Intelligent oral cavity protection and complication monitoring system for trachea cannula

Abstract

The invention belongs to the technical field of medical equipment, and discloses an intelligent oral cavity protection and complication monitoring system for tracheal intubation, which is characterized in that a flexible adaptation module adopts a composite support substrate of medical-grade flexible memory silica gel and a coaxial electrostatic spinning core-shell structure nanofiber membrane, and a titanium-nickel memory alloy deformable skeleton is embedded in the flexible adaptation module; a double-stage adaptation mode of preoperative oral cavity three-dimensional modeling and intraoperative electronic skin dynamic correction is combined, so that the oral cavity of a patient can be accurately fitted, and dynamic changes such as swallowing and dysphoria can be flexibly handled; by matching with a federal learning pressure adaptive model of the pressure adaptive module, pressure, deformation and body position data can be integrated in real time, and when the pressure of a mucous membrane area exceeds 15kPa, the pressure of a gingival area exceeds 20kPa or the body position of a patient is changed, the micro driving unit is automatically triggered to accurately reduce pressure and adjust the form, so that the problems of pressure sores, gomphiasis and the like caused by a hard bite block are avoided; the hidden danger that the supporting force of soft stuffing is insufficient is solved, and double guarantee of safe supporting and dynamic adaptation is achieved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of medical equipment, and particularly relates to an intelligent oral cavity protection and complication monitoring system for tracheal intubation. BACKGROUND

[0002] Oral tracheal intubation is a core treatment method for establishing an artificial airway and ensuring the ventilation of a patient in a critical care unit, an operating room and the like, and plays an irreplaceable role in the clinical processes such as the treatment of respiratory failure and the ventilation of anesthetic surgery. However, in actual clinical application, the oral cavity protection and complication monitoring scheme matched with the traditional tracheal intubation still has the following technical problems: Most of the existing oral cavity protection devices are hard dental pads or simple soft fillers: the long-term compression of the oral cavity mucosa, gums and teeth by the hard dental pad can easily cause facial pressure sores, lip bleeding, tooth loosening and even tooth loss and the like; the soft filler can reduce the compression feeling, but lacks sufficient supporting strength, and the patient may bite the tracheal catheter when agitated, thereby inducing fatal risks such as ventilation disorders.

[0003] Common oral cavity-related complications of tracheal intubation patients include oral cavity infection, mucosal ulcer, ventilator-associated pneumonia (VAP) and unplanned extubation, among which the occurrence of VAP is closely related to the breeding of bacteria and the accumulation of secretions in the oral cavity. The current clinical monitoring of these complications mainly relies on manual observation by medical staff, which is not only low in efficiency and strong in subjectivity, but also cannot realize real-time continuous monitoring; and part of the improved schemes can only monitor a single index, such as pressure sore pressure, and lack comprehensive monitoring of multi-dimensional key parameters such as oral cavity humidity, bacterial concentration and mucosal state, which leads to untimely early warning of complications and easy missed best intervention opportunity.

[0004] The oral cavity protection device and the complication monitoring equipment in the prior art are mostly independent components, and the monitoring data cannot be directly used to guide the dynamic adjustment of the protection device. When the local pressure is too high, the medical staff need to manually adjust the protection device, and the response has obvious lag; at the same time, the nursing operation such as oral cavity cleaning is disconnected with the monitoring data, and precise nursing cannot be carried out according to the actual oral cavity environment state of the patient, which leads to blindness of nursing intervention. SUMMARY

[0005] The purpose of the present application is to provide an intelligent oral cavity protection and complication monitoring system for tracheal intubation to solve the problems raised in the background.

[0006] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme: an intelligent oral cavity protection and complication monitoring system for tracheal intubation, comprising a flexible adaptation module, a multi-modal monitoring module, a central processing module, a pressure self-adaptive module, a cleaning and drug delivery module, a complication prediction module and a wireless power supply module. Preferably, the support base of the flexible adaptation module is a composite structure of medical-grade flexible memory silicone and coaxial electrospinning core-shell structure nanofiber membrane, the edge array type embedded lead zirconate titanate piezoelectric ceramic sheet of the piezoelectric power generation unit is used for mechanical energy recovery, the nanofiber membrane has flexibility support and biocompatibility, the main body is a hollow spherical structure, and a deformable titanium-nickel memory alloy skeleton is embedded inside; the distributed micro drive unit is built in the skeleton, a two-stage adaptation mechanism of preoperative modeling and intraoperative dynamic correction is constructed; the nanofiber membrane is a coaxial electrospinning core-shell structure nanofiber membrane, which has flexibility support and biocompatibility.

[0007] The core layer of the nanofiber membrane is polycaprolactone material with a diameter of 500-800 nm, which provides flexible support strength; the shell layer is a composite layer of gelatin and hydroxyapatite with a thickness of 100-200 nm, which improves biocompatibility and cell adhesion inhibition effect; the core-shell structure is integrally formed by coaxial electrospinning process, and the fiber porosity is ≥70%, which not only guarantees the flexibility of the support body with an elongation at break ≥180%, but also reduces the foreign body reaction of the oral mucosa, while providing a channel for gas exchange.

