Nursing device for improving lung function after thoracic surgery

Abstract

The invention belongs to the technical field of lung function nursing, and discloses a nursing device for improving lung function after thoracic surgery, comprising: a back plate, two sides of the lower end of the back plate are respectively provided with an air filter box and a waste liquid inbox; an outlet of the air filter box is connected with a breathing mask through a hose; a mouthpiece is fixedly arranged on the breathing mask; a suction pump is arranged on the waste liquid inbox, the input end of the suction pump is communicated with a sputum suction pipe, and the sputum suction pipe is in sliding connection with the breathing mask; elastic ropes are fixedly arranged on the two sides of the lower end of the back plate, handles are fixedly arranged at the outer ends of the elastic ropes, and a monitoring mechanism is fixedly arranged on the back plate. The device can be conveniently used when a patient lies in bed, the device can be fixed to the back through the shoulder buckle belt and the waist buckle belt, the patient can hold the handles with two hands to stretch the elastic rope to conduct chest expanding training, the comprehensive evaluation module comprehensively evaluates the physical condition and the lung function condition of the patient, the drainage monitoring module monitors the drainage condition, and the risk prediction module conducts risk prediction.

Description

Technical Field

[0001] This application relates to the technical field of pulmonary function care, and more specifically, to a nursing device for improving pulmonary function after thoracic surgery. Background Art

[0002] After thoracic surgery, pulmonary function care is a crucial and indispensable important link in the entire rehabilitation process. Thoracic surgery, due to the surgical operation area being adjacent to the lungs, inevitably has a certain impact on lung tissue and related functions. At this time, effective pulmonary function care is particularly important, and its core goal is to help patients recover pulmonary function as soon as possible and minimize the occurrence probability of postoperative complications.

[0003] The prior art document with the publication number CN118001697A provides a nursing device for improving pulmonary function after thoracic surgery; the device includes a nursing gown, and further includes: a fixed box fixedly connected to the nursing gown, and a purification chamber is arranged inside the fixed box; through the setting of cleaning fluff, when the patient inhales, the cleaning fluff can adsorb dust-like pollutants in the air, improving the cleanliness of the air, thereby preventing the dust-like pollutants in the air from contaminating the patient's lungs and affecting the efficiency of pulmonary function care for the patient, thus improving the quality of pulmonary function care for the patient. When the person exhales, since the exhaled gas contains some water vapor, the water vapor will be adsorbed by the cleaning fluff, enabling the cleaning fluff to adsorb more dust-like impurities, and can humidify the gas when the person inhales, thereby further improving the quality of pulmonary function care for the patient.

[0004] Although the existing technical solutions mentioned above can achieve relevant beneficial effects through the structures of the existing technologies, they still have the following defects: 1. Most traditional rehabilitation training devices can only be used in specific areas within the ward and are difficult to move. It is inconvenient for patients to use when lying in bed, and it is even impossible to carry when getting out of bed or undergoing outdoor rehabilitation training, greatly restricting the flexibility of patients' rehabilitation training. 2. The existing monitoring means mainly focus on the collection of physiological data and lack the ability to accurately collect and analyze images of abnormal conditions on the patient's face and upper body, such as pain expressions and abnormal breathing postures. It is difficult for medical staff to timely and accurately judge the patient's physical condition through intuitive physical manifestations, which may lead to some potential problems not being discovered and dealt with in a timely manner. 3. The assessment of the patient's physical condition and lung function status after thoracic surgery mostly relies on single physiological data and fails to integrate multi-dimensional information such as physiological data and image analysis results. It is impossible to comprehensively judge the patient's rehabilitation progress from multiple angles, easily resulting in one-sidedness of the assessment results. 4. It is difficult for the existing technologies to combine various data such as the patient's condition, physiological data, body appearance images, and drainage monitoring to predict the risks of the rehabilitation training situation. When the patient is undergoing rehabilitation training, it is impossible to identify in advance the possible risks faced, and it is difficult for medical staff to formulate countermeasures in advance, and the safety of the patient during the rehabilitation training process cannot be fully guaranteed.

[0005] In view of this, we propose a nursing device for improving lung function after thoracic surgery. Summary of the Invention

[0006] 1. Technical problems to be solved

[0007] The purpose of this application is to provide a nursing device for improving lung function after thoracic surgery, which solves the technical problems mentioned in the above background technology and achieves the technical effects that it can be hung on the hospital bed armrest through a hook, facilitating the use of patients when lying in bed; and can be fixed on the back with the help of shoulder straps and waist straps to meet the needs of patients getting out of bed or undergoing outdoor rehabilitation training. The patient can perform chest expansion training by holding the handle with both hands to stretch the elastic rope. The comprehensive evaluation module integrates physiological data and image analysis results to comprehensively evaluate the patient's physical condition and lung function status after thoracic surgery. The drainage monitoring module real-time collects and analyzes data such as the pressure, drainage volume, and color of the drainage bag, and can timely detect abnormal conditions during the wound drainage process. The risk prediction module combines the patient's condition, physiological data, image analysis results, and drainage monitoring data to predict the risks of the rehabilitation training situation and identify in advance the possible risks faced by the patient during the rehabilitation training process.

[0008] 2. Technical solutions

[0009] The technical solution of this application provides a nursing device for improving lung function after thoracic surgery, including: a backboard, a breathing mask, an air filtration box, an elastic rope, a handle, a waste liquid collection box, a suction pump, and a monitoring mechanism;

[0010] Shoulder straps are fixedly arranged on both sides of the upper end of the backboard, and waist straps are fixedly arranged on both sides of the lower end of the backboard; a hook is fixedly arranged on the upper end of the backboard.

[0011] An air filtration box and a waste liquid collection box are respectively fixedly arranged on both sides of the lower end of the backboard;

[0012] An air filter screen and activated carbon are placed in the air filtration box for adsorbing, dust-removing, and filtering the air; the outlet of the air filtration box is connected to the breathing mask through a hose; a bite valve is detachably and fixedly arranged on the breathing mask; the housing of the breathing mask is made of transparent plastic material.

[0013] A suction pump is fixedly arranged on the waste liquid collection box, the input end of the suction pump is communicated with a sputum suction tube, and the sputum suction tube is slidably connected to the breathing mask; the output end of the suction pump extends into the waste liquid collection box. The waste liquid collection box is made of transparent medical-grade plastic material.

[0014] Channels are opened on both sides of the lower end of the backboard, elastic ropes are fixedly arranged in the channels, and handles are fixedly arranged at the outer ends of the elastic ropes, and chest expansion training can be carried out by stretching the elastic ropes and the handles. The elastic ropes are made of high-strength elastic materials.

[0015] A monitoring mechanism is fixedly arranged on the backboard, and the monitoring mechanism monitors the physical condition of the patient after thoracic surgery and discovers abnormal conditions in time.

[0016] Through the above technical solution, when the patient lies on the hospital bed, the backboard can be hung on the armrest of the hospital bed through the hook. When getting out of bed or going outdoors for rehabilitation training, the backboard can be fixed to the back through the shoulder straps and waist straps, and then the patient can get out of bed and carry out rehabilitation training in the ward or outdoors. The patient can hold the handles with both hands and carry out chest expansion training by stretching the elastic ropes. This training method can effectively exercise the chest muscles and respiratory muscles of the patient, enhance the respiratory function, and promote the recovery of lung function. By adjusting the stretching intensity and frequency, the patient can carry out personalized training according to their own physical condition and rehabilitation progress.

[0017] As an alternative solution of the present invention, a drainage bag placement box is fixedly arranged on one side of the backboard, a color sensor is fixedly arranged on the drainage bag placement box, the drainage bag placement box is used for placing a drainage bag, and a pressure sensor is arranged on the drainage bag. The drainage bag placement box is made of transparent plastic material.

[0018] Through the above technical solution, the drainage bag placement box is used to place the drainage bag, and a color sensor is installed to monitor the color of the drainage fluid; a pressure sensor is used to monitor the drainage pressure.

[0019] As an alternative solution of the present invention, the monitoring mechanism includes:

[0020] Data collection module: Collect the surgical data, historical medical conditions and physical data of the patient, label the data, and use it as a reference sample;

[0021] Data acquisition module: Includes a heart rate sensor, a respiratory rate sensor, a blood pressure sensor, a blood oxygen saturation sensor, etc., and collects data such as the patient's heart rate, respiratory rate, blood pressure and blood oxygen saturation in real time;

[0022] Image acquisition module: Includes a high-definition camera. High-definition cameras are fixedly installed near the forehead and chin of the breathing mask; high-definition images of the patient's face and upper body are collected; the expression changes of the patient's eyes, eyebrows, forehead and other parts are captured emphatically. The two cameras cooperate with each other to achieve the all-round image acquisition of the patient's face and upper body, ensuring that the subtle state changes of the patient at different times can be captured.

