Abstract
Microplastic (MP) pollution has emerged as a global environmental concern due to its ubiquity and harmful impacts on ecosystems and human health. MP assessment has therefore become increasingly necessary and common in environmental and experimental samples. Microscopy and spectroscopy are widely employed for the physical and chemical characterization of MPs. However, these analytical methods often require time-consuming pretreatments of samples or expensive instrumentation. In this work, we develop a portable and cost-effective polarization holographic imaging system that prominently incorporates deep learning techniques, enabling efficient, high-throughput detection and dynamic analysis of MPs in aqueous environments. The integration enhances the identification and classification of MPs, eliminating the need for extensive sample preparation. The system simultaneously captures holographic interference patterns and polarization states, allowing for multimodal information acquisition to facilitate rapid MP detection. The characteristics of light waves are registered, and birefringence features are leveraged to classify the material composition and structures of MPs. Furthermore, the system automates real-time counting and morphological measurements of various materials, including MP sheets and additional natural substances. This innovative approach significantly improves the dynamic monitoring of MPs and provides valuable information for their effective filtration and management.
Similar content being viewed by others
Introduction
Microplastics (MPs) are small plastic particles with a size typically less than 5mm1,2. They originate from various sources such as fragmentation of larger plastic debris, shedding of microfibers from synthetic textiles during washing, and utilization of microbeads in personal care products3. These minute particles are ubiquitous in the environment and have been detected in diverse ecosystems, including oceans, rivers, soil, and even air4,5. Due to their small size and persistence, MPs pose a significant threat to the environment, wildlife, and potentially human health, as they can transfer through the food chain and serve as carriers for harmful pollutants, such as heavy metals and persistent organic pollutants6,7. Furthermore, the presence of MPs in aquatic systems can negatively affect the growth, reproduction, and survival of aquatic organisms, causing disruptions in ecosystems and potentially leading to a decline in biodiversity8. The increasing prevalence of MPs has raised global concerns, prompting calls for more sustainable practices in plastic production, consumption, and waste management9,10.
Identifying MPs and assessing water quality have gained increasing public attention11. Chemical and thermal analysis of MPs are the most commonly adopted methods in the laboratory. For instance, Raman spectroscopy can distinguish different types of MPs from the Raman spectra of the samples based on the interaction of light with the chemical bonds of a substance12. Microscopic techniques are also widely utilized for the physical characterization of MPs, providing detailed structural information for identification purposes13. These techniques include stereo, fluorescence, scanning electron, transmission electron, and atomic force microscopy, each with varied imaging capabilities suited to particular applications14. However, while most of the existing analytical methods are advantageous in characterizing MPs in the laboratory, they are generally unsuitable for in in situ detection and high-throughput monitoring of MPs suspended in natural water, as they cannot readily track changes in these particles15. Furthermore, these typical methods often require either complex and expensive facilities, or labor-intensive and time-consuming pretreatments of samples, such as Nile red staining for fluorescent imaging16. To avoid extensive sample preparation in the laboratory, portable devices or technologies need to be developed for fieldwork, and the imaging system should be capable of distinguishing MPs in the aqueous phase, where most of them are found.
The increasing concern for MP pollution has prompted the development of advanced techniques to effectively detect and quantify these harmful particles in various environments17,18. Digital holographic imaging is a versatile method that encodes the wavefront information of objects into interference patterns, facilitating rapid and precise identification of MP particles19,20,21. The interference pattern, also known as a digital hologram, is generated through a process that involves the interaction of two coherent light waves: the reference wave and the object wave22,23. Upon recombination, the light waves create a digital hologram that encodes both the amplitude and phase information of the scattered light at any adjustable depth, providing a comprehensive characterization of the MP samples. Subsequently, the recorded digital hologram is processed using numerical algorithms to reconstruct the complex optical field of the scattered light. Furthermore, digital holographic imaging offers a label-free, non-destructive, non-invasive approach that eliminates the need for extensive laboratory sample preprocessing24. This technique also offers a compact and cost-effective solution by utilizing fewer optical components, rendering it an ideal choice for field-portable applications, especially for in situ environmental monitoring25,26.
