An artificial airway endoballoon pressure monitoring system based on machine learning

Abstract

The application discloses an artificial airway balloon pressure monitoring system based on machine learning, which comprises the following modules: a server, a repeater, an artificial airway balloon provided with a sensing module and a handheld terminal device; wherein the handheld terminal device is used for inflating, holding breath and deflating the artificial airway balloon through a tracheal cannula; and is also used for reading multi-source data detected by the sensing module; the repeater is used for reading balloon pressure, patient temperature and body position change data detected by a plurality of sensing modules at a preset frequency, and transmitting the data to the server; the server is used for acquiring data detected by all sensing modules, analyzing and processing the data by using a machine learning model, and displaying processing results. The application realizes real-time monitoring and intelligent regulation and control of balloon pressure by analyzing and detecting data by using a machine learning model, and improves nursing efficiency and safety.

Description

Technical Field

[0001] This invention relates to the field of pressure monitoring technology for medical devices, and more specifically to a machine learning-based artificial airway intra-inflicted airway pressure monitoring system. Background Technology

[0002] Currently, in clinical practice, some critically ill patients who are unable to maintain normal spontaneous breathing require the establishment of an artificial airway to ensure unobstructed breathing. One common method is intubation using a tube with an inflatable cuff; this type of tube seals the airway by inflating and deflating the cuff, preventing air leakage or aspiration and ensuring effective mechanical ventilation.

[0003] To ensure safety and effectiveness, the pressure inside the airbag needs to be controlled within an ideal range (usually 25-30 cmH2O). Although there are airbag pressure monitoring systems available, factors such as air leakage in the airbag itself, patient coughing, and changes in body position can cause the airbag pressure to fluctuate frequently between measurements, deviating from the target range or triggering false alarms.

[0004] When the cuff pressure is too low, it may cause aspiration in patients, increasing the risk of ventilator-associated pneumonia (VAP); while excessive pressure may cause excessive pressure on the airway mucosa, leading to local capillary ischemia, damage or even necrosis, causing discomfort and complications for patients.

[0005] Therefore, accurate and timely artificial airway cuff pressure management to improve nursing efficiency and safety is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides an artificial airway airbag pressure monitoring system based on machine learning, which analyzes the detection data through a machine learning model to manage the airbag and realize real-time monitoring and intelligent control of the airbag pressure.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] This invention provides a machine learning-based artificial airway insufflation pressure monitoring system, comprising the following modules: server, repeater, artificial airway insufflation with sensing module, and handheld terminal device;

[0009] The server is connected to multiple repeaters via Ethernet cable or 5G; each repeater is connected to multiple sensor modules via Bluetooth; and the handheld terminal device communicates with the sensor modules via NFC.

[0010] The handheld terminal device is used to inflate, hold, and deflate the airbag in the artificial airway via endotracheal intubation; it is also used to read multi-source data detected by the sensor module.

[0011] The repeater is used to read data on intraballoon pressure, patient temperature and position changes detected by multiple sensor modules at a preset frequency, and transmit the data to the server.

[0012] The server is used to acquire data detected by all sensing modules, analyze and process it using a machine learning model, and display the processing results.

[0013] Furthermore, the server includes a data receiving module, a data storage module, a data analysis module, an alarm module, and a display module;

[0014] The data receiving module is used to receive multi-source data detected by the pressure detection module in real time via a repeater;

[0015] The data storage module is used to store the raw data received by the data receiving module and the data analyzed and processed by the data analysis module, forming a dedicated database for each patient, which helps with historical data backtracking and batch analysis.

[0016] The data analysis module is used to dynamically process the multi-source data, generate high-dimensional feature vectors of pressure, body temperature and body position characteristics, and draw analysis conclusions and identify abnormal fluctuations through machine learning models.

[0017] The alarm module is used to issue graded warning signals when there are abnormal fluctuations;

[0018] The display module integrates a visual interface for real-time display of patient status, statistical charts, and analysis conclusions.

[0019] Furthermore, the dynamic processing of multi-source data to generate high-dimensional feature vectors of pressure, body temperature, and body position characteristics specifically includes:

[0020] The mean, fluctuation range, trend and periodic characteristics of the intraballoon pressure are calculated based on the intraballoon pressure over a preset time period.

