Intelligent management system and method for bronchoscope cleaning and tracing based on Internet of Things technology

Abstract

The present invention discloses an intelligent management system and method for tracing cleaning of bronchoscopes based on the Internet of Things technology, and relates to the field of medical equipment technology. The intelligent management system for tracing cleaning of bronchoscopes based on the Internet of Things technology includes: a cleaning data collection module, a cleaning data reverse analysis module, and a cleaning data backtracking module. The present invention realizes comprehensive monitoring and data collection of the bronchoscope cleaning process by setting a cleaning equipment set and a cleaning operation model, improves the accuracy and efficiency of the cleaning work, and performs cleaning judgments through dense connections and twin networks of the model; enhances the analysis capability of cleaning samples, and obtains cleaning feedback results through multi-scale feature extraction; realizes the traceability of cleaning data, and stores the cleaning data and feedback results in a database for easy management and query, ensuring continuous monitoring and improvement of cleaning quality.

Description

Technical Field

[0001] The present invention relates to the technical field of medical equipment, and in particular to an intelligent management system and method for bronchoscope cleaning and tracing based on Internet of Things technology. Background Art

[0002] Bronchoscope cleaning traceability ensures that the bronchoscope is strictly cleaned and disinfected after each use to avoid cross infection and ensure patient safety; by tracing the process and results of each cleaning and providing clear operation records, it helps to improve the transparency and standardization of medical equipment management, meet hygiene standards, and improve medical quality and safety; bronchoscope cleaning traceability usually manually records the cleaning time, personnel, cleaning method and other information, or uses barcodes, RFID tags and intelligent systems for automatic tracking. Common methods include paper records and electronic records, but there may be human recording errors, incomplete data, system incompatibility and difficulty in updating information in real time, which can easily lead to data loss or incomplete monitoring of the cleaning process, affecting the cleaning quality and accuracy of traceability, and it is impossible to record various parameters in the cleaning process. At the same time, there is no reverse analysis of contaminants in the cleaning process, and tracking and backtracing cannot be performed when problems or failures occur. Summary of the invention

[0003] The present invention aims to provide an intelligent management system and method for bronchoscope cleaning and tracing based on Internet of Things technology, which can comprehensively analyze and manage all operation data in the cleaning process.

[0004] The intelligent management method for bronchoscope cleaning traceability based on Internet of Things technology includes the following steps:

[0005] Set the cleaning equipment set Q, Q = {Q n |n=1, 2, ..., N}, where N is the total number of cleaning devices used to clean bronchoscopes; Q n represents a cleaning device in the cleaning device set Q; when any cleaning device Q in the cleaning device set Q n When a bronchoscope to be cleaned is received, the cleaning work is started, and all cleaning work is completed based on the bronchoscope cleaning operation model to obtain bronchoscope cleaning data; the bronchoscope cleaning data includes the bronchoscope identification number, bronchoscope cleaning process time, cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D n ;

[0006] The bronchoscope cleaning operation model includes a label recognition layer, a cleaning data collection layer, a cleaning judgment layer, a cleaning record layer, and a data output layer, wherein the cleaning judgment layer is constructed based on dense connections and twin networks;

[0007] Collect cleaning equipment Q during cleaning work n The corresponding cleaning sample is obtained to obtain the cleaning sample Y to be analyzed n ; Based on the cleaning sample Y to be analyzed n , bronchoscope cleaning data and bronchoscope cleaning sample tracing model to obtain bronchoscope cleaning feedback results;

[0008] The bronchoscope cleaning sample tracing model includes a sample data extraction layer, a feature extraction layer, an attention layer, and a feedback result output layer, which are used to perform multi-scale feature extraction to obtain feedback results;

[0009] The bronchoscope cleaning data and the bronchoscope cleaning feedback results are combined to obtain the bronchoscope cleaning traceability data; the bronchoscope cleaning traceability data is stored in a preset database to complete a bronchoscope cleaning traceability intelligent management.

[0010] As a preferred technical solution of the present invention, the label identification layer is used to obtain the bronchoscope identification number of the bronchoscope to be cleaned; the start time of identifying the bronchoscope identification number is recorded as the start time of bronchoscope cleaning;

[0011] The cleaning data collection layer is used to collect data based on the cleaning equipment Q n Get the image P of the bronchoscope to be cleaned before cleaning n And the cleaned image H n ; The image P before cleaning n And the cleaned image H n Combine to obtain the bronchoscope cleaning image set X n , and record the cleaning equipment Q n Cleaning parameters C n ;

[0012] The cleaning judgment layer is used for the bronchoscope cleaning image set X to be cleaned n Make a judgment and obtain the bronchoscope cleaning completion degree D n ;

[0013] The cleaning judgment layer is constructed based on dense connections and twin networks;

[0014] The cleaning record layer is used to obtain records of all cleaning devices Q in the cleaning device set Q. n The corresponding bronchoscope cleaning image set X nThe time is recorded as the end cleaning time of the bronchoscope; based on the bronchoscope start cleaning time and the bronchoscope end cleaning time, the bronchoscope cleaning process time is obtained; it is judged whether the bronchoscope cleaning process time meets the preset bronchoscope cleaning time threshold, if it meets, no operation is performed; if the bronchoscope starts cleaning time and exceeds the preset bronchoscope cleaning time threshold but still does not receive the bronchoscope end cleaning time, it is not in compliance, and a reminder signal is sent during the cleaning operation;

[0015] The data output layer is used to convert the bronchoscope identification number, bronchoscope cleaning process time, and cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D n Combine to obtain bronchoscope cleaning data; output the bronchoscope cleaning data.