[0023] Image preprocessing module: Preprocess the collected images, including filtering and denoising, image enhancement, grayscale conversion and normalization, etc.;

[0024] Feature extraction module: Extract features from the preprocessed images. The extracted features include color, texture and shape;

[0025] Image analysis module: Identify and analyze the images after feature extraction, and promptly identify the abnormal conditions of the patient.

[0026] Comprehensive evaluation module: Integrate the data such as the patient's heart rate, respiratory rate, blood pressure and blood oxygen saturation with the analysis results of the image analysis module, comprehensively evaluate the physical condition of the patient after thoracic surgery, and evaluate the lung function. By establishing a multi-parameter comprehensive evaluation model and using a data fusion algorithm, the physical condition of the patient after thoracic surgery is comprehensively and accurately evaluated.

[0027] Drainage monitoring module: Collect data such as the pressure of the drainage bag, the volume and color of the drainage fluid, and monitor the wound drainage situation.

[0028] Risk prediction module: According to the patient's condition, combined with the data of the patient's heart rate, respiratory rate, blood pressure and blood oxygen saturation and the analysis results of the image analysis module and the drainage monitoring module, predict the risk of the patient's rehabilitation training after thoracic surgery.

[0029] Alarm module: Includes an alarm. When an abnormal situation is detected, an alarm is issued in a timely manner.

[0030] Intelligent control module: Network-connected to the data acquisition module, image acquisition module, image preprocessing module, feature extraction module, image analysis module, drainage monitoring module, comprehensive evaluation module and alarm module.

[0031] As an alternative solution of the present invention, the image analysis module performs recognition and analysis on the image after feature extraction, and promptly identifies the abnormal conditions of the patient; including the following steps:

[0032] 1. Data collection: Widely collect facial expression and body posture image data of patients after thoracic surgery from the hospital's case database and past clinical research projects. For each image, the annotation team carefully observes the features of the patient's facial expression, body posture, etc.

[0033] 2. Data preprocessing: Perform unified preprocessing operations on the collected original image data. Use the image scaling algorithm to adjust all images to the same size. Perform image normalization processing. Divide the labeled data into a training set, a validation set and a test set. The training set is used for parameter learning of the model, the validation set is used to adjust the hyperparameters of the model during training to prevent overfitting, and the test set is used to evaluate the performance of the finally trained model.

[0034] 3. Build a convolutional neural network (VGG16) model: Select the VGG16 model. Before starting training, initialize all parameters of the VGG16 model (such as the weights of the convolutional kernels, the weights and biases of the fully connected layers). Select an appropriate loss function according to the type of image analysis task (such as binary classification, multi-classification). Select the cross-entropy loss function. Use the Adam optimizer, which has good convergence speed and stability when dealing with large-scale datasets and complex models. Input the divided training set data into the model in batches according to the batch size. In each round of training, the model performs forward propagation calculation based on the input image features, and obtains the prediction result through the operations of the convolutional layer, pooling layer and fully connected layer. Compare the prediction result with the labeled true label, and calculate the value of the loss function. Then, through the backpropagation algorithm, according to the gradient information of the loss function, backpropagate from the output layer to the input layer, and update the parameters of the model to make the prediction result of the model gradually approach the true label. Use the validation set data to evaluate the model, and calculate indicators such as the loss value and accuracy of the model on the validation set. According to the evaluation results of the validation set, adjust the hyperparameters of the model to prevent the model from overfitting and improve the generalization ability of the model.

[0035] 4. Model Evaluation: After the model has completed training for a specified number of rounds on the training set, the test set data is used to comprehensively evaluate the model. Calculate metrics such as the accuracy, recall, and F1-score of the model on the test set to measure the model's ability to identify different types of abnormal conditions (such as pain expressions and abnormal breathing postures). Optimize the model based on the results of the model evaluation.

[0036] 5. Practical Application: The image acquisition module continuously collects facial expressions and body posture images of patients. After these images are processed by the feature extraction module to extract features such as color, texture, and shape, the feature data is input into the trained VGG16 model. After receiving the feature data, the model performs inference calculations according to the feature patterns learned during the training process. Through the operations of the convolutional layer, pooling layer, and fully connected layer, the prediction results of the patient's current state are output, that is, to determine whether there are abnormal conditions in the patient and the possible types of abnormalities.

[0037] 6. Result Feedback: Once the model identifies that there are abnormal conditions in the patient, the abnormal information is promptly fed back to the comprehensive evaluation module.

[0038] As an alternative solution of the present invention, the comprehensive evaluation module fuses data such as the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation with the analysis results of the image analysis module to comprehensively evaluate the patient's physical condition after thoracic surgery and evaluate the pulmonary function. The steps are as follows:

[0039] 1. Data Collection and Integration: Real-time obtain the patient's physiological data such as heart rate, respiratory rate, blood pressure, and blood oxygen saturation from the monitoring device. Receive the analysis result data regarding the patient's facial expressions, body postures, etc. from the image analysis module, including information such as whether there are pain expressions and abnormal breathing postures. Retrieve the patient's surgical data. Collect the patient's historical medical condition data.

[0040] 2. Data Preprocessing: Clean the collected physiological data to check the integrity and accuracy of the data. Remove obvious error or abnormal data points; organize the image analysis result data to ensure the uniformity of the data format for subsequent fusion analysis. Organize the surgical data and historical medical condition data and convert them into a format convenient for analysis, such as encoding the surgical types.

[0041] 3. Construct a Multi-parameter Comprehensive Evaluation Model: Select the Bayesian network model and construct a multi-parameter comprehensive evaluation model based on a large amount of clinical data and medical research results. The model contains the association rules and weight distributions between various parameters to reflect the relative importance of different data in evaluating the patient's physical condition. The structural design of the model takes into account the characteristics of different data types and can effectively process numerical physiological data and descriptive image analysis results, surgical, and historical medical condition data.

[0042] 4. Data fusion analysis: The pre-processed physiological data, image analysis result data, surgical data and historical disease data are integrated according to the requirements of the comprehensive evaluation model. In the data fusion process, the data fusion algorithm is used. Common algorithms include weighted average method, which assigns different weights to each data according to the importance of each data to the patient's physical condition, and then performs weighted calculation. Consider the mutual influence relationship between different data.

[0043] 5. Physical condition assessment: The comprehensive assessment model conducts a comprehensive assessment of the patient's physical condition after thoracic surgery based on the fused data, according to the preset assessment rules and algorithms. During the assessment process, the model will conduct a probability analysis of various possible situations, give the probability of each situation occurring, and provide a reference for subsequent diagnosis and treatment.

[0044] 6. Generate a report on the patient's physical condition: Based on the evaluation results, the comprehensive evaluation module generates a detailed and reliable report on the patient's physical condition. The report includes the patient's current physiological indicators, whether there are abnormal conditions, the type and severity of the abnormal conditions, and the evaluation conclusions on lung function.

[0045] As an optional solution of the present invention, the risk prediction module predicts the risk of the patient's post-thoracic surgery rehabilitation training according to the patient's condition, combined with the patient's heart rate, respiratory rate, blood pressure and blood oxygen saturation data and the analysis results of the image analysis module. The following steps are included:

[0046] 1. Data collection: Extract detailed information from the patient's medical records, including the type of surgery, degree of surgical trauma, postoperative complications, and the patient's preoperative underlying diseases. Collect the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation data in real time. Obtain the image analysis module's analysis results on the patient's facial expressions, body posture, etc.

[0047] 2. Data integration: Clean, standardize and normalize all collected data; integrate the cleaned and standardized physiological data, disease data and image analysis results. With time as the key correlation factor, different types of data at the same time point are associated to form a complete patient status data set.

[0048] 3. Build a risk prediction model: Select the multi-layer perceptron (MLP) model. It has powerful nonlinear fitting capabilities, can mine complex potential relationships in data, and accurately predict rehabilitation training risks.

[0049] 4. Feature Engineering: Extract features that have an important impact on risk prediction from the integrated dataset. For the condition data, the surgical type can be one-hot encoded and transformed into multiple binary features; the degree of surgical trauma can be used as a numerical feature through a quantification index. The heart rate, respiratory rate, blood pressure, and blood oxygen saturation in the physiological data are directly used as features, and some derivative features can be calculated to reflect the regulatory function of the cardiac autonomic nervous system. In the image analysis results, the intensity of the facial pain expression, the duration of abnormal breathing postures, etc. are used as features. Through reasonable feature engineering, improve the accuracy and effectiveness of the model for risk prediction.

[0050] 5. Model Training and Optimization: Divide the preprocessed data into a training set, a validation set, and a test set. Use the training set to train the selected model and adjust the model parameters so that the model can accurately fit the data on the training set. During the training process, use the validation set to monitor the model performance and prevent overfitting. If the performance of the model on the validation set starts to decline, regularization methods (such as L1, L2 regularization) or adjusting the model structure (such as reducing the number of layers or neurons in the neural network) can be used for optimization. Finally, use the test set to evaluate the generalization ability of the model to ensure that the model can also accurately predict the rehabilitation training risk on new data.

