Abnormal early warning method, system and equipment under continuous monitoring of abdominal borborygmus

Abstract

The invention discloses an abnormal early warning method, system and equipment under continuous monitoring of abdominal borborygmus, and relates to the technical field of medical monitoring equipment, the method comprises the following steps: collecting a borborygmus signal in real time through a sound collector, extracting features through a signal processing unit to generate first early warning information, and sending the first early warning information to a server; meanwhile, the data are stored, and second early warning information is generated through time sequence correlation analysis. The technical problem of separation of real-time early warning and long-term trend analysis in the prior art is solved, and the technical effect of remarkably improving the timeliness and accuracy of anomaly detection is achieved.

Description

Technical Field

[0001] This application relates to the technical field of medical monitoring devices, and particularly to an abnormal warning method, system and device for continuous monitoring of abdominal bowel sounds. Background Art

[0002] With the progress of medical technology, abdominal bowel sounds, as an important physiological signal reflecting intestinal health status, are gradually used in disease monitoring and diagnosis. Bowel sounds can provide valuable information about intestinal peristalsis and digestive function, helping doctors detect problems such as gastrointestinal discomfort, indigestion and intestinal obstruction in a timely manner, and then carry out early intervention. Especially for target user groups such as postoperative patients and patients with chronic intestinal diseases, bowel sound monitoring has important diagnostic significance and can provide important reference data during the disease recovery process. However, the existing technology relies on manual auscultation or short-term recording devices (such as patent CN106328150A), which can only achieve single monitoring and cannot capture the dynamic changes of bowel sounds; while the real-time warning system proposed in patent CN2845731Y lacks the ability of long-term data analysis, resulting in a high false alarm rate. In addition, the existing technology mostly focuses on real-time monitoring and single abnormal warning, lacking the ability to combine real-time warning with long-term trend analysis, resulting in insufficient warning accuracy. Therefore, there is an urgent need for an innovative technology that can combine real-time warning and long-term trend analysis to improve the timeliness and accuracy of abnormal detection.

[0003] In the current related technologies, there is a technical problem of the separation between real-time warning and long-term trend analysis. Summary of the Invention

[0004] This application solves the technical problem of the separation between real-time warning and long-term trend analysis in the existing technology by providing an abnormal warning method, system and device for continuous monitoring of abdominal bowel sounds.

[0005] This application provides an abnormal warning method for continuous monitoring of abdominal bowel sounds, including:

[0006] Continuously collecting bowel sound signals of a target user by using a sound collector to obtain real-time bowel sound signals; transmitting the real-time bowel sound signals to a signal processing unit for feature recognition to obtain a real-time bowel sound signal processing result; generating a first warning information according to the real-time bowel sound signal processing result, and displaying the first warning information on a display warning unit; storing the real-time bowel sound signal processing result in a data storage unit, and sending it to a terminal through a data transmission unit for time series correlation analysis to obtain a target time series correlation analysis result, and generating a second warning information according to the target time series correlation analysis result.

[0007] This application provides an abnormal warning system for continuous monitoring of abdominal bowel sounds, including:

[0008] A real-time bowel sound signal acquisition module, which is used to continuously acquire the bowel sound signals of the target user by using a sound collector to obtain real-time bowel sound signals; a feature recognition module, which is used to transmit the real-time bowel sound signals to a signal processing unit for feature recognition to obtain a real-time bowel sound signal processing result; a first warning information generation module, which is used to generate a first warning information according to the real-time bowel sound signal processing result and display the first warning information on a display warning unit; a timing correlation analysis module, which is used to store the real-time bowel sound signal processing result in a data storage unit and send it to a terminal through a data transmission unit for timing correlation analysis to obtain a target timing correlation analysis result, and generate a second warning information according to the target timing correlation analysis result.

[0009] The present application also provides an electronic device, including:

[0010] A memory for storing executable instructions; a processor, which is used to implement an abnormal warning method under continuous monitoring of abdominal bowel sounds when executing the executable instructions stored in the memory.

[0011] It is intended to propose an abnormal warning method, system and device under continuous monitoring of abdominal bowel sounds through the present application. First, the bowel sound signals are collected in real time by a sound collector, and the features are extracted by a signal processing unit to generate a first warning information. At the same time, the data is stored and a second warning information is generated through timing correlation analysis, achieving the technical effect of significantly improving the timeliness and accuracy of abnormal detection. Description of the Drawings

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the operations before or below do not necessarily need to be executed precisely in sequence. On the contrary, according to the need, they can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several steps can be removed from these processes.

[0013] Figure 1 It is a schematic flowchart of an abnormal warning method under continuous monitoring of abdominal bowel sounds provided by an embodiment of the present application;

[0014] Figure 2 It is a schematic structural diagram of an abnormal warning system under continuous monitoring of abdominal bowel sounds provided by an embodiment of the present application;

[0015] Figure 3 It is a schematic structural diagram of an exemplary electronic device of the present application;

[0016] Figure 4 This is a flowchart for the supervised training of the construction of a feedforward neural network in the abnormal warning method under continuous monitoring of abdominal bowel sounds provided by this application.

[0017] Explanation of reference numerals: Real-time bowel sound signal acquisition module 10, feature recognition module 20, first warning information generation module 30, time series correlation analysis module 40, input device 201, processor 202, memory 203, output device 204. Detailed implementation manners

[0018] The above description is only an overview of the technical solution of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented according to the content of the description. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the following specifically gives the detailed implementation manners of this application.

[0019] In order to make the purpose, technical solution and advantages of this application clearer, the following will further describe this application in detail with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

[0020] In the following description, "some embodiments" are involved, which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The terms "first\second" involved are only used to distinguish similar objects and do not represent a specific order for the objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application.