Polarization imaging capitalizes on the unique optical properties of MPs, which interact differently with polarized light compared to other natural materials27,28. By analyzing the changes in the polarization state of light upon interaction with MP particles, this technique enables the selective identification and differentiation of MPs from other particulate matter based on their distinguishable optical anisotropy of MPs29,30. Therefore, combining polarization and holographic imaging results in the formation of a polarization-sensitive digital hologram, which holds promise for achieving multimodal information acquisition from MPs, including morphological, optical, and chemical features. Moreover, integrated polarization holographic imaging has a relatively simple optical configuration and requires fewer sample preparation procedures, making it ideally suitable as a field-portable system that can realize in situ environmental monitoring. With the addition of a fluidic channel design, it is possible to perform high-throughput analysis of MPs in various scenarios, including aqueous environments. Concurrently, computational methods are gradually being incorporated into MP assessment, promoting rapid and accurate approaches. Zhu et al. and Huang et al. are also integrating polarization-related computational imaging techniques with more compact configurations31,32,33,34. In their works, they consider several complex natural environments, such as turbid water. However, the samples are still undergoing assessment, and high-throughput, efficient detection has not yet been achieved.
In addition to these advancements, deep learning (DL)-based models, such as Mask R-CNN and U-Net, have been used in the analysis of MPs35,36. Compared with traditional visual inspection and counting, DL-based models facilitate feature extraction to achieve automated and intelligent MP identification and quantification. Researchers have applied sophisticated DL models capable of analyzing large volumes of data, such as images and spectroscopic information, to accurately identify and quantify MP particles in various environmental samples37,38. DL models have shown remarkable success in recognizing complex patterns and features within holographic interference fringes, enabling rapid and reliable detection of MPs28,39,40. The application of artificial intelligence for multidimensional MP assessment is an emerging field.
In this work, we present a novel approach that emphasizes high-throughput detection and DL models for the dynamic monitoring of MPs. Our method employs a fluidic channel to enable in situ detection of flowing MPs and utilizes polarized digital holography to record and extract multimodal information, such as light intensity, complex holographic patterns, and polarization state, capturing the characteristics of the light wave. By leveraging birefringence features, we can classify material composition and structures. The integration of DL models automates real-time counting, morphological quantification, and classification of different plastic sheets based on large quantities of data. This comprehensive system significantly improves the dynamic monitoring of MPs, providing valuable information for their efficient filtration and management.
Results
Underwater MP imaging and scattering removal
The investigation of MPs typically involves a multi-stage process in which the particles are gathered using sieves, processed, and ultimately positioned in Petri dishes41,42. Furthermore, the examination and analysis of static MP samples is a labor-intensive process, frequently involving a significant degree of reliance on the operator, as it heavily depends on visual inspection. In contrast, imaging dynamic MPs offers a more efficient method for assessment compared to the time-consuming static MP analysis41. However, the light scattering and absorption of suspending particles often result in a veiling effect for underwater imaging, reducing the contrast on holographic fringes43. To address this issue, we demonstrate a compact polarization holographic device that can analyze MP samples dispersed in water and flowing through a customized glass fluidic channel, as shown in Fig. 1.
The fluidic channel is made of clear quartz glass with 95% light transmission and has dimensions of 5mm (inner thickness), 10 mm (width), and 50mm (length). This design prevents any potential disruption to the sample’s holographic fringes and is suitable for samples of various sizes. The detailed configuration of the device is illustrated in Fig. 1a. A laser source with a continuous wave and 632.8nm wavelength is used, passing through a linear polarizer (LP) and a quarter-wave plate (QWP) to generate circularly polarized light. A neutral density filter (NDF) is placed to ensure a suitable light intensity.
Polarization holographic imaging is a non-invasive technique that requires no sample preparation and offers single-shot evaluations. This characteristic enables the dynamic collection of data from flowing samples. A polarization camera is employed to simultaneously record holograms with four different polarization states at \({0}^{\circ }\), \({45}^{\circ }\), \({90}^{\circ }\), and \({135}^{\circ }\), respectively. By acquiring images at four different angles, we can effectively measure all four Stokes parameters, enabling a comprehensive characterization of the polarization state of the scattered light from the MP samples44. These polarization states are used to determine the sample optical anisotropy and calculate birefringence-related features (i.e. phase retardance, angle of polarization). Additionally, this system registers the holographic interference patterns between the light passing through the object and a plane reference wave. As illustrated in Fig. 1a, the laser beam is collimated by a convex lens and circularly polarized by a linear polarizer and a quarter wave plate. The fast axis of the quarter wave plate is oriented at \(45^{\circ }\) with respect to that of the polarizer.
