Microimaging method particle Raman spectrum in-situ identification system and method

Abstract

The invention is applicable to the technical field of particle detection, and provides a microimaging particle Raman spectrum in-situ identification system and method, and the system comprises a light path module, an image acquisition module, a Raman spectrum acquisition module, an image processing module and a Raman spectrum processing module; the light path module is used for focusing the visible light beam on the particle filter membrane or focusing the laser beam on the to-be-detected particles; the image acquisition module is used for acquiring a visible light signal image of the particle filter membrane; the Raman spectrum acquisition module is used for acquiring original Raman spectrum data of particles to be detected; and the image processing module is used for identifying, marking, classifying, generating and screening the visible light signal image to obtain particles to be detected and a particle statistical data table. The system solves the problems that chemical components cannot be obtained and in-situ detection cannot be carried out, and achieves the effects of avoiding transfer interference and improving the identification accuracy of particles.

Description

Technical Field

[0001] The present invention relates to the technical field of particle detection, and more particularly to a system and method for in-situ identification of particles using Raman spectroscopy using a microscopic imaging method. Background Art

[0002] It's widely acknowledged that particulate matter in pharmaceuticals can trigger thrombosis, inflammation, or allergic reactions. This risk is particularly pronounced in injectables. Typical examples include fibers, glass shavings, rubber stoppers, and drug aggregates. Global pharmacopoeias set strict standards for the number and size of particles. While traditional light obscuration methods can count and size particles, they fail to provide compositional information, making it difficult to trace the source of contamination and optimize production processes.

[0003] Traditional microscopy methods classify and judge particles based on morphology, which is highly subjective and cannot distinguish particles with similar morphologies, such as silicone oil and protein aggregates. Scanning electron microscopy-energy dispersive spectrometry (SEM-EDS) requires destructive pretreatment such as gold spraying and can only provide elemental signals, without molecular structure / group information. Traditional infrared and Raman spectroscopy, although they can provide molecular group information and fingerprint information, require the particles to be picked and transferred to a specific substrate for characterization, which is difficult to do and cannot accurately characterize particles with specific morphologies in situ.

[0004] Therefore, a microparticle Raman spectroscopy in-situ identification system and method are proposed to solve the above problems. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a microscopic imaging Raman spectroscopy in situ identification system and method for microparticles, which avoids picking and transferring microparticles and improves the accuracy of microparticle identification.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solutions: a microscopic imaging method Raman spectroscopy in-situ identification system for microparticles, comprising an optical path module, an image acquisition module, a Raman spectrum acquisition module, an image processing module, and a Raman spectrum processing module; the optical path module is used to focus a visible light beam on a microparticle filter membrane or a laser beam on the microparticles to be detected; the image acquisition module is used to acquire visible light signal images of the microparticle filter membrane; the Raman spectrum acquisition module is used to acquire raw Raman spectrum data of the microparticles to be detected; the image processing module is used to identify, label, classify, generate, and screen the visible light signal images to obtain the microparticles to be detected and a microparticle statistical data table; the Raman spectrum processing module is used to preprocess, match, and supplement the microparticle statistical data table for the raw Raman spectrum data.

[0007] The present invention is further configured as follows: the image processing module includes a particle recognition unit, a particle marking unit, a particle classification unit, a particle screening unit and a table generation unit; the particle recognition unit is used to identify particles in the visible light signal image, and calculate the shape and size parameters of the particles to obtain the morphology of the particles; the particle marking unit is used to number the particles according to the shape or size parameters of the particles to obtain the particle number; the particle classification unit is used to perform artificial intelligence classification based on the morphology to obtain the particle category; the particle screening unit is used to screen out the particles to be detected according to the shape, size parameters or particle category of the particles; the table generation unit is used to generate a particle statistical data table based on the shape, size parameters, particle number and particle category of the particles.

