An environmental microplastic living four-dimensional raman monitoring system and dynamic quantitative method

Abstract

The application discloses an environmental microplastic living body four-dimensional Raman monitoring system and a dynamic quantitative method, and relates to the technical field of environmental ecological monitoring and analysis. The system comprises a confocal spectrum acquisition module, a controller and a data processing module. Three-dimensional coordinates of a living body sample are obtained through coordinated spatial scanning, and Raman signals are synchronously acquired to construct four-dimensional spectral data. A multi-dimensional convolution instance segmentation network is used to process the four-dimensional Raman spectral data, identify and classify endogenous biological matrix and exogenous multiple mixed microplastics in the living body, output voxel-level semantic classification results, and then quantitatively calculate the absolute trapped volume and abundance of microplastics in the living body. The application overcomes the defect that the prior art cannot be labeled and quantified in a complex biological background, realizes in-situ accurate identification of mixed microplastics in the living body, and provides a standardized quantitative monitoring means for evaluating the biological enrichment, metabolic law and ecological toxicological effect of environmental microplastics.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring and ecological environmental protection technology, specifically to a technology for detecting environmental pollutants using spectral analysis, and more particularly to a four-dimensional Raman monitoring system and dynamic quantitative method for environmental microplastics. Background Technology

[0002] Microplastics, as environmental pollutants, exist in natural ecosystems such as water and soil, and are transferred and accumulated step by step through the food chain. Microplastics in the natural environment are usually not a single component, but a complex mixture of various polymers, including polyamide (PA), polyethylene (PE), polystyrene (PS), polyethylene terephthalate (PET), and polymethyl methacrylate (PMMA). To accurately assess the ecotoxicological effects and environmental risks of such pollutants, it is necessary to move beyond the limitations of traditional in vitro detection methods targeting only water / soil samples and specifically identify, in situ, the types of mixed microplastics that accumulate in living organisms, their three-dimensional spatial distribution, and their metabolic accumulation patterns in ecosystems.

[0003] Currently, existing microplastic in vivo detection technologies mainly rely on fluorescently labeled microscopic imaging. However, during fluorescent staining of environmental samples, dye shedding and quenching, as well as changes in the surface chemical properties of microplastics caused by fluorescent groups, easily introduce false positive errors, leading to distorted environmental toxicology assessments. Raman spectroscopy, as a label-free material analysis and detection technique, can obtain molecular fingerprints and is theoretically very suitable for in-situ monitoring of pollutants. However, when existing Raman techniques are directly applied to environmental model organisms (such as zebrafish larvae and water fleas), they are easily interfered with by the strong autofluorescence background of biological tissues and the overlapping Raman peaks of complex biochemical molecules. At the same time, existing in vivo detection techniques such as stimulated Raman scattering are limited by narrow-band spectra, usually only able to track a single pure microplastic polymer in a single channel, making it difficult to simultaneously monitor multi-component mixed microplastics in the real environment.

[0004] Furthermore, existing in vivo optical detection technologies are mostly limited to qualitative two-dimensional or three-dimensional image display, lacking mathematical quantification methods to convert discrete optical scanning signals into physical "absolute cutoff volumes." Due to the lack of high-precision in-situ quantitative calculation models, existing technologies struggle to accurately assess the dynamic uptake, accumulation, and excretion of environmentally mixed microplastics in vivo, including their biodynamic processes. This not only hinders a deeper understanding of the differences in microplastic accumulation and clearance within organisms but also severely limits the development of environmental toxicology assessment systems towards standardization and absolute quantification. Therefore, there is an urgent need for a label-free, multiplexer, and three-dimensional absolute quantification Raman monitoring system and method for environmental microplastics in complex in vivo biological settings. Summary of the Invention

[0005] The purpose of this invention is to provide a four-dimensional Raman monitoring system and dynamic quantitative method for environmental microplastics in vivo, in order to solve the technical problems in the prior art, such as the tendency of traditional fluorescent labeling methods to produce false positives, the difficulty of existing Raman technology in simultaneously identifying multi-component mixed microplastics in complex biological backgrounds, and the lack of three-dimensional absolute volume quantitative assessment methods for the dynamic accumulation and metabolic processes of environmental microplastics in vivo.