[0051] 6. Risk Prediction: During the patient's rehabilitation training process, the risk prediction module obtains the patient's physiological data, image analysis results, and condition update information in real time. Through the data interface, the latest heart rate, respiratory rate, blood pressure, blood oxygen saturation data, as well as the latest facial expression and body posture analysis results, are transmitted to the risk prediction model in a timely manner together with the patient's current condition information. The risk prediction model calculates the probability of the patient having a risk at the current rehabilitation training stage according to the input real-time data, the trained model parameters, and algorithms. Output the risk probability prediction result.

[0052] According to the calculated risk probability, set different risk level thresholds to determine the risk level and divide the risk into three levels: low, medium, and high.

[0053] 7. Result Feedback: Timely feedback the risk prediction results to medical staff and patients. Through the hospital information system, display the patient's real-time risk level, the specific content of the risk prediction, and the recommended measures on the work terminals of medical staff (such as computers, tablets). At the same time, the patient's rehabilitation training risk situation can also be informed to the patient through the mobile application or display screen on the patient side, so that the patient can understand their own situation and actively cooperate with the treatment.

[0054] As an alternative solution of the present invention, the drainage monitoring module collects data such as the pressure of the drainage bag, the volume and color of the drainage fluid, and monitors the wound drainage situation, including the following steps:

[0055] 1. Data collection: The pressure sensor uses a high-precision strain gauge pressure sensor, which is closely attached to the outer side of the bottom of the drainage bag to monitor the pressure of the drainage fluid in real time. The liquid level sensor monitors the volume of the drainage fluid in real time; the color sensor is installed near the drainage bag in the drainage bag placement box to monitor the color of the drainage fluid in real time.

[0056] 2. Data integration: Integrate the collected pressure data, liquid level data, and color data. Using time as the key index, associate different types of data at the same time point to form a complete dataset of the drainage situation.

[0057] 3. Data analysis: Use data analysis algorithms to deeply analyze the integrated drainage situation dataset. For the pressure data, analyze its change trend to determine whether the drainage pressure fluctuates within the normal range and whether there are abnormal situations such as sudden increases or decreases in pressure. For the liquid level data, analyze the change rate of the drainage fluid volume to determine whether the drainage is unobstructed and whether there are abnormal situations such as too much or too little drainage fluid volume. For the color data, by comparing with the normal color range of the drainage fluid, determine whether there are abnormal changes in the color of the drainage fluid, such as whether the color turns red due to blood mixing or abnormal color due to infection, etc.

[0058] 4. Abnormality judgment: According to the preset thresholds and judgment rules, judge the analyzed drainage data for abnormalities to determine whether there are abnormal situations. Once an abnormal situation is judged, the data processing module immediately triggers the warning mechanism.

[0059] 5. Monitoring result evaluation: Generate a drainage situation monitoring report in real time. The report content includes the current drainage pressure, drainage fluid volume, drainage fluid color, and the comparison with the normal range, and at the same time shows whether the drainage situation is normal. Combining the real-time report and the comparison result of historical data, the data processing module comprehensively evaluates the wound drainage situation of the patient. The evaluation results are divided into four levels: normal, mild abnormality, moderate abnormality, and severe abnormality. When the drainage situation is at the normal level, it indicates that the patient's wound drainage situation is good and the rehabilitation process is normal; when the drainage situation is mildly abnormal, it prompts medical staff to closely monitor the changes in the drainage situation; when the drainage situation is moderately abnormal, it is recommended that medical staff promptly examine and evaluate the patient and take corresponding treatment measures; when the drainage situation is severely abnormal, an emergency warning is immediately issued to notify medical staff to perform emergency treatment on the patient.

[0060] 6. Abnormality warning: Once the data processing module determines that there is an abnormality in the drainage situation, it immediately issues a warning through multiple methods.

[0061] 3. Beneficial effects

[0062] One or more technical solutions provided in the technical solution of this application have at least the following technical effects or advantages:

[0063] 1. The nursing device of the present invention can be hung on the armrest of the bed by a hook, so that it is convenient for the patient to use when lying in bed; it can also be fixed to the back with the help of shoulder straps and waist straps to meet the patient's needs for getting out of bed or outdoor rehabilitation training, breaking the limitation that traditional rehabilitation training is limited to a specific area of ​​the ward, providing patients with a more flexible rehabilitation environment choice, and improving the convenience and accessibility of rehabilitation training.

[0064] 2. Patients can hold the handles with both hands and stretch the elastic rope to perform chest expansion training, and can adjust the stretching strength and frequency according to their own physical condition and rehabilitation progress to implement personalized rehabilitation training plans, which helps to improve the pertinence and effectiveness of rehabilitation training and avoid extra burden on the body or affect the rehabilitation effect due to improper training intensity.

[0065] 3. The image acquisition module accurately identifies abnormal conditions on the patient's face and upper body, such as painful expressions, abnormal breathing postures, etc. These intuitive physical manifestations can help determine the patient's physical condition.

[0066] 4. The comprehensive evaluation module integrates physiological data with image analysis results to comprehensively evaluate the patient's physical condition and lung function after thoracic surgery, and comprehensively judge the patient's recovery process from multiple angles, making the evaluation results more accurate and comprehensive, and providing strong support for medical staff to formulate scientific and reasonable treatment and rehabilitation plans.

[0067] 5. The drainage monitoring module collects and analyzes data such as the pressure, drainage fluid volume and color of the drainage bag in real time, which can promptly detect abnormal conditions during wound drainage and provide accurate information for medical staff to deal with wound-related issues in a timely manner, thereby ensuring the safety and effectiveness of wound drainage and promoting wound healing.

[0068] 6. The risk prediction module combines the patient's condition, physiological data, image analysis results and drainage monitoring data to make risk predictions for rehabilitation training, identify in advance the risks that patients may face during rehabilitation training, and help medical staff formulate response measures in advance, reduce rehabilitation training risks, and ensure patient safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is an overall schematic diagram of the care device disclosed in this application for improving lung function after thoracic surgery.

[0070] Reference numerals:

[0071] 1. Backboard; 2. Respiratory mask; 3. Air filtration box; 4. Elastic cord; 5. Handle; 6. Waste liquid collection box; 7. Suction pump; 8. Drainage bag placement box; 9. Bite valve; 10. Suction catheter; 11. Shoulder strap; 12. Waist strap. Detailed implementation manners

[0072] The present application will be further described in detail below with reference to the accompanying drawings of the specification.

[0073] Referring to Figure 1 , the embodiment of the present application provides a nursing device for improving lung function after thoracic surgery, including: a backboard 1, a respiratory mask 2, an air filtration box 3, an elastic cord 4, a handle 5, a waste liquid collection box 6, a suction pump 7 and a monitoring mechanism;

[0074] On both sides of the upper end of the backboard 1, shoulder straps 11 are fixedly arranged; on both sides of the lower end of the backboard 1, waist straps 12 are fixedly arranged; the backboard 1 is made of a strong and lightweight material, ensuring that while providing a stable support for the patient, it will not cause too much burden on the patient. Its ergonomic design can closely fit the back curve of the patient, providing comprehensive support from the shoulders to the waist. The shoulder strap 11 is made of a soft and elastic material, such as high-quality nylon elastic band, which can not only closely surround the patient's shoulders, but also effectively disperse the pressure, avoiding leaving marks or discomfort on the shoulders caused by long-term wearing. Through an adjustable buckle design, it can adapt to patients of different body shapes, ensuring the stability and comfort of wearing. The waist strap 12 is made of nylon elastic band; they can closely fit the patient's waist, further enhancing the fit between the entire device and the patient's body, preventing the device from shifting during use. A hook is fixedly arranged at the upper end of the backboard 1.