[0021] The embodiments of this application provide an abnormal warning method under continuous monitoring of abdominal bowel sounds, as Figure 1 shown, the method includes:

[0022] Step S100: Continuously collect the bowel sound signals of the target user using a sound collector to obtain real-time bowel sound signals. Specifically, first select a high-sensitivity audio sensor such as that in an intelligent abdominal bowel sound counter as the sound collector. Before use, check and calibrate its performance and confirm the power supply status. Then, wear the collector near the umbilical area of the target user's abdomen, ensuring a tight fit with the skin. After activation, the collector samples the bowel sound signals at a preset sampling frequency, converts the analog signals into digital signals, and transmits them to the subsequent unit in real time. During the collection process, the collector monitors the signal quality in real time, analyzes the signal strength and stability through an internal algorithm, automatically adjusts the parameters when interference or signal quality degradation occurs, and records information and prompts the operator if it cannot be solved, so as to ensure the accuracy and reliability of the collected real-time bowel sound signals.

[0023] Step S200: Transmit the real-time bowel sound signals to a signal processing unit for feature recognition to obtain the processing results of the real-time bowel sound signals. Specifically, first obtain a training data set composed of a sample bowel sound signal set covering different health status populations and their processing results, and evenly divide it into n sub-training data sets. Taking the first group as an example, divide it into a training set and a validation set according to a preset ratio, train a framework constructed based on a feedforward neural network with the training set, evaluate the accuracy and generalization ability with the validation set, and update the network parameters accordingly. Repeat this process until the training converges to obtain a trained feature recognizer. The specific structure of the feedforward neural network includes: an input layer connected to a first hidden layer, which uses the ReLU (Rectified Linear Unit) activation function to help accelerate training and improve the non-linear expression ability; then there is a second hidden layer, which also uses the ReLU activation function to further enhance the learning ability of the model; finally, the output layer maps the processing results to the numerical values of the bowel sound frequency and intensity. The training of this network adjusts the weights between layers through the backpropagation algorithm to ensure that the model can accurately predict the bowel sound characteristics in different states. After training is completed, it is embedded in the signal processing unit. When the real-time bowel sound signals are transmitted here, the recognizer calculates the bowel sound frequency by analyzing the signal peaks and valleys, and determines the intensity based on the signal amplitude, energy distribution, etc., so as to obtain the processing results of the real-time bowel sound signals containing the bowel sound frequency and intensity, providing data support for the follow-up.

[0024] In a possible implementation, the real-time bowel sound signal is transmitted to a signal processing unit for feature recognition to obtain a processing result of the real-time bowel sound signal. Step S200 further includes step S210 of obtaining a sample bowel sound signal set and a processing result of the sample bowel sound signal as a training data set. Specifically, to obtain the data set for training, samples are first collected from patients with intestinal diseases and healthy volunteers in the digestive medicine ward and outpatient department of the hospital. Using a device similar to a professional intelligent abdominal bowel sound counter's sound acquisition module, bowel sound signals are collected near the umbilicus of the human abdomen for 3-5 minutes, and the basic information, collection time, and physiological status of the person being collected are recorded. After collection, multiple professional personnel manually listen and count the number of bowel sounds, and label the intensity level according to the sound size, audio amplitude, etc. or divide it with the help of professional software to ensure accurate and consistent labeling. Finally, the sample bowel sound signals and the labeling results are sorted and stored in a database in a format, and backups are made to ensure data security and accessibility, thereby constructing a training data set.

[0025] Step S220: The training data set is evenly divided into n groups to obtain n sub-training data sets, where n is an integer greater than or equal to 1. Specifically, when constructing a bowel sound signal feature recognition model, evenly dividing the training data set into n sub-training data sets is of great significance, and its purpose is to prevent the model from overfitting and enhance the generalization ability. When determining the value of n, factors such as the scale of the training data, the complexity of the model, and the training resources need to be considered comprehensively. For example, for a training data set of 1000 samples, n can be selected between 5 and 10. When dividing, first number the samples, and then sequentially assign them to n groups. Automated operations can be achieved with the help of programming tools and the division results can be checked to ensure that the number of samples and the feature distribution in each group are uniform. After the division is completed, labels are made for each sub-training data set and stored separately to ensure data security and accessibility for subsequent model training and calling.

[0026] Step S230, such as Figure 4As shown, the framework constructed based on the feedforward neural network is supervised and trained successively using the n sets of sub-training data sets, and the network parameters are updated according to the training results until the training converges, obtaining a trained feature recognizer. Specifically, first determine the feedforward neural network framework according to the task and data, such as setting the number of neurons in the input layer, hidden layer, and output layer, selecting activation functions such as ReLU and linear functions, and randomly initializing the weights and biases. Then take the first set from the n sets of sub-training data sets, divide it into a training set and a validation set in a ratio of 7:3, train the network with the training set, calculate the predicted value through forward propagation, calculate the error with the mean square error, find the gradient through backpropagation, and update the weights and biases with algorithms such as stochastic gradient descent. Then evaluate the performance with the validation set. Repeat the operation successively with the remaining sub-training data sets. After each training and validation, judge whether the error of the validation set meets the convergence condition that the decline amplitude is extremely small (such as less than 0.001) within a certain number of training cycles. If it meets, obtain a trained feature recognizer; if it does not meet, adjust the network structure, algorithm parameters, or check the data and then retrain until convergence, preparing for processing real-time bowel sound signals.

[0027] Step S240, embed the feature recognizer in the signal processing unit to recognize the real-time bowel sound signal and obtain the processing result of the real-time bowel sound signal. Specifically, before embedding the trained feature recognizer in the signal processing unit, first package it to ensure that the code structure and interfaces are adapted. At the same time, check the hardware performance and software environment of the signal processing unit to ensure that it can support the operation of the recognizer. After the sound collector collects the real-time bowel sound signal, it is transmitted to the signal processing unit, and the unit inputs the signal into the feature recognizer. The recognizer first preprocesses the signal by filtering and noise reduction, and then calculates the number of bowel sounds according to the learned feature pattern, and determines the intensity by analyzing the signal amplitude and energy distribution. Finally, the obtained results of the number of bowel sounds and intensity are output in a specific format, transmitted to the display and warning module for real-time display, or stored and wirelessly sent to the doctor's terminal through the data storage and transmission module to assist in diagnosis and treatment.