[0016] As a preferred technical solution of the present invention, in the cleaning judgment layer, the cleaning image set X of the bronchoscope to be cleaned is n The specific steps to make a judgment include:

[0017] The cleaning judgment layer includes an image preprocessing layer, a feature learning layer, and a result output layer;

[0018] In the image preprocessing layer, the cleaned image P n And the cleaned image H n Perform preprocessing to obtain the image P before preprocessing and cleaning n ' and the pre-processed cleaned image H n ';

[0019] Mining the pre-processed cleaned image P in the feature learning layer n ' and the pre-processed cleaned image H n 'Change characteristics, get the cleaning contrast change characteristics T n ;

[0020] The result output layer is used to compare the change features based on cleaning T n Perform feature extraction to obtain the bronchoscope cleaning completion degree D n ;

[0021] In the feature learning layer, two parallel densely connected convolutional blocks are constructed to pre-process the cleaned image P. n ' and the pre-processed cleaned image H n 'Downsample to obtain the downsampled convolution feature W (P n ) and the downsampled convolutional features W(H n ); At the same time, based on the twin network, the downsampled convolution feature W (P n ) and the downsampled convolutional features W(H n) to perform feature fusion and obtain the down-sampled fused convolution feature R n ; Downsample and fuse convolution features R n After upsampling through densely connected convolutional blocks, the cleaning contrast change feature T is obtained. n ;

[0022] Collecting several groups of cleaning variation training samples; each group of cleaning variation training samples includes a preprocessed image set and a corresponding degree of variation; combining several groups of cleaning variation training samples to obtain a cleaning variation training set;

[0023] The cleaning judgment layer is trained using the cleaning variation training set to obtain an initial cleaning judgment layer; the initial cleaning judgment layer is evaluated to obtain an initial cleaning judgment layer model evaluation result; if the initial cleaning judgment layer model evaluation result is passed, the initial cleaning judgment layer is used as the cleaning judgment layer in the bronchoscope cleaning operation model; otherwise, the model training is continued using the cleaning variation training set.

[0024] As a preferred technical solution of the present invention, the specific steps of determining the preset bronchoscope cleaning time threshold include:

[0025] Construct K preset cleaning time threshold individuals S k , preset cleaning time threshold individual S k Generate based on the historical average cleaning time; set K preset cleaning time threshold individuals S k Combine to obtain a preset cleaning time threshold iterative population; set the maximum number of iterations;

[0026] A bronchoscope cleaning digital twin model is constructed based on the historical cleaning data of bronchoscopes; a pre-set cleaning time threshold individual S k The bronchoscope cleaning digital twin model was used to simulate cleaning operations multiple times, and the change values ​​of the bronchoscope cleaning completion degree were recorded during the multiple simulated cleaning operations to obtain the mean value of the simulated secondary contamination change. The mean value of the simulated secondary contamination change was used as the preset cleaning time threshold individual S k The fitness V k ;

[0027] When the maximum number of iterations is reached, the preset cleaning time threshold individual corresponding to the maximum fitness is output, which is the optimal preset cleaning time threshold individual; the preset bronchoscope cleaning time threshold is set based on the optimal preset cleaning time threshold individual.

[0028] As a preferred technical solution of the present invention,

[0029] The sample data extraction layer is used to analyze the clean sample Y n Perform data extraction to obtain the component data B of the cleaned sample to be analyzed n; The component data of the cleaned sample to be analyzed B n , Cleaning parameters C n Combine to obtain the feedback data F of the cleaning sample to be analyzed n ;

[0030] The feature extraction layer is used to analyze the clean sample feedback data F n Perform feature extraction to obtain the feedback data feature F of the cleaning sample to be analyzed n ';

[0031] The attention layer is used to analyze the feedback data features F of the cleaned samples n 'Weighted feature extraction to obtain the weighted feature J of the feedback data of the cleaned sample to be analyzed n ;

[0032] The feedback result output layer is used to feedback data features F based on the cleaned samples to be analyzed n 'Perform feature classification and obtain bronchoscope cleaning feedback results.

[0033] As a preferred technical solution of the present invention, in the feature extraction layer, the feedback data F of the cleaned sample to be analyzed is n The specific steps for feature extraction include:

[0034] The feature extraction layer includes convolution layer, batch normalization layer and pooling layer; Convolution kernels of different sizes are used for feature extraction;

[0035] Using the formula Represents a convolution operation; where represents the convolution kernel Layer outputs, Indicates Tier The convolution kernel Weight value, =1, 2, …, , Indicates Tier A bias, is the convolution operation;

[0036] All Input to the batch normalization layer and pooling layer for feature extraction to obtain the feedback data feature F of the cleaned sample to be analyzed n '.

[0037] As a preferred technical solution of the present invention, in the attention layer, the feedback data feature F of the cleaned sample to be analyzed is n The specific steps of weighted feature extraction include:

[0038] Feedback data feature F of the cleaned sample to be analyzed n 'Converted into the characteristic sequence F of the feedback data of the cleaning sample to be analyzed n ''; F n ''=[F n1 '', F n2 '', ..., F nU ''], U is the characteristic sequence F of the feedback data of the cleaning sample to be analyzed n '' length;

[0039] The characteristic sequence F of the feedback data of the cleaning sample to be analyzed n ''Perform average pooling and normalization in the channel dimension to obtain the channel attention weight α of the feature sequence of the feedback data of the cleaned sample to be analyzed; α=[α 1 , α 2 , …, α U ]; The channel attention weight α of the feedback data feature sequence of the cleaned sample to be analyzed and the feature sequence F of the feedback data of the cleaned sample to be analyzed n ''Perform element-by-element multiplication to obtain the weighted feature J of the feedback data of the cleaned sample to be analyzed n .