[0008] The present invention is further configured as follows: the Raman spectrum processing module includes a preprocessing unit, a data matching unit and a data filling unit; the preprocessing unit is used to preprocess the original Raman spectrum data to generate preprocessed Raman spectrum data; the data matching unit is used to match the Raman spectrum library and the corresponding preprocessed Raman spectrum data of the particles to be detected to obtain the chemical composition; the data filling unit is used to add the corresponding chemical composition of the particles to be detected and the preprocessed Raman spectrum data into the particle statistical data table and display them in the visible light signal image.

[0009] By adopting the above technical solution, Raman spectroscopy is performed on microparticles to obtain the chemical composition of the corresponding microparticles to be detected, so that the source of the microparticles (exogenous or endogenous) can be intuitively understood, the generation and introduction of microparticles can be prevented to the greatest extent, and a systematic, scientific, and targeted microparticle control system can be established, thereby improving the system's ability to optimize the production process.

[0010] A microscopic imaging method for in-situ identification of microparticles by Raman spectroscopy, using the microscopic counting method for in-situ identification of microparticles by Raman spectroscopy as described above, comprises the following steps: S1. Prepare a particle filter membrane according to the microscopic counting method; S2, focusing the visible light beam on the particle filter and collecting visible light signal images; S3. Identify particles in the visible light signal image, number the particles according to their shape and size parameters to obtain particle numbers, classify the particles according to their morphology to obtain particle categories, and summarize the particle shapes, size parameters, particle numbers, and particle categories to generate a particle statistical data table; S4. Screening the particles to be tested and the corresponding particle numbers according to the shape, size parameters and particle types of the particles; S5, irradiating the particles to be detected with a laser beam and collecting raw Raman spectrum data; S6. Preprocessing the original Raman spectrum data to generate preprocessed Raman spectrum data; S7, obtaining chemical composition based on matching between the Raman spectrum library and the corresponding pre-processed Raman spectrum data of the particles to be detected; S8. Summarize and add the particle statistical data table according to the particle number, particle type, chemical composition and pre-processed Raman spectrum data to be detected, and display it in the visible light signal image.

[0011] By adopting the above technical solution, the original Raman spectrum data is matched to obtain a particle statistical data table, which can display the relevant information of the particles to be detected in the table, thereby improving the efficiency of operators in obtaining detection information.

[0012] The present invention is further configured as follows: S21, partitioning the particle filter membrane; S22, collecting images of each area; S23, stitching the collected images; S24: Generate a visible light signal image.

[0013] The present invention is further configured as follows: in S3, the classification according to the morphology of the particles refers to classifying the identified particles into different types of particle sources based on artificial intelligence.

[0014] The present invention is further configured as follows: in S6, the preprocessing is performed based on the original Raman spectrum data to generate preprocessed Raman spectrum data, including the following steps: performing peak removal, smoothing, baseline correction, and intensity normalization on the original Raman spectrum data in sequence to generate preprocessed Raman spectrum data.

[0015] The present invention is further configured as follows: in S7, the chemical composition is calculated by a spectrum peak matching coefficient algorithm.

[0016] The present invention is further configured as follows: the formula of the spectrum peak matching coefficient algorithm is: ; Among them, C pmc is the peak matching coefficient, M is the total number of matched peaks, ω peak (m) is the weight of the m-th peak.

[0017] By adopting the above technical solution, the original Raman spectrum data of the particles to be detected is placed in the constructed Raman spectrum library for retrieval, and one or more samples with the highest similarity to the particles to be detected are calculated, thereby achieving qualitative analysis and improving the efficiency and accuracy of particle composition detection.

[0018] In summary, this application includes at least one of the following beneficial technical effects: 1. Perform Raman spectroscopy on microparticles to obtain the chemical composition of the corresponding microparticles to be tested, so as to intuitively understand the source of the microparticles (exogenous or endogenous), prevent the generation and introduction of microparticles to the greatest extent, establish a systematic, scientific and targeted microparticle control system, and thus improve the system's ability to optimize the production process.

[0019] 2. The original Raman spectrum data is matched to obtain a spectrum identification table, which can display the relevant information of the particles to be detected in the table, thereby improving the efficiency of operators in obtaining detection information.