[0028] In a possible implementation, the framework constructed based on the feedforward neural network is supervised and trained in turn using the n sets of sub-training data sets, and the network parameters are updated according to the training results until the training converges, obtaining a trained feature recognizer. Step S230 further includes step S231 of extracting the first set of sub-training data sets from the n sets of sub-training data sets. Specifically, when training the feature recognizer using multiple sets of sub-training data sets, to extract the first set of sub-training data sets from the n sets of sub-training data sets, it is first necessary to clarify its storage structure and location, whether it is stored in a database (such as data distinguished by "group_id" stored in a specific table) or a file (such as a CSV file stored in a specified folder according to a specific naming rule). Then, select a suitable extraction tool or method according to the storage form. For a database, the corresponding query language (such as SQL) can be used, and for a file, a library related to the programming language (such as the pandas library in Python) can be used. After extraction, it is also necessary to check the integrity and accuracy of the data. Check the integrity by viewing the number of rows and columns, and write code to verify whether the data values meet the expected range and format to ensure reliable data for subsequent model training.

[0029] Step S232 is to divide the first set of sub-training data sets into a training set and a validation set according to a preset division ratio. Specifically, in model training, dividing the first set of sub-training data sets into a training set and a validation set is crucial. First, determine the preset division ratio based on experience, data scale, and model complexity, such as the common 7:3 or 8:2. For example, for a set of 1000 data, when divided in a 7:3 ratio, the training set has 700 data and the validation set has 300 data. Then, adopt the random division method. Taking the pandas and numpy libraries in Python as an example, after reading the data, use the sample function to shuffle the order and divide it according to the ratio by slicing. After the division, check whether the data distributions are similar, such as comparing statistical features such as the mean and standard deviation of the number of bowel sounds, and also check for data leakage. If there are problems, adjust the division again to ensure reliable data for subsequent model training and evaluation.

[0030] Step S233: Train the framework constructed based on the feedforward neural network using the training set, and verify the trained framework using the validation set to obtain the first training result. Specifically, first build a feedforward neural network framework according to the task and data, determine the number of neurons in the input, hidden, and output layers, select activation functions such as ReLU and linear, and initialize the weights and biases with normal distribution random numbers and 0 respectively. Subsequently, input the bowel sound signals in the training set into the network, predict the number and intensity of bowel sounds through linear and nonlinear transformations, calculate the loss using the mean square error, backpropagate to obtain the gradient, and update the weights and biases using algorithms such as stochastic gradient descent until the preset number of rounds or the loss converges. After training is completed, input the signals in the validation set into the trained network, obtain the prediction result through forward propagation, evaluate the performance using indicators such as mean square error and mean absolute error, and obtain the first training result by synthesizing various indicators. If the error is large and there is underfitting, increase the network complexity; if there is overfitting, use methods such as regularization to solve it, providing a basis for subsequent optimization.

[0031] Step S234: Update the network parameters based on the first training result, and extract the second set of sub-training data sets from the n sets of sub-training data sets again for supervised training, and so on until the training converges to obtain a trained feature recognizer. Specifically, after obtaining the first training result, judge whether the model is underfitting or overfitting by analyzing indicators such as mean square error and mean absolute error. In the case of underfitting, increase the network complexity and adjust the parameters of the optimization algorithm; in the case of overfitting, use regularization, Dropout technology, or increase the amount of training data. After completing the parameter update, extract the second set of sub-training data sets, divide them into a training set and a validation set, train and verify the updated network, and obtain a new round of training results. After each round of training and verification, judge whether to converge based on whether the decrease amplitude of the validation set error within multiple training cycles is less than a threshold (such as 0.001). If it converges, obtain a trained feature recognizer; if it does not converge, repeat the above steps until convergence to ensure that real-time bowel sound signals can be accurately recognized.

[0032] Step S300: Generate a first warning message based on the processing result of the real-time bowel sound signal, and display the first warning message on the display warning unit. Specifically, medical experts and technicians formulate the normal range standard of bowel sounds according to clinical data and medical research for different populations and physiological states, and input it into the warning rule library. The number and intensity data of the real-time bowel sound signal processing result are transmitted to the analysis module, compared with the standard in the rule library, and it is judged whether it is abnormal through a specific algorithm. For example, if the bowel sound is more than 10 times per minute and the intensity exceeds 150% of the normal upper limit, a warning will be triggered. The information generation module generates a first warning message in a preset format, including details of the abnormality, specific values, time, and warning level. The warning message is transmitted to the display warning unit through an encrypted network and is displayed with different colors, flashing frequencies, and warning sounds according to the degree of abnormality. The interface also displays the historical warning records for easy viewing and analysis.

[0033] In a possible implementation manner, when generating a first warning message based on the processing result of the real-time bowel sound signal and displaying the first warning message on the display warning unit, step S300 further includes step S310: Determine whether the processing result of the real-time bowel sound signal satisfies that the number of bowel sounds per unit time exceeds a preset number threshold and / or the intensity of the bowel sound exceeds a preset bowel sound intensity threshold. If so, generate a first warning message. Specifically, medical professionals determine preset number thresholds and preset bowel sound intensity thresholds such as 6 bowel sounds per minute and an intensity of 0 - 50 dB when a healthy adult is quiet and fasting, according to clinical experience and medical research data for different population characteristics and physiological states. The sound collection device collects the real-time bowel sound signal, transmits it to the signal processing unit for analysis and processing by the feature recognizer, obtains the number and intensity data of the bowel sound and stores them in the cache area. Read the data from the cache, and use a special algorithm to compare the number and intensity of the bowel sound with the threshold according to judgment logics such as "OR" and "AND". For example, if the number of bowel sounds is 8 times or the intensity is 60 dB, a warning will be triggered. When the warning condition is met, the warning information generation module generates a first warning message according to the format requirements, including the abnormal situation, specific data, time, device number, identity identification of the detected person, etc., stores it and waits for transmission and processing to notify relevant personnel.