[0040] The intelligent management system for bronchoscope cleaning and tracing based on Internet of Things technology includes:

[0041] The cleaning data collection module includes a data preparation unit and a data collection unit; the data preparation unit is used to set the cleaning equipment set Q, Q={Q n |n=1, 2, ..., N}, where N is the total number of cleaning devices used to clean bronchoscopes; Q n represents a cleaning device in the cleaning device set Q; the data collection unit is used when any cleaning device Q in the cleaning device set Q n When a bronchoscope to be cleaned is received, the cleaning work is started, and all cleaning work is completed based on the bronchoscope cleaning operation model to obtain bronchoscope cleaning data; the bronchoscope cleaning data includes the bronchoscope identification number, bronchoscope cleaning process time, cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D n The bronchoscope cleaning operation model includes a label recognition layer, a cleaning data collection layer, a cleaning judgment layer, a cleaning record layer, and a data output layer, wherein the cleaning judgment layer is constructed based on dense connections and twin networks;

[0042] The cleaning data reverse analysis module includes a feedback judgment unit; the feedback judgment unit is used to collect the Q of the cleaning equipment during the cleaning process. nThe corresponding cleaning sample is obtained to obtain the cleaning sample Y to be analyzed n ; Based on the cleaning sample Y to be analyzed n , bronchoscope cleaning data and bronchoscope cleaning sample tracing model are analyzed to obtain bronchoscope cleaning feedback results; the bronchoscope cleaning sample tracing model includes a sample data extraction layer, a feature extraction layer, an attention layer and a feedback result output layer, which are used to perform multi-scale feature extraction to obtain feedback results;

[0043] The cleaning data backtracking module includes a data management unit; the data management unit is used to combine the bronchoscope cleaning data and the bronchoscope cleaning feedback results to obtain the bronchoscope cleaning tracing data; the bronchoscope cleaning tracing data is stored in a preset database to complete a bronchoscope cleaning tracing intelligent management.

[0044] The present invention has the following advantages:

[0045] The present invention realizes comprehensive monitoring and data collection of the bronchoscope cleaning process by setting a cleaning equipment set and a cleaning operation model, improves the accuracy and efficiency of the cleaning work, and makes cleaning judgments through dense connections and twin networks of the model; enhances the analysis capability of cleaning samples, and obtains cleaning feedback results through multi-scale feature extraction; realizes the traceability of cleaning data, and stores the cleaning data and feedback results in a database for easy management and query, thereby ensuring continuous monitoring and improvement of cleaning quality.

[0046] The present invention can significantly improve the efficiency and quality of bronchoscope cleaning and ensure the standardization and regularization of cleaning work through automated cleaning process and real-time monitoring; by collecting and storing cleaning data, the whole tracing of the bronchoscope cleaning process is achieved, which helps to quickly locate and solve problems when they occur, and improves the management level and safety of medical equipment; by ensuring high standards and traceability of the cleaning process, the hygiene standards of medical equipment are improved, thereby ensuring the safety of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a schematic diagram of the structure of the bronchoscope cleaning and tracing intelligent management system based on the Internet of Things technology adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0049] Example 1, a bronchoscope cleaning and tracing intelligent management method based on Internet of Things technology, comprising the following steps:

[0050] Set the cleaning equipment set Q, Q = {Qn |n=1, 2, ..., N}, where N is the total number of cleaning devices used to clean bronchoscopes; Q n represents the cleaning equipment in the cleaning equipment set Q;

[0051] The cleaning equipment for bronchoscopes includes a series of specialized equipment and tools to ensure the thorough cleaning, disinfection and sterilization of bronchoscopes; generally including ultrasonic cleaning machines, high temperature and high pressure disinfection and sterilization machines, automated cleaning and disinfection machines, drying equipment, etc.; the Internet of Things technology is used on the cleaning equipment for data transmission and image acquisition operations for subsequent analysis operations;

[0052] When any cleaning device Q in the cleaning device set Q n When a bronchoscope to be cleaned is received, the cleaning work is started, and all cleaning work is completed based on the bronchoscope cleaning operation model to obtain bronchoscope cleaning data; the bronchoscope cleaning data includes the bronchoscope identification number, bronchoscope cleaning process time, cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D n ;

[0053] The bronchoscope cleaning operation model includes a label recognition layer, a cleaning data collection layer, a cleaning judgment layer, a cleaning record layer, and a data output layer, wherein the cleaning judgment layer is constructed based on dense connections and twin networks;

[0054] The tag identification layer is used to obtain the bronchoscope identification number of the bronchoscope to be cleaned; the start time of identifying the bronchoscope identification number is recorded as the start time of bronchoscope cleaning; the bronchoscope identification number can be obtained through RFID technology, and the bronchoscope to be cleaned is automatically received and the cleaning task is started, thereby avoiding manual intervention and reducing operational errors;

[0055] The cleaning data collection layer is used to collect data based on the cleaning equipment Q n Get the image P of the bronchoscope to be cleaned before cleaning n And the cleaned image H n ; The image P before cleaning n And the cleaned image H n Combine to obtain the bronchoscope cleaning image set X n , and record the cleaning equipment Q n Cleaning parameters C n ;

[0056] The cleaning judgment layer is used for the bronchoscope cleaning image set X to be cleaned n Make a judgment and obtain the bronchoscope cleaning completion degree D n ;

[0057] The cleaning judgment layer is constructed based on dense connections and twin networks;

