CN-122016675-A - Multispectral imaging detection method and multispectral imaging detection system for judging activity of probiotics based on metabolite monitoring

Abstract

The invention discloses a multispectral imaging detection method and a multispectral imaging detection system for judging the activity of probiotics based on metabolite monitoring, wherein the method comprises the steps of collecting a probiotic liquid preparation sample into a multispectral image data cuboid; the method comprises the steps of carrying out dimension reduction on a multispectral image data cuboid by using a principal component analysis method to obtain a characteristic spectrum data matrix, carrying out clustering on principal components in the characteristic spectrum data matrix to obtain an initial cluster, obtaining a target spectrum similarity cluster in the initial cluster by using a characteristic spectrum vector in the characteristic spectrum data matrix, obtaining an intensity value of a pixel point under a characteristic wavelength for the target spectrum similarity cluster, calculating a comprehensive activity score, presetting an activity threshold, and judging through the comprehensive activity score and the activity threshold to obtain a detection result of a probiotic liquid preparation sample.

Inventors

• JIA ZHIDAN

Assignees

• 美亚特医(北京)营养科技有限公司

Dates

Publication Date

20260512

Application Date

20260130

Claims (9)

1. 1. A multispectral imaging detection method for judging the activity of probiotics based on metabolite monitoring is characterized by comprising the following operation steps: Collecting a multispectral image data cuboid of a probiotic liquid preparation sample, wherein the multispectral image data cuboid comprises a plurality of pixel point coordinates under preset characteristic wavelengths and spectral vectors corresponding to the pixel point coordinates, and the preset characteristic wave band is a characteristic spectral response wave band of short-chain fatty acid; The method comprises the steps of obtaining a characteristic spectrum data matrix, obtaining a target spectrum similarity cluster in the initial cluster by utilizing a principal component analysis method, obtaining an intensity value of a pixel point under characteristic wavelength for the target spectrum similarity cluster, and calculating a comprehensive activity score; And judging the comprehensive activity score and the activity threshold to obtain a detection result of the probiotic liquid preparation sample.
2. 2. The multispectral imaging detection method for determining the activity of probiotics based on metabolite monitoring according to claim 1, wherein the multispectral image data cuboid is subjected to dimension reduction by using a principal component analysis method to obtain a characteristic spectrum data matrix, and the specific operation steps are as follows: Converting the multispectral image data cuboid into a two-dimensional matrix by using a principal component analysis method, and normalizing the two-dimensional matrix to obtain a standardized matrix; Calculating covariance of each row and each column in the standardized matrix, and converting the standardized matrix into a covariance matrix; Decomposing the covariance matrix by a principal component analysis method to obtain a plurality of characteristic values; The method comprises the steps of sorting all characteristic values in a descending order, simultaneously rearranging characteristic vectors corresponding to the characteristic values in the same order, calculating the sum of the first k characteristic values in the descending order, further calculating the sum of all the characteristic values, and calculating the cumulative contribution rate of the sum of the first k characteristic values and the sum of all the characteristic values; judging whether each accumulated contribution rate is larger than or equal to the accumulated contribution rate threshold value or not; If yes, taking all characteristic values corresponding to the accumulated contribution rate as main components, extracting characteristic vectors corresponding to the main components and forming a projection matrix; and reducing the dimension of the projection matrix by using the standardized matrix to obtain a characteristic spectrum data matrix.
3. 3. The multispectral imaging detection method for judging the activity of probiotics based on metabolite monitoring according to claim 2 is characterized in that the primary components in the characteristic spectrum data matrix are used for clustering to obtain an initial cluster, and the characteristic spectrum vector in the characteristic spectrum data matrix is used for obtaining a target spectrum similarity cluster in the initial cluster, wherein the specific operation steps are as follows: Extracting a first principal component of a first column and a last principal component of a last column from the characteristic spectrum data matrix, and constructing an initial cluster center candidate set for all characteristic spectrum vectors of the corresponding rows of the first principal component and the last principal component; Determining characteristic spectrum vectors of center points in the initial clustering center candidate set, and calculating Euclidean distances from all the characteristic spectrum vectors in the initial clustering center candidate set to the characteristic spectrum vectors of the center points; Calculating Euclidean distance again for the first initial clustering center and all the characteristic spectrum vectors in the initial clustering center candidate set, screening the characteristic spectrum vector