CN-122023295-A - Medical instrument sterilization effect verification method and system based on multispectral imaging

Abstract

The invention discloses a medical instrument sterilization effect verification method and system based on multispectral imaging. In the sterilization process, a multispectral camera is used for synchronously collecting reflection spectrum images of a bacillus stearothermophilus indicator region under six spectrum channels to obtain a time sequence multispectral image set, the image set is constructed into four-dimensional spectrum characteristic tensors, an improved Tucker tensor decomposition algorithm is adopted for decomposition, sparsity constraint is applied, a dominant spectrum mode matrix related to spore inactivation is extracted, a spore inactivation index function considering the nonlinear coupling influence of temperature and pressure is constructed based on the dominant spectrum mode matrix, model parameters are optimized through a self-adaptive gradient descent algorithm, and inactivation indexes at the moment of sterilization end are calculated and sterilization effects are judged. The invention realizes the rapid and accurate prediction of the spore inactivation index, provides a real-time and reliable quality control means for sterilizing medical instruments, and effectively improves the verification efficiency and accuracy.

Inventors

• ZHA SHIHONG
• XU GANG

Assignees

• 江阴滨江医疗设备有限公司

Dates

Publication Date

20260512

Application Date

20260120

Claims (6)

1. 1. The medical instrument sterilization effect verification method based on multispectral imaging is characterized by comprising the following steps of: In the sterilization process of medical instruments, synchronously acquiring reflection spectrum images of a bacillus stearothermophilus indicator region coated on the surfaces of the medical instruments by using a multispectral camera under six spectrum channels of 365nm ultraviolet band, 450 nm blue light band, 525 nm green light band, 620 nm red light band, 780 nm near infrared band and 850 nm near infrared band, and acquiring time sequence multispectral image sets according to preset time sampling intervals, wherein the time sequence multispectral image sets comprise multispectral images with different wavelengths and different acquisition moments; D2, constructing a time sequence multispectral image set as a four-dimensional spectral feature tensor, decomposing the four-dimensional spectral feature tensor by adopting an improved Tucker tensor decomposition algorithm to obtain a core tensor, a space factor matrix, a spectral factor matrix and a time factor matrix, and extracting a dominant spectral pattern matrix related to the inactivation of bacillus stearothermophilus by applying sparsity constraint to the spectral factor matrix; Constructing a spore inactivation index function based on the dominant spectrum mode matrix, wherein the spore inactivation index function comprehensively considers the nonlinear coupling influence of the weight coefficients, the inactivation rate constants and the temperature and pressure parameters of different spectrum modes on the inactivation process, and optimizing the model parameters of the spore inactivation index function by using an adaptive gradient descent algorithm through minimizing the prediction error; And D4, calculating a spore inactivation index at the end time of the sterilization process, and judging that the sterilization effect of the medical instrument is qualified when the spore inactivation index is smaller than a preset inactivation threshold value.
2. 2. The method for verifying sterilization effect of medical instrument based on multispectral imaging according to claim 1, wherein the step D1 comprises: D101, before the sterilization process of the medical instrument is started, uniformly coating a bacillus stearothermophilus indicator on a preset monitoring area on the surface of the medical instrument; d102, placing a medical instrument coated with a bacillus stearothermophilus indicator in a sterilization device, installing a multispectral camera at the observation window position of the sterilization device, and adjusting the shooting angle and focal length of the multispectral camera so that the bacillus stearothermophilus indicator area is positioned at the center of an imaging view field; D103 set time sampling interval as Second, during the sterilization process, the multispectral camera is sequentially switched to 365 nm ultraviolet band, 450 nm blue band, 525 nm green band, 620 nm red band, 780 nm near infrared band and 850 nm near infrared band according to the time sampling interval, a frame of reflection spectrum image is collected under each spectrum channel, and six band image collection of one sampling period is completed; d104, repeatedly executing step D103 until the sterilization process is finished to obtain the product containing Time-series multispectral image set with sampling moments Wherein The wavelength of the spectrum channel is represented by one of 365 nm, 450 nm, 525 nm, 620 nm, 780 nm and 850 nm, The acquisition time is represented, and the value range is To the point of 。
3. 3. The method for verifying the sterilization effect of a medical instrument based on multispectral imaging according to claim 2, wherein the step D2 comprises: D201. Collecting the time series multispectral images Recombining according to the space dimension, the spectrum channel dimension and the time dimension to construct a four-dimensional spectrum characteristic tensor Wherein Representing the number of high-level pixels of the image, The number of width pixels representing an image, The number of spectral channels is indicated and, Representing the number of time samples and , Representing a real number, the four-dimensional spectral feature tensor Elements of (2) Represent the first Sampling time, the first Spectral channel, image No Line 1 Pixel gray values at column positions; D202 tensor of the four-dimensional spectral features The improved Tucker tensor decomposition algorithm is adopted for decomposition, and the decomposition form is as follows: Wherein, the The core tensor is represented as a function of the core tensor, 、 、 、 Respectively representing rank parameters of four dimensions, 、 、 、 Respectively representing modular multiplication operations along a first dimension, a second dimension, a third dimension and a fourth dimension, A first spatial factor matrix is represented and, A second spatial factor matrix is represented and, The matrix of spectral factors is represented, Representing a time factor matrix; d203 for the spectral factor matrix Applying an L1 norm sparsity constraint, wherein the constraint conditions are as follows: Wherein, the Representing a spectral factor matrix Is calculated by a spectral factor matrix The absolute values of all the elements in (a) are summed, Representing a preset sparsity threshold; d204, under the condition that the sparsity constraint condition is met, iteratively optimizing a core tensor through an alternate least squares method A first space factor matrix A second space factor matrix Spectral factor matrix And a time factor matrix So that reconstruct error minimization; D205 from the optimized spectral factor matrix Extraction of dominant spectral pattern matrix associated with Bacillus stearothermophilus inactivation The dominant spectral pattern matrix From an optimized spectral factor matrix Front of maximum L1 norm Column vector composition, the calculation formula is: Wherein, the Representing a matrix of dominant spectral patterns, 、 、...、 Representing a spectral factor matrix The middle is arranged in descending order of L1 norm of column vector The index of the individual column vectors is used, Representing the number of dominant spectral patterns extracted; 、 、...、 Representing a spectral factor matrix The middle is arranged in descending order of L1 norm of column vector The index of each column vector corresponds to the column.