[0058] The cleaning record layer is used to obtain records of all cleaning devices Q in the cleaning device set Q. n The corresponding bronchoscope cleaning image set X n The time is recorded as the end cleaning time of the bronchoscope; based on the bronchoscope start cleaning time and the bronchoscope end cleaning time, the bronchoscope cleaning process time is obtained; it is judged whether the bronchoscope cleaning process time meets the preset bronchoscope cleaning time threshold, if it meets, no operation is performed; if the bronchoscope starts cleaning time and exceeds the preset bronchoscope cleaning time threshold but still does not receive the bronchoscope end cleaning time, it is not in compliance, and a reminder signal is sent during the cleaning operation;

[0059] The data output layer is used to convert the bronchoscope identification number, bronchoscope cleaning process time, and cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D n Combine to obtain bronchoscope cleaning data; output bronchoscope cleaning data; the cleaning data includes bronchoscope identification number, cleaning process time, cleaning parameters, cleaning image set and cleaning completion degree and other information. The data of each cleaning task is recorded in detail to ensure that the process and effect of each device and each cleaning operation can be traced;

[0060] The cleaning judgment layer based on dense connections and twin networks can intelligently analyze the cleaning status of the bronchoscope to determine whether the cleaning meets the standards, thus avoiding the subjectivity and errors of manual judgment. Through the cleaning data collection layer and the cleaning record layer, each operation data can be recorded and stored in real time, and the cleaning process and effect can be viewed and analyzed at any time, providing a basis for subsequent operations. The cleaning completion degree D n As a quantitative indicator of cleaning quality, it can accurately evaluate the effect of each cleaning; through comprehensive analysis of cleaning parameters and image sets, it can determine whether the cleaning effect meets the standard, discover potential problems and optimize them in time; based on the bronchoscope cleaning image set X n ,The twin network can perform image analysis to identify whether the cleaning is thorough, avoid incomplete cleaning or missing parts, and further improve the cleaning quality;

[0061] For example, when the cleaning completion degree is 80%, the Internet of Things technology can be used to automatically prompt that cleaning should be continued and automatically adjust the cleaning parameters or processes; cleaning parameter C nThe system contains cleaning parameters set for different cleaning equipment, such as cleaning solution concentration, cleaning time, temperature, pressure, etc., which can be used for subsequent analysis operations. Through complete cleaning records, cleaning parameter tracking and image analysis, the system can provide hospitals with complete compliance reports to ensure that each cleaning meets the cleaning and disinfection standards of medical devices.

[0062] In the cleaning judgment layer, the bronchoscope cleaning image set X is to be cleaned n The specific steps to make a judgment include:

[0063] The cleaning judgment layer includes an image preprocessing layer, a feature learning layer, and a result output layer;

[0064] In the image preprocessing layer, the cleaned image P n And the cleaned image H n Perform preprocessing to obtain the image P before preprocessing and cleaning n ' and the pre-processed cleaned image H n ';

[0065] Mining the pre-processed cleaned image P in the feature learning layer n ' and the pre-processed cleaned image H n 'Change characteristics, get the cleaning contrast change characteristics T n ;

[0066] The result output layer is used to compare the change features based on cleaning T n Perform feature extraction to obtain the bronchoscope cleaning completion degree D n ;

[0067] In the feature learning layer, two parallel densely connected convolutional blocks are constructed to pre-process the cleaned image P. n ' and the pre-processed cleaned image H n 'Downsample to obtain the downsampled convolution feature W (P n ) and the downsampled convolutional features W(H n ); At the same time, based on the twin network, the downsampled convolution feature W (P n ) and the downsampled convolutional features W(H n ) to perform feature fusion and obtain the down-sampled fused convolution feature R n ; Downsample and fuse convolution features R n After upsampling through densely connected convolutional blocks, the cleaning contrast change feature T is obtained. n ;

[0068] Collecting several groups of cleaning variation training samples; each group of cleaning variation training samples includes a preprocessed image set and a corresponding degree of variation; combining several groups of cleaning variation training samples to obtain a cleaning variation training set;

[0069] The cleaning judgment layer is trained using the cleaning variation training set to obtain an initial cleaning judgment layer; the initial cleaning judgment layer is evaluated to obtain an initial cleaning judgment layer model evaluation result; if the initial cleaning judgment layer model evaluation result is passed, the initial cleaning judgment layer is used as the cleaning judgment layer in the bronchoscope cleaning operation model; otherwise, the model training is continued using the cleaning variation training set;

[0070] By comparing and analyzing the images before and after cleaning, the cleaning effect of the bronchoscope can be automatically evaluated. Image preprocessing, feature extraction, and change feature mining can deeply capture the subtle differences between the images before and after cleaning, so as to accurately judge the cleaning completion degree; through the processing of the feature learning layer, the image differences before and after cleaning are accurately extracted as change features, which provides strong data support for the quantitative evaluation of cleaning quality; in the image preprocessing layer, the images before and after cleaning are processed to obtain standardized images, which enables subsequent feature learning to focus on the key changes in the image and improve the learning efficiency and accuracy of the model; downsampling through densely connected convolutional blocks can extract important feature information from the image, which helps to reduce noise, increase sensitivity to key changes, and provide support for subsequent feature It provides accurate data for feature fusion and cleaning effect evaluation; through the feature fusion method of the twin network, the system can efficiently fuse the image features before and after cleaning to obtain more accurate change features, thereby improving the automation level of cleaning traceability management; through the design of downsampling convolution and upsampling convolution, the system can extract features at multiple scales and capture subtle changes in the image. This multi-scale feature learning not only enhances the expressiveness of the model, but also improves the ability to recognize complex cleaning changes; the densely connected convolution block set ensures efficient flow of information in the network through dense connections between feature layers, avoids information loss or degradation, and improves the expressiveness of image features and the accuracy of cleaning judgment;

[0071] Specific steps to determine the preset bronchoscope cleaning time threshold include:

[0072] Construct K preset cleaning time threshold individuals S k , preset cleaning time threshold individual S k Generate based on the historical average cleaning time; set K preset cleaning time threshold individuals S k Combination, to obtain a preset cleaning time threshold iteration population; set the maximum number of iterations; the maximum number of iterations is set by professional technicians according to actual conditions;

[0073] A bronchoscope cleaning digital twin model is constructed based on the historical cleaning data of bronchoscopes; a pre-set cleaning time threshold individual S kThe bronchoscope cleaning digital twin model was used to simulate cleaning operations multiple times, and the change values ​​of the bronchoscope cleaning completion degree were recorded during the multiple simulated cleaning operations to obtain the mean value of the simulated secondary contamination change. The mean value of the simulated secondary contamination change was used as the preset cleaning time threshold individual S k The fitness V k ;

[0074] When the maximum number of iterations is reached, the preset cleaning time threshold individual corresponding to the maximum fitness is output, which is the optimal preset cleaning time threshold individual; the preset bronchoscope cleaning time threshold is set based on the optimal preset cleaning time threshold individual;

[0075] The construction of the digital twin model enables the system to simulate cleaning operations multiple times in a virtual environment. The simulated mean of secondary contamination changes provides a dynamic evaluation of the cleaning effect, thereby helping to further optimize the cleaning time. In actual operation, if the cleaning time is too long, it may be secondary contaminated by pollutants in the air during the cleaning process, so the total duration of the cleaning time should be controlled within a reasonable range. If the cleaning time is too short, the cleaning may not be thorough, and the cleaning fluid may not completely remove the dirt or microorganisms on the surface of the bronchoscope. If the cleaning time is too long, it may lead to waste of resources and increase the potential risk of infection. By optimizing the cleaning time, the cleaning efficiency can be improved and unnecessary exposure can be reduced while ensuring thorough cleaning.

[0076] Collect cleaning equipment Q during cleaning work n The corresponding cleaning sample is obtained to obtain the cleaning sample Y to be analyzed n ; Based on the cleaning sample Y to be analyzed n , bronchoscope cleaning data and bronchoscope cleaning sample tracing model to obtain bronchoscope cleaning feedback results;

[0077] The bronchoscope cleaning sample tracing model includes a sample data extraction layer, a feature extraction layer, an attention layer, and a feedback result output layer, which are used to perform multi-scale feature extraction to obtain feedback results;

[0078] The sample data extraction layer is used to analyze the clean sample Y n Perform data extraction to obtain the component data B of the cleaned sample to be analyzed n ; The component data of the cleaned sample to be analyzed B n , Cleaning parameters C n Combine to obtain the feedback data F of the cleaning sample to be analyzed n ;

[0079] The feature extraction layer is used to analyze the clean sample feedback data F n Perform feature extraction to obtain the feedback data feature F of the cleaning sample to be analyzed n ';

[0080] The attention layer is used to analyze the feedback data features F of the cleaned samples n 'Weighted feature extraction to obtain the weighted feature J of the feedback data of the cleaned sample to be analyzed n ;

[0081] The feedback result output layer is used to feedback data features F based on the cleaned samples to be analyzed n 'Perform feature classification and obtain bronchoscope cleaning feedback results;

[0082] By extracting data from the cleaning samples to be analyzed through the sample data extraction layer, key component data in the cleaning process can be effectively obtained, providing accurate input for subsequent analysis. This step ensures a comprehensive record of the cleaning process, allowing the cleaning samples to be traced in detail, providing a scientific basis for analytical decision-making;

[0083] During the use of the bronchoscope, organic substances such as cell fragments, blood, mucus, etc. from the patient may accumulate. These organic pollutants are usually the main target of the cleaning process; the quantity or quality of pollutants washed away needs to be quantified in the cleaning sample, which can be achieved by analyzing the cleaning water, residues, pollutant deposition, etc. The residual amount of pollutants is an important indicator of the cleaning effect; the cleaning sample can be used to analyze whether there are residual pathogens on the surface of the bronchoscope, especially bacteria, fungi and viruses. Microbial detection is a very critical part of the bronchoscope cleaning process. Preventing secondary contamination and cross infection is the main goal of the cleaning process. The cleaning samples to be analyzed are the cleaning contaminant samples of different equipment during the cleaning process, which are used for reverse analysis of the cleaning effect, or for subsequent cleaning traceability management;

[0084] The component data of the cleaning samples to be analyzed are combined with the cleaning parameters to form the feedback data of the cleaning samples to be analyzed. This data not only covers the characteristics of the samples themselves, but also includes the control parameters in the cleaning process, making the feedback data more representative, comprehensive and accurate. The feature extraction layer can mine information of different scales in the data and capture multi-level change characteristics by extracting features from the feedback data of the cleaning samples to be analyzed. This feature extraction not only helps to identify key factors in the cleaning process, but also discovers potential influencing factors and helps to comprehensively evaluate the cleaning effect. With the introduction of the attention layer, the model can perform weighted feature extraction on the features of the feedback data of the cleaning samples to be analyzed, and assign more appropriate weights to different features. This helps the model to automatically focus on features that have a greater impact on the cleaning results during training, avoid interference from irrelevant information, and thus improve the model's ability to identify key factors. The feedback results of bronchoscope cleaning are not only based on the analysis results of image data, but also combine multi-dimensional data such as cleaning parameters and changes in components during the cleaning process. This comprehensive analysis can comprehensively evaluate the cleaning process, ensure multi-angle verification of the cleaning effect, and improve the credibility of the cleaning quality.