The device is designed to submerge the sample in water, which eliminates significant light reflection and scattering issues and aids in real-time classification. The intensity holograms of MP sheets in the air or aquatic environment are shown in Fig. 1b and c. Both samples perform distinguishable holographic patterns, and they can be clearly photographed with the fluidic channel. Although water may result in a disturbance compared with the air environment, it does not significantly drop the recognition performance of the patterns. Fully submerging the sample in water could reduce the interference of surface reflection.
Real-time counting and morphological analysis
In the experimental setup, the flow rate of water is crucial for accurate fringe recording and sample movement. A water pump is applied to precisely control the speed of water flow. If the system has a high flow rate, it becomes challenging to record fringes precisely. On the other hand, if the system has a low flow rate, the samples struggle to move forward. The volume flow rate of water is optimized at 8 ml/min, allowing for stable and high-throughput detection. The device features a microfluidic channel that is capable of reaching a high throughput of 480 mL/h with accurate images. This design enables efficient MP detection in the aqueous environment.
The optimization ensures that samples of various shapes and materials can be effectively analyzed as they flow through the experimental device. For real-time tracking, the YOLO v5 model, a lightweight convolutional neural network (CNN), is combined with a Strong-SORT network to segment, measure, and count the MP samples45,46,47. A total of 1000 images are captured using the polarization holographic setup, as well as videos with continuous image recording. Each frame of the video is a single hologram, representing the interference pattern resulting from the superposition of object and reference waves. To further support the analysis process, the captured 1000-image dataset is divided into training-validation-test segments with a ratio of 8:1:1. This division of data facilitates the effective training of the DL model and enables reliable evaluation of its performance. As a result, the samples are counted and classified with an accuracy of up to 96\(\%\). As illustrated in Fig. 2a, the number of MP samples during a certain time period can be real-time counted.
(a) Real-time MP counting with YOLO v5 and Strong-SORT. (b) Generated masks for two kinds of MPs, PC and PET. (c) Morphological measurement of Max Feret diameter (MFD) and area for PC and PET. (d) The Bland-Altman analysis comparing the MFD measurements obtained by the estimation and reference values (n = 39); the dashed lines represent the 95% limits of agreement (LOAs). (e) Mixed MP samples with random shapes and varied degrees of transparency. (f) Classification results of varied shapes shown in (e). Scale bar: 1mm.
MP morphological analysis is essential for understanding the distribution, characteristics, and potential impacts of these tiny plastic particles on the environment and human health48. Individual masks are generated for the identified MP sheets, as shown in Fig. 2b. The Maximum Feret diameter (MFD, length) and area are employed to measure the samples’ morphology and provide valuable information for MP management (Fig. 2c). The Feret diameter is a measure of an object’s size along a specified direction and can be defined as the distance between the two parallel planes restricting the object perpendicular to that direction In this study, we successfully demonstrate a novel approach to assess MP morphological parameters and material types based on polarization holographic imaging features using DL models. The results obtained from our experiments and analyses show the effectiveness and potential of this method in identifying and categorizing different types of MP materials. Our research introduces a real-time, high-throughput MP analyzer that leverages multidimensional information to enhance the quantification of MPs and provide a reference for treatment strategies. Moreover, the high classification accuracy, robustness, and generalizability of this approach make it a promising tool for future research and practical applications in assessing and managing MP pollution. MP pollutants have become a global environmental concern, leading to the development and application of various detection methods, including imaging-based, thermal, and chemical analysis techniques58. Practical MP detection often requires the use of multiple methods to achieve reliable results. Combining thermal and chemical techniques with imaging-based methods can provide a more comprehensive approach to MP detection. It is highly desirable to identify and quantify MP particles together with their chemical compositions and morphological distributions. In recent years, several emerging imaging techniques have been applied to the field of MP detection. These imaging techniques can provide valuable information for MP identification and analysis and potentially lead to new insights into the distribution and impact of MPs in the environment. For example, a combination of neutron and X-ray tomography has been tested to detect MP particles in sandy sediments60. Moreover, flow