[0006] To achieve the above objectives, the present invention provides a four-dimensional Raman monitoring system for living environmental microplastics, the system comprising: The confocal spectral acquisition module is configured to perform XYZ layer-by-layer spatial scanning on live samples exposed to a mixed microplastic environment to obtain three-dimensional spatial coordinates (x, y, z), and simultaneously acquire wavelengths covering 500 to 3200 cm⁻¹. -1 Raman spectral signals within the range; The controller, which is communicatively connected to the confocal spectral acquisition module, is configured to coordinate the synchronous acquisition of the spatial scanning and Raman spectral signals. The data processing module, communicatively connected to the controller, is configured to fuse the acquired three-dimensional spatial coordinates (x, y, z) with the Raman scattering signal to construct four-dimensional Raman spectral data (x, y, z, λ); and input the four-dimensional Raman spectral data into a pre-trained microplastic multidimensional convolutional instance segmentation network to generate voxel-level semantic segmentation and classification results, thereby achieving simultaneous label-free in-situ classification of endogenous biological matrices and exogenous multi-mixed microplastics in living organisms.

[0007] Furthermore, the confocal spectral acquisition module specifically includes an optical acquisition unit and a spectral detection module connected in sequence via optical paths; the optical acquisition unit includes, in sequence via optical paths: a laser source, a beam expander lens, a first collimating lens, a first reflecting mirror, a second reflecting mirror, a dichroic mirror, a two-dimensional galvanometer combined with a stage for loading a live sample, a scanning lens, a tube lens, and a microscope objective; the backscattered signal generated by the live sample returns to the dichroic mirror via its original path, passes through a notch filter and a second collimating lens in sequence, and is then focused at a confocal pinhole for spatial filtering before entering the spectral detection module; the spectral detection module includes a first concave reflecting mirror, a grating, a second concave reflecting mirror, and a camera arranged in sequence, for synchronously acquiring the Raman scattering signal.

[0008] This invention also provides a method for the dynamic quantification of environmental microplastics in vivo based on the above system. This method is used for assessing the environmental microplastic pollution effects for non-disease diagnostic purposes and includes the following steps: Step S1, in-situ preparation of live organisms: The live organism sample exposed to the mixed microplastic environment is embedded and fixed, and placed on the displacement stage of the confocal spectroscopy acquisition module; Step S2, Four-dimensional Raman spectral data scanning and fusion: The controller coordinates and executes XYZ layer-by-layer spatial scanning, simultaneously acquiring three-dimensional spatial coordinates (x, y, z) and 500 to 3200 cm⁻¹. -1 Raman spectral signals within a certain range are fused together to construct four-dimensional Raman spectral data (x, y, z, λ). Step S3, Multi-channel component decoding: Input the four-dimensional Raman spectral data into a pre-trained microplastic multidimensional convolutional instance segmentation network, extract high-dimensional spectral features, perform endogenous biological matrix recognition and exogenous multi-mixed microplastic recognition, and output voxel-level semantic segmentation and recognition classification results; Step S4, Qualitative material distribution visualization: Based on voxel-level semantic classification, different endogenous biological matrices and different exogenous multi-mixed microplastics are given pseudo-color to visualize the distribution of endogenous biological matrices and the in vivo spatial distribution of exogenous multi-mixed microplastics. Step S5, quantitative three-dimensional dynamic integration: extract the number of voxels of each type of microplastic in each level of the Z-axis; perform three-dimensional spatial integration on the full scan volume to calculate the absolute retention volume and abundance of each type of microplastic; and evaluate the metabolic kinetics of dynamic uptake, accumulation and excretion of mixed microplastics in vivo.

[0009] Furthermore, the microplastic multidimensional convolutional instance segmentation network described in step S3 adopts an encoder-decoder architecture, specifically characterized by: the encoder part integrating a three-dimensional convolutional module, a residual module, and a three-dimensional dilated convolutional module; the network extracts data from 500 to 3200 cm⁻¹. -1 The Raman spectral information will include specific Raman peaks containing the low-frequency fingerprint region and 3000 cm⁻¹. -1 The overlapping peaks of CH stretching vibrations of nearby lipids and proteins are used to verify cross-spectral region features, thereby decoupling the autofluorescence background of biological tissues from microplastic signals. Simultaneously, the three-dimensional dilated convolution module is used to expand the receptive field of three-dimensional features while maintaining physical spatial resolution, enabling the segmentation of multiple mixed microplastics exhibiting irregular aggregation or entanglement. The decoder part maps the extracted high-dimensional features back to the original three-dimensional physical space through upsampling operations, directly assigning classification labels corresponding to endogenous biological matrices and various types of exogenous microplastics.