[0085] In the feature extraction layer, the clean sample feedback data F is analyzed n The specific steps for feature extraction include:

[0086] The feature extraction layer includes convolution layer, batch normalization layer and pooling layer; Convolution kernels of different sizes are used for feature extraction;

[0087] Using the formula Represents a convolution operation; where represents the convolution kernel Layer outputs, Indicates Tier The convolution kernel Weight value, =1, 2, …, , Indicates Tier A bias, is the convolution operation;

[0088] All Input to the batch normalization layer and pooling layer for feature extraction to obtain the feedback data feature F of the cleaned sample to be analyzed n '.

[0089] In the convolutional layer, Convolution kernels of different sizes are used to extract features of various scales of the input data. This multi-scale feature extraction can help the model capture the change information of different levels and ranges in the cleaned data. Convolution kernels of different sizes are suitable for extracting local features of different sizes, thereby increasing the expressiveness of the model; convolution kernels of different sizes can perform in-depth analysis on the details and overall structure of the feedback data of the cleaned samples to be analyzed. Large convolution kernels can extract global and relatively macro features, while small convolution kernels can pay more attention to the slight changes in the data. This multi-scale feature extraction can more comprehensively capture the detailed changes in the cleaning process and improve the model's ability to understand the cleaning process; the weights and biases of the convolution kernels are adaptively optimized through back propagation during the training process, so that the model can automatically learn the most effective feature expression according to the characteristics of the cleaned samples. In this way, the model can adapt to the changes of different cleaned samples and extract the most useful features;

[0090] In the attention layer, the feedback data features F of the cleaned samples to be analyzed n The specific steps of weighted feature extraction include:

[0091] Feedback data feature F of the cleaned sample to be analyzed n 'Converted into the characteristic sequence F of the feedback data of the cleaning sample to be analyzed n ''; F n ''=[F n1 '', F n2 '', ..., F nU ''], U is the characteristic sequence F of the feedback data of the cleaning sample to be analyzed n '' length;

[0092] The characteristic sequence F of the feedback data of the cleaning sample to be analyzed n ''Perform average pooling and normalization in the channel dimension to obtain the channel attention weight α of the feature sequence of the feedback data of the cleaned sample to be analyzed; α=[α 1 , α 2 , …, α U ]; The channel attention weight α of the feedback data feature sequence of the cleaned sample to be analyzed and the feature sequence F of the feedback data of the cleaned sample to be analyzed n ''Perform element-by-element multiplication to obtain the weighted feature J of the feedback data of the cleaned sample to be analyzed n ;

[0093] By performing average pooling and normalization operations on the feature sequence in the channel dimension, the channel attention weight is calculated. The pooling operation helps capture the global information in the data, and normalization helps eliminate the dimensional differences between different channels; average pooling can reduce unnecessary local fluctuations and noise, and normalization ensures that the weights are compared under the same standard, enhancing the stability and robustness of the model when processing different feature dimensions; by calculating the channel attention weight, the model can selectively focus on features that are more valuable to the task, avoiding the processing of irrelevant features, thereby improving the efficiency of feature selection; through weighted feature extraction, the model can use more informative features to make decisions, so that the model does not rely solely on a single input feature, but makes comprehensive judgments through weighted features, thereby improving task performance;

[0094] The bronchoscope cleaning data and the bronchoscope cleaning feedback results are combined to obtain the bronchoscope cleaning traceability data; the bronchoscope cleaning traceability data is stored in a preset database, and a bronchoscope cleaning traceability intelligent management is completed;

[0095] Through automated cleaning processes and real-time monitoring, the efficiency and quality of bronchoscope cleaning can be significantly improved, ensuring standardization and regularization of cleaning work; by collecting and storing cleaning data, the entire bronchoscope cleaning process can be traced, which helps to quickly locate and solve problems when they arise, improving the management level and safety of medical equipment; by ensuring high standards and traceability of the cleaning process, the hygiene standards of medical devices are improved, thereby ensuring patient safety.

[0096] Example 2, intelligent management system for bronchoscope cleaning and tracing based on Internet of Things technology, see Figure 1 As shown, including:

[0097] The cleaning data collection module includes a data preparation unit and a data collection unit; the data preparation unit is used to set the cleaning equipment set Q, Q={Q n |n=1, 2, ..., N}, where N is the total number of cleaning devices used to clean bronchoscopes; Q n represents a cleaning device in the cleaning device set Q; the data collection unit is used when any cleaning device Q in the cleaning device set Q n When a bronchoscope to be cleaned is received, the cleaning work is started, and all cleaning work is completed based on the bronchoscope cleaning operation model to obtain bronchoscope cleaning data; the bronchoscope cleaning data includes the bronchoscope identification number, bronchoscope cleaning process time, cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D nThe bronchoscope cleaning operation model includes a label recognition layer, a cleaning data collection layer, a cleaning judgment layer, a cleaning record layer, and a data output layer, wherein the cleaning judgment layer is constructed based on dense connections and twin networks;

[0098] The cleaning data reverse analysis module includes a feedback judgment unit; the feedback judgment unit is used to collect the Q of the cleaning equipment during the cleaning process. n The corresponding cleaning sample is obtained to obtain the cleaning sample Y to be analyzed n ; Based on the cleaning sample Y to be analyzed n , bronchoscope cleaning data and bronchoscope cleaning sample tracing model are analyzed to obtain bronchoscope cleaning feedback results; the bronchoscope cleaning sample tracing model includes a sample data extraction layer, a feature extraction layer, an attention layer and a feedback result output layer, which are used to perform multi-scale feature extraction to obtain feedback results;

[0099] The cleaning data backtracking module includes a data management unit; the data management unit is used to combine the bronchoscope cleaning data and the bronchoscope cleaning feedback results to obtain the bronchoscope cleaning tracing data; the bronchoscope cleaning tracing data is stored in a preset database to complete a bronchoscope cleaning tracing intelligent management.