[0075] On both sides of the lower end of the backboard 1, an air filtration box 3 and a waste liquid collection box 6 are respectively fixedly arranged;

[0076] Inside the air filtration box 3, there is an air filter screen and activated carbon, which are used for adsorbing dust and filtering the air. The outlet of the air filtration box 3 is connected to a breathing mask 2 through a hose. A mouthpiece 9 is detachably and fixedly arranged on the breathing mask 2. The main body of the breathing mask 2 is made of soft silicone material, which is light and elastic, and can closely fit the facial contour of the patient, effectively preventing air leakage. At the same time, the silicone material has little irritation to the skin and can reduce the damage to the facial skin caused by long-term wearing. The mouthpiece 9 is made of medical-grade non-toxic plastic, and its shape and size are carefully designed to conform to the natural biting habit of the human oral cavity. The surface of the mouthpiece is specially treated with certain anti-slip texture, which is convenient for the patient to bite and can effectively prevent the mouthpiece from sliding during breathing. The detachable design enables the mouthpiece to be easily cleaned and replaced, ensuring the hygiene and safety of each use. The air filter screen adopts a high-precision filtering material, such as a HEPA (High Efficiency Particulate Air) filter screen, which can effectively filter tiny particles in the air, including dust, pollen, bacteria, etc. If these tiny particles are inhaled by the patient, they may irritate the fragile respiratory tract after surgery and cause complications such as infection. Activated carbon has a strong adsorption capacity and can adsorb odors, harmful gases, and some volatile organic compounds (VOCs) in the air. This can not only provide fresh air for the patient but also further purify the substances in the air that may be harmful to the respiratory tract, ensuring the purity and safety of the air inhaled by the patient. The housing of the breathing mask 2 is made of transparent plastic materials (such as polystyrene PS, polyurethane TPU, etc.).

[0077] A suction pump 7 is fixedly arranged on the waste liquid receiving box 6. The input end of the suction pump 7 is connected in communication with a sputum suction tube 10, and the sputum suction tube 10 is slidably connected to the breathing mask 2. The output end of the suction pump 7 extends into the waste liquid receiving box 6. The waste liquid receiving box 6 is made of transparent medical-grade plastic material, which is convenient for medical staff to observe the collection situation of the internal waste liquid at any time. The suction pump 7 adopts a silent design and generates extremely low noise during operation, which will not cause additional interference to the patient. The suction pump has an adjustable negative pressure function, and medical staff can accurately adjust the negative pressure size according to the actual situation of the patient through the adjustment knob on the pump body.

[0078] On both sides of the lower end of the back plate 1, there are channels. Inside the channels, an elastic cord 4 is fixedly arranged, and the outer end of the elastic cord 4 is fixedly provided with a handle 5. Chest expansion training can be carried out by stretching the elastic cord 4 and the handle 5. The elastic cord 4 is made of a high-strength elastic material, such as a mixture of rubber and fiber, and has good elasticity and durability.

[0079] A monitoring mechanism is fixedly arranged on the backboard 1. The monitoring mechanism monitors the physical condition of the patient after thoracic surgery and can promptly detect abnormal situations. The monitoring mechanism integrates a variety of advanced sensor technologies and can comprehensively and real-time monitor the physical condition of the patient after thoracic surgery. It can monitor key physiological indicators of the patient, such as heart rate, respiratory rate, blood oxygen saturation, etc. By continuously monitoring these indicators, abnormal situations that may occur in the patient's body, such as arrhythmia, shortness of breath, hypoxemia, etc., can be promptly detected. Once abnormal data is detected, the monitoring mechanism will immediately issue an alarm to remind medical staff to take corresponding treatment measures in a timely manner, providing a strong guarantee for the patient's life safety.

[0080] In this technical solution, when the patient lies on the hospital bed, the backboard 1 can be hung on the armrest of the hospital bed through the hook. When getting out of bed or going outdoors for rehabilitation training, the backboard 1 can be fixed to the back through the shoulder strap 11 and the waist strap 12, and then the patient can get out of bed to carry out rehabilitation training in the ward or outdoors. The patient can hold the handle 5 with both hands and perform chest expansion training by stretching the elastic cord 4. This training method can effectively exercise the patient's chest muscles and respiratory muscles, enhance the respiratory function, and promote the recovery of lung function. By adjusting the stretching strength and frequency, the patient can carry out personalized training according to their own physical condition and rehabilitation progress.

[0081] Furthermore, a drainage bag placement box 8 is fixedly arranged on one side of the backboard 1. A color sensor is fixedly arranged on the drainage bag placement box 8. The drainage bag placement box 8 is used to place the drainage bag, and a pressure sensor is arranged on the drainage bag. The drainage bag placement box 8 is made of transparent plastic material.

[0082] In this technical solution, the drainage bag placement box 8 is used to place the drainage bag, and the color sensor monitors the color of the drainage fluid; the drainage pressure is monitored through the pressure sensor. At the opening of the placement box, a soft and elastic rubber gasket is arranged. When the drainage bag is placed, the gasket can closely fit the drainage bag, which can not only prevent foreign dust and other impurities from entering and affecting the purity of the liquid in the drainage bag, but also play a buffering role to a certain extent and reduce the damage to the drainage bag caused by external collision.

[0083] Furthermore, the monitoring mechanism includes:

[0084] Data collection module: Collect the surgical data, historical medical conditions, and physical data of patients, label the data, and use it as a reference sample; Extract the detailed surgical data of patients from the hospital's information management system, covering key information such as the type of surgery, the duration of the surgery, the intraoperative blood loss, and whether there are intraoperative complications. Collect the historical medical conditions of patients, including past diseases, whether there are long-term chronic medical histories (such as chronic obstructive pulmonary disease, cardiovascular diseases, etc.), allergy histories, and previous treatment records, etc. These information helps to judge the basic condition of the patient's body. For example, for patients with chronic cardiopulmonary diseases, the recovery process of pulmonary function and cardiovascular function after surgery may be more complex and requires key attention. Collect physical data such as height, weight, and various preoperative physiological indicators (such as the baseline values of heart rate, blood pressure, respiratory rate, and blood oxygen saturation). Classify and label according to the type, source, and degree of association with the patient's current condition of the data, and build a comprehensive and orderly reference sample library.

[0085] Data acquisition module: Include a heart rate sensor, a respiratory rate sensor, a blood pressure sensor, a blood oxygen saturation sensor, etc., and collect data such as the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation in real time;

[0086] Image acquisition module: Include a high-definition camera, and high-definition cameras are fixedly arranged near the forehead and chin of the breathing mask 2; Collect high-definition images of the patient's face and upper body; Focus on capturing the expression changes in parts such as the patient's eyes, eyebrows, and forehead, such as the patient frowning due to pain, and the pain or anxiety shown in the eyes. The camera at the chin shoots from an elevation angle, and can clearly record the movements of the patient's lips, the movement of the mandible, and the postural changes of the neck, such as the degree of opening of the lips when breathing rapidly and the involuntary tremor of the mandible. The two cameras cooperate with each other to achieve full-range image acquisition of the patient's face and upper body, ensuring that the subtle state changes of the patient at different moments can be captured.

[0087] Image preprocessing module: Preprocess the collected images, including filtering and denoising, image enhancement, grayscale conversion, and normalization, etc.;

[0088] Feature extraction module: Extract features from the preprocessed images, and the extracted features include color, texture, and shape;

[0089] Color feature extraction: Mainly analyze the color distribution and changes in specific areas of the image (such as the facial skin color area). By establishing a color space model, such as RGB, HSV and other color spaces, statistical features such as the mean and variance of the skin color are extracted. For example, when the patient has conditions such as hypoxia and blood loss, the facial skin color will change. By analyzing the color features, these abnormal changes can be detected in time, providing a basis for judging the patient's physical condition.

[0090] Texture feature extraction: Algorithms such as gray-level co-occurrence matrix and local binary pattern are used to extract features such as facial skin texture and wrinkles. These texture features reflect the patient's emotional state and physical fatigue to a certain extent. For example, when a patient is in pain or anxiety for a long time, facial wrinkles may deepen. Through texture feature extraction, these changes can be quantified to assist medical staff in evaluating the patient's physical and mental state.

[0091] Shape feature extraction: Analyze the patient's facial contour, eye shape, lip shape, and limb movements. Using algorithms such as edge detection and contour extraction, obtain the shape parameters of key facial parts, such as the degree of eye opening and closing, the degree of lip opening and closing, and limb movements. These shape features are of great significance for judging the patient's expression and breathing state. For example, when breathing rapidly, the lips may involuntarily open wide. Through shape feature extraction, these changes can be accurately captured.

[0092] Image analysis module: Identify and analyze the images after feature extraction to promptly identify the patient's abnormal conditions; use deep learning algorithms, such as convolutional neural network (CNN), to identify and analyze the images after feature extraction. The CNN model is trained with a large amount of labeled facial expression and body posture image data so that it can learn the image feature patterns in normal and various abnormal states. In actual applications, the extracted patient image features are input into the trained model. Through matching and analyzing the features, the model can quickly identify whether the patient has abnormal conditions, such as pain expressions and abnormal breathing postures. For example, when the model identifies that the patient's face shows typical pain expression features and the duration exceeds a certain threshold, it can be judged that the patient may be in pain.

[0093] Comprehensive evaluation module: Integrate data such as the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation with the analysis results of the image analysis module to comprehensively evaluate the patient's physical condition after thoracic surgery and evaluate the lung function. By establishing a multi-parameter comprehensive evaluation model and using data fusion algorithms, comprehensively and accurately evaluate the patient's physical condition after thoracic surgery. For example, when the heart rate increases, the respiratory rate accelerates, and at the same time, the image analysis shows that the patient's face shows a painful expression and the breathing posture is abnormal, the comprehensive evaluation module can judge that the patient may have relatively severe pain or respiratory dysfunction, and then combine the patient's surgical data and historical condition to more accurately evaluate the lung function and provide a detailed and reliable report on the patient's physical condition for medical staff.