[0008] The distributed micro drive unit is a combination structure of micro pneumatic air bag array and piezoelectric ceramic driver, and each drive unit corresponds to 1 cm 2 The oral contact area; the pneumatic air bag array controls inflation and deflation through a micro electromagnetic valve, the gas source is a built-in micro high-pressure gas pump with a rated pressure of 0.3 MPa, the piezoelectric ceramic driver responds to a voltage of 3-5 V, and the hardness of the support body is steplessly adjusted to 0.5-5 Shore A; the drive unit and the central processing module are connected through flexible wires, and the control signal transmission delay is ≤30 ms; the drive unit itself responds with a delay of ≤50 ms, ensuring the real-time and accuracy of pressure adjustment.

[0009] The three-dimensional topography data of the patient's oral cavity is obtained by an oral digital impression instrument before the operation to generate an initial support shape; the mucosal fit and deformation data are collected in real time by the electronic skin integrated on the surface of the composite base during the operation, and the curvature, diameter and occlusal surface hardness of the support body are dynamically adjusted to adapt to the dynamic changes of the patient's oral cavity such as swallowing and restlessness; The composite support base surface is covered with a biomimetic mucosa layer of TPU and silk fibroin core-shell structure, a highly conductive flexible nanocomposite coating, and an outermost electronic skin sensing layer, which together form a support, conductive and sensing integrated structure; the spherical surface is uniformly distributed with 6 mm diameter air-permeable sputum suction holes, and the hole wall adopts a gradual flexible structure to realize the three goals of dynamic adaptation, safe support and sensing integration.

[0010] The gradual change flexible structure is a gradual change design of hardness of the hole wall from the outside (oral cavity contact side) to the inside (hollow channel side), the hardness of the outside is 1 Shore A, the hardness of the inside is 8 Shore A, and the middle is composed of 5 layers of gradient hardness silica gel (each layer has a hardness difference of 1.4 Shore A); a spiral flow guide groove is arranged on the inside of the hole wall, the groove width is 0.5 mm, the pitch is 2 mm, the sputum is guided to flow to the center of the sputum suction hole quickly, meanwhile, the hole wall collapse is avoided to affect ventilation, and the performance indexes of ventilation efficiency ≥ 50 mL / (min·cm 2 ), and sputum suction flow rate ≥ 10 mL / s are realized.

[0011] Preferably, the multi-modal monitoring module relies on the electronic skin layer on the surface of the flexible adaptive module as an installation carrier, integrates the fusion monitoring structure of the arrayed electrochemical sensor and the multi-spectral sensing unit, realizes the synchronous acquisition of multi-dimensional parameters, the arrayed electrochemical sensor includes a pressure sensor, a humidity sensor, a pH sensor and a specific biological sensor, is used for detecting the concentration of VAP related pathogenic bacteria, and simultaneously integrates a posture sensor to acquire body position information; wherein the lower limit of pathogenic bacteria detection is ≤10 2 CFU / mL, and the pressure monitoring accuracy is ≤1 kPa; The multi-spectral sensing unit integrates near-infrared and visible light dual-spectral modules, acquires optical characteristic parameters of oral mucosa tissue in real time, realizes visual early warning of early mucosal ischemia and ulcer; the sensor adopts a three-mode battery-free power supply design of a wireless power supply module, is powered by a radio frequency base station in a conventional scene, is powered by NFC when operated bedside, and supplements electric energy through a piezoelectric energy recovery unit and stores the electric energy in a micro super capacitor; after data conversion, the data are transmitted in real time to a central processing module through a low-delay transmission protocol.

[0012] Preferably, the pressure adaptive module takes real-time data transmitted by the multi-modal monitoring module as input, constructs a closed-loop control logic of multi-source perception, intelligent decision and adjustment, and the central processing module integrates a pressure adaptive model based on federated learning, and integrates pressure sensor array data, electronic skin deformation data and patient body position information in real time. The central processing module adopts a low-power ARM Cortex-M4 processor, has a built-in 1 MB flash memory and 128 KB RAM, and supports multi-threaded data processing; the data processing procedure is as follows: first, filter sensor noise through a digital filtering algorithm, then, according to the priority order of pressure data>body position data>deformation data>other parameters, preferentially guarantee the real-time response of pressure regulation and body position adaptation, and non-critical data adopts a batch processing mode to reduce power consumption, and the processing delay is ≤10 ms.

[0013] The federated learning pressure adaptive model adopts a five-layer architecture of an input layer, a feature fusion layer, a local decision layer, a federated aggregation layer and an output layer. The input layer receives standardized pressure sensor array data, electronic skin deformation data, and body position feature vectors. The pressure sensor array data has a dimension of M×N, where M is the number of sensors and N is the number of data points within a 10-second sampling period. The electronic skin deformation data has a dimension of P×Q, where P is the number of deformation monitoring points and Q is 10-second sampling period × 50Hz sampling frequency = 500 points. The body position feature vector has a dimension of 3, corresponding to the x / y / z axis attitude angles. The feature fusion layer uses a 1D convolutional neural network to extract local spatiotemporal features. The kernel size is 3×3, the number of convolutional layers is 2, the activation function is ReLU, and the output feature dimension is 64. Then, the global features are fused through a fully connected layer. The local decision-making level uses a gradient boosting tree to construct a local pressure prediction model and outputs the predicted pressure adjustment value for the center. The federated aggregation layer uses a federated averaging algorithm to aggregate local model parameters according to the sample size weight of each participating center. The weight coefficient is equal to the sample size of the center / the total sample size. The aggregation period is 24 hours. The output layer maps the aggregated model parameters into specific drive unit control signals, including inflation / deflation flow rates of 0.1-1 mL / s and adjustment durations of 0.5-3 seconds.