[0034] In a possible implementation, a first warning message is generated according to the processing result of the real-time bowel sound signal, and the first warning message is displayed on the display warning unit. Step S300 further includes step S320. The display warning unit includes an indicator light and a buzzer icon. Among them, the indicator light and the buzzer icon are used for sound and light reminders after receiving the first warning message. Specifically, in terms of hardware, the indicator light and the buzzer are respectively connected to the drive circuit and then connected to the main board of the display warning unit. At the software level, a control program is developed to initialize the display warning unit, set the initial states of the indicator light and the buzzer, and establish a communication interface with the signal receiving module. After receiving the first warning message, the signal receiving module transmits it to the control program. The program parses the key information and calls the corresponding sound and light reminder strategies according to the warning level and the type of abnormality. In the case of a mild warning, the indicator light flashes yellow once every 3 seconds, and the buzzer emits a soft "beep" sound once every 5 seconds. In the case of a moderate warning, the indicator light flashes orange once every 1 second, and the buzzer sounds once every 2 seconds with an increasing volume. In the case of a severe warning, the indicator light is constantly on in red, and the buzzer emits a high-frequency continuous strong sound. The sound and light reminder continues until the user confirms it. After confirmation, the reminder stops and the warning information is stored. If there is a new warning message, the reminder method is updated.

[0035] Step S400, store the processing result of the real-time bowel sound signal in the data storage unit, and send it to the terminal through the data transmission unit for time series correlation analysis to obtain the target time series correlation analysis result, and generate a second warning message according to the target time series correlation analysis result. Specifically, to realize the analysis and warning of the processing result of the real-time bowel sound signal, first, the processing result is sorted in a predetermined format. For example, structured data such as the number of bowel sounds, intensity, and acquisition time are stored in a data storage unit with high reliability and large-capacity storage capabilities, such as a relational database, using the time stamp as the primary key for convenient query. Then, the data transmission unit sends the data to the terminal safely and accurately through transmission methods such as Ethernet, Wi-Fi, 4G / 5G, etc., using the SSL / TLS encryption protocol and the CRC check algorithm. The terminal uses professional tools or self-developed algorithm models, such as the moving average method, ARIMA model, to perform time series correlation analysis on the received data and explore the laws and trends of bowel sound data in the time dimension. The time series correlation analysis includes the dynamic time warping (DTW) algorithm or the autoregressive moving average (ARMA) model, which is used to explore the long-term trend characteristics of the bowel sound signal. Finally, warning rules are set according to the analysis results. If the number and intensity of bowel sounds exceed the normal range within a certain period of time, the warning generation mechanism is triggered to generate a second warning message including abnormal details, cause analysis, and recommended measures to help detect intestinal problems in a timely manner.

[0036] In a possible implementation, the processing result of the real-time bowel sound signal is stored in the data storage unit and sent to the terminal through the data transmission unit for time series correlation analysis to obtain the target time series correlation analysis result. According to the target time series correlation analysis result, a second warning message is generated. Step S400 further includes step S410 of extracting data from the terminal according to a preset analysis window to obtain a sequence of processing results of the real-time bowel sound signal. Specifically, to obtain the sequence of processing results of the real-time bowel sound signal, it is necessary to first clarify the parameters of the preset analysis window, determine the time span according to the application scenario and analysis purpose. For example, for rapid monitoring of abnormalities, set it to 5 - 15 minutes; for understanding chronic intestinal problems, set it to 1 - 2 hours. The starting time is traced back by the corresponding duration from the current moment. Then, establish a connection according to the terminal type and data storage method. For a local database, use the built-in tools or programming language connection libraries, such as Python combined with the mysql-connector-python library; for network transmission, use the requests library. After the connection is successful, perform data extraction. For a database, use SQL statements for screening; for file-form storage, use tools for reading and screening, such as the pandas library in Python. Finally, obtain a sequence of processing results arranged in chronological order, providing a basis for subsequent analysis and warning.

[0037] Step S420, perform time series correlation analysis on the sequence of processing results of the real-time bowel sound signal based on the chronological order to obtain the target time series correlation analysis result. Specifically, before performing time series correlation analysis on the sequence of processing results of the real-time bowel sound signal, first perform data preprocessing. Check for missing values. For a small amount of missing values, fill them with the mean value; for a large amount of missing values, fill them in combination with relevant data or model predictions. Then, smooth the data using the moving average method to reduce noise interference. Next, select a method model according to the data characteristics and analysis purpose. For example, use autocorrelation and partial autocorrelation analysis to determine periodicity and autocorrelation. When the data has linear and seasonal characteristics, select the seasonal autoregressive integrated moving average model (SARIMA), and use the grid search method combined with the mean squared error to select the optimal parameters. Subsequently, divide the training set and the validation set, train the model with the training set, and evaluate the performance with the validation set. If the performance is not satisfactory, adjust the parameters or change the model until the performance is satisfactory. Finally, use the trained model to analyze the sequence to obtain results such as the change trend of the bowel sound frequency and intensity and the time point correlation, providing a basis for intestinal health assessment and warning.

[0038] Step S430: Determine whether the target temporal correlation analysis result satisfies that the number of bowel sounds per unit time exceeds a preset number threshold and / or the intensity of bowel sounds exceeds a preset bowel sound intensity threshold. If so, generate a second warning message. Specifically, medical experts and data analysts jointly determine the preset number and intensity thresholds based on clinical cases, research results, and practical experience for different populations and physiological states. For example, in the normal state of healthy adults, the number of bowel sounds per minute is 3 - 5 times, and the intensity is 0 - 40 dB. The thresholds are adjusted accordingly for children, the elderly, and different physiological states. Extract the number and intensity data of bowel sounds per unit time from the target temporal correlation analysis result, and organize and store them in chronological order. Use a judgment algorithm to compare the data with the thresholds according to the "or" or "and" logic. For example, if the number of bowel sounds in a certain minute is 7 times or the intensity is 45 dB, triggering a warning when the "or" logic is satisfied. When the warning condition is met, the warning message generation module generates a second warning message in a specific format, covering the abnormal situation, occurrence time, specific value, degree of exceeding the threshold, and relevant suggestions. After storage, it is sent to relevant personnel through channels such as text messages and pop - ups to help pay attention to intestinal health.