[0021] Analyze the short-term fluctuations and long-term trends of patients' body temperature, and extract the postural-pressure correlation features by combining postural change data;

[0022] By using an attention mechanism and a weighted fusion algorithm, stress, body temperature, and body position features are integrated into a unified high-dimensional feature vector, which serves as the input to the machine learning model.

[0023] Furthermore, the machine learning model includes a temporal neural network LSTM and a Transformer, as well as a hybrid model CNN-LSTM and a classifier;

[0024] The LSTM is used to capture the long-term dependencies and short-term fluctuations of intraballoon pressure data, and selectively memorizes key temporal features through a gating mechanism.

[0025] The Transformer uses a self-attention mechanism to dynamically calculate the correlation between data at different time points for multi-source heterogeneous data of pressure, body temperature, and body position, and automatically identifies the lag effect of body position changes on pressure.

[0026] The hybrid model CNN-LSTM extracts the morphological features of the stress signal through local filtering in a one-dimensional CNN convolutional layer, and inputs them into an LSTM layer to process temporal correlations, thereby improving the sensitivity to sudden anomalies.

[0027] The classifier uses an SVM classifier or a random forest to detect anomalies in the output of the hybrid model CNN-LSTM.

[0028] Furthermore, the repeater includes a power module, a Bluetooth data receiving module, and a repeater body;

[0029] The power module is used to supply power to the repeater;

[0030] The Bluetooth receiving data module is used to match the Bluetooth transmitting data module of multiple sensing modules and transmit data simultaneously.

[0031] The repeater body is used to acquire data detected by multiple sensor modules and to perform data management and preprocessing.

[0032] Furthermore, the repeater body is specifically used to acquire data on intraballoon pressure, patient temperature and position changes detected by multiple sensor modules at a preset frequency; and to perform noise filtering, outlier removal and time alignment on the acquired data;

[0033] The noise filtering employs either moving average filtering or wavelet denoising.

[0034] Furthermore, the sensing module includes a pressure sensor, a temperature sensor, a device mounting tilt magnetoresistive sensor, an NFC writing module, a Bluetooth data transmission module, and a positioning module.

[0035] The pressure sensor, temperature sensor, and device-mounted tilt magnetoresistive sensor are used to detect intraballoon pressure, patient body temperature, and positional changes, respectively.

[0036] The Bluetooth transmitting data module is used to match the Bluetooth receiving data module of the repeater and to transmit data.

[0037] The positioning module is used to acquire bed information and transmit it to the handheld terminal device;

[0038] The NFC writing module is used to acquire data on the pressure inside the balloon, the patient's body temperature and position changes, and to communicate with the NFC reading module of the handheld terminal device to transmit data.

[0039] Furthermore, the handheld terminal device includes an air pump, an NFC reading module, a processor, and a display screen;

[0040] The air pump is used to inflate, hold, and deflate the airbag in the artificial airway through endotracheal intubation.

[0041] The NFC reading module is used to communicate with the NFC writing module to transmit data and obtain multi-source data from the NFC writing module; it is also used to obtain bed information from the sensing module.

[0042] The processor is used to analyze and organize the multi-source data and bed information;

[0043] The display screen is used to display the information processed by the processor.

[0044] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a machine learning-based artificial airway cuff pressure monitoring system. The system uses a sensor module to detect the pressure of the artificial airway cuff in real time and feeds the information back to the server. The server analyzes trend changes through a machine learning model and provides timely warnings of abnormal situations. Based on the data analysis, the system formulates a cuff management plan for the patient, reducing misjudgments. At the same time, it facilitates medical staff to remotely monitor the patient's status, improving nursing efficiency and safety. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0046] Figure 1 This is a structural diagram of an artificial airway airbag pressure monitoring system based on machine learning, provided as an embodiment of the present invention.

[0047] Figure 2This is a field layout diagram of an artificial airway airbag pressure monitoring system based on machine learning, provided for an embodiment of the present invention. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] This invention discloses a machine learning-based artificial airway insufflation pressure monitoring system, referring to... Figure 1 As shown, it includes the following modules: server, repeater, sensor module and handheld terminal device;

[0050] The server is connected to multiple repeaters via Ethernet cable or 5G; each repeater is connected to multiple sensor modules via Bluetooth; and the handheld terminal device communicates with the sensor modules via NFC.