[0014] The specific steps for model training are as follows: Each center initialized its model parameters based on 500 local labeled data points. The labeled data included pressure, deformation, body position, and corresponding optimal adjustment values. The learning rate was set to 0.01, and the batch size was 32. A stochastic gradient descent optimizer was used, with 100 iterations of local training. Validation was performed every 20 iterations using a validation set, and local training was stopped when the validation error was below 0.5 kPa. Only the model weight parameters are uploaded to the federated server, and the parameter transmission uses an encrypted protocol; The server generates a global model by weighted averaging of parameters from each center. Global parameters are distributed to each center to update the local model and complete the iteration.

[0015] When the pressure value in a region exceeds a preset threshold, the corresponding micro-drive unit built into the flexible adaptation module is automatically triggered to reduce pressure by local inflation or deflation, so that the pressure is evenly distributed on the oral cavity contact surface. When the patient's position changes are detected, the pressure distribution changes are predicted in advance based on the pre-trained position and pressure correspondence model, and the overall shape and support strength of the support are actively adjusted.

[0016] Preferably, the cleaning drug delivery module integrates a micro-drug delivery unit and a flexible adaptive cleaning unit inside the flexible support body based on the monitoring results of pathogen concentration, humidity, etc. output by the multimodal monitoring module. The micro-drug delivery unit achieves targeted drug delivery through an electro-responsive drug delivery electrode based on the pathogen identification results of the multimodal monitoring module, releasing corresponding antibacterial agents for different pathogens, and the drug dosage is dynamically adjusted according to the pathogen concentration. The flexible adaptive cleaning unit includes a retractable micro-spray head and a flexible rotating wiping pad. When the oral humidity exceeds the standard or the concentration of pathogens reaches the threshold, the central processing module automatically triggers the cleaning program. First, the spray head evenly sprays the appropriate antibacterial cleaning solution. After standing for a preset time, the flexible wiping pad plans the optimal wiping trajectory based on the preoperative 3D model and intraoperative dynamic scanning data of the flexible adaptation module, and gently wipes the oral mucosa and tooth surface. The cleaning waste liquid is discharged by the negative pressure suction device through the suction hole of the flexible adaptation module.

[0017] The antibacterial cleaning solution is a neutral compound preparation, dynamically matched according to the type of pathogen: when Gram-positive bacteria (e.g., Staphylococcus aureus) are detected, a cleaning solution containing 0.1% chlorhexidine is used; when Gram-negative bacteria (e.g., Pseudomonas aeruginosa) are detected, a cleaning solution containing 0.05% polyhexamethylene biguanide is used; when no specific pathogen is detected but humidity is excessive, sterile saline is used; the single spray dose of the cleaning solution is 0.5-2 mL, automatically adjusted according to the oral cavity volume, which is calculated through preoperative three-dimensional modeling.

[0018] The trajectory planning adopts a combination of zone coverage and obstacle avoidance optimization logic, dividing the oral cavity into four regions: teeth, gums, buccal mucosa, and tongue surface. The path for each region is planned in the order of "from inside to outside and from left to right". Obstacle avoidance points include endotracheal tube interfaces, oral wounds, etc., and are marked based on preoperative modeling and intraoperative dynamic scanning data. The rotation speed of the wiping cotton is 50-150 r / min, which can be dynamically adjusted according to the sensitivity of the mucosa. The wiping pressure is controlled by a micro pressure sensor in a closed loop and maintained at 0.5-1 kPa, which is lower than the mucosal damage threshold. When the pressure exceeds 1 kPa, the rotation speed is automatically reduced and the contact depth is reduced.

[0019] Preferably, the complication prediction module integrates real-time parameters from the multimodal monitoring module, body position and regulation data from the pressure adaptation module, and patient clinical information to construct a multi-task complication prediction model based on federated learning. It adopts a mode of local model training and federated parameter aggregation, and integrates multi-center clinical data. The model input includes real-time parameters from the multimodal monitoring module, patient clinical information, and dynamic body position data. This multi-task prediction model adopts a hard-shared network architecture, with a 3-layer fully connected network sharing layer, containing 256, 128, and 64 neurons respectively. The activation function is Leaky ReLU, and the dropout rate is 0.2. Independent task branches are set for the three tasks of oral pressure ulcers, infection, and unplanned extubation. Each branch consists of a 1-layer fully connected layer and a Sigmoid activation function. Input feature preprocessing method: Real-time data from multimodal monitoring such as pressure and humidity are standardized using Z-score with mean μ=0 and standard deviation σ=1. Patient clinical information such as age and underlying diseases are encoded using one-hot encoding. Underlying diseases are classified according to ICD-10 encoding. Dynamic body position data are extracted using a 5-minute sliding window to extract time-series statistical features, including mean, variance, and rate of change. Model training parameters: Local training uses the Adam optimizer with an initial learning rate of 0.001. The learning rate decay strategy is to decrease to 0.9 times the original value every 50 rounds. The batch size is 64, and the local training iterations are 150 rounds. The cross-entropy loss function is used. Federated aggregation uses a weighted federated average algorithm. The weight coefficient = effective sample size of the center × local model validation accuracy. The validation set accounts for 25%, and the aggregation period is 48 hours. The predicted probability of each complication is P∈[0,1]. P<0.3 is considered low risk, 0.3≤P<0.7 is considered medium risk, and P≥0.7 is considered high risk. When multiple complications occur, the highest risk level is used. Among them, the infection risk is additionally associated with the pathogen concentration threshold, ≥10 3 When the pressure ulcer risk is CFU / mL, the predicted probability is adjusted by 0.1. When the pressure ulcer risk is associated with the duration of pressure exceeding the standard, and the single exceedance is ≥30 minutes, the predicted probability is adjusted by 0.15.