[0039] In a possible implementation, perform temporal correlation analysis on the real - time bowel sound signal processing result sequence based on the chronological order to obtain the target temporal correlation analysis result. Step S420 further includes step S421: Extract the first real - time bowel sound signal processing result and the second real - time bowel sound signal processing result from the real - time bowel sound signal processing result sequence in the order from front to back in terms of time. Specifically, to extract the real - time bowel sound signal processing result, first, it is necessary to figure out the data storage form. If it is stored in a database, the data mostly exists in the form of a table, and the table has different fields to record important information such as the number of bowel sounds, intensity, and acquisition time. If it is stored in the form of a file, in a common CSV file, different characteristic data are placed in different columns, and each row corresponds to the bowel sound record at a time point. Next, determine the acquisition time as the basis for time sorting. In actual operation, use a data processing tool. First, read the data file storing the real - time bowel sound signal processing result, and then use the sorting function in the tool to sort the data in ascending order according to the acquisition time to ensure that the data is sorted in chronological order. After the data sorting is completed, the first and second real - time bowel sound signal processing results can be extracted from the sorted data. Use the indexing function of the data processing tool to select the first piece of data ranked at the front as the first real - time bowel sound signal processing result, and select the second piece of data immediately following it as the second real - time bowel sound signal processing result, so as to provide initial data for subsequent analysis.

[0040] Step S422, perform time series correlation analysis on the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result to obtain the first time series correlation analysis result. Specifically, when performing time series correlation analysis on the first and second real-time bowel sound signal processing results, data preprocessing is performed first. Carefully check the data integrity. When there are missing values, use linear interpolation for continuous data and moving average for data with large fluctuations; directly remove outliers such as negative bowel sound times that are obviously wrong, and then normalize data of different dimensions to the [0,1] interval to ensure data quality. Then determine the indicators and methods based on the purpose of the analysis. If you are concerned about the similarity of changes, select the mutual correlation coefficient as the indicator and use Pearson correlation analysis; if you want to explore periodicity, use Fourier transform. Then use Python's pandas and numpy libraries to import the preprocessed data for analysis, such as using numpy's corrcoef function to calculate the mutual correlation coefficient and using fft function to perform Fourier transform. Finally, interpret the analysis results. For example, a correlation coefficient close to 1 indicates that the trend of change is highly similar, close to -1 indicates the opposite, and close to 0 indicates weak correlation. Frequency domain data can determine the main periodic components. The results are organized into a table containing analysis indicators, calculation results, and bowel sound characteristics to obtain the first time series correlation analysis result, laying the foundation for subsequent analysis.

[0041] Step S423, extract the third real-time bowel sound signal processing result from the real-time bowel sound signal processing result sequence again, perform a time series correlation analysis on the third real-time bowel sound signal processing result and the first time series correlation analysis result, and obtain a second time series correlation analysis result. Specifically, when obtaining the second time series correlation analysis result, the operation should first be performed according to the storage form of the real-time bowel sound signal processing result sequence. If the sequence is stored in an array, such as the Python list data_list, the third real-time bowel sound signal processing result is found by indexing third_result=data_list[2] (index starts from 0); if it is stored in the database, the third record is filtered out after sorting by time. Then, the format and dimension of the first time series correlation analysis result and the third real-time bowel sound signal processing result are unified, and the third result is organized into a dictionary form that matches the first result. Then, a suitable algorithm is selected to perform the analysis, such as using the dynamic time warping (DTW) algorithm to calculate the optimal matching path of the time dimension of the two to obtain a similarity measure, or using the Granger causality test to determine the causal relationship. Finally, the second time series correlation analysis results are generated based on the algorithm output, such as the similarity value of DTW and the causal relationship conclusion of the Granger causality test, and are interpreted in depth to explore new correlation characteristics, providing strong data support for subsequent research.

[0042] Step S424, and so on. Based on the second time-series correlation analysis result, perform time-series correlation analysis on the remaining real-time bowel sound signal processing results in the real-time bowel sound signal processing result sequence to obtain the target time-series correlation analysis result. Specifically, after obtaining the second time-series correlation analysis result, extract subsequent data from the real-time bowel sound signal processing result sequence in order, such as the fourth real-time bowel sound signal processing result. According to the storage method, query and obtain it through indexing or chronological order. At the same time, carefully check indicators such as timestamps, the number of bowel sounds, and intensity to ensure the accuracy and integrity of the data and its effective connection with the previous analysis results. Then, integrate the new data with the second time-series correlation analysis result, convert the format of the new data features according to the analysis requirements and match the dimensions, and incorporate them into the existing analysis system. Then, use methods such as the dynamic time warping (DTW) algorithm and the autoregressive moving average (ARMA) model, and flexibly adjust the parameters and strategies for analysis in combination with the characteristics of the new data and the previous analysis feedback. Continuously repeat the steps of extraction, integration, and analysis. Each iteration incorporates new data based on the previous result, and evaluates the analysis result after each completion until all the remaining data is processed. Finally, summarize and collate the results of each iteration to form a comprehensive and systematic target time-series correlation analysis result, providing strong data support for intestinal health assessment and diagnosis.