[0020] 3. The original Raman spectrum data of the particles to be detected is placed in the constructed Raman spectrum library for retrieval, and one or more samples with the highest similarity to the particles to be detected are calculated, thereby achieving qualitative analysis and improving the efficiency and accuracy of particle component detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a schematic diagram of the relationship between a microscopic imaging method and Raman spectroscopy in-situ identification system for microparticles of the present invention.

[0022] Figure 2 Schematic diagram of steps S1-S4 of a microscopic imaging Raman spectroscopy in-situ identification method for microparticles according to the present invention.

[0023] Figure 3 Schematic diagram of steps S4-S8 of the in-situ identification method of microparticles using Raman spectroscopy using a microscopic imaging method of the present invention. DETAILED DESCRIPTION

[0024] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0025] It should be noted that, unless otherwise specified, all technical and scientific terms used in this application have the same meaning as commonly understood by ordinary technicians in the technical field to which this application belongs.

[0026] See also Figure 1-3 , the present invention provides the following technical solutions: Example 1, see Figure 1 , a microparticle Raman spectroscopy in-situ identification system using a microscopic imaging method, comprising an optical path module, an image acquisition module, a Raman spectrum acquisition module, an image processing module and a Raman spectrum processing module; An optical path module, used to focus a visible light beam on a particle filter or a laser beam on particles to be detected; Among them, the visible light path includes a visible light beam output path composed of a visible light source, a spectroscope, a microscope objective lens, etc.; the laser beam is focused on the particles to be detected and generates a Raman spectrum signal. The laser light path includes a laser beam output path composed of a single-mode laser, a spectroscope, a Raman filter, a microscope objective lens, etc. The image acquisition module is used to collect visible light signal images of the particle filter membrane; The image acquisition module is composed of a visible light signal input path consisting of a microscope objective lens, a relay lens, an imaging camera, etc. The Raman spectrum acquisition module is used to collect the original Raman spectrum data of the particles to be detected; The Raman spectrum acquisition module consists of a microscope objective, Raman filter, confocal pinhole, focusing lens group, monochromator, signal detector, etc., which form the Raman signal input path. The Raman signal is split inside the monochromator, and the signal detector obtains Raman spectrum data at different wavelength positions. The image processing module is used to identify, mark, classify, generate and filter the visible light signal image to obtain the particles to be detected and the particle statistical data table; The image processing module includes a particle recognition unit, a particle marking unit, a particle classification unit, a particle screening unit and a table generation unit; The particle recognition unit is used to identify particles in the visible light signal image and calculate the shape and size parameters of the particles to obtain the morphology of the particles; The particle marking unit is used to number the particles according to the shape or size parameters of the particles to obtain the particle numbers; The particle classification unit is used to perform artificial intelligence classification based on the morphology to determine the particle category; The particle screening unit is used to screen out particles to be tested based on the shape, size parameters or particle types of the particles; The table generating unit is used for generating a particle statistical data table according to the shape, size parameters, particle number and particle category of the particles.

[0027] The Raman spectrum processing module is used to preprocess, match and supplement the particle statistics table of the original Raman spectrum data; The Raman spectrum processing module includes a pre-processing unit, a data matching unit and a data filling unit; The preprocessing unit is used to preprocess the original Raman spectrum data to generate preprocessed Raman spectrum data; The data matching unit is used to match the Raman spectrum library with the corresponding pre-processed Raman spectrum data of the particles to be detected to obtain the chemical composition; The data filling unit is used to add the chemical composition of the corresponding particles to be detected and the pre-processed Raman spectrum data into the particle statistical data table and display them in the visible light signal image.

[0028] By performing Raman spectroscopy on microparticles, the chemical composition of the corresponding particles to be tested can be obtained, so that the source of the particles (exogenous or endogenous) can be intuitively understood, the generation and introduction of particles can be prevented to the greatest extent, and a systematic, scientific and targeted particle control system can be established, thereby improving the system's ability to optimize the production process.

