Postoperative intelligent monitoring rehabilitation auxiliary instrument for laryngeal pharynx and control method

Abstract

The invention belongs to the technical field of medical rehabilitation equipment, and discloses a postoperative intelligent monitoring rehabilitation auxiliary instrument for the laryngeal pharynx and a control method. The method includes: acquiring pharyngeal physiological data of the pharyngeal of a patient; the method comprises the following steps: constructing a physical information neural network model, inputting a four-channel tensor formed by processing pharyngeal physiological data into the physical information neural network model for analysis, outputting a laryngeal closure degree and a pharyngeal wall tissue elastic modulus, and determining a patient pain level based on the laryngeal closure degree and the pharyngeal wall tissue elastic modulus; on the basis of analysis of mucous membrane dynamic displacement sequence data and pressure data, swallowing mode characteristics of the patient are determined, rehabilitation decision analysis is conducted in combination with the laryngeal closure degree, the pharyngeal wall tissue elasticity modulus and the pain level, and a rehabilitation intervention instruction is generated; the drug administration operation is executed based on the rehabilitation intervention instruction, the execution feedback record after the drug administration operation is completed is obtained, the rehabilitation intervention instruction is adjusted based on the execution feedback record, and the timeliness and effectiveness of a treatment decision are remarkably enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical rehabilitation equipment, more particularly, the present application relates to a kind of intelligent monitoring rehabilitation auxiliary instrument and control method after laryngopharyngeal operation. BACKGROUND

[0002] The rehabilitation monitoring of patient after laryngopharyngeal operation is a complex and critical process, the precise assessment of swallowing function recovery, tissue healing status and pain degree is the core link of clinical management, which is directly related to the prevention of postoperative complications and the optimization of rehabilitation process.

[0003] In the prior art, electronic laryngoscope is combined with patient's subjective report to evaluate the mode after laryngopharyngeal operation, the mucosal swelling, bleeding and wound healing are observed by electronic laryngoscope, and the pain sensation and swallowing difficulty degree described by the patient are combined with clinical experience to judge, but the monitoring of this mode mainly depends on intermittent and point observation, and it is difficult to realize continuous quantitative evaluation of swallowing function, healing process and pain degree, for example, in clinical practice, the physician checks and evaluates the laryngopharynx of the patient after operation by electronic laryngoscope combined with pain scale, but the pain score is affected by the communication ability of the patient, and the dynamic change of swallowing function cannot be continuously tracked, which leads to the need for a long time to adjust the rehabilitation plan, and it is difficult to quantitatively evaluate the dynamic correlation between pain degree and swallowing function, thereby affecting the timeliness and accuracy of rehabilitation intervention.

[0004] In view of this, the present application provides a kind of intelligent monitoring rehabilitation auxiliary instrument and control method after laryngopharyngeal operation to solve the above problems. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a kind of intelligent monitoring rehabilitation auxiliary instrument and control method after laryngopharyngeal operation, comprising: S1, obtaining the physiological data of the pharynx of the patient larynx, the physiological data of the pharynx includes mucosal image data, temperature data, pressure data and mucosal dynamic displacement sequence data; S2, construct a physical information neural network model, the four-channel tensor formed after processing the physiological data of the pharynx is input into the physical information neural network model for analysis, and the output is the degree of laryngeal opening and the elastic modulus of pharyngeal wall tissue, and the pain level of the patient is determined based on the degree of laryngeal opening and the elastic modulus of pharyngeal wall tissue; S3, based on the analysis of mucosal dynamic displacement sequence data and pressure data, determine the swallowing mode characteristics of the patient, and combine the laryngeal opening degree, the elastic modulus of pharyngeal wall tissue and the pain level to analyze the rehabilitation decision, and generate the rehabilitation intervention instruction; S4, based on the rehabilitation intervention instruction, execute the drug administration operation, and obtain the execution feedback record after completing the drug administration operation, and adjust the rehabilitation intervention instruction based on the execution feedback record.

[0006] Further, the method for acquiring the physiological data of the pharynx of the patient includes: Synchronously acquiring mucosal image data, temperature data and pressure data of the pharynx of the patient; Extracting a sequence of continuous image frames corresponding to a swallowing event from the mucosal image data, positioning an anatomical anchor point in the first image, calculating the motion trajectory of the anatomical anchor point in the sequence of continuous image frames, and obtaining mucosal dynamic displacement sequence data; Splicing the mucosal image data, temperature data, pressure data and mucosal dynamic displacement sequence data to form the physiological data of the pharynx.

[0007] Further, the method for constructing a physical information neural network model includes: Constructing a physical information neural network model including an input layer, a hidden layer, a fusion layer and an output layer; the input layer is used to receive a four-channel tensor; the hidden layer is provided with feature extraction channels corresponding to one channel data in the four-channel tensor; the output layer includes two independent fully connected paths.

[0008] Further, the method for outputting the glottal closure degree and the elastic modulus of the pharyngeal wall tissue includes: Processing the mucosal image data, temperature data, mucosal dynamic displacement sequence data and pressure data in the physiological data of the pharynx to obtain a four-channel tensor; Inputting the four-channel tensor into the input layer of the physical information neural network model, and performing convolution operation on the corresponding channel data in the four-channel tensor by the feature extraction channels in the hidden layer to obtain image structure features, displacement behavior features, pressure change features and temperature change features respectively; The fusion layer inputs the image structure features, displacement behavior features, pressure change features and temperature change features after splicing to the two fully connected paths of the output layer respectively, and the two fully connected paths perform fully connected operation respectively to output the glottal closure degree and the elastic modulus of the pharyngeal wall tissue.