[0051] The handheld terminal device is used to inflate, hold, and deflate the airway cuff through endotracheal intubation; it is also used to read multi-source data detected by the sensor module.

[0052] The repeater is used to read data on balloon pressure, patient temperature and position changes detected by multiple sensor modules at a preset frequency, and transmit the data to the server.

[0053] The server is used to acquire data detected by all sensor modules, analyze and process it using machine learning models, and display the processing results.

[0054] This embodiment demonstrates the application of the system of this invention in a typical intensive care unit scenario of a hospital, significantly improving the efficiency and safety of artificial airway management for patients. The scenario setup is as follows: Figure 2As shown, the server of this invention is located in the hospital's data center or a specific server room, connected to multiple repeaters via wired network cable or 5G wireless network. It is responsible for storing and processing data collected from each repeater and performing data analysis using a pre-trained machine learning model. Repeaters are distributed throughout the ward area, each capable of connecting to multiple sensor modules via Bluetooth. These modules automatically read data collected by each sensor module at a preset frequency (e.g., once every 10 seconds), including the pressure inside the balloon, the patient's body temperature, and changes in body position, and send this information to the server. The sensor modules are directly installed near the endotracheal intubation cuff to monitor the cuff pressure in real time; they can also sense important parameters such as patient temperature and changes in body position, ensuring a comprehensive understanding of the patient's physiological state. The handheld terminal device is carried by medical staff and is mainly used for performing intubation, breath-holding, and deflation operations; simultaneously, this device can communicate with the sensor modules via NFC technology to quickly obtain the latest monitoring data, facilitating on-site judgment and adjustments by medical staff.

[0055] For each patient in the ward, once endotracheal intubation is completed, the connected sensor module continuously monitors key indicators such as the pressure inside the cuff, patient temperature, and changes in position. All data is received by a repeater and forwarded to the server. Machine learning algorithms on the server analyze the received data, identifying any factors that might cause abnormal cuff pressure, such as leaks, coughing, or changes in position. If the cuff pressure deviates from the normal range (25-30 cmH2O), the system immediately issues an alarm to notify medical staff. Based on the analysis results, the system can not only provide adjustment suggestions to medical staff but also, in some cases, guide nurses to perform necessary operations directly through handheld terminals, such as appropriately increasing or decreasing the amount of gas in the cuff to maintain optimal pressure levels. Furthermore, medical staff can check the patient's current status while making rounds or passing by, as long as they carry handheld terminals, facilitating timely responses to emergencies and greatly improving the quality of care.

[0056] The components of the system of the present invention are described in detail below.

[0057] The server, as the core component of the machine learning-based artificial airway cuff pressure monitoring system, integrates multiple functional modules to achieve efficient monitoring and management of the patient's condition. In this embodiment, the server includes a data receiving module, a data storage module, a data analysis module, an alarm module, and a display module.

[0058] The data receiving module receives multi-source data detected by the pressure detection module in real time. This module is responsible for receiving multi-source data (such as pressure inside the airbag, patient temperature, and changes in body position) collected by the sensor module from repeaters distributed in different wards. This data is continuously sent to the server's data receiving module at a certain frequency (e.g., once every 10 seconds).

[0059] The data storage module preserves all received data as is, forming a database containing detailed timestamps and source information. This provides a complete data foundation for subsequent data analysis. In addition to the raw data, the results processed by the data analysis module are also stored. Each patient has their own independent database, facilitating long-term tracking of their health status trends and supporting historical data review and batch analysis.

[0060] The data analysis module dynamically processes multi-source data to generate high-dimensional feature vectors for pressure, body temperature, and body position characteristics. First, it calculates the mean, fluctuation range, trend, and periodicity of the intraballoon pressure based on the intraballoon pressure over a preset time period. Second, it analyzes the short-term fluctuations and long-term trends of the patient's body temperature and extracts the body position-pressure correlation features by combining the body position change data. Finally, through an attention mechanism and a weighted fusion algorithm, it integrates the pressure, body temperature, and body position characteristics into a unified high-dimensional feature vector, which serves as the input to the machine learning model.