[0010] Furthermore, in step S5, the process of calculating the absolute retention volume and abundance of various microplastics is as follows: The physical volume of a single step during scanning is defined as the basic voxel unit v0 = Δ. x ×Δ y ×Δ z， Where Δ x and Δ y Δ is the scanning step resolution of the two-dimensional galvanometer (7).z The axial step resolution of the displacement stage (11); Define a classification mask function C k ( x, y, z i ): Regarding the first of the multiple mixed microplastics k Based on the identification results output by the multidimensional convolutional instance segmentation network for the microplastics, the classification is determined for microplastics at the depth slice level. z i And the plane coordinates are ( x,y A specific three-dimensional voxel, when the voxel is recognized by the network as the first... k When considering a category of microplastics, the first... k Classification mask function for microplastics C k ( x, y, z i The value is 1 if it is positive and 0 otherwise. The first step is to address the aforementioned first... k Microplastics are categorized into different types, based on their corresponding classification mask functions. C k ( x, y, z i ), calculate each independent depth slice level on the Z-axis z i Inside, the first k The cumulative added value of two-dimensional feature surfaces occupied by different microplastic categories S k ( z i The calculation formula is: ; The second step is to process all independent depth slice levels. z i The extracted two-dimensional feature surface accumulation value S k ( z i Discrete integration is performed along the Z-axis depth direction to calculate the first... k Absolute cutoff volume of various microplastic categories within the full scan volume Volume k The calculation formula is: ; Where X, Y, and Z represent the total number of step points in the three-dimensional scanning space along the length, width, and depth dimensions, respectively; by traversing all category indices of multiple hybrid microplastics... kThe absolute retention volume of each microplastic in vivo was obtained separately, enabling the separate quantification of multiple mixed microplastics.

[0011] Furthermore, the assessment of the dynamic uptake, accumulation, and excretion of mixed microplastics in vivo as described in step S5 is achieved using a time-series scanning scheme, specifically including: extracting data at each time point. t The absolute retention volume of each type of microplastic calculated below Volume k (t) We plotted dynamic accumulation-clearance kinetics curves of microplastics of each component over time to quantitatively characterize the differences in bioaccumulation and metabolic clearance of different microplastics in living samples.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) Overcoming the bottleneck of live label-free identification of multiple pollutants in complex ecological samples: The system of this invention has a range of 500 to 3200 cm -1 The spectral coverage range includes specific Raman peaks, including the low-frequency fingerprint region, and 3000 cm⁻¹. -1 The overlapping peaks of CH stretching vibrations of nearby lipids and proteins were used to verify the cross-spectral region characteristics, thereby decoupling the strong biological tissue autofluorescence background from the microplastic signal and solving the problem that traditional single-channel detection is difficult to accurately identify multi-component microplastics in complex biological backgrounds. (2) Significantly improves the fine segmentation capability of micro-pollutant clusters under complex physiological environments: This invention integrates three-dimensional dilated convolution and residual modules in the encoder-decoder architecture. While maintaining sub-micron physical spatial resolution, it greatly expands the receptive field of the algorithm for three-dimensional features, and realizes voxel-level semantic segmentation of multiple components of polyamide (PA), polyethylene (PE), polystyrene (PS), polyethylene terephthalate (PET) and polymethyl methacrylate (PMMA), especially improving the classification accuracy of dense cluster edges; (3) A dynamic quantitative method and metabolic assessment system for in vivo monitoring of environmental microplastics based on voxels were constructed: This invention defines a classification mask function. C k ( x, y, z i The discrete four-dimensional Raman spectral signal is converted into an absolute cutoff volume using the basic voxel unit v0. Volume kThis method not only enables visualization of spatial distribution, but also quantitatively characterizes the bioaccumulation, dynamic accumulation, and metabolic clearance kinetics of different types of microplastics in living organisms (such as zebrafish larvae and water fleas) through time-series scanning. It overcomes the limitations of traditional fluorescence or existing Raman imaging, which can only provide qualitative descriptions, and provides a precise and standardized quantitative means for environmental toxicology assessment. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the optical path structure of the four-dimensional Raman monitoring system for living microplastics in the environment provided in an embodiment of the present invention. Figure 1 In the configuration, there are: laser source 1, beam expander lens 2, first collimating lens 3, first reflecting mirror 4, second reflecting mirror 5, dichroic mirror 6, two-dimensional galvanometer 7, scanning lens 8, tube lens 9, microscope objective lens 10, displacement stage 11, notch filter 12, second collimating lens 13, confocal pinhole 14, first concave reflecting mirror 15, grating 16, second concave reflecting mirror 17, camera 18, optical acquisition unit 19, spectral detection module 20, controller 21, and data processing module 22.