[0094] Drainage monitoring module: Collect data such as the pressure of the drainage bag, the volume and color of the drainage fluid to monitor the wound drainage situation.

[0095] Risk prediction module: Based on the patient's condition, combining the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation data with the analysis results of the image analysis module and the drainage monitoring module, predict the risk of the patient's rehabilitation training after thoracic surgery.

[0096] Alarm module: Includes an alarm, which issues an alarm in a timely manner when an abnormal situation is detected.

[0097] Intelligent control module: Network-connected to the data acquisition module, image acquisition module, image preprocessing module, feature extraction module, image analysis module, drainage monitoring module, comprehensive evaluation module, and alarm module.

[0098] Furthermore, the image analysis module performs recognition and analysis on the images after feature extraction to promptly identify the patient's abnormal conditions; it includes the following steps:

[0099] 1. Data collection: Widely collect facial expression and body posture image data of patients after thoracic surgery from the hospital's case database and past clinical research projects. These data sources are diverse, covering patients of different ages, genders, surgical types, and rehabilitation stages. For each image, the annotation team carefully observes the patient's facial expressions, body postures, and other features. For facial expressions, mark whether it is a pain expression, an anxious expression, a calm expression, etc., and record the intensity level of the expression in detail. For example, a pain expression can be further divided into mild, moderate, and severe pain expressions according to the degree of frowning and the downward pull of the corners of the mouth. For body postures, mark abnormal breathing postures (such as the gasping posture during rapid breathing and the posture of accessory respiratory muscles participating during dyspnea), normal postures, and other abnormal postures (such as the body twisting due to discomfort). At the same time, record the time stamp of the image capture and the corresponding physiological index data of the patient at that time (such as heart rate, respiratory rate, blood pressure, etc.).

[0100] 2. Data preprocessing: Perform unified preprocessing operations on the collected original image data. Use an image scaling algorithm to adjust all images to the same size. Perform image normalization processing to unify the pixel value range of the images to the interval [0,1] or [-1,1], eliminating differences in brightness, contrast, etc. between different images, and improving the stability and accuracy of model training. Divide the labeled data into a training set, a validation set, and a test set. The training set is used for parameter learning of the model, the validation set is used to adjust the hyperparameters of the model during training to prevent overfitting, and the test set is used to evaluate the performance of the finally trained model.

[0101] 3. Build a Convolutional Neural Network (VGG16) model: Select the VGG16 model. Before starting training, initialize all the parameters of the VGG16 model (such as the weights of the convolutional kernels, the weights and biases of the fully connected layers). Common initialization methods include random initialization (such as generating random numbers using Gaussian distribution or uniform distribution), Xavier initialization, and Kaiming initialization, etc. Select an appropriate loss function according to the type of image analysis task (such as binary classification, multi-classification). Select the cross-entropy loss function. Adopt the Adam optimizer. The Adam optimizer has better convergence speed and stability when dealing with large-scale datasets and complex models. Input the divided training set data into the model in batches according to the batch size. In each round of training, the model performs forward propagation calculations based on the input image features. Through the operations of the convolutional layer, pooling layer, and fully connected layer, the prediction results are obtained. Compare the prediction results with the labeled true labels and calculate the value of the loss function. Then, through the backpropagation algorithm, according to the gradient information of the loss function, backpropagate from the output layer to the input layer to update the model parameters, so that the prediction results of the model gradually approach the true labels. Use the validation set data to evaluate the model and calculate metrics such as the loss value and accuracy of the model on the validation set. According to the evaluation results of the validation set, adjust the hyperparameters of the model (such as the learning rate, Dropout ratio, etc.) to prevent the model from overfitting and improve the generalization ability of the model.

[0102] 4. Model evaluation: When the model completes the specified number of rounds of training on the training set, use the test set data to comprehensively evaluate the model. Calculate metrics such as the accuracy, recall rate, and F1 value of the model on the test set to measure the model's ability to identify different types of abnormal situations (such as pain expressions, abnormal breathing postures). Optimize the model according to the results of the model evaluation.

[0103] 5. Practical application: The image acquisition module continuously acquires the facial expressions and body posture images of patients. After these images are extracted with features such as color, texture, and shape by the feature extraction module, the feature data is input into the trained VGG16 model. After receiving the feature data, the model performs inference calculations according to the feature patterns learned during the training process. Through the operations of the convolutional layer, pooling layer, and fully connected layer, the prediction results of the patient's current state are output, that is, to judge whether there are abnormal situations in the patient and the possible types of abnormalities (such as pain expressions, abnormal breathing postures, etc.).

[0104] The predicted result output by the model is a probability value, indicating the likelihood that the patient is in a certain abnormal condition. To accurately determine whether the patient is in an abnormal condition, a probability threshold is set. When the probability value output by the model is greater than this threshold, it is determined that the patient has the corresponding abnormal condition. At the same time, to avoid misjudgment caused by instantaneous changes in facial expressions or postures, a time threshold is set. For example, when the model continuously recognizes (such as 5 times, corresponding to a certain time interval. Assuming that the image acquisition interval is 1 second each time, the time threshold is 5 seconds) that the patient's face shows typical pain expression characteristics (the probability value is greater than the threshold), it is determined that the patient may be in a pain state.

[0105] 6. Result feedback: Once the model identifies that the patient has an abnormal condition, the abnormal information is promptly fed back to the comprehensive evaluation module.

[0106] Furthermore, the comprehensive evaluation module integrates data such as the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation with the analysis results of the image analysis module to comprehensively evaluate the patient's physical condition after thoracic surgery and evaluate the lung function. The steps include:

[0107] 1. Data collection and integration: Real-time obtain the patient's physiological data such as heart rate, respiratory rate, blood pressure, and blood oxygen saturation from monitoring devices (such as electrocardiogram monitors, respiratory rate monitors, sphygmomanometers, blood oxygen saturation monitors, etc.). Receive the analysis result data on the patient's facial expressions, body postures, etc. from the image analysis module, including information such as whether there are pain expressions, abnormal respiratory postures, etc., and detailed contents such as the frequency and duration of these abnormal conditions. Retrieve the patient's surgical data, such as key information on the type of surgery, surgical duration, intraoperative blood loss, and whether there are complications. These data come from the hospital's surgical record system. Collect the patient's historical medical condition data, covering past medical history (such as whether there are chronic pulmonary diseases, cardiovascular diseases, etc.), allergy history, previous treatment conditions, etc., and obtain them from the hospital's medical record database.

[0108] 2. Data preprocessing: Clean the collected physiological data to check the integrity and accuracy of the data. Remove obviously incorrect or abnormal data points; organize the image analysis result data to ensure that the data format is unified for subsequent fusion analysis. Organize the surgical data and historical medical condition data and convert them into a format convenient for analysis, such as encoding the type of surgery.

[0109] 3. Construct a multi-parameter comprehensive evaluation model: Select the Bayesian network model and construct a multi-parameter comprehensive evaluation model based on a large amount of clinical data and medical research results. The model includes the association rules and weight assignments among various parameters to reflect the relative importance of different data in evaluating the patient's physical condition. The structural design of the model takes into account the characteristics of different data types and can effectively process numerical physiological data and descriptive image analysis results, surgical and historical condition data. The multi-parameter comprehensive evaluation model is: y = {∑ n i＝1 [w i *x i *f(x i ) / (σ i 2 )]} / {∑ n i＝1 [w i / (σ i 2 )]}; ∑ n i＝1 (w i ) = 1;

[0110] f(x i ) = log b (x i ); where y is the fused comprehensive evaluation value. x i is the value of the i-th physiological data or analysis result. w i is the weight corresponding to x i . These weights reflect the importance of different data in the comprehensive evaluation. σ i is the standard deviation of the i-th data. f(x i ) is a non-linear function used to capture the complex relationships between data. This function can be selected and trained according to the characteristics and requirements of the data. In practical applications, f(x i ) can be a polynomial function, exponential function, logarithmic function or any other function suitable for the data characteristics. b is the base of the logarithm (b > 0 and b ≠ 1).

[0111] 4. Data fusion analysis: Integrate the preprocessed physiological data, image analysis result data, surgical data and historical condition data according to the requirements of the comprehensive evaluation model. For example, correlate physiological data such as heart rate, respiratory rate, blood pressure and blood oxygen saturation with the results of pain expression and respiratory posture in image analysis. In the data fusion process, use data fusion algorithms. Common algorithms include the weighted average method. According to the importance of each data in affecting the patient's physical condition, different weights are assigned to each data, and then weighted calculations are performed.