[0061] When extracting intraballoon pressure characteristics, the mean value can be calculated based on intraballoon pressure readings over a preset time period (e.g., hourly or daily). This helps to understand the patient's overall pressure level over a period of time.

[0062] Regarding fluctuation analysis, statistical methods (such as standard deviation and range) can be used to assess the fluctuation of intraballoon pressure. Large fluctuations may indicate underlying problems, such as balloon leakage or frequent changes in patient position.

[0063] Regarding trend identification, time series analysis techniques can be used to determine the trend of pressure changes within the balloon over time (rising, falling, or remaining stable) and predict potential future pressure anomalies.

[0064] Regarding the detection of periodic characteristics, Fourier transform or other frequency domain analysis tools can be used to look for periodic patterns in the intraballoon pressure. Certain physiological activities (such as breathing) may cause regular fluctuations in the intraballoon pressure.

[0065] When analyzing patient body temperature characteristics, to address short-term fluctuations and long-term trends, body temperature data can be subdivided to examine changes over short periods (e.g., within a day) and long periods (e.g., within a week or month). Short-term fluctuations can reflect acute conditions, while long-term trends tend to be indicators of chronic health. Combining this with postural changes data, considering the potential relationship between patient temperature and position (e.g., temperature may rise when the patient turns over), helps to more accurately explain the causes of temperature fluctuations.

[0066] In extracting the postural-pressure correlation features, the study explored the relationship between postural changes and intraballoon pressure through the analysis of a large number of data samples. For example, does a specific postural position lead to an increase or decrease in pressure? The study tracked patients' postural changes in real time and simultaneously recorded the corresponding intraballoon pressure changes, thus constructing a postural-pressure correlation database.

[0067] Before integrating data from different sources, an attention mechanism is used to weight the importance of each feature. The stress, body temperature, and body position features, adjusted by the attention mechanism, are then merged into a unified high-dimensional feature vector. This vector is then fed as input to the machine learning model for further classification, regression, and prediction tasks.

[0068] This embodiment uses a machine learning model to derive analytical conclusions and identify abnormal fluctuations. If the intensity of a sudden pulse exceeds a first threshold and is not related to body position, it is determined to be an abnormal event, such as coughing or airway displacement. If the pressure baseline drops more than a second threshold and there is no change in body temperature, it is determined to be chronic air leakage; otherwise, it is considered normal fluctuation. The machine learning model in this embodiment includes temporal neural networks LSTM and Transformer, as well as a hybrid model CNN-LSTM and a classifier.

[0069] The Long Short-Term Memory (LSTM) network is used to capture long-term dependencies and short-term fluctuations in balloon pressure data, such as pressure decay caused by slow leakage and instantaneous pressure spikes caused by coughing. It also selectively memorizes key temporal features through gating mechanisms (input gate, forget gate, output gate), thereby accurately capturing short-term fluctuation characteristics.

[0070] Transformer utilizes a self-attention mechanism to compute the correlations between different time points in the input sequence, focusing on all parts of the entire time series, rather than just limited to local regions. This helps in understanding complex time series data. For multi-source heterogeneous data such as stress, body temperature, and body position, Transformer can dynamically adjust the importance weights of each time point and automatically identify the lag effect of positional changes on stress.

[0071] In this embodiment, the one-dimensional convolutional neural network (CNN) layer of the hybrid model CNN-LSTM is used to perform local filtering operations to extract morphological features of the pressure signal, which helps to detect local patterns or anomalies in the signal. The extracted features are then fed into the LSTM layer, where the LSTM is responsible for processing the time-series correlations between these features, further improving the sensitivity to sudden anomalies.

[0072] By combining the spatial feature extraction capabilities of CNNs with the time series modeling capabilities of LSTMs, this hybrid model excels at detecting sudden changes in pressure signals, such as airbag deflation or other emergencies.

[0073] Finally, a classifier such as Support Vector Machine (SVM) or Random Forest is used to perform final anomaly detection on the output of the CNN-LSTM model. Both methods have good classification performance and can determine whether the current state is outside the normal range based on the model parameters trained on historical data.