[0014] Figure 2 This is a schematic diagram of the process for a method for dynamic quantification of living microplastics in the environment provided in an embodiment of the present invention.

[0015] Figure 3 Raman spectral curves of different types of microplastics and zebrafish live tissue background detected by the four-dimensional Raman monitoring system for environmental microplastics provided in this embodiment of the invention.

[0016] Figure 4 This is a schematic diagram of the microplastic multidimensional convolution instance segmentation network architecture provided in an embodiment of the present invention.

[0017] Figure 5 This invention provides a 3D visualization comparison of the qualitative spatial distribution of substances in live zebrafish exposed to a mixed microplastic environment for 3 hours and 8 hours, as shown in the embodiments of the present invention. Figure 5 Image (a) in the image is from an image taken after 3 hours of exposure. Figure 5 (b) in the image is the image after 8 hours of exposure.

[0018] Figure 6 A quantitative statistical comparison chart of the absolute retention volume (voxels) of microplastics in different depth layers (D1 to D7) of live zebrafish after 3 hours of exposure, provided for embodiments of the present invention.

[0019] Figure 7 A quantitative statistical comparison chart of the absolute retention volume (voxels) of microplastics in different depth layers (D1 to D7) of live zebrafish after 8 hours of exposure, provided for embodiments of the present invention. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1: A four-dimensional Raman monitoring system for living environmental microplastics Figure 1 As shown. The system includes: a confocal spectral acquisition module, a controller 21, and a data processing module 22; The confocal spectral acquisition module specifically includes an optical acquisition unit 19 and a spectral detection module 20 connected in sequence via optical paths. The optical acquisition unit 19 includes, in sequence, a laser source 1, a beam expander lens 2, a first collimating lens 3, a first reflecting mirror 4, a second reflecting mirror 5, a dichroic mirror 6, a two-dimensional galvanometer 7, a scanning lens 8, a tube lens 9, a microscope objective lens 10, and a displacement stage 11; wherein, the two-dimensional galvanometer 7, in conjunction with the displacement stage 11 which is loaded with a live sample, is configured to perform XYZ layer-by-layer optical slicing scans to obtain high-resolution three-dimensional spatial coordinates (x, y, z). The backscattered signal generated by the live sample returns via the original path to the dichroic mirror 6, and then passes sequentially through the notch filter 12 and the second collimating lens 13 before being focused onto the confocal pinhole 14. The confocal pinhole 14 is used to perform physical spatial filtering to remove stray biological background light from the defocused surface. The signal passing through the confocal pinhole 14 enters the spectral detection module 20. The spectral detection module 20 includes a first concave mirror 15, a grating 16, a second concave mirror 17, and a camera 18 arranged sequentially, for simultaneously acquiring spectral data in the spectral range of 500 to 3200 cm⁻¹. -1 Raman spectral signal; The controller 21 is communicatively connected to the two-dimensional galvanometer 7, the displacement stage 11 and the camera 18, and is configured to: coordinate the two-dimensional galvanometer 7 and the displacement stage 11 to perform XYZ layer-by-layer optical slicing scanning, and synchronously trigger the camera 18 to acquire data, so as to obtain three-dimensional spatial coordinate information and Raman spectral information. The data processing module 22 is communicatively connected to the controller 21 and is configured to: fuse the acquired three-dimensional spatial coordinate information with the Raman spectral signal to construct four-dimensional Raman spectral data, and input it into a pre-trained microplastic multidimensional convolutional instance segmentation network to generate voxel-level semantic segmentation and classification results, thereby achieving synchronous in-situ classification of endogenous biological matrix and exogenous multi-mixed microplastics.