[0009] Further, the method for obtaining a four-channel tensor includes: Target region cropping is performed on the sequence of continuous image frames corresponding to the swallowing event, the gray mean sequence of the target region is calculated to form first channel data; the displacement change rate sequence is extracted from the mucosal dynamic displacement sequence data to form second channel data; the resting segment data before the swallowing event occurs is extracted from the pressure data, the average value of the resting segment data is taken as the pressure baseline, and the pressure data is corrected based on the pressure baseline to obtain the dynamic pressure response component corresponding to the corrected swallowing event, forming third channel data; the temperature data is normalized to form fourth channel data; The first channel data, the second channel data, the third channel data and the fourth channel data are integrated to form a four-channel tensor based on the time axis of the first channel data.

[0010] Further, the method for determining the pain level of the patient comprises: The preset closing threshold value and the elastic threshold value are preset, the closing threshold value comprises a first closing threshold value and a second closing threshold value, and the elastic threshold value comprises a first elastic threshold value and a second elastic threshold value. The laryngeal inlet closing degree is compared with the closing threshold value to determine the closing risk level corresponding to the laryngeal inlet closing degree of the patient, and the closing risk level comprises a low closing risk, a medium closing risk and a high closing risk level. The pharyngeal wall tissue elastic modulus is compared with the elastic threshold value to determine the elastic risk level corresponding to the pharyngeal wall tissue elastic modulus of the patient, and the elastic risk level comprises a low elastic risk, a medium elastic risk and a high elastic risk level. The closing risk level corresponding to the laryngeal inlet closing degree and the elastic risk level corresponding to the pharyngeal wall tissue elastic modulus are analyzed by adopting the principle of taking the high value of the risk level to determine the pain level of the patient.

[0011] Further, the method for determining the swallowing mode characteristics of the patient comprises: The mucosa dynamic displacement sequence data is analyzed in units of swallowing events to obtain the displacement velocity peak value, displacement trigger delay time length and contraction duration corresponding to each swallowing event to form a displacement mechanics characteristic sequence; the pre-labeled pharyngeal pressure measurement region is selected from the pressure data, and the peak pressure, pressure rising slope and residual pressure ratio of each pharyngeal pressure measurement region are calculated to form a pressure characteristic sequence. The displacement mechanics characteristic sequence and the pressure characteristic sequence corresponding to the same swallowing event are combined to generate the swallowing mode characteristics.

[0012] Further, the method for generating the rehabilitation intervention instruction comprises: The laryngeal inlet closing degree, the pharyngeal wall tissue elastic modulus, the pain level and the swallowing mode characteristic vector are integrated to form a joint input vector; a preset rehabilitation decision rule library is called, the rehabilitation decision rule library comprises a plurality of decision rules, each decision rule defines a set of threshold conditions and rehabilitation intervention parameters; the joint input vector and the threshold conditions of each decision rule in the rehabilitation decision rule library are matched in sequence, when a first decision rule satisfying all the threshold conditions is matched, the rehabilitation intervention parameters corresponding to the decision rule are extracted to generate the rehabilitation intervention instruction comprising the drug type, the drug dose and the drug timing control parameter.

[0013] Further, the method for adjusting the rehabilitation intervention instruction based on the execution feedback record comprises: The rehabilitation intervention instruction is converted into visual administration guidance information, and a confirmation signal is received after the administration operation is completed;With the receiving time of the confirmation signal as the starting point, the new laryngeal inlet closure degree, pharyngeal wall tissue elastic modulus and pain grade of the patient are collected after reaching the preset evaluation time interval to form an execution feedback record; The laryngeal inlet closure degree, pharyngeal wall tissue elastic modulus and pain grade before the administration operation are compared and analyzed with the execution feedback record, and the laryngeal inlet closure degree change, pharyngeal wall tissue elastic modulus change and pain grade change are calculated and obtained; If all three types of changes reach the corresponding preset improvement threshold, the current rehabilitation intervention instruction is maintained;If any change does not reach the preset improvement threshold, the next decision rule in the rehabilitation decision rule library that meets all threshold conditions is called, a new rehabilitation intervention instruction is generated, and the administration guidance information is updated.

[0014] Further, an intelligent postoperative laryngopharyngeal rehabilitation monitoring and assisting instrument, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement any one of the intelligent postoperative laryngopharyngeal rehabilitation control methods.

[0015] The technical effects and advantages of the intelligent postoperative laryngopharyngeal rehabilitation monitoring and assisting instrument and control method are as follows: 1. The present application collects data such as mucosal image, temperature, pressure and mucosal dynamic displacement in real time through the acquisition system, performs modal fusion and feature extraction on four-channel tensors through the construction of a physical information neural network model, fully captures the postoperative physiological state of the patient, and solves the indirectness and point observation limitations of the existing monitoring;By setting two independent fully connected paths, the laryngeal inlet closure degree and the pharyngeal wall tissue elastic modulus are output respectively, and the pain grade mapping and swallowing mode characteristics are combined to accurately quantify and evaluate the dynamic correlation between pain degree and swallowing function, thereby forming objective, quantifiable and repeatable evaluation indexes, laying a foundation for earlier intervention and more accurate rehabilitation programs, and significantly enhancing the timeliness and effectiveness of treatment decisions; 2. The present application combines the swallowing mode characteristics, and combines the laryngeal inlet closure degree, pharyngeal wall tissue elastic modulus and pain grade as a joint input vector, and dynamically matches with the rehabilitation decision rule library to tailor the exclusive rehabilitation program for the patient, realizes individualized and controllable rehabilitation management, improves safety and rehabilitation efficiency, determines the rehabilitation intervention instruction, and adjusts and updates the rehabilitation intervention instruction combined with the execution feedback record, which can optimize the rehabilitation program in real time according to the recovery of the patient, significantly improves the timeliness and effectiveness of the rehabilitation intervention, effectively reduces the occurrence of postoperative complications, shortens the rehabilitation time, and provides more personalized and thoughtful medical services for the patient. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1A method flow chart of a laryngopharyngeal postoperative intelligent monitoring rehabilitation control method of the present application; Figure 2 A physical information neural network model workflow diagram of the present application; Figure 3 A process schematic diagram for obtaining a four-channel tensor of the present application; Figure 4 A laryngopharyngeal postoperative intelligent monitoring rehabilitation auxiliary instrument of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0018] Embodiment 1 Please refer to Figures 1-3 The main design content of the laryngopharyngeal postoperative intelligent monitoring rehabilitation control method described in the present embodiment is as follows: The existing technology adopts an evaluation mode of electronic laryngoscope combined with patient's subjective report, observes the mucosal swelling, bleeding and wound healing through the electronic laryngoscope, and simultaneously makes a judgment according to the patient's self-described pain sensation and the degree of difficulty in swallowing, and comprehensively judges according to the clinical experience. However, the monitoring of this mode mainly depends on intermittent (such as evaluation at fixed time points or fixed periods), point-like observation (single time point captured information, such as a snapshot of a certain image, a certain pain score), and it is difficult to provide continuous real-time data and realize continuous quantitative evaluation of swallowing function, healing process and pain degree, resulting in a blind area in monitoring the whole process of postoperative recovery of the patient.