with the farthest Euclidean distance of the first initial clustering center as the next initial clustering center until k characteristic spectrum vectors are screened out as the initial clustering centers; calculating the distance between the characteristic spectrum vector of each pixel point in the characteristic spectrum data matrix and the k initial clustering centers, screening the initial clustering center with the minimum distance, and clustering the characteristic spectrum vector of all the pixel points in the minimum distance range to the initial clustering center to obtain an initial clustering cluster; Calculating an average vector of the characteristic spectrum vectors in the initial cluster, searching a new centroid in the corresponding initial cluster according to the average vector, calculating a vector distance between the new centroid and an initial cluster center with the minimum distance, presetting a convergence threshold value, and judging whether the vector distance is larger than the convergence threshold value; If yes, judging the initial cluster as a final target spectrum similarity cluster; if not, repeating the steps according to the new centroid until the final target spectrum similarity cluster is obtained through screening.
4. 4. The multispectral imaging detection method for determining the activity of probiotics based on metabolite monitoring according to claim 3, wherein the intensity values of the pixels under the characteristic wavelength are obtained for the target spectrum similarity cluster, and the comprehensive activity score is calculated, and the specific operation steps are as follows: Determining a potential metabolite suspected enrichment area through a characteristic spectrum vector, acquiring an active closed contour from the potential metabolite suspected enrichment area, determining seed point growth from pixel points in the active closed contour to obtain an extended growth area, acquiring an intensity value of the pixel points under characteristic wavelength from the extended growth area, calculating a sample activity level index, further acquiring a sample activity level index of the same batch of preparation samples of the probiotic liquid preparation sample, and calculating a comprehensive activity score.
5. 5. The multispectral imaging detection method for determining the activity of the probiotics based on metabolite monitoring according to claim 4, wherein the potential metabolite suspected enrichment area is determined by the target spectrum similarity cluster through a characteristic spectrum vector, and the activity closed contour is obtained for the potential metabolite suspected enrichment area, and the specific operation steps are as follows: Calculating the difference vector of the mean value vector for each characteristic spectrum vector in the target spectrum similarity cluster, and further calculating the square norm of the difference vector; determining whether the scalar variance value is less than a compact threshold; if yes, judging the target spectrum similarity cluster as a potential metabolite suspected enrichment area; Determining two-dimensional coordinate positions of pixel points corresponding to the characteristic spectrum vectors in the target spectrum similarity cluster, determining a pixel point coordinate set of a suspected enrichment region according to the two-dimensional coordinate positions of the corresponding pixel points, clustering all the suspected enrichment regions to form a spectrum uniform region And carrying out morphological operation on the spectrum uniform region to obtain an active closed contour.
6. 6. The multispectral imaging detection method for determining the activity of the probiotics based on metabolite monitoring according to claim 5, wherein the method is characterized in that the pixel points inside the active closed contour are subjected to seed point growth determination to obtain an extended growth area, the intensity value of the pixel points under the characteristic wavelength is obtained for the extended growth area, a sample activity level index is calculated, further a sample activity level index of the same batch of preparation samples of the probiotic liquid preparation sample is obtained, and a comprehensive activity score is calculated, wherein the specific operation steps are as follows: randomly screening pixel points in the active closed contour to serve as seed points; Calculating the spectrum similarity of the characteristic spectrum vectors corresponding to the seed points and the neighborhood pixel points; judging whether the spectrum similarity is larger than a similarity threshold value or not; If yes, growing the adjacent pixel points, and repeating the growing step by taking the adjacent pixel points as new seed points until the growth is stopped, so as to obtain an extended growth area; The method comprises the steps of performing binarization masking on an extended growth area to obtain a mask of the extended growth area, obtaining intensity values of pixel points under characteristic wavelengths, calculating sample activity level indexes, obtaining sample activity level indexes of preparation samples of the same batch of probiotic liquid preparation samples, calculating batch uniformity indexes, judging an allowable fluctuation range of the batch uniformity indexes, screening a final metabolite enrichment area, and obtaining intensity values of the pixel points under the characteristic wavelengths for the final metabolite enrichment area to calculate comprehensive activity scores.