[0100] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. An intelligent management method for bronchoscope cleaning and tracing based on Internet of Things technology, characterized in that: The following steps are involved: Set the cleaning equipment set Q, Q = {Q n |n=1, 2, ..., N}, where N is the total number of cleaning devices used to clean bronchoscopes; Q n represents a cleaning device in the cleaning device set Q; when any cleaning device Q in the cleaning device set Q n When a bronchoscope to be cleaned is received, the cleaning work is started, and all cleaning work is completed based on the bronchoscope cleaning operation model to obtain bronchoscope cleaning data; the bronchoscope cleaning data includes the bronchoscope identification number, bronchoscope cleaning process time, cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D n ; The bronchoscope cleaning operation model includes a label recognition layer, a cleaning data collection layer, a cleaning judgment layer, a cleaning record layer, and a data output layer, wherein the cleaning judgment layer is constructed based on dense connections and twin networks; Collect cleaning equipment Q during cleaning work n The corresponding cleaning sample is obtained to obtain the cleaning sample Y to be analyzed n ; Based on the cleaning sample Y to be analyzed n , bronchoscope cleaning data and bronchoscope cleaning sample tracing model to obtain bronchoscope cleaning feedback results; The bronchoscope cleaning sample tracing model includes a sample data extraction layer, a feature extraction layer, an attention layer, and a feedback result output layer, which are used to perform multi-scale feature extraction to obtain feedback results; The bronchoscope cleaning data and the bronchoscope cleaning feedback results are combined to obtain the bronchoscope cleaning traceability data; the bronchoscope cleaning traceability data is stored in a preset database, and a bronchoscope cleaning traceability intelligent management is completed; The label identification layer is used to obtain the bronchoscope identification number of the bronchoscope to be cleaned; the start time of identifying the bronchoscope identification number is recorded as the start time of bronchoscope cleaning; The cleaning data collection layer is used to collect data based on the cleaning equipment Q n Get the image P of the bronchoscope to be cleaned before cleaning n And the cleaned image H n ; The image P before cleaning n And the cleaned image H n Combine to obtain the bronchoscope cleaning image set X n , and record the cleaning equipment Q n Cleaning parameters C n ; The cleaning judgment layer is used for the bronchoscope cleaning image set X to be cleaned n Make a judgment and obtain the bronchoscope cleaning completion degree D n ; The cleaning judgment layer is constructed based on dense connections and twin networks; The cleaning record layer is used to obtain records of all cleaning devices Q in the cleaning device set Q. n The corresponding bronchoscope cleaning image set X n The time is recorded as the end cleaning time of the bronchoscope; based on the bronchoscope start cleaning time and the bronchoscope end cleaning time, the bronchoscope cleaning process time is obtained; it is judged whether the bronchoscope cleaning process time meets the preset bronchoscope cleaning time threshold, if it meets, no operation is performed; if the bronchoscope starts cleaning time and exceeds the preset bronchoscope cleaning time threshold but still does not receive the bronchoscope end cleaning time, it is not in compliance, and a reminder signal is sent during the cleaning operation; The data output layer is used to convert the bronchoscope identification number, bronchoscope cleaning process time, and cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D n Combine to obtain bronchoscope cleaning data; output the bronchoscope cleaning data.

2. The method for tracing and intelligent management of bronchoscope cleaning based on Internet of Things technology according to claim 1 is characterized in that: In the cleaning judgment layer, the bronchoscope cleaning image set X is to be cleaned n The specific steps to make a judgment include: The cleaning judgment layer includes an image preprocessing layer, a feature learning layer, and a result output layer; In the image preprocessing layer, the cleaned image P n And the cleaned image H n Perform preprocessing to obtain the image P before preprocessing and cleaning n ' and the pre-processed cleaned image H n '; Mining the pre-processed cleaned image P in the feature learning layer n ' and the pre-processed cleaned image H n 'Change characteristics, get the cleaning contrast change characteristics T n ; The result output layer is used to compare the change features based on cleaning T n Perform feature extraction to obtain the bronchoscope cleaning completion degree D n ; In the feature learning layer, two parallel densely connected convolutional blocks are constructed to pre-process the cleaned image P. n ' and the pre-processed cleaned image H n 'Downsample to obtain the downsampled convolution feature W (P n ) and the downsampled convolutional features W(H n ); At the same time, based on the twin network, the downsampled convolution feature W (P n ) and the downsampled convolutional features W(H n ) to perform feature fusion and obtain the down-sampled fused convolution feature R n ; Downsample and fuse convolution features R n After upsampling through densely connected convolutional blocks, the cleaning contrast change feature T is obtained. n ; Collecting several groups of cleaning variation training samples; each group of cleaning variation training samples includes a preprocessed image set and a corresponding degree of variation; combining several groups of cleaning variation training samples to obtain a cleaning variation training set; The cleaning judgment layer is trained using the cleaning variation training set to obtain an initial cleaning judgment layer; the initial cleaning judgment layer is evaluated to obtain an initial cleaning judgment layer model evaluation result; if the initial cleaning judgment layer model evaluation result is passed, the initial cleaning judgment layer is used as the cleaning judgment layer in the bronchoscope cleaning operation model; otherwise, the model training is continued using the cleaning variation training set.