[0029] Example 2, see Figure 2-3 A microscopic imaging method for in-situ identification of microparticles by Raman spectroscopy, using the above-mentioned microscopic counting method for in-situ identification of microparticles by Raman spectroscopy, comprises the following steps: S1. Prepare the microparticle filter membrane according to the second method of microscopic counting method in Part IV of the 2020 edition of the Chinese Pharmacopoeia 0903; Specifically, the operator takes a single sample, washes the outer surface of the sample with inspection water, carefully flips the sample 20 times on a clean workbench to mix the sample evenly, opens the sample cover, and uses a pipette to extract all the solution in the sample, slowly injecting it into the filter along the inner wall. Let it stand for 1 minute, then slowly filter it until the filter membrane is nearly dry. Then, take another 25mL of inspection water and slowly inject it along the inner wall of the filter, washing and filtering it until the filter membrane is nearly dry. Then, use flat-tipped tweezers to transfer the filter membrane to a petri dish. Slightly open the lid to allow the filter membrane to dry properly, close the petri dish, and place it on the microscope stage.

[0030] S2, focusing the visible light beam on the particle filter and collecting visible light signal images; The more specific steps of S2 are: S21, partitioning the particle filter membrane; S22, collecting images of each area; S23, stitching the collected images; S24: Generate a visible light signal image.

[0031] S3. Identify particles in the visible light signal image, number the particles according to their shape and size parameters to obtain particle numbers, classify the particles according to their morphology to obtain particle categories, and summarize the particle shapes, size parameters, particle numbers, and particle categories to generate a particle statistical data table; Specifically, the shape and size parameters of each particle include: major diameter, area, minor diameter, circumference, diameter and aspect ratio; the identified particles can be sorted, numbered and labeled according to the shape and size parameters; classification is completed by using the deep convolutional neural network model ResNet18 in the artificial intelligence algorithm. By training and learning 13 types of particle images such as rubber, metal, protein aggregates, silicone oil, glass, fiber, microsphere aggregates, cell aggregates, magnetic beads, endogenous particles, drug shedding, insects, and culture medium clumps in the model, and verifying the optimal parameters, it can be classified according to morphology; by classifying the particle morphology, it can facilitate subsequent preliminary screening, divide obvious particle categories into particles to be detected, and improve subsequent screening efficiency.

[0032] S4. Screening the particles to be tested and the corresponding particle numbers according to the shape, size parameters and particle types of the particles; Specifically, screening is to further screen the particles to be detected by parameters such as length diameter, area, and shape ratio for one or several classified particles; for example, for endogenous drug particles classified by artificial intelligence algorithms, particles with a length diameter ≥50μm are screened as particles to be detected, and their corresponding particle numbers are extracted.

[0033] The particles to be detected with particle numbers are obtained through the visible light signal image, which is convenient for subsequent in-situ detection of the particles to be detected, that is, subsequent detection is performed without moving the particle filter membrane, and the results of the subsequent detection can correspond to the particle numbers detected by the visible light signal image, and the corresponding particle numbers can also correspond to the positions of the particles to be detected on the visible light signal image, so that the operator can intuitively understand the relevant information of the particles to be detected on the particle filter membrane.

[0034] S5, irradiating the particles to be detected with a laser beam and collecting raw Raman spectrum data; The laser beam needs to be configured. For example, the excitation wavelength needs to be preset to 785nm, the laser intensity needs to be 50%, the grating needs to be 1200gr / mm, the pinhole needs to be 50μm, and the spectral range needs to be 100~3000cm -1 , the integration time is 10s, the number of acquisitions is 3 average spectra, automatic confocal, and the laser can be turned on for laser beam irradiation only after the configuration is completed, and the original Raman spectrum data of each particle to be detected can be automatically collected.