[0020] It enables simultaneous risk assessment of various complications such as oral pressure ulcers, infections, and unplanned extubation, generating tiered early warning information and intervention suggestions for different risk levels. Low-risk cases prompt enhanced observation of key areas; medium-risk cases automatically trigger the adaptation procedure of the clean drug delivery module; high-risk cases use AR visualization technology to overlay the warning area with real-time oral images, simultaneously pushing the data to medical terminals and issuing audible and visual alarms; it also supports data trend analysis, realizing the upgrade from passive monitoring to proactive prediction and intervention.

[0021] The complication prediction module is equipped with a built-in miniature high-definition camera to collect real-time oral images and support AR visualization early warning functions. The AR visualization technology uses the built-in miniature high-definition camera to collect real-time oral images and combines them with the sensor array coordinates of the multimodal monitoring module to locate the two-dimensional pixel position and three-dimensional spatial coordinates of the warning area. Through augmented reality rendering algorithms, a red highlighted border and risk level text label are superimposed on the warning area. The images and warning information are transmitted to the medical terminal via 5G. The terminal APP supports image magnification, screenshot archiving, and historical image comparison functions, which facilitates medical staff to accurately locate the lesion area.

[0022] Preferably, the wireless power supply module adopts a three-mode power supply architecture of NFC wireless power supply, radio frequency power supply, and piezoelectric energy recovery. In conventional clinical scenarios, continuous power is provided by an external base station through radio frequency wireless power supply technology; during close-range bedside operations, power supply and data synchronization can be achieved through NFC technology; when the patient moves, mechanical energy is converted into electrical energy and stored in a miniature supercapacitor through a piezoelectric power generation unit. The radio frequency base station is wall-mounted, installed 1.5-2m above the bed surface, with a single base station coverage radius of ≥3m, which can meet the power supply needs of a single ward; the base station output power is ≤10W, which meets medical electromagnetic safety standards, and the electromagnetic radiation is ≤1mW / cm². 2 It will not interfere with other medical equipment such as ventilators and electrocardiogram monitors; when multiple base stations are deployed in the same area, it will automatically switch to the power supply link with the strongest signal strength to avoid power outages.

[0023] Data transmission adopts a three-mode, three-channel design of low-power Bluetooth, 5G, and NFC. Real-time monitoring data is transmitted locally with low latency via BLE; historical data and trend analysis results are transmitted to the cloud server via 5G to support the iterative optimization of the complication prediction model; medical staff can achieve data interaction between the device and the mobile terminal via NFC during bedside inspections; and it supports interconnection with ventilators, negative pressure suction devices, and hospital information systems.

[0024] The beneficial effects of this invention are as follows: 1. The flexible adaptation module of this invention adopts a composite support substrate of medical-grade flexible memory silicone and coaxial electrospun core-shell nanofiber membrane, with an embedded titanium-nickel memory alloy deformable skeleton. Combining a two-stage adaptation method of preoperative oral 3D modeling and intraoperative electronic skin dynamic correction, it can accurately fit the patient's oral cavity and flexibly respond to dynamic changes such as swallowing and agitation. With the federated learning pressure adaptation model of the pressure adaptive module, it can integrate pressure, deformation and body position data in real time. When the pressure in the mucosal area exceeds 15kPa, the pressure in the gingival area exceeds 20kPa, or the patient's body position changes, the micro-drive unit is automatically triggered to precisely depressurize and adjust the shape, avoiding problems such as pressure sores and tooth loosening caused by hard dental pads, solving the hidden danger of insufficient support of soft fillings, and achieving dual protection of safe support and dynamic adaptation.

[0025] 2. The multimodal monitoring module of this invention integrates multiple sensors based on electronic skin, which can simultaneously collect data on pressure, humidity, pH value, and concentrations of VAP-related pathogens such as Staphylococcus aureus and Pseudomonas aeruginosa. Combined with a dual-spectrum sensing unit, it can promptly capture early signals such as mucosal ischemia and ulcers. The complication prediction module integrates case data and constructs a multi-task model through federated learning, which can simultaneously assess the risks of oral pressure ulcers, infections, and unplanned extubation, and provide early warnings and intervention suggestions according to low, medium, and high risk levels. In the case of high risk, AR visualization technology is used to accurately indicate the warning area. This improves the efficiency and accuracy of monitoring and realizes the transformation from passive monitoring to active prediction and precise intervention.

