A medical instrument air tightness detection system and detection method

Abstract

The application provides a medical instrument air tightness detection system and a detection method, the method comprising: collecting an image of a medical instrument to be detected based on a data acquisition module; determining the category of the medical instrument to be detected based on a medical instrument recognition module according to the collected medical instrument image; placing the medical instrument to be detected inside a sealed cavity module; according to the category of the medical instrument to be detected, filling the inside of the sealed cavity module with gas of a corresponding pressure through a pressure control module; collecting pressure change data and surrounding environment information at the corresponding moment through the data acquisition module; pre-processing the collected pressure change data based on a data preprocessing module; judging leakage according to the category of the medical instrument to be detected, the surrounding environment information and the pre-processed pressure change data through an air tightness detection module; and outputting the judgment result based on an output module. The application can more sensitively realize medical instrument air tightness detection leakage, has higher detection precision and is convenient to use.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical instrument detection, in particular to a medical instrument air tightness detection system and a detection method. BACKGROUND

[0002] The air tightness of medical instruments is an important indicator to ensure their safety and effectiveness, especially in medical devices involving in vivo implantation, closed delivery or fluid control, the air tightness requirement is particularly strict. At present, the existing air tightness detection equipment usually has the following problems:

[0003] 1. The detection precision is insufficient, which cannot meet the accurate measurement demand of small leakage;

[0004] 2. The degree of automation is low, manual operation is more, and the efficiency is low;

[0005] 3. The detection environment adaptability is poor, which is difficult to meet the complex and diverse medical instrument detection demand.

[0006] Therefore, it is very necessary to design a medical instrument air tightness detection system and a detection method. SUMMARY

[0007] In order to overcome the shortcomings of the prior art, the purpose of the present application is to provide a medical instrument air tightness detection system and a detection method.

[0008] In order to achieve the above-mentioned purpose, the present application provides the following scheme:

[0009] The present application provides a medical instrument air tightness detection system, comprising: a sealed cavity module, a pressure control module, a data acquisition module, a medical instrument identification module, a data preprocessing module, an air tightness detection module and an output module;

[0010] The inside of the sealed cavity module is provided with a medical instrument;

[0011] The pressure control module is connected with the sealed cavity module, which is used for restarting the sealed cavity and detecting the pressure;

[0012] The data acquisition module is used for acquiring the pressure change inside the sealed cavity, the image of the medical instrument and the surrounding environment information;

[0013] The medical instrument identification module is used for identifying the medical instrument according to the image of the medical instrument;

[0014] The data preprocessing module is used for preprocessing the pressure change data inside the sealed cavity;

[0015] The air tightness detection module is used for judging the leakage based on the preprocessed pressure change data and the surrounding environment information according to the medical instrument identification result.

[0016] The output module is configured to output the determination result.

[0017] Preferably, the sealed cavity module comprises a cavity made of transparent pressure-resistant material and a quick sealing joint arranged on the cavity, the pressure control module comprises a gas source, a pressure adjusting device and a pressure sensor, the cavity is connected with the pressure adjusting device through the quick sealing joint, the pressure adjusting device is connected with the gas source, the pressure sensor is arranged in the cavity, and the pressure sensor is connected with the data acquisition module.

[0018] The application further provides a medical instrument air tightness detection method, comprising the following steps:

[0019] Step 100: acquiring an image of a medical instrument to be detected based on a data acquisition module, and determining the category of the medical instrument to be detected based on a medical instrument recognition module according to the acquired image of the medical instrument;

[0020] Step 200: placing the medical instrument to be detected in the sealed cavity module, filling the inside of the sealed cavity module with gas of corresponding pressure through a pressure control module according to the category of the medical instrument to be detected, and acquiring pressure change data and surrounding environment information at the corresponding moment through a data acquisition module;

[0021] Step 300: pre-processing the acquired pressure change data based on a data preprocessing module;

[0022] Step 400: performing leakage judgment based on a gas tightness detection module according to the category of the medical instrument to be detected, the surrounding environment information and the pre-processed pressure change data;

[0023] Step 500: outputting the determination result based on an output module.

[0024] Preferably, in step 100, the category of the medical instrument to be detected is determined based on the medical instrument recognition module according to the acquired image of the medical instrument, and specifically:

[0025] a medical instrument recognition model is constructed based on the improved ResNet34 network structure;

[0026] the medical instrument recognition model is trained based on a preset data set;

[0027] the medical instrument to be detected is input into the trained medical instrument recognition model to obtain the category of the medical instrument to be detected.