[0043] In a possible implementation, perform time-series correlation analysis on the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result to obtain the first time-series correlation analysis result. Step S422 further includes step S4221, using the inner product mapping formula to perform similarity recognition on the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result to obtain the first time-series correlation similarity set. Specifically, when processing the first and second real-time bowel sound signal processing results, first represent them as vectors A‘=(a1,a2,…,a n ) and B‘=(b1,b2,…,b n ), where each component corresponds to different eigenvalue, and \(n\) is the number of feature dimensions. Then use the inner product mapping formula, that is, the similarity calculation formula The numerator reflects the sum of the products of the vectors in each dimension and the direction consistency, and the denominator is used for normalization to ensure the comparability of the similarities of vectors with different lengths. Then, for each feature dimension, such as bowel sound frequency, intensity, duration, etc., substitute into the formula to calculate the similarity value. After calculating all dimensions, a series of similarity values obtained form the first time-series correlation similarity set, which comprehensively reflects the similarity relationship between the two signal processing results from multiple dimensions.

[0044] Step S4222: Normalize the first temporal correlation similarity set and add the processing result to an initially empty matrix to obtain the first temporal correlation matrix. Specifically, after obtaining the first temporal correlation similarity set, since the similarity value ranges calculated for different feature dimensions vary due to differences in the original data dimension and value range (for example, the similarity ranges calculated for bowel sound frequency and intensity are different), normalization processing is required. Usually, the min-max normalization method is used, and the formula is where x is the original similarity value, x min and x max are the minimum and maximum values in the set, and x norm is the normalization result. After this processing, all similarity values will be mapped to the interval [0, 1]. After completing the normalization, start constructing the first temporal correlation matrix. First, prepare an initially empty matrix with a size determined by the number of feature dimensions. Assuming there are m feature dimensions, an m×m square matrix can be constructed. Then, place the normalized similarity values into the matrix according to the rules. For example, place the normalized similarity value calculated based on the i-th and j-th feature dimensions in the i-th row and j-th column. Finally, obtain the first temporal correlation matrix that shows the similarity relationships between different feature dimensions and provides a data basis for subsequent analysis.

[0045] Step S4223: Perform a convolution operation on the first temporal correlation matrix and the processing result of the second real-time bowel sound signal using a graph convolutional network to obtain the first temporal correlation analysis result. Specifically, when using a graph convolutional network to obtain the first temporal correlation analysis result, first, a suitable graph convolutional network model needs to be built, which includes a convolutional layer, a pooling layer, and a fully connected layer. The convolutional layer extracts local features, the pooling layer reduces the dimension, and the fully connected layer integrates the features. The number and parameters of each layer need to be reasonably configured according to the task and data characteristics. Then, set the key parameters. For example, the convolution kernel size determines the neighborhood range, the stride controls the sliding interval, the padding parameter affects the data size, and there are also the number of network layers, the number of neurons, etc. These all need to be debugged multiple times to find the optimal combination. Then, process the input data. Treat the first temporal correlation matrix as the adjacency matrix and the processing result of the second real-time bowel sound signal as the node features. Perform preprocessing in advance to make the data format and dimension adapt to the network, and normalization may also be required. Finally, perform the convolution operation. The convolution kernel slides on the data for weighted summation, and the features are gradually abstracted and compressed through multiple layers of processing. After integration by the fully connected layer, the first temporal correlation analysis result is output, providing a basis for subsequent analysis.

[0046] In the embodiment of the present application, bowel sound signals are collected in real time through a sound collector, and the first warning information is generated by extracting features through a signal processing unit. At the same time, the data is stored and the second warning information is generated through temporal correlation analysis, achieving the technical effect of significantly improving the timeliness and accuracy of anomaly detection.

[0047] In the above text, with reference toFigure 1 A method for abnormal early warning under continuous monitoring of abdominal bowel sounds according to an embodiment of the present invention is described in detail. Next, reference will be made to Figure 2 describe an abnormal early warning system under continuous monitoring of abdominal bowel sounds according to an embodiment of the present invention.

[0048] The abnormal early warning system under continuous monitoring of abdominal bowel sounds according to an embodiment of the present invention is used to solve the technical problem of the separation of real-time early warning and long-term trend analysis in the prior art, and achieves the technical effect of significantly improving the timeliness and accuracy of abnormal detection. The abnormal early warning system under continuous monitoring of abdominal bowel sounds includes: a real-time bowel sound signal acquisition module 10, a feature recognition module 20, a first early warning information generation module 30, and a time series correlation analysis module 40.

[0049] The real-time bowel sound signal acquisition module 10 is used to continuously collect the bowel sound signals of the target user by using a sound collector to obtain real-time bowel sound signals.

[0050] The feature recognition module 20 is used to transmit the real-time bowel sound signal to a signal processing unit for feature recognition to obtain a real-time bowel sound signal processing result.

[0051] The first early warning information generation module 30 is used to generate a first early warning information according to the real-time bowel sound signal processing result and display the first early warning information on a display early warning unit.

[0052] The time series correlation analysis module 40 is used to store the real-time bowel sound signal processing result in a data storage unit, and send it to a terminal through a data transmission unit for time series correlation analysis to obtain a target time series correlation analysis result, and generate a second early warning information according to the target time series correlation analysis result.

[0053] Next, the specific configuration of the feature recognition module 20 will be described in detail. As described above, the real-time bowel sound signal is transmitted to the signal processing unit for feature recognition to obtain the processing result of the real-time bowel sound signal. The feature recognition module 20 further includes: a training data set acquisition unit for acquiring a sample bowel sound signal set and a sample bowel sound signal processing result as a training data set; a sub-training data set division unit for evenly dividing the training data set into n groups to obtain n groups of sub-training data sets, where n is an integer greater than or equal to 1; a supervised training unit for sequentially using the n groups of sub-training data sets to perform supervised training on a framework constructed based on a feedforward neural network and updating network parameters according to the training results until the training converges to obtain a trained feature recognizer; and a real-time bowel sound signal recognition unit for embedding the feature recognizer in the signal processing unit to recognize the real-time bowel sound signal and obtain the processing result of the real-time bowel sound signal.