4. 4. The method for verifying the sterilization effect of a medical instrument based on multispectral imaging according to claim 3, wherein the step D3 comprises: D301 calculating dominant spectral pattern matrix The L2 norm of each column vector of (a) to obtain a spectrum mode intensity sequence, the first of the spectrum mode intensity sequence The calculation formula of the individual elements is: Wherein, the Represent the first The intensity values of the individual dominant spectral patterns, Representing dominant spectral pattern matrix Is the first of (2) The column vector is used to determine the position of the column, Representation of the first pair The column vector computes the L2 norm, Is in the range of 1 to ; D302 temperature time sequence in synchronous acquisition and sterilization process And pressure time series Wherein Indicating any time during the sterilization process, Is in the range of 0 to , Indicating the end time of the sterilization process; D303 nonlinear coupling function defining temperature and pressure parameters to spore inactivation rate The calculation formula is as follows: Wherein, the Representation of The coupling influence factor of the time temperature and pressure parameters on the deactivation rate, Indicating the temperature-sensitive coefficient of the temperature, Representation of The sterilization temperature at the moment of time is equal to the sterilization temperature, The reference temperature is indicated as such, Representing the coefficient of pressure sensitivity and, Representation of The sterilization pressure at the moment of time is equal to the sterilization pressure, Representing a reference pressure; Is a natural constant; D304 construction of a spore inactivation exponential function The calculation formula is as follows: Wherein, the Representation of The spore inactivation index at the moment of time, Represent the first The weight coefficients of the individual dominant spectral modes, Represent the first The deactivation rate constants corresponding to the individual dominant spectral modes, Indicating the time from the sterilization start time 0 to the current time For nonlinear coupling functions Performing time integration; d305 obtaining a reference inactivation index time series in a Standard Sterilization experiment The reference inactivation index time sequence is obtained by measuring the number of viable spores at different sterilization moments through a culture method and normalizing; D306 define prediction error As a spore inactivation index function Time series with reference inactivation index The mean square error between the two is calculated as follows: Wherein, the The total prediction error is indicated and, Represent the first The time of the sampling is the same as the time of the sampling, Represent the first The predicted value of the inactivation index is calculated according to the spore inactivation index function at each sampling moment, Represent the first The reference inactivation index for each sampling instant, Representing the total number of sampling moments; D307 minimizing prediction error using adaptive gradient descent algorithm Iterative optimization weight coefficient set And a set of deactivation rate constants The parameter updating formula of the adaptive gradient descent algorithm is as follows: Wherein, the Represent the first The weight coefficient of the dominant spectral mode is at The updated value after a number of iterations, Represent the first The weight coefficient of the dominant spectral mode is at The value after the number of iterations is, The learning rate of the weight coefficient is represented, Representing prediction error For the first Post-iteration first The weighting coefficients of the individual dominant spectral modes are partial derivatives, Represent the first The deactivation rate constant corresponding to the dominant spectral mode is at The updated value after a number of iterations, Represent the first The deactivation rate constant corresponding to the dominant spectral mode is at The value after the number of iterations is, The learning rate indicative of the deactivation rate constant, Representing prediction error For the first Post-iteration first Obtaining partial derivatives of deactivation rate constants corresponding to the dominant spectrum modes; D308, repeating step D307 until the prediction error And converging or reaching the preset maximum iteration times to obtain an optimized weight coefficient set and an optimized inactivation rate constant set.
5. 5. The method for verifying sterilization effect of medical instruments based on multispectral imaging according to claim 4, wherein the step D4 comprises: D401 when the sterilization process is finished Substituting the optimized spore inactivation index function Calculating spore inactivation index at the end of the sterilization process ; D402 set the deactivation threshold as When (when) When the sterilization effect of the medical instrument is judged to be qualified, when And judging that the sterilization effect of the medical instrument is unqualified.
6. 6. Medical instrument sterilization effect verification system based on multispectral imaging, which is characterized by comprising: The sequence acquisition module is used for acquiring a reflection spectrum image of a bacillus stearothermophilus indicator region coated on the surface of the medical instrument by using a multispectral camera in the sterilization process of the medical instrument, and acquiring a time sequence multispectral image set according to a preset time sampling interval; The characteristic decomposition module is used for constructing a time sequence multispectral image set into a four-dimensional spectral characteristic tensor and decomposing the four-dimensional spectral characteristic tensor to obtain a core tensor, a space factor matrix, a spectral factor matrix and a time factor matrix, and extracting a dominant spectrum mode matrix related to the inactivation of the bacillus stearothermophilus by applying sparsity constraint to the spectral factor matrix; The model construction module is used for constructing a spore inactivation index function based on the dominant spectrum mode matrix, and optimizing model parameters of the spore inactivation index function by using an adaptive gradient descent algorithm through minimizing a prediction error; The judging module is used for calculating a spore inactivation index at the end time of the sterilization process and judging that the sterilization effect of the medical instrument is qualified when the spore inactivation index is smaller than a preset inactivation threshold value; to realize the medical instrument sterilization effect verification method based on multispectral imaging as claimed in any one of claims 1 to 5.