[0026] 3. The cleaning drug delivery module of this invention, based on multimodal monitoring data, achieves targeted drug delivery through an electrically responsive drug delivery electrode, releasing corresponding antibacterial agents for different pathogens, with the dosage precisely adjustable to within 1 μL. When oral humidity exceeds the standard or the concentration of pathogens reaches the threshold, the cleaning program is automatically initiated, spraying antibacterial cleaning solution through a retractable spray head, followed by gentle wiping with a flexible wiping cotton pad along an optimal trajectory, and waste liquid is discharged through the suction hole. The three-mode power supply design of the wireless communication module can ensure stable battery life of the system for more than 72 hours. The three-mode three-channel transmission enables low-latency data interaction and can also interconnect with ventilators and hospital information systems. This allows nursing operations to accurately match the patient's real-time oral condition without manual intervention, improving the accuracy of nursing care, reducing the risk of infection, and alleviating the workload of medical staff. Attached Figure Description

[0027] Fig. 1 This is a flowchart of the overall system of the present invention; Fig. 2 This is a flowchart of the pressure adaptive adjustment process of the present invention; Fig. 3 This is a flowchart illustrating the clean drug delivery linkage process of the present invention. Detailed Implementation

[0028] 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.

[0029] like Figs. 1 to 3 As shown in the figure, this invention provides an intelligent oral cavity protection and complication monitoring system for endotracheal intubation, including a flexible adaptation module, a multimodal monitoring module, a central processing module, a pressure adaptive module, a clean drug delivery module, a complication prediction module, and a wireless communication module. The specific implementation of each module is as follows: The flexible adaptation module uses a medical-grade flexible memory silicone and a coaxial electrospun core-shell nanofiber membrane composite substrate. The main body is a hollow spherical structure that allows the endotracheal tube to pass through. A deformable titanium-nickel memory alloy skeleton is embedded inside. The skeleton has a built-in distributed micro-drive unit, with one set configured per square centimeter. The response delay is ≤50ms, which constructs a two-stage personalized adaptation mechanism of precise preoperative modeling and dynamic intraoperative correction.

[0030] Preoperatively, three-dimensional oral cavity morphology data were acquired using a digital oral impression device with an accuracy of ±10μm to generate an initial support morphology. Intraoperatively, an ultra-thin, breathable electronic skin integrated onto the composite substrate was used to collect real-time data on mucosal adhesion and deformation, dynamically adjusting the curvature, diameter, and occlusal surface hardness of the support to adapt to dynamic oral cavity changes such as swallowing and agitation. The ultra-thin, breathable electronic skin has a thickness of 1-5μm and a water vapor permeability ≥350g·m³. -2 ·h -1 .

[0031] The electronic skin sensing layer uses flexible polyimide as a substrate and integrates distributed capacitive sensing units with a unit spacing of 500μm. It senses the degree of adhesion by measuring the change in capacitance value during mucosal adhesion (range 10-100pF, resolution 0.1pF). Deformation data is acquired through piezoresistive strain gauges with a sensitivity ≥2.0, which can capture minute deformations of 0.1%. The acquired capacitance and resistance signals are transmitted to the central processing module through a highly conductive flexible nanocomposite coating to achieve real-time quantitative feedback of adhesion status and deformation.

[0032] The composite substrate surface is covered with a biomimetic mucosa layer with a TPU and silk fibroin core-shell structure, and the outer layer is coated with a highly conductive flexible nanocomposite coating to achieve the integration of conductivity and sensing functions. It has both biocompatibility, high tensile strength with an elongation rate of ≥250%, and sensing function. The spherical surface is uniformly distributed with 6mm diameter breathable suction holes. The hole walls adopt a gradient flexible structure to take into account the convenience of suction operation, oral ventilation, and support stability, achieving the triple goals of dynamic adaptation, safe support, and sensing integration.

[0033] The multimodal monitoring module utilizes the electronic skin layer on the surface of the flexible adapter module as a mounting carrier, integrating an array of electrochemical sensors and a multispectral sensing unit into a fusion monitoring structure. This enables simultaneous acquisition of multi-dimensional parameters. The electrochemical sensor array includes a pressure sensor, a humidity sensor, a pH sensor, and a specific biosensor. The pressure sensor has a monitoring accuracy of ≤1 kPa, the humidity sensor has a range of 20%-100% RH, and the pH sensor has a range of 3.5-8.5. It can simultaneously detect the concentration of VAP-related pathogens such as Staphylococcus aureus and Pseudomonas aeruginosa in the oral cavity, with a detection limit ≤10. 2 CFU / mL; it also integrates a posture sensor to collect patient position information in real time; The specific biosensor employs the principle of antigen-antibody specific binding, immobilizing specific antibodies corresponding to VAP-related pathogens such as Staphylococcus aureus, Pseudomonas aeruginosa, and Klebsiella pneumoniae on the surface of the sensor's sensitive membrane. When oral pathogens bind to the antibodies, they cause changes in the electrochemical signal (impedance / current) on the sensor surface. Through signal amplification and feature extraction, qualitative identification and quantitative detection of the target pathogens are achieved. Each pathogen corresponds to an independent antibody-sensitive unit, avoiding cross-interference.

[0034] pH data is used to determine the acid-base balance of the oral microenvironment: when pH ≤ 4.5 is detected, it indicates that the oral cavity is too acidic, which may be due to the accumulation of secretions or bacterial metabolism. This triggers the cleaning and drug delivery module to prioritize the spraying of neutral antibacterial cleaning solution and increase the cleaning frequency. When pH ≥ 8.0 is detected, it indicates that the alkalinity is abnormal, which may be related to drug residues or infection. This triggers the use of specific biosensors to enhance the detection of pathogens and ensure that nursing interventions are targeted.

