Intelligent stomach tube based on optical fiber sensing and image monitoring and intelligent early warning method thereof

Abstract

The invention provides an intelligent stomach tube based on optical fiber sensing and image monitoring, and the stomach tube comprises a stomach tube body which is internally integrated with an optical fiber and a miniature camera assembly; at least three FBG (Fiber Bragg Grating) sensing nodes which are respectively positioned in the stomach, the esophagus and the throat section are inscribed on the optical fiber and are used for acquiring pressure and temperature signals; the miniature camera assembly is arranged on the inner wall of the throat section and used for collecting images. A demodulation-edge AI processing terminal is arranged outside the body, connected with the sensing nodes and the camera shooting assembly and used for demodulating signals and conducting multi-mode fusion analysis on the basis of pressure signals and image data, real-time judgment of the situation that the catheter enters the airway by mistake, intelligent evaluation of the risk of aspiration before catheter drawing and early warning output when the risk exceeds the standard are achieved. According to the invention, the FBG sensing node and the miniature camera shooting assembly are integrated in the stomach tube, and the in-vitro demodulation-edge AI processing terminal is combined, so that precise and automatic monitoring and early warning of the whole process from tube indwelling to tube drawing evaluation are realized, and the safety and the medical quality of a patient are fundamentally improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent gastric tube technology, and in particular to an intelligent gastric tube based on fiber optic sensing and image monitoring, and its intelligent early warning method. Background Technology

[0002] As my country's population ages, the number of elderly patients with dysphagia caused by cerebrovascular accidents, intracranial tumors, neurodegenerative diseases, or various causes of impaired consciousness continues to rise. For patients with severe dysphagia who are unable to eat safely orally, nasogastric tubes are routinely used in clinical practice to establish an enteral nutrition pathway to ensure their basic nutritional intake and medication needs.

[0003] Currently, the traditional nasogastric tubes widely used in clinical practice rely primarily on the operator's experience for judgment and assessment during insertion and long-term management, which has many limitations and risks, as detailed below: Traditional tube placement procedures rely primarily on empirical methods such as auscultation of airflow, observation of coughing, or aspiration of gastric fluid to determine if the tube tip has entered the stomach. These methods lack objective, real-time evidence for tube placement. The accuracy of this approach is limited, and it cannot reliably identify cases where a gastric tube has been mistakenly inserted into the trachea. If nutritional fluid is administered through a mistakenly inserted gastric tube, it can lead to fatal aspiration pneumonia, posing a serious medical safety hazard.

[0004] After successful placement, the gastric tube may shift due to vomiting, coughing, changes in position, or unconscious removal, and may even dislodge into the esophagus or pharynx. Traditional methods cannot continuously and in real-time monitor the position of the gastric tube; confirmation can only be achieved by periodically aspirating gastric fluid or performing another X-ray. This creates blind spots in monitoring and makes it difficult to detect shift events in a timely manner, which may lead to misinjection or reflux of the nutrient solution.

[0005] Before removing the nasogastric tube, if the patient's swallowing function may recover, multiple water or food intake tests are necessary to assess the risk of aspiration. This process relies entirely on the subjective observation of nursing staff (e.g., whether coughing or voice changes occur), is labor-intensive, lacks standardized procedures, and has limited sensitivity, making it difficult to identify occult aspiration. This may result in premature removal of the tube, leading to aspiration pneumonia; or premature removal, prolonging the patient's discomfort during tube placement and increasing the risk of complications.

[0006] Based on the above issues, there is an urgent clinical need for a new type of intelligent gastric tube that integrates real-time, multimodal information perception capabilities and has intelligent data analysis and risk warning capabilities, so as to achieve precise and automated monitoring and early warning throughout the entire process from tube placement to tube removal assessment, thereby fundamentally improving patient safety and medical quality. Summary of the Invention

[0007] This invention proposes an intelligent gastric tube based on fiber optic sensing and image monitoring, and its intelligent early warning method, which solves the problem that existing gastric tubes cannot achieve accurate and automated monitoring and early warning throughout the entire process from tube insertion and continuous monitoring to tube removal assessment.