[0028] Preferably, the medical instrument recognition model is divided into 9 parts, namely a convolution layer, a maximum pooling layer Maxpool, Stage1, Stage2, SE, Stage3, Stage4, an average pooling layer Avgpool and a full connection layer FC.

[0029] Preferably, in step 300, the collected pressure change data is preprocessed based on the data preprocessing module, specifically:

[0030] Obtain pressure change data, and perform denoising processing on the pressure change data based on an IWOA-VMD algorithm.

[0031] According to the specific embodiments provided by the present application, the following technical effects are disclosed:

[0032] The present application provides a kind of medical instrument air tightness detection system and detection method, which comprises the image of medical instrument to be detected based on data acquisition module acquisition, based on medical instrument recognition module according to the image of medical instrument collected The category of medical instrument to be detected, the medical instrument to be detected is placed in the inside of sealed cavity module, according to the category of medical instrument to be detected, the corresponding pressure gas is filled into the inside of sealed cavity module by pressure control module, and the pressure change data and the surrounding environment information of corresponding time are collected by data acquisition module, the collected pressure change data is preprocessed based on data preprocessing module, leakage is judged by air tightness detection module according to the category of medical instrument to be detected, surrounding environment information and preprocessed pressure change data, and the output module is outputted based on output module. The present application can more sensitively realize medical instrument air tightness detection leakage, and detection precision is higher, convenient to use. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0034] Figure 1 The method flow chart provided by the embodiments of the present application is provided.

[0035] Figure 2 The medical instrument recognition model structure diagram is provided.

[0036] Figure 3 The decomposition schematic diagram of convolution kernel is provided. DETAILED DESCRIPTION

[0037] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0038] The present application aims to provide a medical instrument air tightness detection system and a detection method, which can more sensitively realize medical instrument air tightness detection leakage, has higher detection precision, and is convenient to use.

[0039] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] The present application provides a medical instrument air tightness detection system, comprising a sealed cavity module, a pressure control module, a data acquisition module, a medical instrument identification module, a data preprocessing module, an air tightness detection module and an output module.

[0041] The inside of the sealed cavity module is provided with a medical instrument, which is used to place the medical instrument to be tested in a closed environment to prevent external interference.

[0042] The pressure control module is connected to the sealed cavity module, which is used to restart the sealed cavity and perform pressure detection.

[0043] The data acquisition module is used to acquire the pressure change inside the sealed cavity, the image of the medical instrument and the surrounding environment information.

[0044] The medical instrument identification module is used to identify the medical instrument according to the image of the medical instrument.

[0045] The data preprocessing module is used to preprocess the pressure change data inside the sealed cavity.

[0046] The air tightness detection module is used to make a leakage judgment based on the preprocessed pressure change data and the surrounding environment information according to the medical instrument identification result.

[0047] The output module is used to output the judgment result.

[0048] It should be noted that the present application also includes an execution mechanism module, which comprises an automatic mechanical arm and a clamping device, and is used to realize the automatic taking and placing of the medical instrument and the sealing operation.

[0049] The sealing cavity module comprises a cavity made of a transparent pressure-resistant material and a quick sealing joint arranged on the cavity, the pressure control module comprises a gas source, a pressure regulating device and a pressure sensor, the cavity is connected with the pressure regulating device through the quick sealing joint, the pressure regulating device is connected with the gas source, the pressure sensor is arranged in the cavity, and the pressure sensor is connected with the data acquisition module.

[0050] All the devices mentioned above can adopt the devices in the prior art, and the present application does not limit the model, and the specific functions can be realized.

[0051] As shown in Figure 1 The present application also provides a medical instrument air tightness detection method, comprising:

[0052] Step 100: based on the data acquisition module, the image of the medical instrument to be detected is collected, and based on the medical instrument recognition module, the category of the medical instrument to be detected is determined according to the collected image of the medical instrument;

[0053] Step 200: placing the medical instrument to be detected in the sealing cavity module, according to the category of the medical instrument to be detected, the pressure control module is used to fill the corresponding pressure gas into the inside of the sealing cavity module, and the data acquisition module is used to collect the pressure change data and the surrounding environment information at the corresponding time;

[0054] Step 300: based on the data preprocessing module, the collected pressure change data is preprocessed;

[0055] Step 400: based on the air tightness detection module, the leakage is judged according to the category of the medical instrument to be detected, the surrounding environment information and the preprocessed pressure change data;

[0056] Step 500: based on the output module, the judgment result is output.