cytometry (FlowCam) can be used to detect and analyze MPs in water samples61,62, combining high-resolution imaging with flow cytometry to capture images of particles. As more technologies are developed, additional MP features can be acquired. We can further explore the possibility of enhancing the integration of new advanced technologies and the current detection methods for MPs. However, there are still complex situations for MP analysis. For instance, multiple scattering and absorption exist in real seawater samples that are usually mixed with a lot of impurities, affecting the imaging effect63. Tracking and detecting MPs in cases of complicated turbid water scattering is our next focus of work64,65. It has been primarily demonstrated that our imaging system is capable of seeing through scattering media and obtaining multimodal information about the object. The results suggest that polarization features can substantially improve image contrast even in highly turbid water31. In addition, an objective lens will be introduced into the system to increase the detection of full-scale samples, such as those on the micrometer scale. Simultaneously, more natural samples will be incorporated to make the results closer to the seawater situation, making the detection results of the system more practical. In this study, polarization holographic imaging technology is applied for feature capture and fieldwork using a compactly designed system. A monochromatic laser light source with a wavelength of 632.8 nm is employed. An LP and a QWP are combined to generate a circularly polarized laser beam. An NDF is placed to ensure the light intensity is suitable. The uniformly distributed circularly polarized light then passed through the sample plate and interacted with the samples, encoding the sample feature information with wavefront deformation and polarization state modification. A polarization CMOS camera (Crevis Tech, MC-A500P-163) is used to record polarization-sensitive digital holograms with four different polarization states. The captured image has a resolution of 2464 pixels \(\times\)2056 pixels. The pixels with size 3.45\(\upmu \!{\textrm{m}}\) are arranged in a group of \(2\times 2\) grid arrays, which simultaneously capture the polarization states at \({0}^{\circ }\), \({45}^{\circ }\), \({90}^{\circ }\), and \({135}^{\circ }\), respectively. These polarization states are used to calculate birefringence-related features. To ensure the accuracy and reliability of the system, a calibration process has been implemented. This system calibration encompasses circularly polarized light calibration and lateral resolution calibration, accounting for the impacts of the medium, the sample, and dispersion. A reference sample, such as glass without birefringence, is utilized to calibrate the polarization holographic imaging system. The light intensity distribution at four distinct angles is adjusted to achieve a uniformly circularly polarized light beam, and a non-birefringent glass sample is employed to validate the DoLP parameter. Reference samples are prepared with known physical parameters, including size and shape, under consistent medium and dispersion conditions. Analyzing the reference sample enables the determination of the relationship between the measured optical phase and the established physical parameters. In this system, the calibrated spatial resolution is 20μm. A USAF-1951 target (Edmund Optics, USA) is used for resolution characterization. The examination of the reference sample allows for the estimation of potential errors and uncertainties in the measurements, thereby ensuring the reliability of the obtained results. Sample particle detection in flow is performed using a customized fluidic channel made of 95% light transmission quartz glass. The dimensions of the channel are designed according to the MP size range: 5 mm (inner thickness), 10 mm (width), and 50 mm (length). The flow speed is precisely controlled by a water pump (Kamoer Tech, NKCP-C-SO4B), with a speed control range starting from 2 ml/min and capable of reaching a maximum volume flow rate of 15 ml/min. The optimal speed should be determined according to the sample situation. In this study, the optimized speed is set at 8 ml/min for stable MP movement. In a typical digital holography system, the formation of the holographic fringe pattern is achieved by the spatial superposition of the complex reference wave (\(U_R\)) and object wave (\(U_O\)). \(U_R\) and \(U_O\) are initially plane waves. A wavefront deformation is introduced in \(U_O\) with the reflection, transmission, and scattering from the object. A photodetector such as a charge-coupled device (CCD) or complementary metal-oxide-semiconductor (CMOS) camera is used to record the intensity distribution (I(x, y)) of the resulting holographic patterns. where \(^*\) denotes the complex conjugate of the referring part. The spatial coordinates (x, y) are omitted in the equation for clarity. The first two parts of the result are the DC-terms and acting as noise. DC-terms can be limited by applying a numerical filter in the frequency domain or subtracting the mean intensity values. The third and fourth terms are the generated interference patterns. The last term will be considered