[0022] Example 2: A method for dynamic quantitative analysis of environmental microplastics based on the above system, such as... Figure 2 As shown, the method is used for the biological uptake and metabolic assessment of environmental microplastic pollution, and the specific steps are as follows: Step S1, In-situ preparation of live specimens: Live specimens used for environmental pollution assessment and exposed to a mixed microplastic environment containing polyamide (PA), polyethylene (PE), polystyrene (PS), polyethylene terephthalate (PET) and polymethyl methacrylate (PMMA) are physically restricted and embedded and fixed, and placed on the displacement stage 11 of the confocal spectroscopy acquisition module. Step S2, Four-dimensional Raman spectral data scanning and fusion: Controller 21 controls the two-dimensional galvanometer 7 and the displacement stage 11 to perform XYZ layer-by-layer optical slice scanning to obtain high-resolution three-dimensional spatial coordinates (x, y, z); the backscattered signal generated by the live sample returns along the original path, passes through the notch filter 12 and the second collimating lens 13 in sequence, and is focused on the confocal pinhole 14. After spatial filtering by the confocal pinhole 14, the spectral range of 500 to 3200 cm⁻¹ is simultaneously acquired using the spectral detection module 20. -1 Raman spectral signals, such as Figure 3 As shown, the Raman spectral curves of different types of microplastics (PA, PE, PS, PET) and zebrafish live tissue background detected by the environmental microplastic live four-dimensional Raman monitoring system are shown. After simultaneously acquiring the above three-dimensional spatial coordinate information and Raman spectral information, the two are fused to construct four-dimensional Raman spectral data. Step S3, Multi-channel Component Decoding: The four-dimensional Raman spectral data is input into a pre-trained microplastic multi-dimensional convolutional instance segmentation network. The network employs an encoder-decoder architecture with skip connections. After the four-dimensional Raman spectral data (x, y, z, λ) is input, it first passes through the first three-dimensional convolutional module for shallow feature extraction. Subsequently, the data stream splits into two paths: one path acts as a skip connection, directly transmitting the shallow spatial and spectral features without residual processing to the connection layer at the decoder end; the other path continues to pass through the residual module and the three-dimensional dilated convolutional module in sequence to extract deep, high-dimensional features with a larger receptive field. Next, the deep, high-dimensional features and the shallow features from the skip connection are fused at the connection layer to avoid the loss of spatial location information of microplastic particles in deep convolution. Finally, the fused features are then passed through the three-dimensional convolutional module and the upsampling layer at the decoder end, ultimately outputting voxel-level semantic segmentation and classification labels for microplastics. The network extracts high-dimensional spectral features and performs recognition of endogenous biological matrices and exogenous multi-mixed microplastics, extracting Raman spectral features of complex biomolecules such as lipids and proteins, and utilizing 500 to 3200 cm⁻¹ spectra. -1 Raman spectroscopy features are used to identify and classify multi-mixed microplastics. Based on the above identification and extraction, voxel-level semantic segmentation and classification results are output. Step S4, Qualitative material distribution visualization: Based on voxel-level semantic classification, different endogenous biological matrices and different exogenous multi-mixed microplastics are given pseudo-color to visualize the distribution of endogenous biological matrices and the in vivo spatial distribution of exogenous multi-mixed microplastics. Step S5, quantitative three-dimensional dynamic integration: extract the number of voxels of each type of microplastic in each level of the Z-axis; perform three-dimensional spatial integration on the full scan volume to calculate the absolute retention volume and abundance of each type of microplastic; and evaluate the metabolic kinetics of dynamic uptake, accumulation and excretion of mixed microplastics in vivo.