[0112] Meanwhile, consider the mutual influence relationships among different data. For example, an elevated heart rate may be related to multiple factors such as pain and respiratory dysfunction. The model will comprehensively analyze the associations among these factors instead of viewing each data in isolation.

[0113] 5. Physical condition assessment and judgment: Based on the fused data, the comprehensive assessment model comprehensively evaluates the patient's physical condition after thoracic surgery according to the preset assessment rules and algorithms. For example, when the heart rate is elevated, the respiratory rate is increased, and the image analysis shows that the patient's face presents a painful expression and the breathing posture is abnormal, the model will judge that the patient may have relatively severe pain or respiratory dysfunction based on this comprehensive information. The model will also combine the patient's surgical data and historical medical conditions to further refine the assessment results. If the patient has undergone lung surgery and has a history of chronic lung disease, when the above abnormal conditions occur, the model will be more inclined to judge that the possibility of respiratory dysfunction is relatively high and can more accurately evaluate the degree to which lung function is affected. During the assessment process, the model will conduct probability analysis on various possible situations and give the likelihood of each situation occurring, providing a reference for subsequent diagnosis and treatment.

[0114] 6. Generate a patient physical condition report: According to the assessment results, the comprehensive assessment module generates a detailed and reliable patient physical condition report. The report content includes the patient's current physiological index status, whether there are abnormal conditions and the types and severity of the abnormal conditions, the assessment conclusion of the lung function situation, etc. The report will also provide suggestions based on the assessment results, such as whether further examinations are needed and whether the treatment plan needs to be adjusted (such as increasing the dose of painkillers, adjusting the oxygen inhalation concentration, etc.). The report is presented in a clear and easy-to-understand format, facilitating medical staff to quickly obtain key information and providing strong support for subsequent medical decisions. The report can be transmitted in real time to the work terminals of medical staff (such as computers, tablets, etc.) through the hospital information system to ensure that medical staff can timely understand the patient's physical condition.

[0115] Furthermore, the risk prediction module predicts the risks of the patient's rehabilitation training after thoracic surgery based on the patient's condition, combining the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation data with the analysis results of the image analysis module. It includes the following steps:

[0116] 1. Data collection: Detailed medical information is extracted from the patient's medical records, including the type of surgery (such as lobectomy, radical esophagectomy, and other different surgical procedures), the degree of surgical trauma (assessed by the size of the surgical incision, the number of organs involved, etc.), postoperative complications (such as whether there is lung infection, poor wound healing, etc.), and the patient's preoperative underlying diseases (such as chronic obstructive pulmonary disease, cardiovascular disease, etc.). The patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation data are collected in real time. Obtain the analysis results of the image analysis module on the patient's facial expressions, body postures, etc. For example, the frequency and intensity of facial expressions of pain reflect the degree of pain that the patient may have; the occurrence of abnormal breathing postures (such as heavy breathing during shortness of breath, and the posture of auxiliary respiratory muscles during dyspnea).

[0117] 2. Data integration: Clean, standardize and normalize all collected data; integrate the cleaned and standardized physiological data, disease data and image analysis results. With time as the key correlation factor, different types of data at the same time point are linked to form a complete patient status data set. For example, the heart rate, respiratory rate, blood pressure, and blood oxygen saturation data at a certain moment are integrated with the facial expression analysis results and body posture analysis results corresponding to that moment, as well as the patient's condition information at that time (such as whether he is in the postoperative complication treatment stage) to provide comprehensive and accurate data input for subsequent risk prediction models.

[0118] 3. Construct a risk prediction model: Select the multi-layer perceptron (MLP) model. It has powerful nonlinear fitting capabilities, can mine complex potential relationships in data, and accurately predict rehabilitation training risks. The risk prediction model is:

[0119] P risk =Sigmoid{∑ n i＝1 [u i *g(x i )+b+a*HRV+β*Pain+C*tdr]};

[0120] g(x i )=log(1+x i );HRV 2 ={∑ N j＝1 [HR j -HR - ] 2}; where x i are input features (such as heart rate, respiratory rate, blood pressure, blood oxygen saturation, etc.). iw is the weight, representing the contribution of each feature to risk prediction. b is the bias term, used to adjust the output of the model. Sigmoid is the activation function (Sigmoid activation function). HRV is the heart rate variability feature, and HR j is the j-th heart rate value, and HR - is the average value of the heart rate, N is the number of heart rate values. Pain is the pain intensity feature output by the facial expression analysis module. a, β, and C are the weight coefficients, used to adjust the contributions of HRV, Pain, and the drainage condition to risk prediction. g(x i ) is the feature transformation function, used to perform a non-linear mapping on the input features to enhance the expressive ability of the model. P risk is the output of the model as the risk probability. tdr is the result of the drainage condition monitored by the drainage monitoring module.

[0121] 4. Feature Engineering: Extract features that have an important impact on risk prediction from the integrated dataset. For the condition data, the surgical type can be one-hot encoded and transformed into multiple binary features; the degree of surgical trauma can be used as a numerical feature through quantification indicators (such as the length of the surgical incision, the number of organs involved, etc.). The heart rate, respiratory rate, blood pressure, and blood oxygen saturation in the physiological data are directly used as features, and some derivative features can be calculated, such as heart rate variability (measured by the standard deviation of the heart rate over a period of time), which reflects the regulatory function of the cardiac autonomic nervous system. In the image analysis results, the intensity of the facial pain expression, the duration of abnormal breathing postures, etc. are used as features. Through reasonable feature engineering, the accuracy and effectiveness of the model for risk prediction are improved.

[0122] 5. Model Training and Optimization: Divide the preprocessed data into a training set, a validation set, and a test set. Use the training set to train the selected model and adjust the model parameters so that the model can accurately fit the data on the training set. During the training process, use the validation set to monitor the model performance to prevent overfitting. If the performance of the model on the validation set starts to decline, regularization methods (such as L1, L2 regularization) or adjusting the model structure (such as reducing the number of layers or neurons in the neural network) can be used for optimization. Finally, use the test set to evaluate the generalization ability of the model to ensure that the model can accurately predict the rehabilitation training risk on new data. The loss function for model training and optimization is:

[0123] L = -(1 / N)Σ M i＝1 {[γ i *y i log(y i ^)+(1 - y i )log(1 - y i ^)] + λ * Pe(y i , y i ^)};

[0124] Pe(y i ,y i ^) = {(y i ^ - y i ) 2 , if y i = 1; 0, if y i = 0}; where y i is the true label (0 represents low risk and 1 represents high risk). y i ^ is the risk probability predicted by the model. M is the number of samples. γ i is the class weight used to balance the uneven distribution of high-risk and low-risk samples. λ is the weight coefficient of the penalty term used to control the intensity of the penalty term. Pe(y i ,y i ^) is the risk level penalty term used to impose a greater penalty on the prediction error of high-risk samples.

[0125] 6. Risk prediction: During the patient's rehabilitation training process, the risk prediction module obtains the patient's physiological data, image analysis results, and condition update information in real time. Through the data interface, the latest heart rate, respiratory rate, blood pressure, blood oxygen saturation data, as well as the latest facial expression and body posture analysis results, are transmitted to the risk prediction model in a timely manner together with the patient's current condition information (such as the number of days of postoperative recovery, progress of complication treatment, etc.).

[0126] The risk prediction model calculates the probability of the patient having a risk at the current rehabilitation training stage according to the input real-time data, the trained model parameters, and the algorithm. The risk probability prediction result is output.

[0127] According to the calculated risk probability, different risk level thresholds are set to determine the risk level and divide the risk into three levels: low, medium, and high.

[0128] 7. Result feedback: Timely feedback the risk prediction results to medical staff and patients. Through the hospital information system, display the real-time risk level of the patient, the specific content of the risk prediction (such as the training may be interrupted due to respiratory function problems), and the recommended measures (such as adjusting the training intensity, increasing the oxygen inhalation time) on the work terminals of medical staff (such as computers, tablets). At the same time, the risk situation of the patient's rehabilitation training can also be informed to the patient through the mobile application or display screen on the patient side, so that the patient can understand their own condition and actively cooperate with the treatment. Record the results and relevant data of each risk prediction and store them in the patient's electronic medical record system. The recorded content includes the risk prediction time, risk level, input physiological data, image analysis results, condition information, and the risk probability value output by the model, etc. These records can not only provide a basis for adjusting the subsequent rehabilitation training plan, but also help track and analyze the entire rehabilitation process of the patient, providing data support for the medical team to summarize experience and improve the treatment plan.