7. 7. The multispectral imaging detection method for determining the activity of the probiotics based on metabolite monitoring according to claim 6, wherein the extended growth region is subjected to a binarization mask to obtain an extended growth region mask, the intensity value of the pixel point under the characteristic wavelength is obtained, and the activity level index of the sample is calculated, wherein the method comprises the following specific operation steps: acquiring a single-band gray level image of reflectivity of the multispectral image data cuboid under the characteristic wavelength of a characteristic band; calculating a difference index image by using the single-band gray level image and the reference image to serve as a reaction intensity image; performing binarization masking on the extended growth area to obtain an extended growth area masking; Calculating the average gray value of all pixel points in the expanded growth area mask in the reaction intensity image, and taking the average gray value as a relative concentration index of short chain fatty acid; The method comprises the steps of obtaining intensity values of all pixel points in an extended growth area mask under characteristic wavelength according to the preset characteristic wavelength, calculating average spectral response intensity of the pixel points in the extended growth area mask by using the intensity values, and obtaining a sample activity level index by weighting and calculating the short chain fatty acid relative concentration index and the average spectral response intensity.
8. 8. The method for multi-spectral imaging detection based on metabolite monitoring and determination of probiotic activity according to claim 7, wherein the method comprises the steps of obtaining sample activity level indexes of the same batch of preparation samples of the probiotic liquid preparation sample, calculating batch uniformity indexes, determining allowable fluctuation range of the batch uniformity indexes, and screening final metabolite enrichment areas, wherein the method comprises the following specific operation steps: The preparation samples of the same batch of the probiotic liquid preparation samples are obtained, and the sample activity level index of each preparation sample of the same batch is obtained through the steps; calculating by using the batch average value and the batch standard deviation to obtain a batch uniformity index Collecting a qualified average uniformity value and an allowable fluctuation range of a historical reference batch; judging whether the batch uniformity index is not in the allowable fluctuation range or not; If so, judging that the sample activity level indexes of each preparation sample in the same batch have differences, returning to the step, and reselecting a characteristic spectrum vector in the characteristic spectrum data matrix by using a K-means algorithm to serve as an initial clustering center; calculating the distance between the initial clustering center and the rest of the characteristic spectrum vectors in the characteristic spectrum data matrix, and further calculating the square of the distance of the characteristic spectrum vectors; Calculating the distance square sum of all the characteristic spectrum vectors according to the distance square sum of the characteristic spectrum vectors to obtain a probability value of the next characteristic spectrum vector serving as an initial clustering center; Screening K initial cluster centers according to the probability values of the initial cluster centers; And continuously executing calculation of the distance between the characteristic spectrum vector of each pixel point in the characteristic spectrum data matrix and the k initial clustering centers, screening the initial clustering centers with the minimum distance, clustering the characteristic spectrum vector of all the pixel points in the minimum distance range to the initial clustering centers to obtain an initial clustering cluster, repeating the steps to obtain a new batch uniformity index, and judging that the new batch uniformity index is in the allowable fluctuation range to obtain an expanded growth area mask as a final metabolite enrichment area.
9. 9. A multispectral imaging detection system for judging the activity of probiotics based on metabolite monitoring is characterized by comprising an acquisition module, an analysis module, an identification module, a detection module and a detection module, wherein the acquisition module is used for acquiring the activity of the probiotics; The acquisition module is used for acquiring a multispectral image data cuboid for the probiotic liquid preparation sample, wherein the multispectral image data cuboid comprises a plurality of pixel point coordinates under preset characteristic wavelengths and spectrum vectors corresponding to the pixel point coordinates; The analysis module is used for reducing the dimension of the multispectral image data cuboid by utilizing a principal component analysis method to obtain a characteristic spectrum data matrix, clustering by utilizing principal components in the characteristic spectrum data matrix to obtain an initial cluster, acquiring a target spectrum similarity cluster in the initial cluster by utilizing a characteristic spectrum vector in the characteristic spectrum data matrix, acquiring the intensity value of a pixel point under a characteristic wavelength for the target spectrum similarity cluster, and calculating a comprehensive activity score; The identification module is used for presetting an activity threshold value, and judging the activity threshold value through the comprehensive activity score to obtain a detection result of the probiotic liquid preparation sample.