[0057] In step 100, based on the medical instrument recognition module, the category of the medical instrument to be detected is determined according to the collected image of the medical instrument, specifically:

[0058] 1. Based on the improved ResNet34 network structure, a medical instrument recognition model is constructed, specifically:

[0059] In order to meet the application requirement of high-precision identification of medical instrument image, an improved residual network model SE-ResNet36 is constructed based on ResNet34 network, and the overall architecture of the network model is as shown in Figure 2 The network structure is divided into 9 parts, namely convolution layer, maximum pooling layer Maxpool, Stage1, Stage2, SE, Stage3, Stage4, average pooling layer Avgpool and full connection layer FC.

[0060] The SE-ResNet36 network model is based on ResNet34, the first layer of the original network model is decomposed, a convolution kernel of 7 is split into three convolution kernels of 3, and an SE attention module is added after Stage2, which can pay attention to the relationship between channels, so that the model can automatically learn the importance of different channel features. The overall network model architecture and internal parameters of SE-ResNet36 are shown in Table 1.

[0061] Table 1 Overall network model architecture and internal parameters of SE-ResNet36

[0062]

[0063] The 7x7 large size convolution kernel of the first layer of ResNet34 extracts coarse-grained features that are not conducive to the recognition of medical devices, affecting the recognition effect of the network model on medical devices. In order to more accurately identify medical devices, more effective fine features need to be extracted, so when designing the network model, the first layer of the convolution kernel of 7 is split into three convolution kernels of 3 in series, replacing the original large size convolution kernel, as shown in Figure 3

[0064] The SE attention module is proposed in the SE-ResNet36 network model, which belongs to one of the channel attention modules, can make the network pay more attention to the channel information, can enhance the learning ability of the model, and the SE module has the characteristics of plug and play, can be directly added to other convolutional neural networks, can effectively improve the recognition accuracy of the model, the SE module can be divided into compression (Squeeze) and excitation (Excitation) two processes, first use Squeeze operation to compress the feature map, then learn the channel features, then perform Excitation operation, give the feature map different channel weights from the channel perspective, finally get the feature map with channel attention, SE attention mechanism can make the model strengthen the channel features of the input feature map, highlight important features and suppress unimportant features, thereby improving the performance of the model;

[0065] 2. Based on the preset data set, the medical device recognition model is trained, wherein the pictures of various medical devices can be directly collected, and image enhancement is performed, for example, image rotation processing, color dithering processing and Gaussian noise processing, and after enhancement, the medical device image is labeled based on the labeling software to obtain the preset data set;

[0066] 3. The medical device to be detected is input into the trained medical device recognition model to obtain the category of the medical device to be detected.​

[0067] In step 200, the medical instrument to be detected is placed in the sealed cavity module, the pressure control module is used to fill the gas with corresponding pressure into the sealed cavity module according to the category of the medical instrument to be detected, and the data acquisition module is used to collect the pressure change data and the surrounding environment information at the corresponding time.

[0068] The medical instrument to be detected is placed in the sealed cavity module, the pressure control module is used to fill the gas with corresponding pressure into the sealed cavity module according to the category of the medical instrument to be detected, and the data acquisition module is used to collect the pressure change data and the surrounding environment information at the corresponding time.

[0069] In step 300, the data preprocessing module is used to preprocess the collected pressure change data, and the specific steps are as follows.

[0070] The pressure change data is obtained, the collected pressure change data is taken as original time series data, and the IWOA-VMD algorithm is used for denoising processing, and the specific steps include:

[0071] The parameter optimization range of the VMD algorithm is initialized, and the IWOA algorithm is used to find the parameter combination corresponding to the minimum value of the composite fitness function;

[0072] The optimal parameter combination is substituted into the VMD decomposition model, and the original time series data is decomposed into a series of IMFs;

[0073] The different IMFs are calculated and identified using the multi-scale permutation entropy-variance contribution rate selection principle, and are classified into noise dominant components and effective signal components, and the noise dominant components are denoised by NLM to obtain denoised components;

[0074] The denoised components and the effective signal components are reconstructed to obtain the denoised original time series data;