[0019] In addition, the pain score and the degree of difficulty in swallowing mainly depend on the patient's self-description, which is easily affected by the communication ability. For example, infants, patients with cognitive disorders and people with limited language ability may not be able to accurately express their feelings, resulting in a lack of objectivity in pain score and affecting the accuracy of the evaluation results.

[0020] Therefore, a laryngopharyngeal postoperative intelligent monitoring rehabilitation control method is designed, which comprises: Step S1, obtaining the physiological data of the pharynx of the patient's larynx, the physiological data of the pharynx including mucosal image data, temperature data, pressure data and mucosal dynamic displacement sequence data.

[0021] The method for obtaining the physiological data of the pharynx of the patient's larynx comprises: Simultaneously, the device collects mucosal imaging data, temperature data, and pressure data from the patient's pharynx. The rehabilitation aid is equipped with a sensor array (including an endoscopic imaging probe, a miniature thermal sensor array, and a flexible pressure-sensing patch). The endoscopic imaging probe is inserted through the mouth, with its tip aimed at the laryngopharynx, and continuously acquires video images of the pharyngeal mucosal surface at a fixed frame rate to form mucosal imaging data. A miniature thermal sensor array is integrated into the outer wall of the endoscopic imaging probe to collect temperature values ​​at each measurement point at a fixed sampling frequency, forming temperature data; a flexible pressure sensing patch is fixed around the outer side of the pharyngeal wall and attached to the inner wall of the pharynx to collect pressure values ​​in each pressure measurement area at a fixed sampling frequency, forming pressure data.

[0022] A sequence of consecutive image frames corresponding to a swallowing event is extracted from mucosal imaging data. A swallowing event refers to a complete swallowing action. For each consecutive image frame in the mucosal imaging data, the structural similarity index between adjacent frames is calculated to obtain a time series of the structural similarity index. The time series of the structural similarity index is then subjected to first-order difference to obtain a sequence of structural similarity change rates. Local minima are detected in the sequence of structural similarity change rates. If the rate of change before and after a local minima changes from negative to positive, and the structural similarity index corresponding to that point is a local minimum, then that frame is determined as the swallowing initiation frame. The frame after the initiation frame where the structural similarity index first reaches a local maximum and the rate of change changes from positive to negative is taken as the swallowing termination frame. Continuous image frames from the swallowing initiation frame to the swallowing termination frame are extracted as the sequence of consecutive image frames corresponding to the swallowing event.

[0023] It should be explained that the structural similarity index is an indicator used to measure the similarity between two images in three dimensions: brightness, contrast, and structural information. Its value ranges from [-1, 1], that is, the closer the value is to 1, the more similar the two images are, and the closer it is to -1, the greater the difference.

[0024] Anatomical anchor points are located in the first frame image, and their motion trajectories in a continuous sequence of image frames are calculated to obtain mucosal dynamic displacement sequence data. Anatomical anchor points refer to key locations with clear anatomical features selected in the mucosal image data (such as the tip of the epiglottis, the apex of the arytenoid cartilage, and the bottom of the piriform fossa). Using the position of the anatomical anchor point in the first frame image as the starting point for tracking, the displacement vector of the anatomical anchor point in each frame image is calculated to form a motion trajectory. The displacement components of the anatomical anchor point in the horizontal and vertical directions are calculated separately, and the average value of the displacement components of the anatomical anchor point in each frame is taken to generate mucosal dynamic displacement sequence data that is time-aligned with the image frame sequence.

[0025] Pharyngeal physiological data are formed by stitching together mucosal imaging data, temperature data, pressure data, and mucosal dynamic displacement sequence data.

[0026] S2. Construct a physical information neural network model. Input the four-channel tensor formed by processing the pharyngeal physiological data into the physical information neural network model for analysis. Output the laryngeal hilum closure degree and the elastic modulus of the pharyngeal wall tissue. Determine the patient's pain level based on the laryngeal hilum closure degree and the elastic modulus of the pharyngeal wall tissue. Methods for constructing physical information neural network models include: Construct a physical information neural network model that includes an input layer, a hidden layer, a fusion layer, and an output layer; the input layer is used to receive a four-channel tensor; the hidden layer is set with feature extraction channels, each corresponding to one channel of data in the four-channel tensor; the output layer contains two independent fully connected paths.

[0027] Methods for measuring laryngeal hilum closure degree and pharyngeal wall tissue elastic modulus include: The mucosal imaging data, temperature data, mucosal dynamic displacement sequence data, and pressure data in the pharyngeal physiological data were processed separately to obtain a four-channel tensor. The four-channel tensor is input into the input layer of the physical information neural network model. The feature extraction channels in the hidden layer perform convolution operations on the corresponding channel data in the four-channel tensor to obtain image structure features, displacement behavior features, pressure change features, and temperature change features, respectively.