[0023] Example 3: A quantitative three-dimensional dynamic integral and metabolic kinetic assessment scheme for in vivo mixed microplastics Based on Example 2, this example further illustrates a scheme for quantitative three-dimensional dynamic integration and metabolic kinetics assessment of in vivo mixed microplastics. In step S5 of Example 2, the quantitative three-dimensional dynamic integration process includes a step-by-step calculation logic: first, extracting the number of voxels at each level along the Z-axis, and then performing three-dimensional spatial integration. Specifically, as follows: The physical volume of a single step during scanning is defined as the basic voxel unit v0 = Δ. x ×Δ y ×Δ z , where Δ x and Δ y For the scanning step resolution of the two-dimensional galvanometer 7, Δ z This is the axial step resolution of the displacement stage 11; Define a classification mask function C k ( x, y, z i ): Regarding the first of the multiple mixed microplastics k Based on the identification results output by the multidimensional convolutional instance segmentation network for the microplastics, the classification is determined for microplastics at the depth slice level. z i And the plane coordinates are ( x,y A specific three-dimensional voxel, when the voxel is recognized by the network as the first... k When considering a category of microplastics, the first... k Classification mask function for microplastics C k ( x, y, z i The value is 1 if the condition is not met, and 0 otherwise. The first step is to address the aforementioned first... k Microplastics are categorized into different types, based on their corresponding classification mask functions. C k ( x, y, z i), calculate each independent depth slice level on the Z-axis z i Inside, the first k The cumulative added value of two-dimensional feature surfaces occupied by different microplastic categories S k ( z i The calculation formula is: ; The second step is to process all independent depth slice levels. z i The extracted two-dimensional feature surface accumulation value S k ( z i Discrete integration is performed along the Z-axis depth direction to calculate the first... k Absolute cutoff volume of various microplastic categories within the full scan volume Volume k The calculation formula is: ; Where X, Y, and Z represent the total number of step points in the three-dimensional scanning space along the length, width, and depth dimensions, respectively; by traversing all category indices of multiple hybrid microplastics... k The absolute retention volume of each microplastic in vivo was obtained separately, enabling the separate quantification of multiple mixed microplastics. The evaluation of the dynamic uptake, accumulation, and excretion of multiple mixed microplastics in a living sample, as described in step S5 of Example 2, is achieved using a time-series scanning scheme, specifically including: The live samples were subjected to periodic three-dimensional qualitative and quantitative scans at multiple set time points during the exposure period containing mixed microplastics and at multiple set time points during the recovery period after being transferred to a clean culture medium. Extract each time node t The absolute retention volume of each type of microplastic calculated below Volume k (t) We plotted dynamic accumulation-clearance kinetics curves of microplastics of each component over time to quantitatively characterize the differences in bioaccumulation and metabolic clearance of different microplastics in living samples.

[0024] Example 4: An experimental evaluation of the dynamic accumulation of mixed microplastics in zebrafish: This invention utilizes the above system and method to dynamically monitor the enrichment of multiple mixed microplastics in zebrafish. Figure 5 The 3D visualization results show that after 3 hours of exposure Figure 5In (a), discrete signals from different multiple mixed microplastics were observed in the gastrointestinal tract of zebrafish, indicating that they were in the initial ingestion stage; 8 hours after exposure... Figure 5 In (b), a large-scale cluster of microplastics was observed in the gastrointestinal tract, indicating that due to continuous presence in the suspension, the intake rate exceeds the excretion rate, resulting in a large net accumulation of microplastics in vivo, which may even cause intestinal obstruction and slowed peristalsis. Furthermore, this invention achieves absolute quantitative statistics on different plastic particles through three-dimensional dynamic integration. For example... Figure 6 As shown, after 3 hours of exposure, the number of voxels at each depth (D1 to D7) revealed that polystyrene (PS) dominated in vivo, exhibiting a relatively rapid initial uptake or adhesion rate; however, as... Figure 7 As shown, after 8 hours of exposure, the number of voxels in polyethylene (PE) surged dramatically (reaching approximately 8,000) and became absolutely dominant. This quantitative data precisely reveals the differences in the accumulation kinetics of microplastics of different materials in vivo; Eight hours after zebrafish were transferred to a clean rearing medium free of microplastics, the Raman signal of microplastics almost completely disappeared, indicating that the ingested microplastics had been metabolized and excreted from the body. These experimental results demonstrate that this invention can accurately assess the dynamic uptake, specific accumulation, and excretion of mixed microplastics in vivo, providing a powerful, standardized, and quantitative method for the ecotoxicological assessment and health risk monitoring of environmental microplastic pollution.

[0025] Preferably, the live samples described in this invention are not limited to zebrafish larvae, but are also applicable to other model organisms with optically transparent or semi-transparent characteristics, such as water fleas and Caenorhabditis elegans, as well as the qualitative and quantitative analysis of three-dimensional cell spheres or environmental microbial membrane substrates cultured in vitro.