[0008] The technical solution of this invention is implemented as follows: The first aspect of this invention provides an intelligent gastric tube based on fiber optic sensing and image monitoring, comprising: The gastric tube itself; An optical fiber installed on the gastric tube body has at least three FBG sensing nodes etched along the axial direction on the optical fiber; the multiple FBG sensing nodes are located in the gastric segment, esophageal segment and pharyngeal inlet segment, respectively, and are used to collect pressure and temperature signals from three different locations. A miniature camera component is installed on the gastric tube body at a position corresponding to the inner wall of the pharyngeal inlet segment, used to acquire image data of the corresponding area; The demodulation-edge AI processing terminal, located outside the body, is connected to multiple FBG sensor nodes and miniature camera components. It is used to demodulate the received signals and perform multimodal fusion analysis based on the demodulated pressure signals and image data to determine whether the tube has been accidentally inserted into the airway and to assess the risk of aspiration before extubation. When a risk is determined to exist, a warning signal is output.

[0009] Specifically, the FBG sensing node is encapsulated in a stainless steel capillary tube with an outer diameter of no more than 0.4 mm, and the side wall of the capillary tube is provided with a pressure-transmitting hole for pressure transmission.

[0010] Preferably, the external end of the gastric tube body is provided with a detachable magnetic-photoelectric hybrid interface, which simultaneously provides a USB-C power and data channel for the miniature camera component, and an FC / APC fiber optic channel for the FBG sensor node.

[0011] Preferably, the gastric tube body integrates optical fibers and multi-core microcables for transmitting signals and power. The optical fibers and multi-core microcables are integrated into the tube wall by co-extrusion in the same groove during the forming process of the gastric tube body, and the wiring is carried out in a spiral pre-twisting manner during the co-extrusion process.

[0012] Specifically, the demodulation of the FBG sensor signal by the demodulation-edge AI processing terminal includes the following steps: The reflectance spectra of each FBG sensor node were collected, and the real-time drift of the center wavelength of each FBG sensor node was obtained by demodulation. : ; in, This represents the actual Bragg wavelength of the FBG sensing node after being affected by an external physical field. The initial Bragg wavelength of the FBG sensing node under stress-free conditions and at the reference temperature. The effective elastic-optical coefficient of the optical fiber. The effective refractive index of the optical fiber, For the strain acting on the FBG sensing node, This is the coefficient of thermal expansion of the optical fiber. This represents the temperature change relative to a reference temperature. The wavelength shift is decoupled using a sensitivity matrix to separate the independent strain. and temperature change The decoupling formula is as follows: ; in, K The sensitivity matrix was obtained through experimental calibration. and The wavelength shift between the two FBG sensing nodes; According to the formula , will strain Convert to pressure value P ,in, k These are the conversion coefficients determined through calibration.

[0013] A second aspect of the present invention provides an intelligent early warning method, comprising the following steps: During the tube placement process, image sequences acquired by the miniature camera component and pressure and temperature timing signals acquired by each FBG sensor node are simultaneously collected. The pressure and temperature time-series signals are processed to extract pressure time-series features, including pressure values, pressure change rates, and inter-channel pressure gradients. These features are then input into a pre-trained long short-term memory (LSTM) network classification model to obtain the first risk confidence score for gastric tube abortion based on pressure signal analysis. ; The image sequences were preprocessed and analyzed using a pre-trained convolutional neural network model to obtain the second risk confidence score for gastric tube aspiration into the airway based on image analysis. ; If the first risk confidence level Greater than the first preset threshold, or the second risk confidence level If the risk of the gastric tube entering the airway is high, the risk is determined to be greater than the second preset threshold, and an audible and visual alarm signal is immediately triggered.

[0014] Optionally, a weighted fusion algorithm is used to calculate the overall risk confidence level. The calculation formula is as follows: ; in, and These are preset weighting coefficients; If we consider the overall risk confidence level If the risk exceeds the preset risk threshold, the risk of the gastric tube being accidentally inserted into the airway is determined to be high, and an audible and visual alarm signal is immediately triggered.