[0028] Among them, image structural features characterize the morphology, contour continuity, and spatial patterns of structural deformation during swallowing of the mucosal surface; displacement behavior features characterize the movement patterns, contraction synchronicity, and dynamic response characteristics of mucosal tissue during swallowing; pressure change features characterize the intensity, duration, and regional coordination of pharyngeal pressure waves, used to reflect swallowing propulsion force and residual risk; temperature change features characterize the dynamics of heat distribution caused by local tissue metabolic state, inflammatory response, or changes in blood perfusion.

[0029] The input layer separates the received four-channel tensor into channels and feeds them into the four feature extraction channels in the hidden layer. The first feature extraction channel performs three convolution operations on the gray-level mean sequence in sequence: the first convolution slides and scans the spatial and temporal neighborhood of the gray-level mean sequence to extract local gray-level change patterns; the second convolution further extracts edge and texture combination features from the local gray-level change patterns; and the third convolution captures higher-order spatial structure associations to form image structure features.

[0030] The second feature extraction channel performs a three-layer convolution operation on the displacement change rate sequence: the first layer convolution detects local abrupt changes in the displacement change rate over time; the second layer convolution identifies continuous regions of displacement acceleration or deceleration; and the third layer convolution integrates the dynamic trends of displacement across multiple frames to form a comprehensive representation of the timing and amplitude of mucosal contraction, outputting displacement behavior features.

[0031] The third feature extraction channel performs three convolution operations on the dynamic pressure response components: the first convolution captures the rising edge and peak region of the pressure response components in the time dimension; the second convolution identifies the duration and fall slope of the pressure waveform; and the third convolution models the spatial distribution characteristics of the coordinated pressure changes in multiple regions and outputs the pressure change characteristics.

[0032] The fourth feature extraction channel performs a three-layer convolution operation on the normalized temperature data: the first layer convolution detects slow drifts or abrupt changes in temperature in the time series; the second layer convolution identifies the continuous trend and local stationary segments of temperature changes; and the third layer convolution fuses temperature consistency information within the spatial neighborhood to enhance sensitivity to areas of inflammation or blood flow changes.

[0033] The fusion layer concatenates image structural features, displacement behavior features, pressure change features, and temperature change features, and then inputs them to two fully connected paths in the output layer. Each of the two fully connected paths performs a fully connected operation, outputting the laryngeal hiatus closure degree and the pharyngeal wall elastic modulus. The laryngeal hiatus closure degree represents the degree to which the laryngeal hiatus closes during swallowing, and is an important indicator of the larynx's ability to prevent food or liquid from entering the airway during swallowing. The pharyngeal wall elastic modulus represents the elastic properties of the pharyngeal wall tissue, that is, the tissue's ability to deform under stress, and is a biomechanical indicator for assessing tissue edema or healing status.

[0034] The fusion layer concatenates the features sequentially to form a fused feature vector, which is then input into two independent fully connected paths. In each of the two fully connected paths, at least one fully connected operation is performed on the fused feature vector, and finally, two scalar values ​​are output. The scalar value output by the first fully connected path is used as the laryngeal hilum closure degree, and the other scalar value is used as the elastic modulus of the pharyngeal wall tissue.

[0035] It should be explained that a higher degree of laryngeal closure indicates more complete closure of the laryngeal opening and a lower risk of aspiration; a lower degree of laryngeal closure indicates incomplete closure of the laryngeal opening and a higher risk of aspiration or choking.

[0036] It should be explained that the higher the value of the elastic modulus of the pharyngeal wall tissue, the harder and less elastic the pharyngeal wall tissue is, which usually indicates the presence of fibrosis or scar formation. The lower the value, the softer and more elastic the tissue is, reflecting that the tissue is in the early stage of edema or healing.

[0037] Methods for obtaining four-channel tensors include: The sequence of consecutive image frames corresponding to a swallowing event is cropped into a target region, and the gray-scale mean sequence of the target region is calculated to form the first channel data. The target region is used to represent a selected specific image region containing the pharyngeal mucosa. The target region must cover anatomical structures closely related to swallowing function (such as the tip of the epiglottis, the apex of the arytenoid cartilage, the bottom of the piriform fossa, etc.). The gray-scale mean of the target region is calculated for each cropped image frame to form a numerical sequence equal to the number of image frames. For example, if the swallowing event contains 25 images, the first channel data is a one-dimensional sequence of length 25, with each element corresponding to the gray-scale mean of one frame.

[0038] The displacement change rate sequence is extracted from the mucosal dynamic displacement sequence data to form the second channel data. The displacement change rate is obtained by performing a first-order difference calculation on the mucosal dynamic displacement sequence data, that is, subtracting the mucosal dynamic displacement values ​​at adjacent time points to obtain the displacement change amount within each time interval. To ensure that the second channel data has the same number of time points as the first channel data, the displacement change rate of the first frame is set to 0, and the remaining time points are successively subtracted to form a displacement change rate sequence with a length consistent with the number of frames of the swallowing event. For example, if the mucosal dynamic displacement sequence data contains mucosal dynamic displacement values ​​at 25 time points, then after the first-order difference, 24 changes are obtained. After setting the rate of the first frame to 0, the length of the second channel data is 25.

[0039] The pressure data is extracted from the resting interval data preceding the swallowing event. The average value of the resting interval data is used as the pressure baseline, and the pressure data is corrected based on this baseline to obtain the corrected dynamic pressure response components corresponding to the swallowing event, forming the third channel data. The resting interval represents a continuous time interval without swallowing action preceding the swallowing initiation frame, and is determined by maintaining stability through a structural similarity index. The pressure baseline is subtracted point-by-point from the pressure data within the time period corresponding to the swallowing event to obtain the dynamic pressure response components; for example, if the swallowing event corresponds to 25 pressure sampling points, 25 dynamic pressure response values ​​are obtained after correction, reflecting the pressure change relative to the resting state.