Description

Multispectral imaging detection method and multispectral imaging detection system for judging activity of probiotics based on metabolite monitoring Technical Field The invention relates to the field of spectrum detection of microbial metabolites, in particular to a multispectral imaging detection method and a multispectral imaging detection system for judging the activity of probiotics based on metabolite monitoring. Background In many beverage products, probiotic formulations, including liquid formulations (e.g., fermentation broths, suspensions) and solid formulations (e.g., lyophilized powders, tablets, capsules), are used, the core quality attributes and nutritional efficacy of which are established on the number of viable bacteria of the probiotic bacteria in the product and their metabolic activity. Along with the continuous improvement of the requirements of the market on the product functionality and quality consistency, the traditional detection method which only takes the number of living bacteria as an index cannot meet the requirements of accurately and rapidly evaluating the actual functional state of the probiotics. Currently, viable count of probiotic formulations is largely dependent on traditional plating methods. The method is used as a gold standard, has the defects that the detection time is as long as 48-72 hours, the operation is complicated, the number of viable bacteria which can be cultivated can be reflected, and the real-time metabolic activity and the functional state of the bacteria can not be represented at all. Furthermore, the result is an overall average measurement of the sample, whether by chromatography or culture, with complete loss of any information about the spatial distribution of the active ingredient (live bacteria or metabolites) in the formulation. For formulations, uniformity of active ingredient distribution (whether aggregation or concentration gradient is present) is a key factor affecting product batch consistency, dose accuracy and functional stability, but the prior art is totally silent on this. To overcome the limitations of the conventional methods, some rapid detection techniques have been attempted to be applied in the field of probiotics, such as viable bacteria detection techniques based on ATP bioluminescence or specific fluorescent dyes. However, these methods still fall under the indirect presumption that they are not specifically associated with direct functional products such as SCFAs, and are susceptible to complex substrate interference, lack of specificity and accuracy. In recent years, spectroscopic analysis techniques (such as near infrared spectroscopy) have been attracting attention because of their rapid and nondestructive characteristics, but conventional spectroscopic techniques acquire an average spectrum of a point or a region on a sample, and signals thereof are superposition and mixing of spectra of all chemical components (living bacteria, dead bacteria, metabolites, carriers/media) in the sample. The mixed spectrum method has two fundamental defects, namely, firstly, poor specificity, difficulty in extracting weak characteristic signals of specific metabolites (such as SCFAs) from complex backgrounds with high sensitivity, and secondly, lack of spatial resolution capability, inability to identify and quantify microscopic spatial distribution and heterogeneity of the metabolites in a sample, which is just an important dimension for evaluating functions of flora and preparation process quality. Disclosure of Invention The invention aims to provide a multispectral imaging detection method and a multispectral imaging detection system for judging the activity of probiotics based on metabolite monitoring, which solve the technical problems pointed out in the prior art. The invention provides a multispectral imaging detection method for judging the activity of probiotics based on metabolite monitoring, which comprises the following operation steps: Collecting a multispectral image data cuboid of a probiotic liquid preparation sample, wherein the multispectral image data cuboid comprises a plurality of pixel point coordinates under preset characteristic wavelengths and spectral vectors corresponding to the pixel point coordinates, and the preset characteristic wave band is a characteristic spectral response wave band of short-chain fatty acid; The method comprises the steps of obtaining a characteristic spectrum data matrix, obtaining a target spectrum similarity cluster in the initial cluster by utilizing a principal component analysis method, obtaining an intensity value of a pixel point under characteristic wavelength for the target spectrum similarity cluster, and calculating a comprehensive activity score; And judging the comprehensive activity score and the activity threshold to obtain a detection result of the probiotic liquid preparation sample. Correspondingly, the invention also provides a multispectral imaging detection syste