[0074] Once an anomaly is detected, the system triggers a corresponding alarm mechanism to alert medical staff to take action. Simultaneously, these classifiers can help doctors better understand which factors are most likely to cause the anomaly, thus providing a reference for treatment plans.

[0075] The alarm module is used to issue tiered warning signals when abnormal fluctuations occur; depending on the degree of abnormality, the alarm module will issue different levels of warning signals. For example, a slight deviation from the standard value may trigger a low-level warning, while a severe deviation will trigger an emergency notification.

[0076] To ensure medical staff receive alerts promptly, the alarm module uses multiple notification methods, including SMS, email, and mobile application push notifications. It also allows for setting a silent mode for specific time periods to avoid unnecessary disturbances at night.

[0077] The display module integrates a visual interface for real-time display of patient status, statistical charts, and analytical conclusions. It features an intuitive user interface that allows medical staff to quickly understand the patient's current condition. The interface not only displays real-time pressure readings, body temperature values, and body position information, but also provides various statistical charts (such as trend graphs and bar charts) to help interpret the meaning behind the data. The display module allows users to perform further operations through simple clicks or drag-and-drop actions, such as viewing historical data for a specific time period and adjusting display parameters, enhancing the user experience and improving work efficiency.

[0078] The repeater in this embodiment includes a power module, a Bluetooth data receiving module, and a repeater body.

[0079] The power module is fundamental to ensuring the normal operation of the repeater, providing it with the necessary power. Depending on the application scenario, the power supply can take various forms, such as a built-in rechargeable battery or direct connection to the hospital's standard AC power. For battery-powered systems, battery life and replacement cycle should be considered to ensure stable operation of the equipment over extended periods. To extend battery life or reduce energy consumption, the power module in this embodiment integrates an intelligent power management system. This system dynamically adjusts power consumption based on actual usage, such as automatically entering a low-power mode when there are no data transmission tasks.

[0080] The Bluetooth receiving data module is specifically designed for pairing and communication with the Bluetooth transmitting data modules on multiple sensor modules. Each sensor module periodically sends data packets containing information such as airbag pressure, patient temperature, and positional changes to the repeater. Because multiple sensor modules simultaneously transmit data to the same repeater, the Bluetooth receiving data module possesses excellent multi-connection management and data synchronization capabilities, ensuring that all data is received accurately. Furthermore, given the potential presence of various wireless signal sources in a hospital environment, the Bluetooth receiving data module also exhibits strong anti-interference capabilities, guaranteeing reliable data transmission.

[0081] The main body of the repeater is responsible for receiving all the raw data from the Bluetooth data receiving module and performing preliminary processing. This includes, but is not limited to, noise removal, missing value filling, outlier removal, time alignment, and format conversion to facilitate more efficient processing when uploading to the server. Noise removal employs a moving average filter, which smooths the signal by averaging a continuous range of data points, reducing the impact of random noise. This method is simple and effective, suitable for removing high-frequency noise. Another noise removal technique uses wavelet denoising, which decomposes the signal into wavelet coefficients of different frequency bands using wavelet transform technology. Then, wavelet coefficients considered noise are removed according to a set threshold before reconstructing the signal. Wavelet denoising better preserves the detailed features of the signal and is suitable for signal purification in complex environments.

[0082] The repeater itself is also responsible for the effective management of the acquired data. For example, it can sort and store the data according to timestamps or other identifiers; it can also filter out important information and send it to the server first according to set rules.

[0083] To alleviate server load and speed up response times, the repeater also performs basic data analysis operations, such as calculating averages and detecting outliers. This completes some necessary preliminary work before forwarding data to the server.

[0084] Through the effective collaboration of these three main components, the repeater not only reliably collects data from multiple sensor modules, but also improves the efficiency and stability of the entire monitoring system by performing preliminary data processing and optimization. This is crucial for ensuring the safety of critically ill patients.

[0085] The sensing module includes a pressure sensor, a temperature sensor, a device mounting tilt magnetoresistive sensor, an NFC module, a Bluetooth data transmission module, and a positioning module;

[0086] Among them, the pressure sensor, temperature sensor and equipment tilt magnetoresistive sensor are used to detect data on intraballoon pressure, patient body temperature and position changes, respectively.