[0040] Temperature data is normalized to form the fourth channel data. The normalization method uses min-max scaling to map the temperature data within the time period corresponding to the swallowing event to the interval between 0 and 1; for example, if the original temperature range is 35.2℃ to 36.8℃, after normalization all values ​​fall within the interval [0, 1], preserving the original trend.

[0041] Using the timeline of the first channel data as a reference, the first, second, third, and fourth channel data are integrated to form a four-channel tensor. The data from each channel are aligned with the timeline of the first channel data, and the aligned four channel data are combined into a four-dimensional vector at each time point, forming a four-channel tensor of size N×4, where N is the number of image frames corresponding to the swallowing event.

[0042] Methods for determining a patient's pain level include: A preset closure threshold and an elastic threshold are defined. The closure threshold includes a first closure threshold and a second closure threshold, and the elastic threshold includes a first elastic threshold and a second elastic threshold. The first closure threshold is less than the second closure threshold; the first elastic threshold is less than the second elastic threshold.

[0043] It should be explained that the closure threshold and elasticity threshold were determined based on clinical observation data, literature data, and the results of large-sample statistical analysis.

[0044] The laryngeal hilum closure degree is compared with the closure threshold to determine the closure risk level corresponding to the patient's laryngeal hilum closure degree. The closure risk level includes low closure risk, medium closure risk, and high closure risk. The comparison rules are as follows: if the laryngeal hilum closure degree is less than the first closure threshold, the closure risk level is high, indicating severe laryngeal hilum closure insufficiency, poor airway protection, and a high risk of pain and coughing; if the laryngeal hilum closure degree is greater than or equal to the first closure threshold but less than the second closure threshold, the closure risk level is medium; if the laryngeal hilum closure degree is greater than or equal to the second closure threshold, the closure risk level is low, indicating adequate laryngeal hilum closure and good function.

[0045] The elastic modulus of the pharyngeal wall tissue is compared with an elastic threshold to determine the elastic risk level corresponding to the patient's pharyngeal wall tissue elastic modulus. The elastic risk level includes low elastic risk, medium elastic risk, and high elastic risk levels. The comparison rules are as follows: if the elastic modulus of the pharyngeal wall tissue is less than the first elastic threshold, the elastic risk level is low elastic risk level, indicating that the tissue is in the early stage of edema or healing, and the risk of pain is relatively low; if the elastic modulus of the pharyngeal wall tissue is greater than or equal to the first elastic threshold and less than the second elastic threshold, the elastic risk level is medium elastic risk level; if the elastic modulus of the pharyngeal wall tissue is greater than or equal to the second elastic threshold, the elastic risk level is high elastic risk level, indicating tissue fibrosis or scar formation, which is closely related to pain.

[0046] Using the principle of selecting the highest risk level, the closure risk level corresponding to the degree of laryngeal closure and the elastic risk level corresponding to the elastic modulus of the pharyngeal wall tissue were analyzed to determine the patient's pain level. A risk level mapping relationship was established, assigning numerical levels to the closure risk level and the elastic risk level: low risk level corresponds to value 1, medium risk level corresponds to value 2, and high risk level corresponds to value 3. The numerical levels corresponding to the closure risk level and the elastic risk level were compared, and the larger value was taken as the final risk level. Based on this, the patient's pain level was determined: value 1 corresponds to mild pain, value 2 corresponds to moderate pain, and value 3 corresponds to severe pain.

[0047] For example, the first closure threshold is set to 0.4, and the second closure threshold is set to 0.7; the first elasticity threshold is set to 8 kPa, and the second elasticity threshold is set to 15 kPa; the laryngeal hilum closure degree is 0.35, which is less than the first closure threshold of 0.4, corresponding to a high closure risk level (mapped risk level value 3); the pharyngeal wall tissue elastic modulus is 6 kPa, which is less than the first elasticity threshold of 8 kPa, corresponding to a low elasticity risk level (mapped risk level value 1); the larger value of 3 is taken to determine the patient's pain level as severe pain.

[0048] S3. Based on the analysis of mucosal dynamic displacement sequence data and pressure data, determine the characteristics of the patient's swallowing pattern, and combine the laryngeal hilum closure degree, pharyngeal wall tissue elastic modulus and pain level to conduct rehabilitation decision analysis and generate rehabilitation intervention instructions.

[0049] Methods for determining a patient's swallowing pattern characteristics include: The mucosal dynamic displacement sequence data is analyzed on a per-swallowing-event basis to obtain the peak displacement velocity, displacement trigger delay, and contraction duration corresponding to each swallowing event, forming a displacement dynamics feature sequence. Each swallowing event corresponds to a continuous segment of mucosal dynamic displacement sequence data, the time range of which is determined by the swallowing start frame to the swallowing end frame. The first-order difference of the mucosal dynamic displacement sequence data corresponding to each swallowing event is performed to calculate the displacement velocity sequence, and the peak displacement velocity (representing the maximum absolute value in the displacement velocity sequence) is obtained from the displacement velocity sequence.

[0050] The displacement trigger delay duration represents the time interval from the swallowing initiation frame to the frame before the first inflection point in the displacement velocity sequence where the local acceleration changes from positive to negative. This inflection point corresponds to the starting moment when the displacement velocity changes from accelerating upward to decelerating, reflecting the initial response of the tissue starting to actively contract. For example, if the swallowing initiation frame is frame 1 and the displacement velocity reaches the local upward inflection point in frame 4, then the displacement trigger delay duration is 3 frames.

[0051] The contraction duration represents the time span from the frame where the velocity rises to an inflection point in the displacement velocity sequence to the starting frame where the velocity falls back to a local steady state; for example, if the velocity rises to an inflection point in frame 4 and the velocity enters a steady state after frame 18, then the contraction duration is 14 frames.