[0015] Furthermore, during the placement of the gastric tube, pressure signals from multiple FBG sensor nodes are continuously collected and analyzed in real time. If an abnormal pressure pattern consistent with gastric tube displacement is detected, a gastric tube displacement warning is triggered.

[0016] A third aspect of the present invention provides an intelligent early warning method, comprising the following steps: Before the planned extubation, at least one feeding test was conducted, during which real-time images acquired by the miniature camera component and pressure signals acquired by multiple FBG sensor nodes were collected simultaneously. Semantic segmentation is performed on the acquired images to identify and segment the food residue areas, and the percentage of pixels in the residue areas relative to the total visible area is calculated. ; Analyze the pressure signal, extract the time domain and frequency domain features of the signal, and determine whether there is an abnormal reflux pressure waveform that matches the characteristics of gastric contents reflux in terms of amplitude, frequency or waveform shape; If the proportion of food residue pixels If the residual safety threshold is lower than the preset threshold and no abnormal backflow pressure waveform is detected, the patient's risk of aspiration is determined to be low, and a suggestion message of "extubation can be considered" is generated. If the proportion of food residue pixels If the residual safety threshold is exceeded, or if an abnormal backflow pressure waveform is detected, it is determined that there is a risk of accidental aspiration, and a "delayed tube removal" warning message is generated.

[0017] Optionally, after acquiring pressure signals and image data, the risk of aspiration can be assessed using the following methods: Extract the temporal features of the pressure signal, input the extracted temporal features into a pre-trained neural network model, and obtain a confidence level indicating that the current pressure signal is in an "abnormal backflow state". The image is semantically segmented to identify areas with food residue. Based on the area, shape, and texture features of these areas, a lightweight convolutional neural network classifier is used to output a risk confidence level indicating the presence of food residue. The aspiration risk score for a single feeding test is obtained by weighted fusion of two confidence levels. If the average aspiration risk score of multiple consecutive tests exceeds the preset risk threshold, it is determined that there is a high risk of aspiration and a warning message of "delayed extubation" is generated.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention integrates fiber Bragg grating sensing nodes and miniature camera components into a conventional-sized gastric tube, and combines them with an external demodulation-edge AI processing terminal to achieve real-time monitoring of the tube insertion path using both mechanical and visual methods, as well as intelligent assessment of the risk of aspiration before tube removal using both image and pressure methods. This fundamentally changes the traditional model of gastric tube insertion and monitoring that relies on subjective experience, and provides objective, continuous, and automated decision support for clinical practice, significantly improving patient safety and nursing quality. (2) During the tube placement process, the present invention uses an AI model to analyze video images of the pharynx and multi-channel pressure time-series signals in real time, and adopts efficient fusion decision logic to quickly identify the risk of the gastric tube being accidentally inserted into the airway and immediately alarm; this greatly reduces the risk of aspiration pneumonia, a serious complication caused by the tube being accidentally inserted into the airway, and ensures the safety of the tube placement operation. (3) During the period of gastric tube placement, the system can automatically and continuously monitor the pressure signal and warn of the risk of gastric tube displacement by identifying abnormal pressure signals. This automated continuous monitoring function makes up for the blind spot of manual periodic examination, can promptly detect and remind medical staff to deal with the risk of displacement, and avoid the resulting misinjection or reflux of nutrient solution; and avoids the frequent use of X-ray examination, reducing the patient's radiation exposure. (4) In the extubation decision-making stage, the present invention provides an objective and quantitative assessment basis for the timing of extubation by quantitatively analyzing the images of food residue in the stomach and the waveform of reflux pressure. It transforms the traditional extubation assessment process that relies on subjective observation and experience into intelligent analysis based on multimodal data, which reduces human error and helps to optimize the timing of extubation while ensuring safety. This avoids both premature extubation causing aspiration and premature extubation causing patient discomfort. Attached Figure Description

[0019] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the internal structure of the intelligent gastric tube based on fiber optic sensing and image monitoring according to the present invention.