for further modulation and reconstruction. Numerical reconstruction of holograms is employed to decode the complex wavefront encoded in the captured holograms, revealing the distributions of amplitude and phase at any desired depth. This process facilitates the extraction of valuable information about the sample, including its morphological characteristics. The angular spectrum function, a widely used method, is applied to calculate the forward or backward propagation of light, allowing for the reconstruction of images across various focal planes. This function is mathematically expressed as where \(\Gamma (x, y, z)\) represents the reconstructed complex wavefront with coordinates (x, y, z), \(\mathscr {F}\) denotes the Fourier transform, \(E_{R}^{*}\) is the complex conjugate of the reference wave, h is the captured hologram, \(\lambda\) is the wavelength of illumination light, and \(f_{x}\) and \(f_{y}\) are the transverse spatial frequencies. Holograms at four different angles are simultaneously recorded to obtain a comprehensive characterization of the polarization states of the MP samples. The intensity differences in the four-angle images indicate their optical anisotropy, which is beneficial for material classification. To facilitate material classification, the SSD of four polarized angle images is calculated using Matlab software (Mathworks, US) to facilitate material classification. By acquiring images at four distinct angles, we can effectively measure Stokes parameters, enabling a complete analysis of the polarization state. Moreover, the polarization states of light are characterized using Linear Stokes parameters where I represents the grayscale intensity of a hologram at a certain polarization angle. Moreover, specific polarization parameters are derived from the captured holographic images based on Stokes parameters to characterize the polarization response of MP sheet samples, such as the degree of linear polarization (DoLP \(\in [0,1]\)) and the angle of polarization (AoP \(\in [-\frac{\pi }{4},+\frac{\pi }{4}]\)) The DoLP represents the magnitude of the linear polarization state, while the AoP indicates the orientation of the polarization state. Furthermore, both DoLP and AoP are independent of particle geometrical parameters, such as shape, area, size, and perimeter, and remain consistent across a wide range of reconstruction distances. This information is crucial for the reliable identification and quantification of MPs. In this study, six different types of MP sheets are applied: PC, PET, PVC, PP, PS, and PMMA. These samples with varying birefringence values are purchased from **nsheng Plastic Material Company, China. To ensure sample diversity, additional materials such as glass, lens paper, and algae are included in the control group. The samples are imaged using a stereo microscope (Nikon Model Eclipse Ni-U, Japan) with a digital camera. The captured MP images by the polarization CMOS camera are accurately labeled using ImageJ software (National Institutes of Health, US), and the dataset is divided into training, validation, and test sets at a ratio of 8:1:1. The YOLO v5 model, a lightweight CNN, is combined with a Strong-SORT network to segment, measure, and count the MP samples. The DL model is further fine-tuned. To fine-tune the hyperparameters, specifically the learning rate, batch size, and whether augmentations should be used, a manual grid search is performed. The area and MFD values are calculated by a self-developed Python script. The scatter diagram and line chart are visualized by GraphPad Prism software (GraphPad Inc, USA). MP particles with an approximate size of 100μm in size, are further processed using a Retsch CryoMill (Haan Tech, Germany). The ultra-low operating temperature is maintained by circulating liquid nitrogen around the grinding chamber, which is made of zirconium oxide. Subsequently, the ground MP particles are analyzed for various morphological parameters using a particle analyzer Morphologi 4 (Malvern Panalytical, UK), including circularity, volume, length, width, and diameter. Raman spectra serve as a reference for MP identification. MP sheets are dried at 40 °C and the Raman spectra are assessed using a Optical-Photothermal InfraRed mIRage (California, USA), by illuminating the sample with a mid-IR pulsed tunable laser of 532nm detection laser for 10 s exposure time. Raman spectra of the MP samples are acquired in the wavenumber range of 200–4170 cm\(^{-1}\). Baseline correction and smoothing of the acquired spectra are performed with the PTIR Studio 4.4 software.Discussion
Methods
Experimental setup, calibration and data collection
Generation of digital holograms and reconstrunction
Polarization holographic feature analysis
Sample preparation and image processing
Reference data acquisition
Data availability
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.
References
Akdogan, Z. & Guven, B. Microplastics in the environment: A critical review of current understanding and identification of future research needs. Environ. Pollut. 254, 113011 (2019).
Thompson, R. C. et al. Lost at sea: Where is all the plastic?. Science 304, 838–838 (2004).
Koelmans, A. A. et al. Risk assessment of microplastic particles. Nat. Rev. Mater. 7, 138–152 (2022).