[0052] The peak displacement velocity, displacement trigger delay, and contraction duration are combined into a triplet as the displacement dynamics feature of the swallowing event. The steps are repeated for all swallowing events to form a displacement dynamics feature sequence.

[0053] Pre-calibrated pharyngeal pressure measurement regions were selected from the pressure data. The peak pressure, pressure rise slope, and residual pressure ratio were calculated for each region to form a pressure characteristic sequence. The pharyngeal pressure measurement regions included three fixed measurement points: the epipharynx, midpharynx, and hypopharynx. For each region, pressure waveform data was extracted within the time period corresponding to the swallowing event. Using the swallowing start frame to swallowing end frame as a time window, pressure sampling values ​​within this window were extracted as the pressure response sequence for that region. The peak pressure was represented by the maximum value in the pressure response sequence.

[0054] The pressure rise slope is expressed as the ratio of the pressure change from the initial pressure rise point to the peak pressure to the time interval; the residual pressure ratio is expressed as the ratio of the pressure value in the next frame after the swallow termination frame to the peak pressure. The displacement dynamics feature sequence and pressure feature sequence corresponding to the same swallowing event are combined, and this combination is performed on all swallowing events to generate swallowing pattern features.

[0055] Methods for generating rehabilitation intervention instructions include: The laryngeal hilum closure degree, pharyngeal wall tissue elastic modulus, pain level, and swallowing pattern feature vectors are integrated to form a joint input vector. A preset rehabilitation decision rule library is invoked, which contains several decision rules. Each decision rule defines a set of threshold conditions and rehabilitation intervention parameters. The threshold conditions include upper and lower limits of constraints on each component in the joint input vector, and the rehabilitation intervention parameters include drug administration type, drug dosage, and drug administration timing control parameters.

[0056] The joint input vector is sequentially matched with the threshold conditions of each decision rule in the rehabilitation decision rule base. The matching order is as follows: starting from the first decision rule, it is determined whether the joint input vector satisfies all the threshold conditions in that decision rule. The determination method is: for each condition component, check whether the value at the corresponding position of the joint input vector falls within the range specified by the condition; for example, if the laryngeal closure degree of 0.35 is ≥0.7 (no), then the first decision rule is not satisfied, and the matching continues to the second decision rule.

[0057] When the first decision rule that satisfies all threshold conditions is matched, the rehabilitation intervention parameters corresponding to that decision rule are extracted, and a rehabilitation intervention instruction containing drug type, dosage, and dosing timing control parameters is generated. The drug type records the drug name or formulation type, the dosage records the volume or mass value of a single dose, and the dosing timing control parameter field records the dosing frequency and time interval; for example, the generated rehabilitation intervention instruction is {Drug type: "Dexamethasone nebulizer solution", Dosage: "2 mL", Dosing timing control parameter: "3 times daily, 6 hours apart"}.

[0058] S4. Execute drug administration based on rehabilitation intervention instructions, obtain execution feedback records after drug administration is completed, and adjust rehabilitation intervention instructions based on execution feedback records.

[0059] Methods for adjusting rehabilitation intervention instructions based on execution feedback records include: The rehabilitation intervention instructions are transformed into visualized drug administration guidance information, and a confirmation signal is received upon completion of the drug administration procedure. The visualized drug administration guidance information is presented in the form of a graphical interface. After the patient or medical staff completes the drug administration procedure using the guidance information, they trigger a confirmation signal by clicking the confirmation button on the operating terminal (the confirmation signal includes a timestamp, recording the specific time the drug administration procedure was completed; for example, if the operator clicks the "Drug Administration Completed" button at 20:05, a confirmation signal with the timestamp "20:05" is generated).

[0060] Starting from the moment the confirmation signal is received, after the preset assessment time interval is reached, the patient's new laryngeal closure degree, pharyngeal wall tissue elastic modulus and pain level are collected to form an execution feedback record.

[0061] It should be explained that the preset assessment interval is determined comprehensively based on the patient's specific condition, type of surgery, and postoperative recovery stage. For patients with milder conditions, less surgical trauma, and faster recovery, the preset assessment interval can be relatively shorter; for patients with more severe conditions, complex surgery, or slow postoperative recovery, the preset assessment interval may be appropriately extended, such as setting it to be assessed once a day or every two days, to avoid placing an unnecessary burden on the patient due to overly frequent assessments, while still effectively monitoring the patient's recovery progress.

[0062] The laryngeal hilum closure degree, pharyngeal wall elastic modulus, and pain level before the drug administration procedure were compared and analyzed with the execution feedback records. The changes in laryngeal hilum closure degree, pharyngeal wall elastic modulus, and pain level were calculated and obtained. Specifically, the difference between the laryngeal hilum closure degree before the drug administration procedure and the laryngeal hilum closure degree in the execution feedback records was calculated to obtain the change in laryngeal hilum closure degree; the difference between the pharyngeal wall elastic modulus before the drug administration procedure and the pharyngeal wall elastic modulus in the execution feedback records was calculated to obtain the change in pharyngeal wall elastic modulus; and the difference between the pain level before the drug administration procedure and the pain level in the execution feedback records was calculated using numerical coding to obtain the change in pain level.

[0063] If all three types of changes reach their corresponding preset improvement thresholds, the current rehabilitation intervention instructions will be maintained. The preset improvement thresholds are the minimum effective improvement amounts determined based on clinical rehabilitation guidelines, postoperative recovery standards, and statistical analysis of historical patient data. Among them, the laryngeal hilum closure improvement threshold is a positive value, indicating the minimum extent to which laryngeal hilum closure function needs to be improved; the pharyngeal wall tissue elastic modulus improvement threshold is a negative value, indicating the minimum degree to which tissue elasticity needs to be softened; and the pain level improvement threshold is a negative value, indicating the minimum level to which the subjective perception of pain needs to be relieved.