[0075] The variational mode decomposition (VMD) algorithm is introduced in detail;

[0076] As an active branch in the field of adaptive signal decomposition, the VMD algorithm has played an important role in the field of signal processing in recent years. The core idea of the VMD algorithm is to adaptively decompose the input signal f(t) into k modal functions u k (t) near the center frequency ω k (t), while minimizing u kThe bandwidth of (t) and VMD algorithm decomposition process is described as follows:

[0077] Let the kth modal function u k The expression of (t) is:

[0078] u k (t) = A k (t) cos [φ k (t)]

[0079] In the formula, A k (t) is the instantaneous amplitude of u k (t) and A k (t) ≥ 0, φ k (t) is the instantaneous phase of u k (t), and is a non-decreasing function.

[0080] The Hilbert transform is calculated for the k modal functions u k (t), so as to obtain the corresponding one-sided spectrum. The relevant analytic signal is:

[0081]

[0082] In the formula, δ(t) is the unit impulse function, and * represents convolution.

[0083] The baseband signal is obtained by multiplying the analytic signal by

[0084]

[0085] In order to obtain the bandwidth of each modal function, signal demodulation is realized through Gaussian smoothing. The constraint variation model is obtained as follows:

[0086]

[0087] In the formula, K is a preset decomposition modal number.

[0088] In order to solve the above constraint variation model, the Lagrange multiplier λ and the quadratic penalty factor α are introduced, and the above formula is transformed into the following non-constraint variation model:

[0089]

[0090] The alternating direction method of multipliers (ADMM), Parseval theorem and gradient descent algorithm are used for solving, and the modal function {u k}, the center frequency {ω k} and the Lagrange multiplier λ are constantly updated and iterated, and the formula is as follows: ​

[0091]

[0092] where τ is the time step.

[0093] The ADMM algorithm converges, and the above process update iteration terminates when the following conditions are met:

[0094]

[0095] where c is the convergence parameter, generally ε = 1 × 10 -6 .

[0096] The whale optimization algorithm (WOA) is introduced in detail:

[0097] The whale optimization algorithm is a new swarm intelligence optimization algorithm based on the iterative search of humpback whale population. It is modeled according to the hunting behavior of humpback whales to realize optimization of complex optimization problems. According to the algorithm characteristics, it is divided into three stages of searching for prey, surrounding prey and spiral bubble net hunting.

[0098] (1) Searching for prey stage

[0099] In this stage, according to the random walk mechanism, each whale participating in hunting updates its position iteratively according to the current positions of each other. According to the search position at the tth time, when the coefficient |A|>1, the whale population randomly searches for high-quality prey in the global range, and the expression of its (t+1)th position update is:

[0100] D = |C·x rand (t)-x(t)|

[0101] x(t+1) = x rand (t)-A·D

[0102] Where t is the current iteration number, x rand (t) is the arbitrary position of the whale, x(t) is the current position, and:

[0103] A = 2ar1-a

[0104] C = 2r2

[0105] Where r1, r2 ∈ rand [0, 1], a = 2-2t / T, T is the selected generation number. According to the linear decreasing property of a, A gradually decreases with a. When |A|≤1, the algorithm enters the surrounding prey stage.

[0106] (2) Surrounding prey stage

[0107] When the whales find the target, the search range will gradually shrink, and the search position will be updated according to the current best individual of the population, which is expressed as:

[0108] D1 = |C x best (t) - x(t)|

[0109] x(t+1) = x best (t) - A D

[0110] where x best (t) is the position of the best individual in the current whale population.

[0111] (3) Spiral bubble net hunting phase

[0112] Whales have a special hunting method, using the spiral rising of the bubble net to continuously shrink and surround the target. Its mathematical model is expressed as follows:

[0113] D2 = |x best (t) - x(t)|

[0114] x(t+1) = D2 e bl cos(2πl) + x best (t)

[0115] where b is a constant coefficient; l is a random number in the interval [-1, 1].