[0054] Among them, when sequentially using the n groups of sub-training data sets to perform supervised training on a framework constructed based on a feedforward neural network and updating network parameters according to the training results until the training converges to obtain a trained feature recognizer, the supervised training unit further includes: a first group of sub-training data set extraction sub-unit for extracting the first group of sub-training data sets from the n groups of sub-training data sets; a first group of sub-training data set division sub-unit for dividing the first group of sub-training data sets into a training set and a validation set according to a preset division ratio; a framework validation sub-unit for using the training set to train a framework constructed based on a feedforward neural network and using the validation set to validate the trained framework to obtain a first training result; and a network parameter update sub-unit for updating network parameters based on the first training result and extracting the second group of sub-training data sets from the n groups of sub-training data sets again for supervised training, and so on until the training converges to obtain a trained feature recognizer.

[0055] Next, the specific configuration of the first warning information generation module 30 will be described in detail. As described above, the first warning information is generated according to the real-time bowel sound signal processing result, and the first warning information is displayed on the display warning unit. The first warning information generation module 30 further includes: a first warning information generation unit, which is used to judge whether the real-time bowel sound signal processing result meets the condition that the number of bowel sounds per unit time exceeds a preset number threshold and / or the bowel sound intensity exceeds a preset bowel sound intensity threshold. If so, the first warning information is generated.

[0056] Among them, the first warning information generation module 30 further includes: a display warning unit composition unit, which is used for the display warning unit to include an indicator light and a buzzer icon. Among them, the indicator light and the buzzer icon are used for sound and light reminders after receiving the first warning information.

[0057] Next, the specific configuration of the timing correlation analysis module 40 will be described in detail. As described above, the real-time bowel sound signal processing result is stored in the data storage unit and sent to the terminal through the data transmission unit for timing correlation analysis to obtain the target timing correlation analysis result. According to the target timing correlation analysis result, the second warning information is generated. The timing correlation analysis module 40 further includes: a terminal data extraction unit, which is used to extract data from the terminal according to a preset analysis window to obtain a real-time bowel sound signal processing result sequence; a target timing correlation analysis result acquisition unit, which is used to perform timing correlation analysis on the real-time bowel sound signal processing result sequence based on the time sequence to obtain the target timing correlation analysis result; a second warning information generation unit, which is used to judge whether the target timing correlation analysis result meets the condition that the number of bowel sounds per unit time exceeds a preset number threshold and / or the bowel sound intensity exceeds a preset bowel sound intensity threshold. If so, the second warning information is generated.

[0058] Among them, a temporal correlation analysis is performed on the sequence of real-time bowel sound signal processing results based on the chronological order to obtain the target temporal correlation analysis result. The target temporal correlation analysis result acquisition unit further includes: a bowel sound signal processing result extraction subunit, which is used to extract the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result from the sequence of real-time bowel sound signal processing results in the order from front to back in terms of time; a first temporal correlation analysis result acquisition subunit, which is used to perform a temporal correlation analysis on the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result to obtain a first temporal correlation analysis result; a second temporal correlation analysis result acquisition subunit, which is used to extract the third real-time bowel sound signal processing result from the sequence of real-time bowel sound signal processing results again, and perform a temporal correlation analysis on the third real-time bowel sound signal processing result and the first temporal correlation analysis result to obtain a second temporal correlation analysis result; a temporal correlation analysis subunit, which is used to continue in this way, and perform a temporal correlation analysis on the remaining real-time bowel sound signal processing results in the sequence of real-time bowel sound signal processing results based on the second temporal correlation analysis result to obtain the target temporal correlation analysis result.

[0059] Among them, a temporal correlation analysis is performed on the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result to obtain a first temporal correlation analysis result. The first temporal correlation analysis result acquisition subunit further includes: a similarity recognition micro-unit, which is used to perform similarity recognition on the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result by using the inner product mapping formula to obtain a first set of temporal correlation similarity degrees; a normalization processing micro-unit, which is used to perform normalization processing on the first set of temporal correlation similarity degrees and add the processing result into an initially empty matrix to obtain a first temporal correlation matrix; a convolution operation micro-unit, which is used to perform a convolution operation on the first temporal correlation matrix and the second real-time bowel sound signal processing result by using a graph convolutional network to obtain a first temporal correlation analysis result.

[0060] The abnormal warning system under continuous monitoring of abdominal bowel sounds provided by the embodiments of the present invention can execute the abnormal warning method under continuous monitoring of abdominal bowel sounds provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0061] Based on the foregoing embodiments, the embodiments of the present application further provide an electronic device, and when the processor of the electronic device is executed, it can implement the method described in any of the previous embodiments.

[0062] Figure 3 It is a schematic structural diagram of an electronic device provided by an embodiment of the present invention, showing a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present invention. Figure 3 The displayed electronic device is only an example and should not impose any limitations on the functions and usage scope of the embodiments of the present invention. The electronic device is presented in the form of a general computing device, and its components may include, but are not limited to, an input device 201, a processor 202, a memory 203, and an output device 204. Among them, the processor 202 may be one or more; the processor 202 executes various functional applications and data processing of the computer device by running software programs, instructions, and modules stored in the memory 203, that is, implements the above-mentioned abnormal warning method under continuous monitoring of abdominal bowel sounds.

[0063] Although this application makes various references to certain modules in the system according to the embodiments of this application, however, any number of different modules can be used and run on the user terminal and / or server. The included individual units and modules are only divided according to functional logic, but are not limited to the above division as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0064] The above specific embodiments do not constitute a limitation on the protection scope of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the protection scope of this application.