[0064] If any change does not reach the preset improvement threshold, the next decision rule in the rehabilitation decision rule base that satisfies all threshold conditions is invoked to generate a new rehabilitation intervention instruction and update the medication guidance information. Starting from the next decision rule corresponding to the current rehabilitation intervention instruction, the joint input vector is sequentially matched with the threshold conditions of each decision rule until the first decision rule that satisfies all threshold conditions is found. The new rehabilitation intervention instruction is then transformed into visualized medication guidance information, and the medication guidance content displayed on the operation terminal is updated.

[0065] For example: Before drug administration {laryngeal hilum closure degree: 0.35, pharyngeal wall elastic modulus: 18, pain level: severe pain}; after executing the feedback record {laryngeal hilum closure degree: 0.52, pharyngeal wall elastic modulus: 14, pain level: moderate pain}; then the change in laryngeal hilum closure degree is +0.17; the change in pharyngeal wall elastic modulus is -4 kPa; and the change in pain level is -1. The threshold for improvement in laryngeal closure is +0.15, the threshold for improvement in pharyngeal wall elastic modulus is −3 kPa, and the threshold for improvement in pain level is −1. Therefore, the change in laryngeal closure is +0.17 ≥ +0.15 (satisfied), the change in pharyngeal wall elastic modulus is −4 ≤ −3 (satisfied), and the change in pain level is −1 ≤ −1 (satisfied). All three changes have reached the preset improvement thresholds, and the current rehabilitation intervention instruction is maintained.

[0066] In this embodiment, the acquisition system monitors mucosal images, temperature, pressure, and dynamic displacement of the mucosa in real time. By constructing a physical information neural network model, modal fusion and feature extraction are performed on the four-channel tensor to comprehensively capture the patient's postoperative physiological state, overcoming the limitations of indirect and point-like observation in existing monitoring. By setting up two independent fully connected paths to output the laryngeal hilum closure degree and pharyngeal wall tissue elastic modulus respectively, and combining pain level mapping and swallowing pattern characteristics, the dynamic correlation between pain intensity and swallowing function is accurately quantified and assessed, thereby forming objective, quantifiable, and repeatable assessment indicators. This lays the foundation for earlier intervention and more precise rehabilitation plans, significantly enhancing the timeliness and effectiveness of treatment decisions. By combining swallowing pattern features with laryngeal hilum closure degree, pharyngeal wall elastic modulus, and pain level as joint input vectors, and dynamically matching them with a rehabilitation decision rule base, a personalized rehabilitation plan can be tailored for each patient. This enables individualized and controllable rehabilitation management, improves safety and rehabilitation efficiency, determines rehabilitation intervention instructions, and adjusts and updates these instructions based on execution feedback records. The rehabilitation plan can be optimized in real time based on the patient's recovery status, significantly improving the timeliness and effectiveness of rehabilitation intervention. This can effectively reduce the occurrence of postoperative complications, shorten rehabilitation time, and provide patients with more personalized and considerate medical services.

[0067] Example 2 Please see Figure 4 As shown, a postoperative intelligent monitoring and rehabilitation aid for laryngopharyngeal surgery includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, it implements a postoperative intelligent monitoring and rehabilitation control method for laryngopharyngeal surgery.

[0068] It should be explained that the intelligent monitoring and rehabilitation aid for postoperative laryngopharyngeal surgery is shaped like a "pacifier". The front end of the intelligent monitoring and rehabilitation aid for postoperative laryngopharyngeal surgery is equipped with multiple types of sensors for data collection, and the middle part is equipped with an empty tube, which can be used to deliver medicine into the mouth.

[0069] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0070] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0071] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

[0072] In conclusion, 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 method for intelligent monitoring and rehabilitation control after laryngopharyngeal surgery, characterized in that, include: S1. Obtain pharyngeal physiological data from the patient's throat, including mucosal imaging data, temperature data, pressure data, and mucosal dynamic displacement sequence data. S2. Construct a physical information neural network model. Input the four-channel tensor formed by processing the pharyngeal physiological data into the physical information neural network model for analysis. Output the laryngeal hilum closure degree and the elastic modulus of the pharyngeal wall tissue. Determine the patient's pain level based on the laryngeal hilum closure degree and the elastic modulus of the pharyngeal wall tissue. S3. Based on the analysis of mucosal dynamic displacement sequence data and pressure data, determine the characteristics of the patient's swallowing pattern, and combine the laryngeal hilum closure degree, pharyngeal wall tissue elastic modulus and pain level to conduct rehabilitation decision analysis and generate rehabilitation intervention instructions. S4. Execute drug administration based on rehabilitation intervention instructions, obtain execution feedback records after drug administration is completed, and adjust rehabilitation intervention instructions based on execution feedback records.

2. The intelligent monitoring and rehabilitation control method for postoperative laryngopharyngeal surgery according to claim 1, characterized in that, The method for obtaining pharyngeal physiological data of a patient's throat includes: Simultaneously collect mucosal imaging data, temperature data, and pressure data of the patient's pharynx; A sequence of consecutive image frames corresponding to swallowing events is extracted from mucosal imaging data. Anatomical anchor points are located in the first image frame. The motion trajectory of the anatomical anchor points in the sequence of consecutive image frames is calculated to obtain mucosal dynamic displacement sequence data. Pharyngeal physiological data are formed by stitching together mucosal imaging data, temperature data, pressure data, and mucosal dynamic displacement sequence data.

3. The intelligent monitoring and rehabilitation control method for postoperative laryngopharyngeal surgery according to claim 1, characterized in that, The method for constructing a physical information neural network model includes: Construct a physical information neural network model that includes an input layer, a hidden layer, a fusion layer, and an output layer; the input layer is used to receive a four-channel tensor; the hidden layer is set with feature extraction channels, each corresponding to one channel of data in the four-channel tensor; the output layer contains two independent fully connected paths.