[0021] Figure 2 This is a schematic diagram of the system architecture of the intelligent gastric tube in an embodiment of the present invention.

[0022] Figure 3 This is an architectural block diagram of the demodulation-edge AI processing terminal in an embodiment of the present invention. Detailed Implementation

[0023] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0024] Reference Figure 1 , 2 As shown, the first aspect of the present invention provides an intelligent gastric tube based on fiber optic sensing and image monitoring, comprising: The gastric tube body is made of medical-grade polyurethane material with an outer diameter of no more than 16 Fr (approximately 5.3 mm). To meet biocompatibility and durability requirements, a Parylene-C coating of approximately 5 μm thickness can be sprayed onto its outer wall.

[0025] An optical fiber mounted on the gastric tube body is coated with polyimide to improve its heat resistance and flexibility. At least three FBG sensing nodes are etched axially on the optical fiber; multiple FBG sensing nodes are located in the gastric segment, esophageal segment, and pharyngeal inlet segment, respectively, and are used to collect pressure and temperature signals from three different locations.

[0026] Specifically, the FBG sensing node is encapsulated in a stainless steel capillary tube with an outer diameter of no more than 0.4 mm. The side wall of the capillary tube has a pressure-transmitting hole with a diameter of about 0.3 mm to achieve high-sensitivity measurement of pressure-temperature dual parameters.

[0027] A miniature camera component is installed on the gastric tube body at a position corresponding to the inner wall of the pharyngeal inlet segment, used to acquire image data of the corresponding area; In this embodiment, the miniature camera assembly includes a CMOS image sensor (such as OVM6948, with an external size ≤0.7mm×0.7mm) and a ring-shaped LED supplementary light source arranged around it. The LED light source has a power of no more than 150mW and a color temperature of 4500-5500K to provide suitable and uniform illumination. The camera assembly outputs MIPI-CSI signals through a four-channel miniature coaxial cable bundle.

[0028] An externally mounted demodulation-edge AI processing terminal connects to multiple FBG sensor nodes and miniature camera components; such as Figure 3As shown, the terminal includes a spectral demodulation module (FBG demodulation module), a camera receiving module (camera interface board), an artificial intelligence processor (edge ​​AI inference unit), an alarm output module (UI and alarm), and a communication module; wherein, the spectral demodulation module can use an arrayed waveguide grating demodulator with a sampling rate of up to 10kHz and an accuracy of ±1pm. The artificial intelligence processor integrates a GPU or NPU with a power consumption of ≤10W, used to run artificial intelligence algorithms specifically trained for this invention.

[0029] Preferably, the external end of the gastric tube body is provided with a detachable magnetic-photoelectric hybrid interface. This interface has a self-aligning magnetic positioning structure and provides a USB-C power and high-speed data channel (supporting Power over Ethernet+) for the miniature camera component, as well as an FC / APC fiber optic channel for the FBG sensor node, and the whole meets the IP67 waterproof rating.

[0030] Preferably, to achieve high integration and reliability, the gastric tube body integrates optical fibers and multi-core microcables for transmitting signals and power. These optical fibers and multi-core microcables are integrated into the tube wall during the tube body molding process using a co-extrusion method within the same groove. Furthermore, the wiring is pre-twisted in a spiral manner during co-extrusion to reduce tensile stress during bending and enhance the tube's flexibility. The entire tube body and internal wiring harness are encapsulated with a double layer of epoxy-silicone, allowing it to withstand at least 25 cycles of low-temperature ethylene oxide sterilization.

[0031] Alternatively, the power supply mode can be 5 V / 3 A wired power supply via USB-C, or switch to 13.56 MHz inductively coupled wireless power supply for long-term inpatient applications.