Claims

1. An abnormal early warning method under continuous monitoring of abdominal bowel sounds, characterized in that: The method comprises: The sound collector is used to continuously collect the bowel sound signals of the target user to obtain real-time bowel sound signals; Transmitting the real-time bowel sound signal to a signal processing unit for feature recognition to obtain a real-time bowel sound signal processing result; Generating first warning information according to the real-time bowel sound signal processing result, and displaying the first warning information on a display warning unit; The real-time bowel sound signal processing result is stored in a data storage unit, and sent to a terminal through a data transmission unit for timing correlation analysis to obtain a target timing correlation analysis result, and a second warning information is generated according to the target timing correlation analysis result.

2. The abnormal early warning method under continuous monitoring of abdominal bowel sounds according to claim 1, characterized in that: The real-time bowel sound signal is transmitted to a signal processing unit for feature recognition to obtain a real-time bowel sound signal processing result, including: Acquire a sample bowel sound signal set and a sample bowel sound signal processing result as a training data set; Dividing the training data set into n groups equally to obtain n groups of sub-training data sets, where n is an integer greater than or equal to 1; Using the n groups of sub-training data sets to sequentially perform supervised training on a framework constructed based on a feedforward neural network, and updating network parameters according to the training results until the training converges to obtain a trained feature identifier; The feature identifier is embedded in a signal processing unit to identify the real-time bowel sound signal and obtain the real-time bowel sound signal processing result.

3. The abnormal early warning method under continuous monitoring of abdominal bowel sounds according to claim 2, characterized in that: The framework constructed based on the feedforward neural network is supervised and trained in sequence using the n sets of sub-training data sets, and the network parameters are updated according to the training results, including: Extracting a first set of sub-training data sets from the n sets of sub-training data sets; Divide the first group of sub-training data sets into a training set and a validation set according to a preset division ratio; Using the training set to train a framework built based on a feedforward neural network, and using the verification set to verify the trained framework to obtain a first training result; The network parameters are updated based on the first training result, and a second group of sub-training data sets are extracted from the n groups of sub-training data sets for supervised training, and so on, until the training converges to obtain a trained feature identifier.

4. The abnormal early warning method under continuous monitoring of abdominal bowel sounds according to claim 1, characterized in that: Generating first warning information according to the real-time bowel sound signal processing result, and displaying the first warning information on a warning display unit, including: It is determined whether the real-time bowel sound signal processing result satisfies the condition that the number of bowel sounds per unit time exceeds a preset number threshold and / or the bowel sound intensity exceeds a preset bowel sound intensity threshold. If so, a first warning message is generated.

5. The abnormal early warning method under continuous monitoring of abdominal bowel sounds according to claim 1, characterized in that: The real-time bowel sound signal processing result is stored in a data storage unit, and sent to a terminal through a data transmission unit for time series correlation analysis to obtain a target time series correlation analysis result, and a second warning information is generated according to the target time series correlation analysis result, including: Extracting data from the terminal according to a preset analysis window to obtain a real-time bowel sound signal processing result sequence; Performing a time series correlation analysis on the real-time bowel sound signal processing result sequence based on the chronological order to obtain the target time series correlation analysis result; It is determined whether the target time series association analysis result satisfies the condition that the number of bowel sounds per unit time exceeds a preset number threshold and / or the intensity of bowel sounds exceeds a preset bowel sound intensity threshold. If so, a second warning message is generated.

6. The abnormal early warning method under continuous monitoring of abdominal bowel sounds according to claim 5, characterized in that: Performing a time series correlation analysis on the real-time bowel sound signal processing result sequence based on the time sequence to obtain the target time series correlation analysis result includes: Extracting a first real-time bowel sound signal processing result and a second real-time bowel sound signal processing result from the real-time bowel sound signal processing result sequence in a time-ordered order; Performing a time series correlation analysis on the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result to obtain a first time series correlation analysis result; extracting a third real-time bowel sound signal processing result from the real-time bowel sound signal processing result sequence again, performing a time series correlation analysis on the third real-time bowel sound signal processing result and the first time series correlation analysis result to obtain a second time series correlation analysis result; By analogy, based on the second time series correlation analysis result, a time series correlation analysis is performed on the remaining real-time bowel sound signal processing results of the real-time bowel sound signal processing result sequence to obtain the target time series correlation analysis result.

7. The abnormal early warning method under continuous monitoring of abdominal bowel sounds according to claim 6, characterized in that: include: Performing similarity recognition on the first real-time bowel sound signal processing result and the second real-time bowel sound signal processing result by using an inner product mapping formula to obtain a first time series correlation similarity set; Normalizing the first time series association similarity set, and adding the processing result into an initially empty matrix to obtain a first time series association matrix; A graph convolutional network is used to perform a convolution operation on the first temporal association matrix and the second real-time bowel sound signal processing result to obtain a first temporal association analysis result.

8. The abnormal early warning method under continuous monitoring of abdominal bowel sounds according to claim 1, characterized in that: The display warning unit includes an indicator light and a buzzer icon, wherein the indicator light and the buzzer icon are used to provide sound and light reminders after receiving the first warning information.

9. Abnormal early warning system under continuous monitoring of abdominal bowel sounds, characterized in that: The system is used to implement the abnormal early warning method under continuous monitoring of abdominal bowel sounds according to any one of claims 1 to 8, and the system comprises: A real-time bowel sound signal acquisition module, which is used to continuously collect bowel sound signals of a target user using a sound collector to obtain real-time bowel sound signals; A feature recognition module, wherein the feature recognition module is used to transmit the real-time bowel sound signal to a signal processing unit for feature recognition to obtain a real-time bowel sound signal processing result; A first warning information generating module, the first warning information generating module is used to generate first warning information according to the real-time bowel sound signal processing result, and display the first warning information on the display warning unit; A timing correlation analysis module is used to store the real-time bowel sound signal processing results in a data storage unit, and send them to a terminal through a data transmission unit for timing correlation analysis to obtain a target timing correlation analysis result, and generate a second warning information according to the target timing correlation analysis result.

10. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; The processor is used to implement the abnormal warning method under continuous monitoring of abdominal bowel sounds according to any one of claims 1 to 8 when executing the executable instructions stored in the memory.

Images