4. The intelligent monitoring and rehabilitation control method for postoperative laryngopharyngeal surgery according to claim 3, characterized in that, The method for outputting the laryngeal hilum closure degree and the elastic modulus of the pharyngeal wall tissue includes: The mucosal imaging data, temperature data, mucosal dynamic displacement sequence data, and pressure data in the pharyngeal physiological data were processed separately to obtain a four-channel tensor. The four-channel tensor is input into the input layer of the physical information neural network model. The feature extraction channels in the hidden layer perform convolution operations on the corresponding channel data in the four-channel tensor to obtain image structure features, displacement behavior features, pressure change features and temperature change features respectively. The fusion layer concatenates image structural features, displacement behavior features, pressure change features, and temperature change features, and then inputs them to two fully connected paths of the output layer. Each of the two fully connected paths performs a fully connected operation and outputs the laryngeal hilum closure degree and the elastic modulus of the pharyngeal wall tissue.

5. The intelligent monitoring and rehabilitation control method for postoperative laryngopharyngeal surgery according to claim 4, characterized in that, The method for obtaining a four-channel tensor includes: The first channel data is generated by cropping the target region from the continuous image frame sequence corresponding to the swallowing event and calculating the gray-scale mean sequence of the target region. The second channel data is generated by extracting the displacement change rate sequence from the mucosal dynamic displacement sequence data. The third channel data is generated by extracting the resting segment data before the swallowing event from the pressure data, using the average value of the resting segment data as the pressure baseline, and correcting the pressure data based on the pressure baseline to obtain the dynamic pressure response component corresponding to the swallowing event after correction. The fourth channel data is generated by normalizing the temperature data. Using the time axis of the first channel data as a reference, the first channel data, the second channel data, the third channel data, and the fourth channel data are integrated to form a four-channel tensor.

6. The intelligent monitoring and rehabilitation control method for postoperative laryngopharyngeal surgery according to claim 4, characterized in that, The methods for determining a patient's pain level include: The system includes a preset closure threshold and an elastic threshold. The closure threshold includes a first closure threshold and a second closure threshold, and the elastic threshold includes a first elastic threshold and a second elastic threshold. The laryngeal hilum closure degree is compared with the closure threshold to determine the closure risk level corresponding to the patient's laryngeal hilum closure degree. The closure risk level includes low closure risk, medium closure risk and high closure risk. The elastic modulus of the pharyngeal wall tissue is compared with the elasticity threshold to determine the elasticity risk level corresponding to the patient's pharyngeal wall tissue elasticity modulus. The elasticity risk level includes low elasticity risk, medium elasticity risk and high elasticity risk. By adopting the principle of taking the higher risk level, the closure risk level corresponding to the laryngeal hilum closure degree and the elastic risk level corresponding to the elastic modulus of the pharyngeal wall tissue were analyzed to determine the patient's pain level.

7. The intelligent monitoring and rehabilitation control method for postoperative laryngopharyngeal surgery according to claim 1, characterized in that, The method for determining the characteristics of a patient's swallowing pattern includes: The mucosal dynamic displacement sequence data is analyzed on a per-swallowing-event basis to obtain the peak displacement velocity, displacement trigger delay time and contraction duration corresponding to each swallowing event, forming a displacement dynamic characteristic sequence; pre-calibrated pharyngeal pressure measurement areas are selected from the pressure data, and the peak pressure, pressure rise slope and residual pressure ratio of each pharyngeal pressure measurement area are calculated to form a pressure characteristic sequence; The displacement dynamics feature sequence and pressure feature sequence corresponding to the same swallowing event are combined to generate swallowing pattern features.

8. The intelligent monitoring and rehabilitation control method for postoperative laryngopharyngeal surgery according to claim 1, characterized in that, The method for generating rehabilitation intervention instructions includes: The laryngeal hilum closure degree, pharyngeal wall tissue elastic modulus, pain level, and swallowing pattern feature vectors are integrated to form a joint input vector. A preset rehabilitation decision rule library is invoked, which contains several decision rules. Each decision rule defines a set of threshold conditions and rehabilitation intervention parameters. The joint input vector is matched sequentially with the threshold conditions of each decision rule in the rehabilitation decision rule library. When the first decision rule that satisfies all threshold conditions is matched, the rehabilitation intervention parameters corresponding to the decision rule are extracted to generate a rehabilitation intervention instruction that includes drug administration type, drug administration dosage, and drug administration timing control parameters.

9. The intelligent monitoring and rehabilitation control method for postoperative laryngopharyngeal surgery according to claim 8, characterized in that, The method for adjusting rehabilitation intervention instructions based on execution feedback records includes: The rehabilitation intervention instructions are transformed into visualized drug administration guidance information, and a confirmation signal is received after the drug administration operation is completed. Starting from the moment the confirmation signal is received, after the preset assessment time interval is reached, the patient's new laryngeal closure degree, pharyngeal wall tissue elastic modulus and pain level are collected to form an execution feedback record. The laryngeal hilum closure degree, pharyngeal wall elastic modulus and pain level before the drug administration were compared and analyzed with the execution feedback records to calculate and obtain the changes in laryngeal hilum closure degree, pharyngeal wall elastic modulus and pain level. If all three types of changes reach the corresponding preset improvement threshold, the current rehabilitation intervention instruction is maintained; if any change does not reach the preset improvement threshold, the next decision rule in the rehabilitation decision rule base that satisfies all threshold conditions is called to generate a new rehabilitation intervention instruction and update the medication guidance information.

10. A postoperative intelligent monitoring and rehabilitation aid for laryngopharyngeal surgery, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a postoperative intelligent monitoring and rehabilitation control method for the laryngopharynx as described in any one of claims 1 to 9.

Images