[0035] S6. Preprocessing the original Raman spectrum data to generate preprocessed Raman spectrum data; The more specific steps of S6 are: S61. Peak removal processing based on cyclic elimination; specifically, the original Raman spectrum data is filtered by a low-pass filter, the peak is identified according to the size of the residual, and the element value at the peak is replaced by the mean value of the elements in the peak neighborhood, while the element values of the non-peak parts in the original spectrum remain unchanged, to obtain a spectrum with preliminary peak removal, and then the spectrum is low-pass filtered to obtain a smoother spectrum. The original spectrum minus the spectrum is obtained to obtain a sequence containing only the peak, and the above process is repeated until all the values in the peak sequence are zero; S62. Smoothing processing based on a cyclic three-point zero-order Savitzky-Golay filtering method; specifically, the original Raman spectrum data after the spike removal processing is filtered by setting the window width to 3 and performing a sliding window average filter. Since the window width is already the minimum, the amount of noise filtered out by each cyclic filtering is also the minimum. By cyclically using the three-point sliding window averaging method to filter the spectrum, the noise in the original Raman spectrum data is slowly removed.

[0036] S63. Baseline correction based on the principle of a small window sliding average method. Specifically, initially, the raw Raman spectral data, which has undergone the above-described smoothing process, is smoothed using the smallest sliding window. Positions in the raw Raman spectral data that are larger than the smoothed spectrum are identified as spectral peaks and removed. This process is repeated continuously, with the window width gradually increasing. The area of the peak eliminated with each cycle generally increases first and then decreases. When the area of the eliminated peak increases again, it indicates that the peak has been substantially eliminated and the increase in area is primarily due to the baseline. The cycle is then terminated.

[0037] S64, intensity normalization processing; specifically, the original Raman spectrum data that has undergone the above-mentioned baseline correction processing is adjusted to the same standard value by adjusting the integrated area or intensity of the signal to generate pre-processed Raman spectrum data.

[0038] S7, obtaining chemical composition based on matching between the Raman spectrum library and the corresponding pre-processed Raman spectrum data of the particles to be detected; The Raman spectrum library is established by pre-processing the learning samples as described above, labeling them and adding them to the Raman spectrum library. The multi-scale peak detection (MSPD) peak search algorithm is used to search for peaks in the pre-processed Raman spectrum data with labels, and the peak discrimination algorithm based on Voigt function fitting is used to obtain the fitted peak height a of the searched peak. i , thereby obtaining the weight ω of each peak in the Raman spectrum of each learning sample peak (i); The formula for the weight of each peak is: ; Where k is the total number of components; The formula for the spectrum peak matching coefficient algorithm is: ; Among them, C pmc is the peak matching coefficient, M is the total number of matched peaks, ω peak (m) weight of the mth peak; In the above two formulas, i represents the relevant parameters of the sample used for learning in the establishment of the Raman spectrum library, and m represents the relevant parameters of the particles to be detected during the detection process; S8. Summarize and add the particle statistical data table according to the particle number, particle type, chemical composition and pre-processed Raman spectrum data to be detected, and display it in the visible light signal image.

[0039] By matching the original Raman spectral data to obtain a spectral identification table, the relevant information of the particles to be detected can be displayed in the table, improving the efficiency of operators in obtaining detection information. The original Raman spectral data of the particles to be detected can be placed in the constructed Raman spectrum library for retrieval through the spectral peak matching coefficient, and one or more samples with the highest similarity to the particles to be detected can be calculated, thereby realizing qualitative analysis and improving the efficiency and accuracy of particle composition detection.

[0040] Obviously, the embodiments described above are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

Claims

1. A microscopic imaging Raman spectroscopy in-situ identification system for microparticles, characterized by: include: An optical path module is used to focus a visible light beam on a particle filter or a laser beam on particles to be detected; An image acquisition module, used for acquiring visible light signal images of the particle filter membrane; Raman spectrum acquisition module, used to collect original Raman spectrum data of the particles to be detected; An image processing module is used to identify, mark, classify, generate and filter visible light signal images to obtain particles to be detected and a particle statistical data table; as well as Raman spectrum processing module is used to preprocess, match and supplement the particle statistics table of raw Raman spectrum data.