[0032] The gastric tube body of this invention is a hollow, flexible tube, with its central main lumen (approximately 2 mm in inner diameter) serving as a continuous nutrition and drug delivery channel. A sensing and wiring system is integrated within the tube wall. The manufacturing and assembly method of the intelligent gastric tube is as follows: First, at least three FBGs are etched axially at intervals on a single-mode optical fiber to form a sensing node. The etched optical fiber is then embedded with a predetermined tension into a groove with a diameter of approximately 0.5 mm pre-reserved in the tube wall during the extrusion molding of the gastric tube body. Polyurethane (PU) adhesive is used for initial positioning to ensure that the FBG nodes are accurately located in the designed gastric segment, esophageal segment, and pharyngeal inlet segment, and that all integration operations do not involve the central nutrition channel.

[0033] For the location of the FBG sensing node, a stainless steel capillary tube with an outer diameter of no more than 0.4 mm is used for local reinforcement. A pressure-permeable hole with a diameter of approximately 0.3 mm is laser-drilled on the side wall of the capillary tube to ensure pressure transmission sensitivity and protect the grating. The fiber optic tail is led out from the proximal end of the gastric tube body and the end face is treated for connection.

[0034] At the point on the gastric tube body corresponding to the pharyngeal inlet segment, a transparent observation window with a diameter of approximately 2.5 mm is formed by local thickening. This observation window is cast using medical-grade optical polyurethane (Optical PU) to ensure light transmittance and biocompatibility. The observation window is located inside the tube wall and does not protrude into the central lumen.

[0035] A miniature CMOS camera module (such as the OVM6948) is precisely fitted to the inner wall of the observation window, and a ring-shaped LED light source (color temperature 4500-5500K) is evenly arranged around the camera to provide shadowless illumination and reduce glare. The flexible printed circuit board (FPC) or miniature coaxial cable bundle of the camera module extends outward along the cable grooves in the tube wall along with the aforementioned optical fiber, and all cable bundles are externally wrapped with a medical silicone sleeve for integrated protection.

[0036] The optical fiber carrying the FBG is spirally pre-twisted with the multi-core micro-cable that powers the camera and transmits data, and then embedded in the co-extruded groove of the tube wall to achieve integrated "optical fiber-cable-tube".

[0037] For the outer wall and end of the gastric tube front functional segment that integrates sensing and camera components, a double-layer potting process is used for encapsulation: first, epoxy resin is injected for structural fixation and sealing, and then a layer of soft medical-grade silicone is covered to ensure that the encapsulation does not block the nutrient channel inlet and to ensure overall biocompatibility, sealing and resistance to multiple sterilizations.

[0038] The magnetic-photoelectric hybrid interface is assembled at the proximal end of the gastric tube body. The side of the interface integrates a fiber optic connector (FC / APC) and a USB-C interface, while the center is still a standard Luer lock interface, which is directly connected to the central nutrition channel of the gastric tube for connecting a syringe or nutrition bag.

[0039] Finally, the smart gastric tube is quickly connected to the external demodulation-edge AI processing terminal via magnetic docking through the hybrid interface. The terminal automatically recognizes the connection and starts up, powering the gastric tube and beginning to receive sensor and image data.

[0040] The component parameters and specifications used in the intelligent gastric tube of this invention are shown in Table 1 below: Table 1 Parameter Table of Intelligent Gastric Tube Component

[0041] Specifically, the demodulation and processing of the FBG sensing signal by the demodulation-edge AI processing terminal includes the following steps: The reflection spectra of each FBG sensing node were acquired at a sampling rate of 1 kHz, and the real-time drift of the center wavelength of each FBG sensing node was obtained using arrayed waveguide grating demodulation technology. ; The wavelength shift of a single FBG is caused by both strain and temperature, following a physical formula: ; in, This represents the actual Bragg wavelength of the FBG sensing node after being affected by an external physical field. The initial Bragg wavelength of the FBG sensing node under stress-free conditions and at the reference temperature. The effective elastic-optical coefficient of the optical fiber. The effective refractive index of the optical fiber, For the strain acting on the FBG sensing node, This is the coefficient of thermal expansion of the optical fiber. This represents the temperature change relative to a reference temperature. To eliminate cross-sensitivity, two FBG sensing nodes with different sensitivity characteristics need to be used, and the sensitivity matrix needs to be obtained through experimental calibration. K Then, its inverse matrix is ​​used for decoupling to separate the independent strain. and temperature change The decoupling formula is as follows: ; in, K The sensitivity matrix was obtained through experimental calibration. and The wavelength shift between the two FBG sensing nodes; According to the formula , will strain Convert to pressure value P ,in, k The conversion coefficients are determined through mechanical calibration. Subsequently, the pressure and temperature time-series signals are digitally filtered (e.g., using a Butterworth low-pass filter) and Kalman filtered to reduce noise.