[0116] The hunting phase is divided by setting a random probability. When the random probability p < 0.5, the whale updates the position of the best individual; when p ≥ 0.5, the whale gradually approaches the target. In the approach process, |A| is constantly decreasing, and when |A| = 0, the optimal solution is obtained. The overall hunting mechanism is described as follows:

[0117]

[0118] Next, the improved strategy of whale optimization algorithm is introduced:

[0119] The traditional whale optimization algorithm adjusts global search and local development through the coefficient A. The value of A is mainly related to a, but the linearly changing convergence factor a has drawbacks in search ability in the early stage and iteration speed in the later stage. In order to improve its influence on the performance of the algorithm, without affecting the overall trend, a segmented nonlinear convergence factor m is proposed, which is expressed as follows:

[0120]

[0121] In the whole iteration process, m shows a nonlinear decreasing trend. In the early stage, the larger parameter m gradually decreases, which can improve the global exploration ability; in the later stage, the smaller parameter m can improve the convergence speed of the algorithm.

[0122] In addition, considering the influence of inertia weight on the optimization ability of algorithm, an adaptive weight strategy is proposed herein, and the formula is as follows:

[0123]

[0124] Therefore, the position updating formula becomes:

[0125]

[0126] The traditional WOA algorithm is easy to fall into local optimum, and the Cauchy operator is introduced herein to give the Cauchy disturbance to the optimal individual. The Cauchy operator can generate larger disturbance to make the algorithm escape from local optimum, and can also generate smaller disturbance to improve the convergence speed.

[0127] Therefore, the Cauchy variation formula is selected as follows, and the position of the best individual is updated as follows:

[0128] x new (t)=x best (t)·[1+Cauchy(0,1)]

[0129] In the formula, x new (t) is the new value of the current optimal value after disturbance, and Cauchy(0,1) is the Cauchy operator.

[0130] The fitness function of the IWOA-VMD algorithm is:

[0131] The energy spread can represent the energy distribution of the signal in the frequency domain, and the smaller the value is, the higher the energy proportion of the corresponding component is. The energy calculation formula of each component is:

[0132]

[0133] In the formula, N is the length of the signal. The energy probability density formula is obtained by normalizing the energy spread:

[0134]

[0135] The energy spread calculation formula is:

[0136]

[0137] In practical application, if the energy spread is used as a single fitness function, the signal may be decomposed into a simple harmonic wave, which is due to the over-decomposition of the signal caused by the pursuit of energy concentration. In order to improve the over-decomposition of the signal, the kurtosis is introduced to describe the waveform characteristics of the signal. The kurtosis is a dimensionless parameter representing the degree of data peak, and is often used in fault diagnosis of nonlinear signals. Its expression is as follows:

[0138]

[0139] where μ and σ denote the mean and variance of the signal respectively; E(x-μ) 4 denotes the fourth-order mathematical expectation.

[0140] For the characteristics of wireless signals, a compound fitness function is constructed by combining energy and kurtosis. By iteratively searching for the minimum value of the compound fitness function, the best combination of (K, a) parameters corresponding to the best decomposition effect is found. The fitness function calculation formula is as follows:

[0141] F=1 / Ku+0.1xH E

[0142] The principle of selecting the optimal modal component of multi-scale permutation entropy-variance contribution rate is introduced:

[0143] Multi-scale permutation entropy (MPE) considers the signal characteristics under multiple scales by introducing permutation entropy of different scales and weighted average to describe the randomness and complexity of the signal. The basic principle of MPE is as follows:

[0144] 1. Multi-scale coarse-grained is performed on the time series x={x1,x2,…,x L}

[0145]

[0146] where s is the scale factor, is the multi-scale time series. When s is set to 1, its is the original time series, and the permutation entropy can be calculated.

[0147] 2. Phase space reconstruction is performed on the time series

[0148]

[0149] where τ is the time delay factor, m is the embedding dimension, is the reconstructed sequence.

[0150] 3. The reconstructed is arranged in ascending order, and m! kinds of permutation combinations can be obtained, and the number of occurrences of each permutation is N t , and the corresponding probability is That is:

[0151]

[0152]

[0153] ​4. Scale factor is s The arrangement is calculated and normalized

[0154]

[0155] As a kind of data statistical theory based on Gaussian hypothesis, principal component analysis can reduce the redundancy and invalid characteristics of data components, eliminate data correlation to highlight the advantages of useful information of original sample data. As a common method for selecting principal components in principal component analysis, variance contribution rate mainly represents the proportion of IMF energy occupying the energy of the signal before decomposition. The greater the variance contribution rate, the stronger the explanatory ability of IMF to the original signal. Therefore, the importance of IMF can be sorted according to the size of variance contribution rate. The calculation formula of variance contribution rate is as follows:

[0156]

[0157] In the formula, D (IMF k ) represents the variance of the kth IMF, V k represents the variance contribution rate of the kth IMF.