3. The method for tracing and intelligent management of bronchoscope cleaning based on Internet of Things technology according to claim 2 is characterized in that: Specific steps to determine the preset bronchoscope cleaning time threshold include: Construct K preset cleaning time threshold individuals S k , preset cleaning time threshold individual S k Generate based on the historical average cleaning time; set K preset cleaning time threshold individuals S k Combine to obtain a preset cleaning time threshold iterative population; set the maximum number of iterations; A bronchoscope cleaning digital twin model is constructed based on the historical cleaning data of bronchoscopes; a pre-set cleaning time threshold individual S k The bronchoscope cleaning digital twin model was used to simulate cleaning operations multiple times, and the change values ​​of the bronchoscope cleaning completion degree were recorded during the multiple simulated cleaning operations to obtain the mean value of the simulated secondary contamination change. The mean value of the simulated secondary contamination change was used as the preset cleaning time threshold individual S k The fitness V k ; When the maximum number of iterations is reached, the preset cleaning time threshold individual corresponding to the maximum fitness is output, which is the optimal preset cleaning time threshold individual; the preset bronchoscope cleaning time threshold is set based on the optimal preset cleaning time threshold individual.

4. The method for tracing and intelligent management of bronchoscope cleaning based on Internet of Things technology according to claim 3 is characterized in that: The sample data extraction layer is used to analyze the clean sample Y n Perform data extraction to obtain the component data B of the cleaned sample to be analyzed n ; The component data of the cleaned sample to be analyzed B n , Cleaning parameters C n Combine to obtain the feedback data F of the cleaning sample to be analyzed n ; The feature extraction layer is used to analyze the clean sample feedback data F n Perform feature extraction to obtain the feedback data feature F of the cleaning sample to be analyzed n '; The attention layer is used to analyze the feedback data features F of the cleaned samples n 'Weighted feature extraction to obtain the weighted feature J of the feedback data of the cleaned sample to be analyzed n ; The feedback result output layer is used to weight the feature J based on the feedback data of the cleaned samples to be analyzed. n Perform feature classification and obtain bronchoscope cleaning feedback results.

5. The method for tracing and intelligent management of bronchoscope cleaning based on Internet of Things technology according to claim 4 is characterized in that: In the feature extraction layer, the clean sample feedback data F is analyzed n The specific steps for feature extraction include: The feature extraction layer includes convolution layer, batch normalization layer and pooling layer; Convolution kernels of different sizes are used for feature extraction; Using the formula Represents a convolution operation; where represents the convolution kernel Layer outputs, Indicates Tier The convolution kernel Weight value, =1, 2, …, , Indicates Tier A bias, is the convolution operation; All Input to the batch normalization layer and pooling layer for feature extraction to obtain the feedback data feature F of the cleaned sample to be analyzed n '.

6. The method for tracing and intelligent management of bronchoscope cleaning based on Internet of Things technology according to claim 5 is characterized in that: In the attention layer, the feedback data features F of the cleaned samples to be analyzed n The specific steps of weighted feature extraction include: Feedback data feature F of the cleaned sample to be analyzed n 'Converted into the characteristic sequence F of the feedback data of the cleaning sample to be analyzed n ''; F n ''=[F n1 '', F n2 '', ..., F nU ''], U is the characteristic sequence F of the feedback data of the cleaning sample to be analyzed n '' length; The characteristic sequence F of the feedback data of the cleaning sample to be analyzed n ''Perform average pooling and normalization in the channel dimension to obtain the channel attention weight α of the feature sequence of the feedback data of the cleaned sample to be analyzed; α=[α1, α2, …, α U ]; The channel attention weight α of the feedback data feature sequence of the cleaned sample to be analyzed and the feature sequence F of the feedback data of the cleaned sample to be analyzed n ''Perform element-by-element multiplication to obtain the weighted feature J of the feedback data of the cleaned sample to be analyzed n .

7. The intelligent management system for bronchoscope cleaning and tracing based on Internet of Things technology is characterized by: The system applies the bronchoscope cleaning and tracing intelligent management method based on Internet of Things technology as described in any one of claims 1 to 6, including: The cleaning data collection module includes a data preparation unit and a data collection unit; the data preparation unit is used to set the cleaning equipment set Q, Q={Q n |n=1, 2, ..., N}, where N is the total number of cleaning devices used to clean bronchoscopes; Q n represents a cleaning device in the cleaning device set Q; the data collection unit is used when any cleaning device Q in the cleaning device set Q n When a bronchoscope to be cleaned is received, the cleaning work is started, and all cleaning work is completed based on the bronchoscope cleaning operation model to obtain bronchoscope cleaning data; the bronchoscope cleaning data includes the bronchoscope identification number, bronchoscope cleaning process time, cleaning parameters C n , bronchoscope cleaning image set X n and bronchoscope cleaning completion D n The bronchoscope cleaning operation model includes a label recognition layer, a cleaning data collection layer, a cleaning judgment layer, a cleaning record layer, and a data output layer, wherein the cleaning judgment layer is constructed based on dense connections and twin networks; The cleaning data reverse analysis module includes a feedback judgment unit; the feedback judgment unit is used to collect the Q of the cleaning equipment during the cleaning process. n The corresponding cleaning sample is obtained to obtain the cleaning sample Y to be analyzed n ; Based on the cleaning sample Y to be analyzed n , bronchoscope cleaning data and bronchoscope cleaning sample tracing model are analyzed to obtain bronchoscope cleaning feedback results; the bronchoscope cleaning sample tracing model includes a sample data extraction layer, a feature extraction layer, an attention layer and a feedback result output layer, which are used to perform multi-scale feature extraction to obtain feedback results; The cleaning data backtracking module includes a data management unit; the data management unit is used to combine the bronchoscope cleaning data and the bronchoscope cleaning feedback results to obtain the bronchoscope cleaning tracing data; the bronchoscope cleaning tracing data is stored in a preset database to complete a bronchoscope cleaning tracing intelligent management.

Images