Rillig, M. C. & Lehmann, A. Microplastic in terrestrial ecosystems. Science 368, 1430–1431 (2020).
Sharma, S. & Chatterjee, S. Microplastic pollution, a threat to marine ecosystem and human health: A short review. Environ. Sci. Pollut. Res. 24, 21530–21547 (2017).
Vethaak, A. D. & Legler, J. Microplastics and human health. Science 371, 672–674 (2021).
Kinigopoulou, V., Pashalidis, I., Kalderis, D. & Anastopoulos, I. Microplastics as carriers of inorganic and organic contaminants in the environment: A review of recent progress. J. Mol. Liq. 350, 118580 (2022).
Luo, H. et al. Environmental behaviors of microplastics in aquatic systems: A systematic review on degradation, adsorption, toxicity and biofilm under aging conditions. J. Hazard. Mater. 423, 126915 (2022).
Chow, C.-F., So, W.-M. W., Cheung, T.-Y. & Yeung, S.-K. D. Plastic waste problem and education for plastic waste management. Emerging Practices in Scholarship of Learning and Teaching in a Digital Era 125–140 (2017).
Shin, S.-K., Um, N., Kim, Y.-J., Cho, N.-H. & Jeon, T.-W. New policy framework with plastic waste control plan for effective plastic waste management. Sustainability 12, 6049 (2020).
Cowger, W. et al. Critical review of processing and classification techniques for images and spectra in microplastic research. Appl. Spectrosc. 74, 989–1010 (2020).
Schymanski, D., Goldbeck, C., Humpf, H. U. & Furst, P. Analysis of microplastics in water by micro-Raman spectroscopy: Release of plastic particles from different packaging into mineral water. Water Res. 129, 154–162 (2018).
Baruah, A., Sharma, A., Sharma, S. & Nagraik, R. An insight into different microplastic detection methods. Int. J. Environ. Sci. Technol. 19, 5721–5730 (2022).
Mariano, S., Tacconi, S., Fidaleo, M., Rossi, M. & Dini, L. Micro and nanoplastics identification: Classic methods and innovative detection techniques. Front. Toxicol. 3, 636640 (2021).
Li, J. et al. Recognition of microplastics suspended in seawater via refractive index by mueller matrix polarimetry. Mar. Pollut. Bull. 188, 114706 (2023).
Shim, W. J., Song, Y. K., Hong, S. H. & Jang, M. Identification and quantification of microplastics using nile red staining. Mar. Pollut. Bull. 113, 469–476 (2016).
Priya, A. et al. Removing microplastics from wastewater using leading-edge treatment technologies: A solution to microplastic pollution-a review. Bioprocess Biosyst. Eng. 46, 309–321 (2023).
Fu, W., Min, J., Jiang, W., Li, Y. & Zhang, W. Separation, characterization and identification of microplastics and nanoplastics in the environment. Sci. Total Environ. 721, 137561 (2020).
Ren, Z., Xu, Z. & Lam, E. Y. End-to-end deep learning framework for digital holographic reconstruction. Adv. Photonics 1, 016004–016004 (2019).
Zhu, Y., Yeung, C. H. & Lam, E. Y. Digital holographic imaging and classification of microplastics using deep transfer learning. Appl. Opt. 60, A38–A47 (2021).
Zhu, Y., Yeung, C. H. & Lam, E. Y. Holographic classifier: Deep learning in digital holography for automatic micro-objects classification. in booktitle2020 IEEE 18th International Conference on Industrial Informatics (INDIN), vol. 1, 515–520 (IEEE, 2020).
Goodman, J. W. Introduction to Fourier Optics (W. H. Freeman, 2017).
Wang, Z. et al. Digital holography as metrology tool at micro-nanoscale for soft matter. Light Adv. Manuf. 3, 151–176 (2022).
Merola, F. et al. Searching and identifying microplastics in marine environment by digital holography. Eur. Phys. J. Plus 133, 1–6 (2018).
Asamoah, B. O. et al. Towards the development of portable and in situ optical devices for detection of micro-and nanoplastics in water: A review on the current status. Polymers 13, 730 (2021).
Schnitzler, L. et al. Lensless digital holographic microscopy as an efficient method to monitor enzymatic plastic degradation. Mar. Pollut. Bull. 163, 111950 (2021).