[0087] The Bluetooth transmitting data module is matched with the repeater's Bluetooth receiving data module, using low-power Bluetooth technology to achieve wireless transmission of data such as pressure, temperature, and changes in body position; the data is efficiently collected and sent to the server for processing without disturbing the patient.

[0088] The positioning module is used to acquire bed information and transmit it to the handheld terminal device;

[0089] The NFC writing module is used to acquire data on the pressure inside the balloon, the patient's body temperature and position changes, and to communicate with the NFC reading module of the handheld terminal device to transmit data.

[0090] The NFC writing module is responsible for acquiring data on intraballoon pressure, patient temperature, and positional changes from pressure sensors, temperature sensors, and the device-mounted tilt magnetoresistive sensor, and storing this data. When healthcare workers approach using a handheld terminal device, the NFC writing module can communicate with it via the NFC reading module to quickly transfer the most recently collected data and access it.

[0091] The sensing module integrates multiple sensors and technologies to comprehensively and accurately monitor various physiological indicators of patients. It transmits data to repeaters and handheld terminal devices via efficient wireless communication, providing strong support for clinical decision-making. The system's design fully considers various needs in practical applications, ensuring the real-time nature, accuracy, and reliability of the data.

[0092] The handheld terminal device in this embodiment includes an air pump, an NFC reading module, a processor, and a display screen;

[0093] The air pump inflates, deflates (maintains constant pressure), and inflates the airbag in the artificial airway through the endotracheal tube. The pump's operation is controlled by function keys (such as the [Inflate] deflation key and the [Hold] breath-hold key). For example, a long press of the [Hold] key activates the breath-hold function; a short press of the [Hold] key toggles the breath-hold time setting.

[0094] The NFC reading module communicates with the NFC writing module in the sensing module to acquire data on balloon pressure, patient temperature and position changes, and bed information. When the latest data from the sensors is needed quickly, simply bringing the handheld terminal close to the sensing module automatically completes data synchronization. Furthermore, NFC is also required to identify the sensing module's code when performing binding or unbinding functions.

[0095] The processor is responsible for processing and analyzing multi-source data and bed information obtained from the NFC reader module. This includes, but is not limited to, data storage, parameter updates, and timeout management. For example, parameters that can be changed on the device (such as ward / room number, bed number, and preset air pressure) are saved immediately after the change and remain even after restarting the device.

[0096] The display screen shows information processed by the processor. It uses an OLED interface with a resolution of 128*64mm, providing a clear view of the current operating status, warning messages, and other important notifications. This embodiment offers multiple functional interfaces, such as a default main page (inflation function), a breath-holding function interface, a deflation function interface, and a ward round function interface. Each interface follows specific design conventions to ensure a consistent and user-friendly experience.

[0097] This handheld terminal device is designed with the practical needs of a medical environment in mind, integrating a variety of convenient functions to improve work efficiency while ensuring patient safety. Through the effective collaboration of the aforementioned components, real-time monitoring and timely response to patient conditions are achieved.

[0098] This invention's system integrates and analyzes multi-source data, combining attention mechanisms and weighted fusion algorithms to create a unified high-dimensional feature vector, thereby better capturing the interactions between different physiological parameters. By combining machine learning algorithms and a human-centered design philosophy, it provides an efficient, accurate, and easy-to-use solution to improve the quality and safety of artificial airway management. This not only advances medical technology but also brings a better nursing experience to patients.