[0026] It should be understood that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications and substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for dynamic quantitative analysis of living environmental microplastics, employing a four-dimensional Raman spectroscopy system for monitoring living environmental microplastics, characterized in that, The system includes: a confocal spectral acquisition module, a controller (21), and a data processing module (22); The confocal spectroscopy acquisition module is configured to perform XYZ layer-by-layer spatial scanning on live samples exposed to microplastic environments to obtain three-dimensional spatial coordinates (x, y, z), and simultaneously acquire wavelengths covering 500 to 3200 cm⁻¹. -1 The range of Raman spectral signals; the microplastic environment is a multi-component mixed microplastic environment containing polyamide, polyethylene, polystyrene, polyethylene terephthalate and polymethyl methacrylate; The controller (21) is connected to the confocal spectral acquisition module and is configured to coordinate the synchronous acquisition of spatial scanning and Raman spectral signals; The data processing module (22) is connected to the communication controller (21) and is configured to fuse the acquired three-dimensional spatial coordinates (x, y, z) with the Raman spectral signal to construct four-dimensional Raman spectral data (x, y, z, λ). The four-dimensional Raman spectral data is input into a pre-trained microplastic multidimensional convolutional instance segmentation network to generate voxel-level semantic segmentation and classification results, so as to realize in-situ identification and label-free classification of endogenous biological matrix and exogenous multi-mixed microplastics in living organisms. Then, based on the voxel-level semantic classification results, the voxel quantity of various microplastics is extracted and three-dimensional spatial integration is performed to calculate the absolute cut-off volume and abundance of various microplastics in living organisms, so as to realize the quantitative analysis of exogenous multi-mixed microplastics. The method is used for assessing the environmental microplastic pollution effects for non-disease diagnostic purposes, and includes the following steps: Step S1, in situ preparation of live organisms: The live organisms exposed to the mixed microplastic environment are embedded and fixed, and placed on the displacement stage of the confocal spectroscopy acquisition module. Step S2, Four-dimensional Raman spectral data scanning and fusion: The controller coordinates and executes XYZ layer-by-layer spatial scanning, simultaneously acquiring three-dimensional spatial coordinates (x, y, z) and 500 to 3200 cm⁻¹. -1 Raman spectral signals within a certain range are fused together to construct four-dimensional Raman spectral data (x, y, z, λ). Step S3, Multi-channel component decoding: Input the four-dimensional Raman spectral data into the pre-trained microplastic multi-dimensional convolutional instance segmentation network, extract high-dimensional spectral features, perform endogenous biological matrix recognition and exogenous multi-mixed microplastic recognition, and output voxel-level semantic segmentation and recognition classification results; Step S4, Qualitative material distribution visualization: Based on voxel-level semantic classification, different endogenous biological matrices and different exogenous multi-mixed microplastics are given pseudo-color to visualize the distribution of endogenous biological matrices and the in vivo spatial distribution of exogenous multi-mixed microplastics. Step S5, quantitative three-dimensional dynamic integration: extract the number of voxels of each type of microplastic in each level of the Z-axis; perform three-dimensional spatial integration on the full scan volume to calculate the absolute retention volume and abundance of each type of microplastic; and evaluate the metabolic kinetics of dynamic uptake, accumulation and excretion of mixed microplastics in vivo. In step S3, the microplastic multidimensional convolutional instance segmentation network adopts an encoder-decoder architecture with skip connections. After the four-dimensional Raman spectral data (x, y, z, λ) is input, it first passes through the first three-dimensional convolutional module for shallow feature extraction. Then, the data stream is divided into two paths: one path acts as a skip connection, directly transmitting the shallow spatial and spectral features without residual processing to the connection layer at the decoder end; the other path continues to pass through the residual module and the three-dimensional dilated convolutional module in sequence to extract deep high-dimensional features with a larger receptive field. Next, the deep high-dimensional features and the shallow features from the skip connection are fused in the connection layer to avoid the loss of spatial location information of microplastic particles in deep convolution. Finally, the fused features pass through the three-dimensional convolutional module and the upsampling layer at the decoder end in sequence, and finally output the voxel-level semantic segmentation and classification label of microplastics. The network extracts data from 500 to 3200 cm⁻¹. -1 The Raman spectral information, including specific Raman peaks containing the low-frequency fingerprint region, is compared with the 3000 cm⁻¹. -1 The overlapping peaks of CH stretching vibrations of nearby lipids and proteins are used to verify the cross-spectral region characteristics, thereby decoupling the strong biological tissue autofluorescence background from the microplastic signal. At the same time, for multiple mixed microplastic clusters with irregular aggregation or entanglement, the three-dimensional hole convolution module expands the receptive field of three-dimensional features while maintaining sub-micron physical spatial resolution, thereby achieving high-precision segmentation of multiple mixed microplastics such as polyamide, polyethylene, polystyrene, polyethylene terephthalate, and polymethyl methacrylate. The decoder part maps the extracted high-dimensional features back to the original three-dimensional physical space through upsampling operations, and directly assigns classification labels corresponding to endogenous biological matrix and various types of exogenous microplastics.