Description

Medical instrument sterilization effect verification method and system based on multispectral imaging Technical Field The invention relates to the technical field of medical instrument sterilization, in particular to a medical instrument sterilization effect verification method and system based on multispectral imaging. Background Sterilization of medical instruments is a key element in ensuring patient safety in the field of medical and health. The medical instruments used in China are large in production and use, various disposable and reusable medical instruments are consumed annually, and the demand for verifying the sterilization effect is urgent. The traditional sterilization effect verification method mainly depends on a biological indicator culture method, namely, a biological indicator containing specific spores is placed in the sterilization process, and whether sterilization is qualified or not is judged by observing the survival condition of the spores through culture of a culture medium after sterilization. However, this method has many limitations, the culture period usually needs 7 to 14 days, real-time monitoring and rapid feedback of the sterilization process cannot be realized, and the operation efficiency and emergency response capability of the medical institution are severely restricted. In recent years, with the rapid development of optical imaging technology and computer vision technology, researchers have begun to explore sterilization effect verification methods based on image analysis. In the prior art, part of methods adopt a single-wavelength visible light imaging system to shoot biological indicators in the sterilization process, and spore activity is judged by analyzing the color change of an image. For example, some commercial products use an RGB camera to capture a color image of a biological indicator, and when spores are deactivated, the media color changes, allowing for rapid detection by comparison of the color differences. However, such methods rely only on color information in the visible light band, have low response sensitivity to spore physiological state changes, and are susceptible to interference from ambient light conditions and color fluctuations of the culture medium itself, resulting in insufficient accuracy and stability of detection results. In addition, multispectral imaging technology is adopted to obtain response information of the biological indicator in a plurality of spectral bands, and detection accuracy is attempted to be improved through spectral feature analysis. Such methods typically set a plurality of characteristic wavelengths in the ultraviolet, visible, and near infrared bands, and perform a simple linear combination or threshold determination after acquiring corresponding spectral images. However, the existing multispectral imaging methods still have obvious defects in the data processing level, on one hand, the methods generally simply flatten multispectral image sequences into a two-dimensional matrix for analysis, neglect the inherent relativity of the multispectral image data in the space dimension and the spectrum dimension, so that a large amount of valuable high-dimensional structural information is lost, and on the other hand, the existing methods lack pertinence in extracting spectral features, fail to effectively identify the feature wave bands highly related to spore inactivation, so that the extracted features have high redundancy and poor interpretability, and further influence the accuracy of the subsequent inactivation index prediction. In addition, the kinetics of inactivation of spores during sterilization is affected by nonlinear coupling of various environmental factors such as temperature, pressure, etc. Most of the existing inactivation index prediction models adopt simplified linear assumption or constant inactivation rate constant, and cannot accurately describe a complex dynamic inactivation rule in the actual sterilization process. For example, in the autoclaving process, the synergism of temperature and pressure can cause the spore inactivation rate to show nonlinear change characteristics, the traditional model tends to overestimate the inactivation rate in the early stage of sterilization, and underestimate the inactivation effect in the later stage of sterilization, and finally, the predicted inactivation index has a significant deviation from an actual value, and the relative error can reach more than 20%. The lack of prediction precision not only affects the accurate evaluation of the sterilization effect, but also can cause energy waste caused by excessive sterilization or potential safety hazards caused by insufficient sterilization. In summary, the existing medical instrument sterilization effect verification method has defects in detection speed, accuracy and practicality, and is difficult to meet the high standard requirements of the modern medical and health system on sterilization quality control. Theref