2. The microparticle Raman spectroscopy in-situ identification system according to claim 1, characterized in that: The image processing module includes a particle recognition unit, a particle marking unit, a particle classification unit, a particle screening unit and a table generation unit; The particle recognition unit is used to identify particles in the visible light signal image and calculate the shape and size parameters of the particles to obtain the morphology of the particles; The particle marking unit is used to number the particles according to the shape or size parameters of the particles to obtain particle numbers; The particle classification unit is used to perform artificial intelligence classification based on the morphology to determine the particle category; The particle screening unit is used to screen out particles to be detected based on the shape, size parameters or particle types of the particles; The table generating unit is used to generate a particle statistical data table according to the shape, size parameters, particle number and particle category of the particles.

3. The microparticle Raman spectroscopy in-situ identification system according to claim 2, characterized in that: The Raman spectrum processing module includes a pre-processing unit, a data matching unit and a data filling unit; The preprocessing unit is used to preprocess the original Raman spectrum data to generate preprocessed Raman spectrum data; The data matching unit is used to match the Raman spectrum library with the corresponding pre-processed Raman spectrum data of the particles to be detected to obtain the chemical composition; The data supplementation unit is used to supplement the chemical composition of the corresponding particles to be detected and the pre-processed Raman spectrum data into the particle statistical data table and display them in the visible light signal image.

4. A microscopic imaging method for in-situ identification of microparticles by Raman spectroscopy, using the microscopic imaging method for in-situ identification of microparticles by Raman spectroscopy as claimed in claim 3, characterized in that: The following steps are involved: S1. Prepare a particle filter membrane according to the microscopic counting method; S2, focusing the visible light beam on the particle filter and collecting visible light signal images; S3. Identify particles in the visible light signal image, number the particles according to their shape and size parameters to obtain particle numbers, classify the particles according to their morphology to obtain particle categories, and summarize the particle shapes, size parameters, particle numbers, and particle categories to generate a particle statistical data table; S4. Screening the particles to be tested and the corresponding particle numbers according to the shape, size parameters and particle types of the particles; S5, irradiating the particles to be detected with a laser beam and collecting raw Raman spectrum data; S6. Preprocessing the original Raman spectrum data to generate preprocessed Raman spectrum data; S7, obtaining chemical composition based on matching between the Raman spectrum library and the corresponding pre-processed Raman spectrum data of the particles to be detected; S8. Summarize and add the particle statistical data table according to the particle number, particle type, chemical composition and pre-processed Raman spectrum data to be detected, and display it in the visible light signal image.

5. The in-situ identification method of microparticles by Raman spectroscopy using microscopic imaging according to claim 4, characterized in that: In S2, the collecting of visible light signal images includes the following steps: S21, partitioning the particle filter membrane; S22, collecting images of each area; S23, stitching the collected images; S24: Generate a visible light signal image.

6. The in-situ identification method of microparticles by Raman spectroscopy using microscopic imaging according to claim 4, characterized in that: In S3, the classification according to the morphology of the particles refers to classifying the identified particles into different types of particle sources based on artificial intelligence.

7. The in-situ identification method of microparticles by Raman spectroscopy using microscopic imaging according to claim 4, characterized in that: In S6, the preprocessing is performed on the original Raman spectrum data to generate preprocessed Raman spectrum data, including the following steps: performing peak removal, smoothing, baseline correction, and intensity normalization on the original Raman spectrum data in sequence to generate preprocessed Raman spectrum data.

8. The in-situ identification method of microparticles by Raman spectroscopy using microscopic imaging according to claim 4, characterized in that: In S7, the chemical composition is calculated by a spectrum peak matching coefficient algorithm.

9. The in-situ identification method of microparticles by Raman spectroscopy using microscopic imaging according to claim 8, wherein the peak matching coefficient algorithm is calculated as follows: ； in, C pmc is the peak matching coefficient, M is the total number of matched peaks, ω peak (m) is the weight of the m-th peak.

Images