[0042] Temporal features (such as instantaneous values, rates of change, and gradients) are extracted from the filtered multi-channel (pharynx, esophagus, and stomach) pressure signals. These features are then input into a pre-trained Long Short-Term Memory (LSTM) network model. This LSTM model can classify the input pressure temporal patterns and output a set of confidence scores, including confidence scores for states such as "accidental airway entry," "normal stomach," and "reflux activity."

[0043] A second aspect of the present invention provides an intelligent early warning method, comprising the following steps: During the tube placement process, image sequences (e.g., 720p resolution, 30 frames / second) acquired by the miniature camera component and pressure and temperature timing signals acquired by each FBG sensor node are simultaneously collected. The pressure and temperature time-series signals are processed to extract pressure time-series features, including pressure values, pressure change rates, and inter-channel pressure gradients. These features are then input into a pre-trained long short-term memory (LSTM) network classification model to obtain the first risk confidence score for gastric tube abortion based on pressure signal analysis. ; The image sequence is preprocessed (including fisheye distortion correction and CLAHE contrast enhancement), and analyzed using a pre-trained convolutional neural network model (e.g., using a lightweight MobileNetV3-Small architecture as the backbone network, combined with a DeepLabV3+ semantic segmentation head to perform real-time segmentation of each frame and identify anatomical structures such as "airway entrance" and "esophageal wall") to obtain the second risk confidence score of gastric tube misplacement into the airway based on image analysis. ; If the first risk confidence level Greater than the first preset threshold (e.g., 0.7), or the second risk confidence level. If the risk of the gastric tube entering the airway is high, the risk is determined to be greater than the second preset threshold (e.g., 0.7), and an audible and visual alarm signal is immediately triggered.

[0044] Alternatively, in another embodiment, a weighted fusion algorithm can be used to calculate the overall risk confidence level. The calculation formula is as follows: ; in, and In this embodiment, a preset weighting coefficient is used to optimize the data based on clinical validation. , ; If we consider the overall risk confidence level If the risk exceeds a preset risk threshold (e.g., 0.5), the risk of the gastric tube being accidentally inserted into the airway is determined to be high, and an audible and visual alarm signal is immediately triggered.

[0045] Furthermore, during the placement of the gastric tube, the system enters a continuous monitoring mode; it continuously collects pressure signals from the FBG sensor nodes in the gastric and esophageal segments and analyzes these pressure signals in real time; by analyzing the stability of the pressure baseline or the disappearance of characteristic waveforms, if an abnormal pressure pattern consistent with gastric tube displacement is detected (e.g., disappearance of gastric pressure characteristics), a gastric tube displacement warning is triggered. Simultaneously, the system can combine images from a pharyngeal camera to observe whether the visualization of the gastric tube in the pharynx is abnormal, providing auxiliary judgment.

[0046] A third aspect of this invention provides another intelligent early warning method specifically for assessing the risk of aspiration before extubation, comprising the following steps: Before planned extubation, at least one feeding test was conducted (three in this implementation), during which real-time images acquired by the miniature camera component and pressure signals acquired by multiple FBG sensor nodes were collected simultaneously. Semantic segmentation is performed on the acquired images (using a different model head than that used in the tube placement stage) to identify and segment the food residue area, and the percentage of pixels in the residue area relative to the total visible area is calculated. ; The pressure signal is analyzed, and its time and frequency domain features are extracted. A trained classifier (such as a classifier head integrated with an LSTM model) is used to determine whether there is an abnormal reflux pressure waveform that matches the characteristics of gastric contents reflux in terms of amplitude, frequency, or waveform morphology.