[0099] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0100] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A machine learning-based artificial airway insufflation pressure monitoring system, characterized in that, It includes the following modules: server, repeater, artificial airway inflator with sensing module and handheld terminal device; The server is connected to multiple repeaters; each repeater is connected to multiple sensor modules; and the handheld terminal device communicates with the sensor modules via NFC. The handheld terminal device is used to inflate, hold, and deflate the airbag in the artificial airway via endotracheal intubation; it is also used to read multi-source data detected by the sensor module. The repeater is used to read data on balloon pressure, patient temperature and position changes detected by multiple sensor modules at a preset frequency, and transmit the data to the server. The server is specifically used to: calculate the mean, fluctuation range, trend and periodic characteristics of the intraballoon pressure based on the intraballoon pressure over a preset time period; analyze the short-term fluctuations and long-term trends of the patient's body temperature, and extract the body position-pressure correlation features by combining body position change data; and integrate the pressure, body temperature and body position features into a unified high-dimensional feature vector through an attention mechanism and a weighted fusion algorithm, which serves as the input to the machine learning model. It is also used to perform analysis and processing using machine learning models and to display the processing results; The machine learning model includes a temporal neural network LSTM and a Transformer, and also includes a hybrid model CNN-LSTM and a classifier; The LSTM is used to capture the long-term dependencies and short-term fluctuations of intraballoon pressure data, and selectively memorizes key temporal features through a gating mechanism. The Transformer uses a self-attention mechanism to dynamically calculate the correlation between data at different time points for multi-source heterogeneous data of pressure, body temperature, and body position, and automatically identifies the lag effect of body position changes on pressure. The hybrid model CNN-LSTM extracts the morphological features of the pressure signal through local filtering in a one-dimensional CNN convolutional layer, and inputs them into an LSTM layer to process temporal correlations, thereby improving the sensitivity to sudden anomalies. The classifier uses an SVM classifier or a random forest to detect anomalies in the output of the hybrid model CNN-LSTM.

2. The artificial airway intracavitary pressure monitoring system based on machine learning as described in claim 1, characterized in that, The server includes a data receiving module, a data storage module, a data analysis module, an alarm module, and a display module; The data receiving module is used to receive multi-source data detected by the pressure detection module in real time via a repeater. The data storage module is used to store the raw data received by the data receiving module and the data analyzed and processed by the data analysis module, forming a dedicated database for each patient, which helps with historical data backtracking and batch analysis. The data analysis module is used to dynamically process the multi-source data, generate high-dimensional feature vectors of pressure, body temperature and body position characteristics, and draw analysis conclusions and identify abnormal fluctuations through machine learning models. The alarm module is used to issue graded warning signals when there are abnormal fluctuations; The display module integrates a visual interface for real-time display of patient status, statistical charts, and analysis conclusions.

3. The artificial airway intracavitary pressure monitoring system based on machine learning as described in claim 1, characterized in that, The repeater includes a power module, a Bluetooth data receiving module, and a repeater body; The power module is used to supply power to the repeater; The Bluetooth receiving data module is used to match the Bluetooth transmitting data module of multiple sensing modules and transmit data simultaneously. The repeater body is used to acquire data detected by multiple sensor modules and to perform data management and preprocessing.

4. The artificial airway intracavitary pressure monitoring system based on machine learning as described in claim 3, characterized in that, The repeater body is specifically used to acquire data on intraballoon pressure, patient temperature and position changes detected by multiple sensor modules at a preset frequency; and to perform noise filtering, outlier removal and time alignment on the acquired data. The noise filtering employs either moving average filtering or wavelet denoising.

5. The artificial airway intra-inflicted airway pressure monitoring system based on machine learning as described in claim 1, characterized in that, The sensing module includes a pressure sensor, a temperature sensor, a device mounting tilt magnetoresistive sensor, an NFC writing module, a Bluetooth data transmission module, and a positioning module. The pressure sensor, temperature sensor, and device-mounted tilt magnetoresistive sensor are used to detect intraballoon pressure, patient body temperature, and positional changes, respectively. The Bluetooth transmitting data module is used to match the Bluetooth receiving data module of the repeater and to transmit data. The positioning module is used to acquire bed information and transmit it to the handheld terminal device; The NFC writing module is used to acquire data on the pressure inside the balloon, the patient's body temperature and position changes, and to communicate with the NFC reading module of the handheld terminal device to transmit data.

6. The artificial airway intracavitary pressure monitoring system based on machine learning as described in claim 5, characterized in that, The handheld terminal device includes an air pump, an NFC reading module, a processor, and a display screen; The air pump is used to inflate, hold, and deflate the airbag in the artificial airway through endotracheal intubation. The NFC reading module is used to communicate with the NFC writing module to transmit data and obtain multi-source data from the NFC writing module; it is also used to obtain bed information from the sensing module. The processor is used to analyze and organize the multi-source data and bed information; The display screen is used to display the information processed by the processor.

Images