[0035] The multispectral sensing unit integrates near-infrared and visible light dual-spectral modules to collect optical characteristic parameters of oral mucosal tissue in real time, such as hemoglobin concentration and tissue oxygen saturation, so as to realize early visual warning of mucosal ischemia and ulceration. Signal interpretation employs a combination of characteristic threshold comparison and trend analysis: when the tissue oxygen saturation detected by near-infrared spectroscopy is ≤60%, or the mucosal redness value extracted by visible light spectroscopy decreases by ≥30% from the baseline, it is judged as an early ischemia warning; when the tissue oxygen saturation fluctuation amplitude is ≤5% and the hemoglobin concentration is ≤100g / L for 30 consecutive minutes, an ulcer risk warning is triggered; the baseline value is the average detection value within 10 minutes of the patient's initial intubation, which is automatically calibrated every 2 hours to avoid individual differences and interference from ambient light.

[0036] The sensor adopts a three-mode battery-free power supply design with a wireless communication module. In normal scenarios, it is powered by an RF base station. During bedside operation, it is powered by NFC. When the patient is active, the power is replenished by a piezoelectric energy recovery unit and stored in a miniature supercapacitor with a capacity of ≥100μF to ensure power supply stability under extreme conditions.

[0037] After analog-to-digital conversion, the data is transmitted to the central processing module in real time via a low-latency transmission protocol (latency ≤ 20ms).

[0038] The pressure adaptive module takes real-time data such as pressure, deformation, and body position information transmitted by the multimodal monitoring module as input to construct a closed-loop control logic of multi-source perception, intelligent decision-making, and precise adjustment. The central processing module integrates a pressure adaptation model based on federated learning, which is trained by integrating data from more than 1,000 cases from multiple centers. It integrates pressure sensor array data, electronic skin deformation data, and patient body position information in real time, and collects data through an integrated posture sensor.

[0039] When the pressure value in a certain area exceeds the preset threshold, the corresponding micro-drive unit built into the flexible adaptation module is automatically triggered. The preset threshold is set according to the different tolerance levels of tissues in different oral regions, such as ≤20kPa in the gingival area and ≤15kPa in the mucosa area. Precise pressure reduction is achieved by local inflation or deflation, with an adjustment accuracy of ≤0.5mm, so that the pressure is evenly distributed on the oral contact surface. When the patient's body position changes, such as from supine to lateral or from lateral to prone, the pressure distribution change is predicted in advance based on the pre-trained body position and pressure correspondence model. The overall shape and support strength of the support are actively adjusted to avoid a sudden increase in local pressure after the body position changes.

[0040] The cleaning drug delivery module integrates a micro-drug delivery unit and a flexible adaptive cleaning unit within a flexible support body, based on the pathogen concentration and humidity monitoring results output by the multimodal monitoring module. The micro-precision drug delivery unit achieves targeted drug delivery through an electrically responsive drug delivery electrode based on the pathogen identification results of the multimodal monitoring module. It releases corresponding antibacterial agents for different pathogens, such as vancomycin for Staphylococcus aureus and ceftazidime for Pseudomonas aeruginosa. The dosage is dynamically adjusted according to the pathogen concentration, with a dosage accuracy of ≤1μL.

[0041] The electroresponsive drug delivery electrode is a platinum-iridium alloy microelectrode array with an electrode diameter of 50 μm and an array spacing of 200 μm. The electrode surface is coated with a polypyrrole-drug composite membrane. When a DC voltage of 1-3V is applied, the polypyrrole membrane undergoes a redox reaction, and the pores expand to release the drug. The voltage and the drug release rate are linearly correlated; for every 1V increase in voltage, the release rate increases by 0.2 μL / min. The central processing module dynamically adjusts the applied voltage based on the pathogen concentration data to achieve precise dose adjustment within the range of 0.1-5 μL. After drug administration is completed, a reverse voltage is applied, and the pores of the polypyrrole membrane contract to stop drug release.

[0042] The flexible adaptive cleaning unit includes a retractable micro-spray head and a flexible rotating wiping pad. The spray angle of the retractable micro-spray head is adjustable, ranging from 30° to 120°. The flexible rotating wiping pad is made of medical-grade microfiber non-woven fabric, and the wiping pressure is adjustable. When oral humidity exceeds the standard (accumulation of secretions) or the concentration of pathogens reaches the threshold, the central processing module automatically triggers the cleaning program: first, the appropriate antibacterial cleaning solution is evenly sprayed through the spray head. After a preset time, the flexible wiping pad plans the optimal wiping trajectory based on the preoperative 3D model and intraoperative dynamic scanning data of the flexible adaptation module, gently wiping the oral mucosa and tooth surface. The cleaning waste liquid is discharged by the negative pressure suction device through the suction hole of the flexible adaptation module. The preset time is set according to the type of cleaning solution, ranging from 5 to 15 seconds.

[0043] The complication prediction module integrates real-time parameters from the multimodal monitoring module, positional and adjustment data from the pressure adaptation module, and patient clinical information to construct a multi-task complication prediction model based on federated learning. It employs a local model training and federated parameter aggregation approach, integrating multi-center clinical data, including over 100,000 cases of endotracheal intubation patients, while protecting patient data privacy. Model inputs include: real-time parameters from the multimodal monitoring module, patient clinical information, and dynamic positional data. Real-time parameters from the multimodal monitoring module include pressure, humidity, pathogen concentration, and tissue optical parameters. Patient clinical information includes age, underlying diseases, and intubation duration.