[0158] The NLM algorithm denoising is introduced in detail:

[0159] Based on the weighted average operation of similar structure of image to improve the signal-to-noise ratio, the non-local mean filtering algorithm realizes the purpose of image denoising. NLM algorithm selects all the information in the image, retains the obvious main features, and eliminates the obviously irrelevant redundant components. By analyzing the similar characteristics, it is found that the NLM algorithm idea can also be applied to the field of signal filtering. Using NLM algorithm for filtering, as many similar values as possible are found in the search range, and the estimated value is obtained by using the weighted average of the weighted coefficient to make it close to the true value in the signal, so as to achieve the purpose of noise reduction. The principle of NLM algorithm is as follows:

[0160] Assume that the noisy signal y(t) is the superposition of the original signal u(t) and the external interference noise n(t), as shown in the following formula:

[0161] y(t) = u(t) + n(t)

[0162] NLM algorithm estimates the original signal u(t) by calculating the weighted average value K(t) of all similar blocks in y(t), and the calculation process is as follows:

[0163]

[0164] Where ω(t,s) is the similarity of two search blocks centered at t and s respectively, N(t) represents the set of all points in the target search region centered at the target sample point t, Z(t) is a normalization constant representing the sum of all search block similarities in the search domain, and ω(t,s) also satisfies and These two basic conditions are used to calculate ω(t,s) as follows:

[0165]

[0166] Where L Δ is the neighborhood block centered at s, L Δ = 2P+1, the parameter P affects the number of similar structure blocks found, △ is the search block centered at t, K is half the length of the △ region, the parameter K affects the calculation amount and calculation time, and λ is the filter bandwidth parameter affecting the smoothness of the filtered signal.

[0167] The selection of the three key parameters P, K, and λ greatly affects the denoising effect of the algorithm on one-dimensional signals. If the similar block parameter P is too small, more similar blocks can be found in the search range, and if P is too large, similar blocks cannot be found. In the present application, P is set to 10<P<20. The search block parameter K can theoretically search the entire domain, but this will increase the calculation difficulty of the algorithm and reduce the calculation efficiency. Literature indicates that when the signal length N>4000, 0.25N<K<0.35N is taken, the search range contains more signal information, and a balance is reached in calculation amount and calculation time. When N<4000, K=0.5N is taken, the search range can cover the entire domain and most signal information can be obtained. If the bandwidth parameter A is too small, noise fluctuations will occur, which will interfere with the weighted average of similar blocks and reduce the filtering effect. If λ is too large, the search block similarity will be incorrectly determined, the signal will be too smooth, and the filtering effect will be too strong. In the present application, the filtering effect is better when 0.3σ<λ<0.6σ(σ is the standard deviation of the noisy signal).

[0168] In step 400, the air-tightness detection module judges the leakage according to the category of the medical instrument to be detected, the surrounding environment information, and the preprocessed pressure change data. Specifically:

[0169] The AI model can be used to analyze the category of the medical instrument to be detected, the surrounding environment information and the pre-processed pressure change data, for example, using a time series analysis model such as LSTM, Transformer, etc. to model the pressure change trend, taking the category of the medical instrument to be detected and the surrounding environment information as additional dimensions of the model. Through the training of a large amount of historical detection data, the model learns the characteristic patterns of different types of leakage (such as small leakage, intermittent leakage, and large leakage). During the detection process, the model analyzes the pressure change data in real time and intelligently judges the leakage rate, leakage location and leakage type to obtain the judgment result.

[0170] In step 500, the output module outputs the judgment result, specifically:

[0171] The detection result is presented in a graphical form by the output module, including the pressure change curve, the leakage rate value and the leakage type determination result, etc.

[0172] The present application can also be provided with self-learning and optimization:

[0173] For example, the AI model can be updated through the cloud, constantly learning and optimizing the model parameters from new detection data to improve the detection accuracy. Special detection samples (such as abnormal leakage cases) can also be labeled to further enhance the model's ability to identify special situations.

[0174] In the present specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other.

[0175] The principles and implementation modes of the present application are described by applying specific examples in this paper. The above description of the embodiments is only to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range will be changed. In summary, the content of the present specification should not be understood as a limitation of the present application.