2. The method for dynamic quantitative analysis of environmental microplastics in vivo according to claim 1, characterized in that: The confocal spectral acquisition module specifically includes an optical acquisition unit (19) and a spectral detection module (20) connected in sequence by optical paths. The optical acquisition unit (19) specifically includes a laser source (1), a beam expander (2), a first collimating lens (3), a first reflecting mirror (4), a second reflecting mirror (5), a dichroic mirror (6), a two-dimensional galvanometer (7), a scanning lens (8), a tube lens (9), a microscope objective (10), and a displacement stage (11) connected in sequence via optical paths. The two-dimensional galvanometer (7), combined with the displacement stage (11) loaded with a live sample, is configured to perform XYZ layer-by-layer optical slicing scans to obtain high-resolution three-dimensional spatial coordinates (x, y, ...). z); The Raman scattering signal generated by the live sample returns to the dichroic mirror (6) via the original path, passes through the notch filter (12) and the second collimating lens (13) in sequence, and is focused on the confocal pinhole (14); the confocal pinhole (14) is used to perform physical spatial filtering to remove biological background stray light from the defocus surface; the signal passing through the confocal pinhole (14) enters the spectral detection module (20); the spectral detection module (20) includes a first concave mirror (15), a grating (16), a second concave mirror (17) and a camera (18) arranged in sequence, and is used to simultaneously acquire spectral ranges from 500 to 3200 cm⁻¹. -1 The Raman spectral signal λ.

3. The method for dynamic quantitative analysis of environmental microplastics according to claim 2, characterized in that, In step S5, the process of calculating the absolute retention volume and abundance of various microplastics is as follows: The physical volume of a single step during scanning in the system is defined as a basic voxel unit v0, where v0 = Δ x ×Δ y ×Δ z , where Δ x and Δ y Δ is the scanning step resolution of the two-dimensional galvanometer (7). z The axial step resolution of the displacement stage (11); Define a classification mask function C k ( x, y, z i ): Regarding the first of the multiple mixed microplastics k Based on the identification results output by the multidimensional convolutional instance segmentation network for the microplastics, the classification is determined for microplastics at the depth slice level. z i And the plane coordinates are ( x,y A specific three-dimensional voxel, when the voxel is recognized by the network as the first... k When considering a category of microplastics, the first... k Classification mask function for microplastics C k ( x, y, z i The value is 1 if the condition is not met, and 0 otherwise. The first step is to address the aforementioned first... k Microplastics are categorized into different types, based on their corresponding classification mask functions. C k ( x, y, z i ), calculate each independent depth slice level on the Z-axis z i Inside, the first k The cumulative added value of two-dimensional feature surfaces occupied by different microplastic categories S k ( z i The calculation formula is: ; The second step is to process all independent depth slice levels. z i The extracted two-dimensional feature surface accumulation value S k ( z i Discrete integration is performed along the Z-axis depth direction to calculate the first... k Absolute cutoff volume of various microplastic categories within the full scan volume Volume k The calculation formula is: ; Where X, Y, and Z represent the total number of step points in the three-dimensional scanning space along the length, width, and depth dimensions, respectively; by traversing all category indices of multiple hybrid microplastics... k The absolute retention volume of each microplastic in vivo was obtained, enabling separate quantification of multiple mixed microplastics.

4. The method for dynamic quantitative analysis of environmental microplastics in vivo according to claim 1, characterized in that, Step S5, which assesses the metabolic kinetics of the mixed microplastics in vivo, specifically includes: The live samples were subjected to periodic three-dimensional qualitative and quantitative scans at multiple set time points during the exposure period containing mixed microplastics and at multiple set time points during the recovery period after being transferred to a clean culture medium. Extract each time node t The absolute retention volume of each type of microplastic was calculated, and the dynamic accumulation-clearance kinetics curves of each type of microplastic over time were plotted to quantitatively characterize the differences in bioaccumulation and metabolic clearance of different microplastics in living samples.

5. The method for dynamic quantitative analysis of environmental microplastics according to claim 1, characterized in that: The live sample mentioned in step S1 is a model organism with optical transparency or translucency, and the model organism is selected from zebrafish larvae, water fleas and Caenorhabditis elegans; or the live sample is a three-dimensional cell sphere or environmental microbial membrane matrix cultured in vitro.

Images