[0129] Furthermore, the drainage monitoring module collects data such as the pressure of the drainage bag, the volume and color of the drainage fluid to monitor the wound drainage situation, including the following steps:

[0130] 1. Data collection: The pressure sensor uses a high-precision strain gauge pressure sensor, which is closely attached to the outer side of the bottom of the drainage bag. It has extremely high sensitivity, and the measurement accuracy can reach ±0.1 kPa, which can accurately capture the subtle changes in the drainage pressure. The liquid level sensor monitors the volume of the drainage fluid in real time; the liquid level sensor is installed near the bottom inside the drainage bag. An ultrasonic liquid level sensor can be selected, which calculates the liquid level height by emitting and receiving ultrasonic signals and using the propagation time of ultrasonic waves in the liquid. The color sensor is installed near the drainage bag in the drainage bag placement box to monitor the color of the drainage fluid in real time. The color sensor has high sensitivity and fast response characteristics, and can capture the color changes of the drainage fluid.

[0131] 2. Data integration: Integrate the collected pressure data, liquid level data and color data, use time as the key index to associate different types of data at the same time point, and form a complete drainage situation data set.

[0132] 3. Data analysis: Use data analysis algorithms to deeply analyze the integrated drainage situation data set. For the pressure data, analyze its change trend to judge whether the drainage pressure fluctuates within the normal range and whether there are abnormal situations such as sudden increase or decrease in pressure. For the liquid level data, analyze the change rate of the drainage fluid volume to judge whether the drainage is unobstructed and whether there are abnormal situations such as too much or too little drainage fluid volume. For the color data, judge whether there are abnormal changes in the color of the drainage fluid by comparing it with the normal drainage fluid color range, such as whether the color turns red due to blood mixing or abnormal color due to infection, etc.

[0133] 4. Abnormality Judgment: According to the preset threshold and judgment rules, the analyzed drainage data is judged for abnormalities to determine whether there are any abnormal situations. For example, when the drainage pressure exceeds 15% of the upper limit of the normal range or is lower than 10% of the lower limit of the normal range, it is determined that the pressure is abnormal; when the drainage volume increases or decreases sharply within a short period of time by more than a certain percentage (such as 30%), it is determined that the drainage volume is abnormal; when the RGB value of the drainage fluid color deviates from the normal range by more than a certain degree (such as the red component R value exceeds 20% of the upper limit of the normal range), it is determined that the color is abnormal. Once an abnormal situation is judged, the data processing module immediately triggers the warning mechanism.

[0134] 5. Monitoring Result Evaluation: A monitoring report on the drainage situation is generated in real time. The report content includes the current drainage pressure, drainage volume, drainage fluid color, and the comparison with the normal range, and at the same time shows whether the drainage situation is normal. Combining the real-time report and the comparison result of historical data, the data processing module comprehensively evaluates the wound drainage situation of the patient. The evaluation results are divided into four levels: normal, slightly abnormal, moderately abnormal, and severely abnormal. When the drainage situation is at the normal level, it indicates that the patient's wound drainage situation is good and the rehabilitation process is normal; when the drainage situation is slightly abnormal, it prompts the medical staff to closely monitor the changes in the drainage situation; when the drainage situation is moderately abnormal, it is recommended that the medical staff promptly check and evaluate the patient and take corresponding treatment measures; when the drainage situation is severely abnormal, an emergency warning is immediately issued to notify the medical staff to perform emergency treatment on the patient.

[0135] 6. Abnormal Warning: Once the data processing module determines that the drainage situation is abnormal, it immediately issues a warning in multiple ways. First, the warning information is sent to the hospital's information management system through a wireless communication module (such as Bluetooth, Wi-Fi, etc.), and a warning prompt is displayed on the work terminals of the medical staff (such as computers, tablets), including the type of abnormality, the time of abnormality occurrence, patient information, etc. At the same time, a loud alarm sound is emitted through a sound alarm to remind the medical staff to pay attention in a timely manner. In addition, the warning information can also be sent to the mobile phones of the medical staff responsible for this patient through text messages or push notifications to ensure that the medical staff can obtain the warning information in a timely manner.

[0136] The present invention provides a nursing device for improving lung function after thoracic surgery, which includes the following steps: When the patient lies on the hospital bed, the backboard 1 can be hung on the armrest of the hospital bed through the hook. When getting out of bed or going outdoors for rehabilitation training, the backboard 1 can be fixed to the back through the shoulder strap 11 and the waist strap 12, and then the patient can get out of bed and carry out rehabilitation training in the ward or outdoors. The patient can hold the handle 5 with both hands and perform chest expansion training by stretching the elastic cord 4. This training method can effectively exercise the chest muscles and respiratory muscles of the patient, enhance the respiratory function, and promote the recovery of lung function. By adjusting the stretching intensity and frequency, the patient can carry out personalized training according to their own physical condition and rehabilitation progress. The drainage bag placement box 8 is used to place the drainage bag. The data acquisition module of the monitoring mechanism collects data such as the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation in real time; the image acquisition module acquires high-definition images of the patient's face and upper body; the image preprocessing module preprocesses the acquired images, including filtering and denoising, image enhancement, grayscale conversion, and normalization, etc.; the feature extraction module extracts features from the preprocessed images, and the extracted features include color, texture, and shape; the image analysis module identifies and analyzes the images after feature extraction to timely identify the abnormal conditions of the patient; the comprehensive evaluation module fuses the data such as the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation with the analysis results of the image analysis module to comprehensively evaluate the physical condition of the patient after thoracic surgery and evaluate the lung function. The drainage monitoring module collects data such as the pressure of the drainage bag, the volume and color of the drainage fluid to monitor the wound drainage situation. The risk prediction module predicts the risk of the patient's rehabilitation training after thoracic surgery according to the patient's condition, combined with the analysis results of the patient's heart rate, respiratory rate, blood pressure, blood oxygen saturation data, the image analysis module, and the drainage monitoring module. When an abnormal condition is detected, the alarm module issues an alarm in time.

[0137] The nursing device of the present invention can be hung on the armrest of the hospital bed through a hook, which is convenient for patients to use when lying in bed; it can also be fixed on the back with the help of shoulder straps and waist straps to meet the needs of patients getting out of bed or outdoor rehabilitation training. Patients can stretch the elastic rope by holding the handle with both hands for chest expansion training, and can adjust the stretching strength and frequency according to their own physical conditions and rehabilitation progress to implement a personalized rehabilitation training plan, which helps to improve the pertinence and effectiveness of rehabilitation training and avoid causing additional burden to the body or affecting the rehabilitation effect due to improper training intensity. The image acquisition module accurately identifies abnormal conditions on the patient's face and upper body, such as pain expressions, abnormal breathing postures, etc. These intuitive physical manifestations can assist in judging the patient's physical condition. The comprehensive evaluation module integrates physiological data and image analysis results to comprehensively evaluate the patient's physical condition and lung function after thoracic surgery, and comprehensively judges the patient's rehabilitation progress from multiple angles, making the evaluation results more accurate and comprehensive, and providing strong support for medical staff to formulate scientific and reasonable treatment and rehabilitation plans. The drainage monitoring module collects and analyzes data such as the pressure, drainage volume, and color of the drainage bag in real time, and can timely detect abnormal conditions during the wound drainage process. For example, an abnormal increase in drainage pressure may indicate a blockage of the drainage tube, and abnormal drainage fluid color or a sharp change in volume may reflect wound healing problems or complications, providing accurate information for medical staff to timely handle wound-related problems, ensuring the safety and effectiveness of wound drainage, and promoting wound healing. The risk prediction module combines the patient's condition, physiological data, image analysis results, and drainage monitoring data to predict the risks of the rehabilitation training situation and identify in advance the risks that patients may face during the rehabilitation training process.

[0138] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A nursing device for improving lung function after thoracic surgery, comprising a backboard, a breathing mask, an air filter box, an elastic rope, a handle, a waste liquid receiving box, a suction pump and a monitoring mechanism, characterized in that: Shoulder straps are fixedly arranged on both sides of the upper end of the back plate, and waist straps are fixedly arranged on both sides of the lower end of the back plate; a hook is fixedly arranged on the upper end of the back plate; An air filter box and a waste liquid receiving box are fixedly arranged on both sides of the lower end of the back plate; an air filter screen and activated carbon are placed in the air filter box; the outlet of the air filter box is connected to a breathing mask through a hose; a mouthpiece is detachably fixedly arranged on the breathing mask; the shell of the breathing mask is made of transparent plastic material; a suction pump is fixedly arranged on the waste liquid receiving box, and a sputum suction tube is connected to the input end of the suction pump, and the sputum suction tube is slidably connected to the breathing mask; the output end of the suction pump extends into the waste liquid receiving box; Channels are opened on both sides of the lower end of the backboard, elastic ropes are fixedly arranged in the channels, and handles are fixedly arranged at the outer ends of the elastic ropes, and stretching and chest expansion training can be performed through the elastic ropes and handles; A monitoring mechanism is fixedly installed on the back panel to monitor the patient's physical condition after thoracic surgery and detect abnormal conditions in time.