Valentino, M. et al. Intelligent polarization-sensitive holographic flow-cytometer: Towards specificity in classifying natural and microplastic fibers. Sci. Total Environ. 815, 152708 (2022).
Zhu, Y., Yeung, C. H. & Lam, E. Y. Microplastic pollution monitoring with holographic classification and deep learning. J. Phys. Photonics 3, 024013 (2021).
Yang, Y., Huang, H. & Guo, C. Polarization holographic microscope slide for birefringence imaging of anisotropic samples in microfluidics. Opt. Express 28, 14762–14773 (2020).
Wang, J. et al. Linear polarization holography. Opto-Electron. Sci. 1, 210009 (2022).
Huang, J., Zhu, Y., Li, Y. & Lam, E. Y. Snapshot polarization-sensitive holography for detecting microplastics in turbid water. ACS Photonics 10, 4483–4493 (2023).
Zhu, Y., Li, Y., Huang, J. & Lam, E. Y. Smart polarization and spectroscopic holography for real-time microplastics identification. Commun. Eng. (2024).
Zhu, Y., Li, Y., Huang, J., Zhang, Y. & Lam, E. Y. Holographic and polarization features analysis for microplastics characterization and water monitoring. In booktitleMultimodal Sensing and Artificial Intelligence: Technologies and Applications III in SPIE Optical Metrology, 12621–33 (SPIE, 2023).
Li, Y., Zhu, Y., Huang, J., Zhang, Y. & Lam, E. Y. Polarization holographic imaging for high-throughput microplastic analysis. In booktitleDigital Holography and Three-Dimensional Imaging, HM1D–6 (Optica Publishing Group, 2023).
Lee, K. S. et al. U-net skip-connection architectures for the automated counting of microplastics. Neural Comput. Appl. 34, 7283–7297 (2022).
Lorenzo-Navarro, J. et al. Deep learning approach for automatic microplastics counting and classification. Sci. Total Environ. 765, 142728 (2021).
Brandt, J., Mattsson, K. & Hassellov, M. Deep learning for reconstructing low quality FTIR and Raman spectra a case study in microplastic analyses. Anal. Chem. 93, 16360–16368 (2021).
Zhu, Y., Yeung, C. H. & Lam, E. Y. Digital holography with deep learning and generative adversarial networks for automatic microplastics classification. In booktitleHolography, Diffractive Optics, and Applications X, vol. 11551, 22–27 (SPIE, 2020).
Zeng, T., Zhu, Y. & Lam, E. Y. Deep learning for digital holography: A review. Opt. Express 29, 40572–40593 (2021).
Zhu, Y., Lo, H. K. A., Yeung, C. H. & Lam, E. Y. Microplastic pollution assessment with digital holography and zero-shot learning. APL Photonics 7, 076102 (2022).
Cacace, T. et al. Compact holographic microscope for imaging flowing microplastics. In booktitle2021 International Workshop on Metrology for the Sea; Learning to Measure Sea Health Parameters (MetroSea), 229–233 (IEEE, 2021).
Prume, J. A., Gorka, F. & Löder, M. G. From sieve to microscope: An efficient technique for sample transfer in the process of microplastics’ quantification. MethodsX 8, 101341 (2021).
Zhu, Y., Zeng, T., Liu, K., Ren, Z. & Lam, E. Y. Full scene underwater imaging with polarization and an untrained network. Opt. Express 29, 41865–41881 (2021).
Otani, Y., Endo, N., Hagen, N. & Shibata, S. Imaging microplastics consumed by water organisms using a full-stokes polarization camera. In booktitleBiomedical Imaging and Sensing Conference 2021, vol. 11925, 88–89 (SPIE, 2021).
Shishkin, I. E. & Grekov, A. N. Implementation of yolov5 for detection and classification of microplastics and microorganisms in marine environment. In booktitle2023 International Russian Smart Industry Conference (SmartIndustryCon), 230–235 (IEEE, 2023).
Jocher, G. et al. ultralytics/yolov5: v7. 0-yolov5 sota realtime instance segmentation. Zenodo (2022).
Du, Y. et al. Strongsort: Make deepsort great again. IEEE Transactions on Multimedia (2023).
Helm, P. A. Improving microplastics source apportionment: A role for microplastic morphology and taxonomy?. Anal. Methods 9, 1328–1331 (2017).