Claims

1. A medical device airtightness testing system, characterized in that, include: Sealed cavity module, pressure control module, data acquisition module, medical device identification module, data preprocessing module, airtightness detection module and output module; The sealed cavity module contains medical devices. The pressure control module is connected to the sealing cavity module and is used to inflate the sealing cavity and detect the pressure. The data acquisition module is used to collect pressure changes inside the sealed cavity, images of the medical device, and information about the surrounding environment. The medical device identification module is used to identify medical devices based on images of the medical devices. The data preprocessing module is used to preprocess the pressure change data inside the sealed cavity, specifically: Acquire pressure change data and perform noise reduction processing based on the IWOA-VMD algorithm. The specific steps include: Initialize the VMD algorithm parameter optimization range, and find the parameter combination corresponding to the minimum value of the composite fitness function according to the IWOA algorithm; Substituting the optimal parameter combination into the VMD decomposition model, the original time series data is decomposed into a series of IMFs; Different IMFs are calculated and identified using the multi-scale permutation entropy-variance contribution rate selection principle. They are classified into noise-dominant components and effective signal components. The noise-dominant components are then denoised using NLM to obtain the denoised components. The denoised components are reconstructed from the effective signal components to obtain the denoised original time series data; The airtightness detection module is used to determine leakage based on the medical device identification results, pre-processed pressure change data, and surrounding environmental information. The output module is used to output the judgment result; The sealed cavity module includes a cavity made of transparent pressure-resistant material and a quick-sealing connector on the cavity. The pressure control module includes a gas source, a pressure regulating device, and a pressure sensor. The cavity is connected to the pressure regulating device via the quick-sealing connector, and the pressure regulating device is connected to the gas source. The pressure sensor is installed inside the cavity and is connected to the data acquisition module. The medical device to be tested is placed inside the sealed cavity module. According to the type of medical device to be tested, gas at a corresponding pressure is injected into the sealed cavity module through the pressure control module. Based on practical experience, different pressure stabilization times, gas types, and pressures are set for each type of medical device. After filling, the set pressure is maintained for a certain period of time, and the pressure change inside the sealed cavity module is detected by the pressure sensor.

2. A method for testing the airtightness of a medical device, applied to the medical device airtightness testing system of claim 1, characterized in that, include: Step 100: The data acquisition module acquires images of the medical device to be tested, and the medical device identification module determines the category of the medical device to be tested based on the acquired images. Step 200: Place the medical device to be tested inside the sealed cavity module. According to the type of medical device to be tested, fill the sealed cavity module with gas at the corresponding pressure through the pressure control module, and collect pressure change data and corresponding ambient environmental information through the data acquisition module. Step 300: Preprocess the collected pressure change data based on the data preprocessing module; Step 400: The airtightness detection module determines leakage based on the type of medical device to be tested, surrounding environmental information, and pre-processed pressure change data; Step 500: Output the judgment result based on the output module.

3. The method according to claim 2, characterized in that, In step 100, the medical device identification module determines the category of the medical device to be detected based on the acquired image of the medical device, specifically as follows: A medical device identification model was constructed based on the improved ResNet34 network structure. The medical device identification model is trained based on a pre-set dataset; The image of the medical device to be detected is input into the trained medical device recognition model to obtain the category of the medical device to be detected.

4. The method according to claim 3, characterized in that, The medical device recognition model is divided into 9 parts, namely, convolutional layer, max pooling layer, stage 1, stage 2, SE attention module, stage 3, stage 4, average pooling layer, and fully connected layer FC.

5. The method according to claim 4, characterized in that, In step 300, the collected pressure change data is preprocessed using the data preprocessing module, specifically as follows: Acquire pressure change data and perform noise reduction processing based on the IWOA-VMD algorithm. The specific steps include: Initialize the VMD algorithm parameter optimization range, and find the parameter combination corresponding to the minimum value of the composite fitness function according to the IWOA algorithm; Substituting the optimal parameter combination into the VMD decomposition model, the original time series data is decomposed into a series of IMFs; Different IMFs are calculated and identified using the multi-scale permutation entropy-variance contribution rate selection principle. They are classified into noise-dominant components and effective signal components. The noise-dominant components are then denoised using NLM to obtain the denoised components. The denoised components are reconstructed from the effective signal components to obtain the original time series data after denoising.

Images