2. The nursing device for improving lung function after thoracic surgery according to claim 1, characterized in that: A drainage bag placement box is fixedly arranged on one side of the back plate, a color sensor is fixedly arranged on the drainage bag placement box, a pressure sensor is arranged on the drainage bag, and the drainage bag placement box is made of transparent plastic material.

3. The nursing device for improving lung function after thoracic surgery according to claim 1, characterized in that: The monitoring agencies include: Data collection module: collects the patient's surgical data, historical medical conditions and physical data, and labels the data as a reference sample; Data acquisition module: including heart rate sensor, respiratory rate sensor, blood pressure sensor, and blood oxygen saturation sensor, which collect the patient's heart rate, respiratory rate, blood pressure, and blood oxygen saturation data in real time; Image acquisition module: HD cameras are fixed near the forehead and chin of the breathing mask to collect HD images of the patient's face and upper body; Image preprocessing module: preprocess the collected images, including filtering and denoising, image enhancement, grayscale and normalization; Feature extraction module: extracts features from the preprocessed image. The extracted features include color, texture and shape. Image analysis module: recognize and analyze the image after feature extraction to identify the patient's abnormal condition in time; Comprehensive evaluation module: integrates the patient's heart rate, respiratory rate, blood pressure and blood oxygen saturation data with the analysis results of the image analysis module to conduct a comprehensive evaluation of the patient's physical condition after thoracic surgery and evaluate the lung function; Drainage monitoring module: collects the pressure of the drainage bag, the amount and color data of the drainage fluid, and monitors the drainage of the wound; Risk prediction module: Based on the patient's condition, combined with the patient's heart rate, respiratory rate, blood pressure and blood oxygen saturation data and the analysis results of the image analysis module and drainage monitoring module, the risk prediction of the patient's post-thoracic surgery rehabilitation training is carried out; Alarm module: including alarm, which will send out alarm in time when abnormal situation is detected; Intelligent control module: connected to the data acquisition module, image acquisition module, image preprocessing module, feature extraction module, image analysis module, drainage monitoring module, comprehensive evaluation module and alarm module through a network.

4. The nursing device for improving lung function after thoracic surgery according to claim 3, characterized in that: The image analysis module promptly identifies abnormal conditions of the patient, including the following: Data collection: We collected a wide range of facial expression and body posture image data from patients undergoing thoracic surgery. For each image, we carefully observed the facial expression and body posture characteristics of the patient. Data preprocessing: Perform unified preprocessing operations on the collected raw image data; divide the labeled data into training set, validation set and test set; Build a convolutional neural network VGG1 model: select the VGG16 model, initialize all parameters of the VGG16 model, select the cross entropy loss function, and use the Adam optimizer; use the training set to train the model, use the validation set data to evaluate the model, and adjust the model's hyperparameters based on the evaluation results of the validation set; Model evaluation: Use the test set data to comprehensively evaluate the model; optimize the model based on the results of the model evaluation; Practical application: Input the feature data into the trained VGG16 model. After receiving the feature data, the model outputs the prediction result of the patient's current status. Result feedback: Once the model identifies an abnormal condition in a patient, the abnormal information will be fed back to the comprehensive evaluation module.

5. The nursing device for improving lung function after thoracic surgery according to claim 3, characterized in that: The comprehensive assessment module conducts a comprehensive assessment of the patient's physical condition after thoracic surgery, including the following contents: Data collection and integration: Obtain the patient's heart rate, respiratory rate, blood pressure and blood oxygen saturation physiological data in real time from the monitoring equipment, receive the analysis result data of the patient's facial expression and body posture from the image analysis module; collect the patient's historical medical data; Data preprocessing: clean the collected physiological data, organize the image analysis results, organize the surgical data and historical disease data, and convert them into a format that is easy to analyze; Constructing a multi-parameter comprehensive evaluation model: Selecting the Bayesian network model, based on a large amount of clinical data and medical research results, constructing a multi-parameter comprehensive evaluation model; the model contains association rules and weight distribution between various parameters to reflect the relative importance of different data to the assessment of the patient's physical condition; Data fusion analysis: The pre-processed physiological data, image analysis result data, surgical data and historical disease data are integrated according to the requirements of the comprehensive evaluation model; Physical condition assessment: Based on the post-fusion data, conduct a comprehensive assessment of the patient's physical condition after thoracic surgery; Generate patient physical condition report: Based on the evaluation results, the comprehensive evaluation module generates a detailed and reliable patient physical condition report.

6. The nursing device for improving lung function after thoracic surgery according to claim 5, characterized in that: The multi-parameter comprehensive evaluation model is: y={∑ n i＝1 [w i *x i *f(x i ) / (σ i 2 )]} / {∑ n i＝1 [w i / (σ i 2 )]};∑ n i＝1 (w i )=1; f(x i )=log b (x i ), where y is the comprehensive evaluation value after fusion; x i is the value of the ith physiological data or analysis result; w i is with x i The corresponding weight; σ i is the standard deviation of the ith data; f(x i ) is a nonlinear function used to capture the complex relationship between data; b is the base of the logarithm, b>0 and b≠1.

7. The nursing device for improving lung function after thoracic surgery according to claim 5, characterized in that: The model training optimization loss function is: L=-(1 / N)Σ M i＝1 {[γ i *y i log(y i ^)+(1-y i )log(1-y i ^)]+λ*Pe(y i ,y i ^)}; Pe(y i ,y i ^)={(y i ^-y i ) 2 , if y i =1;0,if y i =0}; where y i is the true label; y i ^ is the risk probability predicted by the model; M is the number of samples; γ i is the category weight, which is used to balance the uneven distribution of high-risk and low-risk samples; λ is the weight coefficient of the penalty term, which is used to control the intensity of the penalty term; Pe(y i ,y i ^) is the risk level penalty item.

8. The nursing device for improving lung function after thoracic surgery according to claim 3, characterized in that: The risk prediction module predicts the risk of patients' rehabilitation training after thoracic surgery, including the following: Data collection: Extract detailed medical information from the patient's medical records, collect the patient's heart rate, respiratory rate, blood pressure and blood oxygen saturation data in real time; obtain the image analysis module's analysis results on the patient's facial expression and body posture; Data integration: clean, standardize and normalize all collected data; Constructing a risk prediction model: Selecting a multi-layer perceptron (MLP) model to construct a risk prediction model; feature Engineering: Extracting features that have important impact on risk prediction from the integrated dataset; Model training optimization: Divide the preprocessed data into training set, validation set and test set, use the training set to train the selected model and adjust the model parameters; Use the validation set to monitor model performance and prevent overfitting, and use the test set to evaluate the generalization ability of the model; Risk prediction: Based on the input real-time data, according to the trained model parameters and algorithms, the probability of the patient experiencing risks in the current rehabilitation training stage is calculated, and the risk probability prediction results are output; Result feedback: Provide timely feedback of risk prediction results to medical staff and patients.

9. The nursing device for improving lung function after thoracic surgery according to claim 1, characterized in that: The risk prediction model is: P risk =Sigmoid{∑ n i＝1 [u i *g(x i )+b+a*HRV+β*Pain+C*tdr]}; g(x i )=log(1+x i );HRV 2 ={∑ N j＝1 [HR j -HR - ] 2 }; where x i is the input feature; u i It represents the contribution of each feature to risk prediction; b is the bias term; Sigmoid is the Sigmoid activation function; HRV is the heart rate variability characteristic, HR j is the jth heart rate value, HR - is the average value of heart rate, N is the number of heart rate values; Pain is the pain intensity feature output by the facial expression analysis module; a, β and C are weight coefficients used to adjust the contribution of HRV, Pain and drainage conditions to risk prediction; g(x i ) is the feature transformation function; P risk The output of the model is the risk probability; tdr is the drainage situation result monitored by the drainage monitoring module.

10. The nursing device for improving lung function after thoracic surgery according to claim 1, characterized in that: The drainage monitoring module monitors the wound drainage, including the following: Data collection: The pressure sensor monitors the drainage fluid pressure in real time; the liquid level sensor monitors the drainage fluid volume in real time; the color sensor monitors the drainage fluid color in real time; Data integration: Integrate the collected pressure data, liquid level data and color data, use time as the key index, associate different types of data at the same time point, and form a complete drainage data set; Data analysis: Use data analysis algorithms to conduct in-depth analysis of the integrated drainage data set; Abnormal judgment: According to the pre-set thresholds and judgment rules, the analyzed drainage data is judged for abnormalities to determine whether there are any abnormal situations; Monitoring result evaluation: Generate drainage monitoring report in real time; Combine the real-time report with historical data comparison results, and the data processing module will conduct a comprehensive evaluation of the patient's wound drainage situation; Abnormal warning: Once the data processing module determines that the drainage situation is abnormal, it will immediately issue a warning in a variety of ways.

Images