Li, Y., Parkinson, D. Y., Feng, J., **a, C.-H. & Gong, X. Quantitative X-ray tomographic analysis reveals calcium precipitation in cataractogenesis. Sci. Rep. 11, 17401 (2021).
Bland, J. M. & Altman, D. G. Comparing methods of measurement: Why plotting difference against standard method is misleading. Lancet 346, 1085–1087 (1995).
Pizzurro, F., Recchi, S., Nerone, E., Salini, R. & Barile, N. B. Accumulation evaluation of potential microplastic particles in Mytilus galloprovincialis from the Goro Sacca (Adriatic sea, Italy). Microplastics 1, 303–318 (2022).
Fang, J. K. H. et al. Adverse impacts of high-density microplastics on juvenile growth and behaviour of the endangered tri-spine horseshoe crab Tachypleus tridentatus. Mar. Pollut. Bull. 187, 114535 (2023).
Severini, E. et al. River-groundwater interaction and recharge effects on microplastics contamination of groundwater in confined alluvial aquifers. Water 14, 1913 (2022).
Mueller, H. Theory of photoelasticity in amorphous solids. Physics 6, 179–184 (1935).
Hong, N., Synowicki, R. A. & Hilfiker, J. N. Mueller matrix characterization of flexible plastic substrates. Appl. Surf. Sci. 421, 518–528 (2017).
Siebourg, W. et al. Birefringence-an important property of plastic substrates for magneto-optical storage disks. Polym. Eng. Sci. 30, 1133–1139 (1990).
Song, Y. K. et al. A comparison of microscopic and spectroscopic identification methods for analysis of microplastics in environmental samples. Mar. Pollut. Bull. 93, 202–209 (2015).
Wang, Z.-M., Wagner, J., Ghosal, S., Bedi, G. & Wall, S. SEM/EDS and optical microscopy analyses of microplastics in ocean trawl and fish guts. Sci. Total Environ. 603, 616–626 (2017).
Leung, M.M.-L. et al. Improved Raman spectroscopy-based approach to assess microplastics in seafood. Environ. Pollut. 289, 117648 (2021).
Tötzke, C., Oswald, S. E., Hilger, A. & Kardjilov, N. Non-invasive detection and localization of microplastic particles in a sandy sediment by complementary neutron and X-ray tomography. J. Soils Sedim. 21, 1476–1487 (2021).
Nelson, H. et al. Use of imaging flow cytometry (FlowCam) in the study of microplastics. In booktitleOcean Sciences Meeting (Oregon Convention Center Portland, 2018).
Woods, M. N., Stack, M. E., Fields, D. M., Shaw, S. D. & Matrai, P. A. Microplastic fiber uptake, ingestion, and egestion rates in the blue mussel (Mytilus edulis). Mar. Pollut. Bull. 137, 638–645 (2018).
Li, H. et al. Underwater active polarization descattering based on a single polarized image. Opt. Express 31, 21988–22000 (2023).
Wei, Y., Han, P., Liu, F. & Shao, X. Estimation and removal of backscattered light with nonuniform polarization information in underwater environments. Opt. Express 30, 40208–40220 (2022).
Zhang, Y., Chan, S. H. & Lam, E. Y. Photon-starved snapshot holography. APL Photonics 8, 056106 (2023).
Acknowledgements
The work is supported in part by the Research Grants Council of Hong Kong (GRF 17201620, RIF R7003-21). This work is partly funded by the Hong Kong Scholars Program (Grant No. XJ2022032). Jianqing Huang appreciates the partial financial support from Shanghai Jiao Tong University.
Author information
Authors and Affiliations
Contributions
Y.L.: writing (original draft), writing (review and editing), sample preparation, experiments performing, formal analysis; Y.Z.: writing (review and editing), sample preparation, experiments performing; J.H.: writing (review and editing), sample preparation, experiments performing; Y.H.: writing (review and editing), sample preparation, experiments performing; J.F.: writing (review and editing), resources; E.Y.L.: writing (review and editing), resources, supervision, funding acquisition.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Li, Y., Zhu, Y., Huang, J. et al. High-throughput microplastic assessment using polarization holographic imaging. Sci Rep 14, 2355 (2024). https://doi.org/10.1038/s41598-024-52762-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-024-52762-5
- Springer Nature Limited
This article is cited by
-
Smart polarization and spectroscopic holography for real-time microplastics identification
Communications Engineering (2024)