[0047] If the proportion of food residue pixels If the residual risk is below the preset safety threshold (e.g., 5%) and no abnormal reflux pressure waveform is detected, the patient's risk of aspiration is determined to be low, and a suggestion message of "extubation can be considered" is generated. If the proportion of food residue pixels If the residual pressure exceeds the safety threshold or an abnormal reflux pressure waveform is detected, it is determined that there is a risk of aspiration and a "delayed extubation" warning message is generated. All warning messages can be uploaded to the hospital information system via Wi-Fi or Ethernet.

[0048] Optionally, after acquiring pressure signals and image data, the risk of aspiration can be assessed using the following methods: Extract the temporal features of the pressure signal, input the extracted temporal features into a pre-trained neural network model (such as an LSTM model), and obtain a confidence level indicating that the current pressure signal is in an "abnormal backflow state". The image is semantically segmented to identify areas with food residue. Based on the area, shape, and texture features of these areas, a lightweight convolutional neural network classifier is used to output a risk confidence level indicating the presence of food residue. The aspiration risk score for a single feeding test is obtained by weighted fusion of two confidence levels. If the average aspiration risk score of multiple consecutive tests exceeds the preset risk threshold, it is determined that there is a high risk of aspiration and a warning message of "delayed extubation" is generated.

[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart gastric tube based on fiber optic sensing and image monitoring, characterized in that, include: The gastric tube itself; An optical fiber installed on the gastric tube body has at least three FBG sensing nodes etched along the axial direction on the optical fiber; the multiple FBG sensing nodes are located in the gastric segment, esophageal segment and pharyngeal inlet segment, respectively, and are used to collect pressure and temperature signals from three different locations. A miniature camera component is installed on the gastric tube body at a position corresponding to the inner wall of the pharyngeal inlet segment, used to acquire image data of the corresponding area; The demodulation-edge AI processing terminal, located outside the body, is connected to multiple FBG sensor nodes and miniature camera components. It is used to demodulate the received signals and perform multimodal fusion analysis based on the demodulated pressure signals and image data to determine whether the tube has been accidentally inserted into the airway and to assess the risk of aspiration before extubation. When a risk is determined to exist, a warning signal is output.

2. The intelligent gastric tube based on fiber optic sensing and image monitoring as described in claim 1, characterized in that, The FBG sensing node is encapsulated in a stainless steel capillary tube with an outer diameter of no more than 0.4 mm, and the side wall of the capillary tube is provided with a pressure-transmitting hole for pressure transmission.

3. The intelligent gastric tube based on fiber optic sensing and image monitoring as described in claim 1, characterized in that, The external end of the gastric tube body is provided with a detachable magnetic-photoelectric hybrid interface, which simultaneously provides a USB-C power and data channel for the miniature camera component, and an FC / APC fiber optic channel for the FBG sensor node.

4. The intelligent gastric tube based on fiber optic sensing and image monitoring as described in claim 1, characterized in that, The gastric tube body integrates optical fibers and multi-core micro-cables for transmitting signals and power. The optical fibers and multi-core micro-cables are integrated into the tube wall by co-extrusion in the same groove during the forming process of the gastric tube body, and the wiring is carried out in a spiral pre-twisting manner during the co-extrusion process.

5. The intelligent gastric tube based on fiber optic sensing and image monitoring as described in claim 1, characterized in that, The demodulation-edge AI processing terminal demodulates the FBG sensor signal by the following steps: The reflectance spectra of each FBG sensor node were collected, and the real-time drift of the center wavelength of each FBG sensor node was obtained by demodulation. : ； in, This represents the actual Bragg wavelength of the FBG sensing node after being affected by an external physical field. The initial Bragg wavelength of the FBG sensing node under stress-free conditions and at the reference temperature. The effective elastic-optical coefficient of the optical fiber. The effective refractive index of the optical fiber, For the strain acting on the FBG sensing node, This is the coefficient of thermal expansion of the optical fiber. This represents the temperature change relative to a reference temperature. The wavelength shift is decoupled using a sensitivity matrix to separate the independent strain. and temperature change The decoupling formula is as follows: ； in, K The sensitivity matrix was obtained through experimental calibration. and The wavelength shift between the two FBG sensing nodes; According to the formula , will strain Convert to pressure value P ,in, k These are the conversion coefficients determined through calibration.

