the Euclidean distance from the sample point to the center of the multiple clusters to generate the allergy pattern clustering results. The electrical impedance time warping submodule is based on the allergy pattern clustering results and uses the autoregressive sliding average model algorithm to perform time series analysis. The p and q parameters of the model are determined by the autocorrelation function graph and partial autocorrelation function graph analysis method. P is the number of autoregressive terms and q is the number of moving average terms. The parameter values ​​are determined according to the autocorrelation characteristics of the data. The regularity of electrical impedance data changes over time is analyzed to generate allergy-related electrophysiological indicators.

6. The electrical impedance analysis children's allergy prediction system according to claim 5, characterized in that: The allergy pattern deep recognition module includes an electrical impedance feature extraction submodule, a deep pattern learning submodule, and an allergy sequence mining submodule; The electrical impedance feature extraction submodule uses a convolutional neural network algorithm to extract features based on allergy-related electrophysiological indicators, sets 32 filters, each with a kernel size of 3×3, and uses a rectified linear unit activation function for feature mapping to extract key features from the electrophysiological indicators and generate feature extraction results; Based on the feature extraction results, the deep pattern learning submodule continues to deepen the learning process by adding convolutional layers and maximum pooling algorithms from the convolutional neural network algorithm, setting 64 filters, maintaining the kernel size at 3×3, and setting the maximum pooling window size to 2×2 to further extract and identify patterns in the electrophysiological data. A random dropout algorithm is added with a dropout ratio of 0.25 to avoid overfitting and generate pattern learning results. The allergy sequence mining submodule is based on the pattern learning results and applies the long short-term memory network algorithm for sequence learning. The number of neurons is set to 50, the adaptive moment estimation algorithm is used as the optimizer, the learning rate is adjusted to 0.01, and the multi-class logarithmic loss function is used for error evaluation. The network is trained by the backpropagation algorithm, and the parameters of the neural network are cyclically adjusted to capture the time series characteristics in the electrophysiological data, mine the sequence patterns associated with allergic reactions, and generate deep recognition allergy patterns.

7. The electrical impedance analysis children's allergy prediction system according to claim 6, characterized in that: The allergy prediction dynamic adjustment module includes an allergy reaction adaptive learning submodule, an electrical impedance feedback control submodule, and an allergy prediction optimization submodule; The allergic reaction adaptive learning submodule uses a recurrent neural network model built on TensorFlow and a long short-term memory network algorithm to process time series data. Parameters include 128 units and a hyperbolic tangent activation function. During training, the early stopping method is used to monitor the loss value on the validation set, with a tolerance of 10 times and a minimum improvement of 0.01 to generate adaptive learning analysis results. The electrical impedance feedback control submodule uses the proportional-integral-differential control algorithm for feedback control based on the results of adaptive learning analysis. It monitors the electrical impedance data in real time and compares it with the preset allergy mode threshold. The PID parameters are dynamically adjusted, including the proportional coefficient P set to 0.1, the integral coefficient I set to 0.01, and the differential coefficient D set to 0.001, to generate the feedback control adjustment results. The allergy prediction optimization submodule optimizes the allergy prediction model based on the feedback control adjustment results using a grid search algorithm and a cross-validation method. Specifically, it tests differentiated long-short-term memory network algorithm parameter combinations, including the number of units and activation functions, determines the optimal parameter configuration through a five-fold cross-validation method, and dynamically adjusts the allergy prediction threshold to generate a dynamically adjusted prediction threshold.

8. The electrical impedance analysis children's allergy prediction system according to claim 7, characterized in that: The electrical impedance deep learning analysis module includes an electrical impedance time series analysis submodule, an electrical impedance feature deepening submodule, and an electrical impedance dynamic modeling submodule; The electrical impedance time series analysis submodule uses a long short-term memory network algorithm to perform time series analysis on the electrical impedance data based on a dynamically adjusted prediction threshold. The number of neurons is set to 100, and the time series features are extracted. The long-term and short-term trends of the electrical impedance data are analyzed. The time step and batch size are adjusted, and the network structure is optimized to match the characteristics of the electrical impedance data to generate preliminary results of the time series analysis. The electrical impedance feature deepening submodule uses a multi-layer long short-term memory network algorithm to perform in-depth feature analysis based on the preliminary results of time series analysis. It increases the number of network layers and neurons, deepens the feature extraction of electrical impedance data through a layer-by-layer learning strategy, adjusts the parameters of the long short-term memory network layer, including increasing the number of neurons to 150, matching the random inactivation ratio, mining the electrical impedance data pattern, and generating feature deepening results. The electrical impedance dynamic modeling submodule constructs a dynamic long-short-term memory network model based on the feature deepening results, adjusts the model parameters, including setting the learning rate to 0.001 and the number of iterations to 50 times, using the adaptive moment estimation algorithm for network optimization, setting the loss function to the mean square error function, and training the model through the back propagation algorithm to optimize the network's ability to capture dynamic changes in the electrical impedance data and generate electrical impedance depth analysis results.

9. The electrical impedance analysis children's allergy prediction system according to claim 8, characterized in that: The electrical impedance spectrum comprehensive analysis module includes a multi-spectrum wavelet analysis submodule, an electrical impedance signal decomposition submodule, and an electrical impedance frequency optimization submodule; Based on the results of the electrical impedance depth analysis, the discrete wavelet transform algorithm is used to decompose the signal. The db4 wavelet in the Daubechies wavelet function is selected and the number of decomposition layers is set to 5 to capture the characteristics of the electrical impedance signal at the differentiated frequency level and generate multi-spectral wavelet analysis results. The electrical impedance signal decomposition submodule uses a wavelet threshold denoising algorithm and a wavelet reconstruction algorithm based on the multi-spectral wavelet analysis results to process the wavelet coefficients, remove noise and enhance the key features of the signal, refine the frequency characteristics of resistance and reactance, and generate an electrical impedance signal refinement result; The electrical impedance frequency optimization submodule uses a power spectrum density estimation algorithm to identify and optimize key frequency components in the electrical impedance signal based on the electrical impedance signal refinement result, extracts high-resolution frequency domain information, and generates electrical impedance spectrum detail data.

10. The electrical impedance analysis children's allergy prediction system according to claim 9, characterized in that: The children's allergy risk assessment module includes a risk data classification submodule, a model cross-validation submodule, and a risk feature selection submodule; The risk data classification submodule uses the support vector machine algorithm to perform classification analysis based on the detailed data of the electrical impedance spectrum. The kernel function is set to the radial basis kernel function, and the parameters C and gamma are adjusted to 1.0 and 0.

1. The electrical impedance data is feature mapped and hyperplane constructed to distinguish differentiated allergy risk levels and generate classification analysis results. The model cross-validation submodule uses the K-fold cross-validation method to test the stability and generalization ability of the support vector machine model based on the classification analysis results. The cross-validation fold is set to 5. By verifying the performance of the differentiated data subset, the generalization effect of the model is evaluated and the model validation results are generated; Based on the model validation results, the risk feature selection submodule uses a univariate feature selection algorithm to optimize the support vector machine model, selects the top 10 feature variables that are highly correlated with allergy risk, improves model performance and explanatory power, and generates child allergy risk assessment results.

Images