[0044] It enables simultaneous risk assessment of various complications such as oral pressure ulcers, infections, and unplanned extubation. The risk levels are divided into low, medium, and high, and tiered early warning information and intervention suggestions are generated for different risk levels: low risk prompts to strengthen observation of key areas; medium risk automatically triggers the adaptation program of the cleaning and drug delivery module; high risk uses AR visualization technology to overlay the warning area with real-time oral images, pushes it to the medical terminal simultaneously, and issues an audible and visual alarm to remind medical staff to handle the emergency.

[0045] Unplanned extubation risk assessment focuses on three core parameters: ①Frequency of body position changes: ≥3 times per hour for drastic body position adjustments; ② The radial force on the conduit detected by the flexible adapter module is ≥5N; ③ When the patient is agitated, the fluctuation range of occlusal pressure is ≥10 kPa and the duration is ≥1 minute; When any two parameters are met simultaneously, the risk level is raised by one level. In high-risk situations, an additional bedside monitor will issue a special alarm to remind medical staff to reinforce the catheter fixation device.

[0046] It also supports data trend analysis, which predicts the development trend of complications in the next 2-4 hours, realizing the upgrade from passive monitoring to active prediction and precise intervention.

[0047] The data trend analysis employs a sliding window time series modeling method, setting three time windows: 1 hour, 2 hours, and 4 hours. Trend characteristics of the monitored parameters within each window are extracted, including the upward / downward slope, fluctuation amplitude, and frequency of peak occurrence. The trend characteristics are modeled using a long short-term memory network, integrating historical monitoring data from the past 72 hours with real-time data to output trend curves of risk values ​​for various complications over the next 2-4 hours. When the slope of the trend curve is ≥0.1 (risk value / hour), the warning level is automatically raised by one level, triggering intervention measures in advance.

[0048] The wireless power supply module adopts a three-mode power supply architecture: NFC wireless power supply, radio frequency power supply, and piezoelectric energy recovery. In conventional clinical scenarios, it is continuously powered by an external base station through radio frequency wireless power supply technology. During close-range bedside operations, it can achieve rapid power supply and data synchronization through NFC technology, and is compatible with mobile terminals such as smartphones. When the patient is active, the piezoelectric power generation unit converts mechanical energy into electrical energy and stores it in a miniature supercapacitor as an emergency backup power source, ensuring the system's endurance and stability in complex clinical environments, and enabling continuous operation for ≥72 hours.

[0049] The piezoelectric power generation unit uses lead zirconate titanate piezoelectric ceramic sheets with a thickness of 0.2 mm and a diameter of 5 mm. These sheets are embedded in an array at the edge of the flexible adapter module. When the patient swallows or turns their head, generating mechanical energy, the piezoelectric ceramic sheets bend and deform, generating an induced voltage of 0.5-5V. This voltage is then processed by a rectifier bridge and a voltage regulator circuit and stored in a miniature supercapacitor. The energy conversion efficiency is ≥15%. Moderate-intensity activities by the patient (such as turning their head 5 times or swallowing 10 times per hour) can supplement 20% of the electrical energy required for system operation, ensuring redundancy in battery life.

[0050] Data transmission adopts a three-mode, three-channel design of low-power Bluetooth, 5G, and NFC: real-time monitoring data is transmitted locally with low latency (≤20ms) via BLE, ensuring rapid response of modules such as pressure regulation and clean drug administration; historical data and trend analysis results are transmitted to the cloud server via 5G to support the iterative optimization of complication prediction models; and NFC enables rapid data interaction between the device and mobile terminal during bedside rounds by medical staff.

[0051] It also supports interconnection with ventilators, negative pressure suction devices, and hospital information systems (HIS), realizing a closed-loop data system for the entire airway management process and improving clinical diagnostic and treatment efficiency.

[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An endotracheal tube smart oral protection and complication monitoring system, comprising: The flexible adaptation module, the multi-modal monitoring module, the central processing module, the pressure self-adaptive module, the cleaning and drug delivery module, the complication prediction module, and the wireless supply module are included. The flexible adaptation module adopts a composite support base, internally embedded with a deformable skeleton, and a distributed driving unit built-in the skeleton, to construct a two-stage adaptation mechanism of preoperative modeling and intraoperative dynamic correction, to collect oral fitting data through an electronic skin sensing layer and dynamically adjust the support shape, to realize dynamic adaptation, safe support and sensing integration. The multi-modal monitoring module relies on the sensing adaptation layer of the flexible adaptation module to construct a fusion monitoring structure, integrates multi-dimensional sensors and dual-spectrum sensing units, synchronously collects oral pressure, humidity, pathogenic bacteria concentration and mucosal tissue optical property parameters, and outputs core monitoring data. The central processing module receives multi-modal monitoring module data, runs intelligent adaptation and prediction models, and controls the pressure self-adaptive module and the cleaning and drug delivery module, to output warning information. The pressure self-adaptive module is based on monitoring data to construct a perception-decision-regulation closed-loop control logic, to integrate pressure, deformation and body position information through an intelligent adaptation model, and to automatically trigger the flexible adaptation module driving unit to reduce pressure and adjust the shape when the pressure exceeds the threshold or the body position changes. The cleaning and drug delivery module integrates cleaning and drug delivery units in the flexible support body, to construct a monitoring-cleaning-treatment closed-loop linkage mechanism, to dynamically adapt cleaning operations and targeted drug delivery, to reduce the risk of infection complications and to optimize the oral monitoring environment. The complication prediction module integrates multi-modal monitoring real-time data, body position and regulation data of the pressure self-adaptive module, and multi-center clinical data, to construct an intelligent prediction model using local training and federated parameter aggregation, to realize multi-complication risk level evaluation, to generate stepwise warning and intervention suggestions. The wireless supply module adopts a multi-mode power supply architecture to ensure stable battery life, and realizes low-latency transmission of monitoring data and interconnection of devices through multi-channel data transmission design.

2. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 1, wherein, The composite support base of the flexible adaptation module adopts a medical-grade flexible memory silicone and nanofiber membrane composite structure, with a hollow spherical main body and a titanium-nickel memory alloy deformable skeleton embedded inside. The distributed driving unit is built-in the skeleton. The two-stage adaptation mechanism is achieved by obtaining three-dimensional topography data through an oral digital impression instrument to generate an initial support shape before surgery, and collecting mucosal fitting degree and deformation data through an electronic skin sensing layer to dynamically adjust the support body curvature, diameter and occlusal surface hardness during surgery. The composite support base surface is covered with a biomimetic mucosa layer and a conductive flexible coating in sequence, with evenly distributed air-permeable sputum suction holes on the spherical surface, and the hole wall adopts a gradual flexible structure. The electronic skin sensing layer is covered on the outside of the conductive flexible coating.

3. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 2, wherein, The multi-modal monitoring module takes the electronic skin layer of the flexible adaptive module surface as the installation carrier, and adopts a fusion monitoring structure of an array type electrochemical sensor and a multi-spectral sensing unit; the electrochemical sensor array includes pressure, humidity, pH, and specific biological sensors for detecting VAP related pathogenic bacteria concentration; the multi-modal monitoring module simultaneously integrates a posture sensor to collect body position information in real time; the multi-spectral sensing unit integrates near-infrared and visible light dual-spectral modules to collect mucosal tissue optical property parameters to realize early visualization warning of mucosal ischemia and ulcer; the sensor adopts a three-mode battery-free design of NFC wireless power supply, radio frequency power supply, and piezoelectric energy recovery, and data is transmitted to the central processing module through a low-delay transmission protocol after analog-to-digital conversion.

4. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 3, wherein, The closed-loop control logic of the pressure adaptive module takes real-time data of the multi-modal monitoring module as input, the central processing module integrates a pressure adaptation model based on federated learning, and integrates pressure sensor array data, electronic skin deformation data, and body position information in real time; when the regional pressure exceeds the preset threshold, the corresponding regional micro driving unit is triggered to reduce pressure through local inflation and deflation; when the body position changes are monitored, the pre-training parameters of the federated learning pressure adaptation model are used to predict the pressure distribution change, and the overall shape and support intensity of the support body are actively adjusted.

5. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 4, wherein, The micro drug delivery unit of the cleaning and drug delivery module is based on the pathogenic bacteria identification results of the multi-modal monitoring module, and realizes targeted drug delivery through electric response drug delivery electrodes for specific pathogenic bacteria, releases corresponding antibacterial agents, and the drug delivery dose is dynamically adjusted according to the pathogenic bacteria concentration; the flexible adaptive cleaning unit includes a retractable micro spray head and a flexible rotating cotton swab, when the humidity exceeds the standard or the pathogenic bacteria concentration reaches the threshold, the central processing module triggers the cleaning program: first spray the antibacterial cleaning liquid, and then wipe the cotton swab according to the preoperative three-dimensional model and intraoperative dynamic scanning data to plan the trajectory to wipe the oral mucosa and tooth surface, and the cleaning waste liquid is discharged by the negative pressure device through the air-permeable sputum suction hole.

6. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 5, wherein, The complication prediction module adopts a multi-task prediction model based on federated learning, which integrates multi-center clinical data through local training and federated parameter aggregation, and the input parameters include multi-modal monitoring real-time data, patient clinical information, and body position dynamic data; The risk level of complications such as oral pressure ulcers, infections, and unplanned extubation can be synchronously evaluated, and ladder type early warning and intervention suggestions are generated: low risk prompts to strengthen observation, medium risk triggers cleaning and drug delivery program, and high risk triggers AR visualization superposition warning area and pushes to the medical terminal; Support data trend analysis to realize passive monitoring to active prediction intervention upgrade.

7. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 6, wherein, The wireless power supply module adopts a three-mode power supply architecture of NFC, radio frequency, and piezoelectric energy recovery: in a conventional scene, the radio frequency base station supplies power, bedside operation is powered by NFC and synchronously data, and the piezoelectric unit converts mechanical energy into electrical energy and stores it in a micro super capacitor when the patient is active; Data transmission adopts BLE, 5G, NFC three-mode three-channel design: real-time data is transmitted through BLE local low-delay, historical data and trend analysis results are uploaded to the cloud to support model iteration, bedside inspection can realize device and mobile terminal interaction through NFC, and interconnection and intercommunication with breathing machines, negative pressure suction devices and hospital information systems are supported.

8. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 7, wherein, The biomimetic mucosa layer adopts a TPU and silk fibroin core-shell structure.

9. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 8, wherein, The specific biosensor has specific antibodies corresponding to VAP-related pathogenic bacteria fixed on the surface.

10. The endotracheal tube smart oral cavity protection and complication monitoring system of claim 9, wherein, The antibacterial cleaning liquid is dynamically matched according to the type of detected pathogenic bacteria.

Images