6. An intelligent early warning method, based on the intelligent gastric tube according to any one of claims 1 to 5, characterized in that, Includes the following steps: During the tube placement process, image sequences acquired by the miniature camera component and pressure and temperature timing signals acquired by each FBG sensor node are simultaneously collected. The pressure and temperature time-series signals are processed to extract pressure time-series features, including pressure values, pressure change rates, and inter-channel pressure gradients. These features are then input into a pre-trained long short-term memory (LSTM) network classification model to obtain the first risk confidence score for gastric tube abortion based on pressure signal analysis. ; The image sequences were preprocessed and analyzed using a pre-trained convolutional neural network model to obtain the second risk confidence score for gastric tube aspiration into the airway based on image analysis. ; If the first risk confidence level Greater than the first preset threshold, or the second risk confidence level If the risk of the gastric tube entering the airway is high, the risk is determined to be greater than the second preset threshold, and an audible and visual alarm signal is immediately triggered.

7. The intelligent early warning method as described in claim 6, characterized in that, The weighted fusion algorithm is used to calculate the overall risk confidence level. The calculation formula is as follows: ； in, and These are preset weighting coefficients; If we consider the overall risk confidence level If the risk exceeds the preset risk threshold, the risk of the gastric tube being accidentally inserted into the airway is determined to be high, and an audible and visual alarm signal is immediately triggered.

8. The intelligent early warning method as described in claim 6, characterized in that, During the placement of the gastric tube, pressure signals from multiple FBG sensor nodes are continuously collected and analyzed in real time. If an abnormal pressure pattern consistent with gastric tube displacement is detected, a gastric tube displacement warning is triggered.

9. An intelligent early warning method, based on the intelligent gastric tube according to any one of claims 1 to 5, characterized in that, Includes the following steps: Before the planned extubation, at least one feeding test was conducted, during which real-time images acquired by the miniature camera component and pressure signals acquired by multiple FBG sensor nodes were collected simultaneously. Semantic segmentation is performed on the acquired images to identify and segment the food residue areas, and the percentage of pixels in the residue areas relative to the total visible area is calculated. ; Analyze the pressure signal, extract the time domain and frequency domain features of the signal, and determine whether there is an abnormal reflux pressure waveform that matches the characteristics of gastric contents reflux in terms of amplitude, frequency or waveform shape; If the proportion of food residue pixels If the residual safety threshold is lower than the preset threshold and no abnormal backflow pressure waveform is detected, the patient's risk of aspiration is determined to be low, and a suggestion message of "extubation can be considered" is generated. If the proportion of food residue pixels If the residual safety threshold is exceeded, or if an abnormal backflow pressure waveform is detected, it is determined that there is a risk of accidental aspiration, and a "delayed tube removal" warning message is generated.

10. The intelligent early warning method as described in claim 9, characterized in that, After acquiring pressure signals and image data, the risk of aspiration is assessed using the following methods: Extract the temporal features of the pressure signal, input the extracted temporal features into a pre-trained neural network model, and obtain a confidence level indicating that the current pressure signal is in an "abnormal backflow state". The image is semantically segmented to identify the food residue area; based on the area, shape and texture features of the area, a lightweight convolutional neural network classifier is used to output a risk confidence score representing the presence of food residue. The aspiration risk score for a single feeding test is obtained by weighted fusion of two confidence levels. If the average aspiration risk score of multiple consecutive tests exceeds the preset risk threshold, it is determined that there is a high risk of aspiration and a warning message of "delayed extubation" is generated.

Images