In Situ Detection of Aquatic Microplastics using Laser-Based Holographic Imaging and Deep Learning

Aquatic ecosystems are paramount to the health of the planet, but they face considerable threats from anthropogenic pollutants –. Notably, microplastics (particles < mm) account for a substantial fraction of marine contamination , . Research indicates their ubiquitous presence in nearly all oceans and freshwater bodies, and suggests they constitute % of all aquatic waste , . Moreover, scientists have detected microplastics in % of marine species, demonstrating the sheer extent of this form of aquatic contamination and the hazard it poses to human health –.

Accurately quantifying microplastics is essential to successful remediation initiatives . However, traditional monitoring techniques are limited in their ability to systematically cover large geographic regions; they require bulky, expensive, and complex laboratory equipment, as well as tedious, discrete sampling protocols, significantly hindering overall assessment frequency , .

Direct in situ microparticle classification is key to enabling routine microplastic monitoring. Digital holography could offer a compelling solution to the challenge of in situ detection. It enables the detailed capture of a three-dimensional volume in a single image, overcoming the size and practicality limitations of other techniques , . When combined with a deep learning classifier, microplastics could be rapidly differentiated from natural particles in water.

Key development objectives for this project include:

Design, construct, and optimize a low-cost digital holography prototype for underwater imaging of microparticles.

Create computational algorithms for hologram reconstruction and custom-trained AI classification models for identifying spherical and irregular microplastics from natural particles in lake water.

Demonstrate the capability of in situ microplastic detection in natural waterways via field trials.

A multi-step methodology was employed to construct and optimize the holographic imaging device for in situ microplastic detection (Figure ). CAD models were developed in Solidworks to support the strategic arrangement of components, and custom parts were fabricated via 3D printing. Simulations in Ansys Zemax were used to model optical configurations and enhance image quality.

A submersible prototype was constructed for proof-of-concept testing. The holographic imaging system consists of: nm laser, beam expanding optics, narrowband filter, CMOS image sensor, and optical-grade windows. A custom-built waterproof chassis houses the components while allowing laser beam propagation through the sampling cavity. Optical resolution was measured using microscope calibration targets.

Numerical algorithms were implemented to reconstruct raw holograms using the Fourier-based angular spectrum method , . Five mathematical focus metrics , were evaluated to autofocus objects in the reconstructed images (Figure ). Detection of microparticles was accomplished by the automatic grouping and segmentation of objects at different distances along the sampling cavity.

To differentiate microplastics from natural aquatic particles, custom-trained AI classification models were developed using microbeads and irregular-shaped polyethylene (PE), polystyrene (PS), and polyethylene terephthalate (PET) reference samples obtained from a leading ecotoxicology researcher. Training was performed using , particles, consisting of microplastics and suspended particulate matter within sample lake water. Model hyperparameters were fine-tuned to enhance detection accuracy and computational processing speed.

The prototype holographic assembly was scaled down in size and implemented on a previously-developed bionic robot, creating a fully integrated system capable of autonomously collecting, processing, and analyzing holograms. Computational fluid dynamics (CFD) simulations in Siemens STAR-CCM+ and empirical testing were used to assess the robot’s kinematics.

The prototype device and bionic robot were field tested at waterways to assess performance under real-world conditions. The optimized computational algorithms and AI detection models were applied to the resultant holograms to generate statistical data and microplastic concentrations.

Optical Performance

Optical simulations performed in Ansys Zemax yielded an optimized Galilean laser beam expander consisting of two lenses with an expansion ratio of x. The final optical design exhibits high beam quality (peak wavefront error < waves) with uniform illumination across the image sensor (Figure ). Tolerancing of the lens spacing ensured the desired magnification was achieved with near diffraction-limited performance. Testing of the prototype demonstrated an optical resolution of m across the entire cm sampling cavity (Figure ), which is key for distinguishing microparticle surface features.

Numerical Algorithms

Dynamic reconstruction of microparticles across arbitrary 2D planes from a single raw hologram was achieved, with autofocus in each plane evaluated using five metrics, including image gradient, intensity probability, and statistical approaches. Overall, the Tenengrad metric exhibited the best accuracy, and was capable of reconstructing images to within m of best focus, resulting in distinct, sharp particle features (Figure ). Additional computational algorithms were developed to extract microparticle characteristics, including size, contour, area, and concentration. The accuracy of the size estimation algorithm was evaluated using measurements from an optical microscope, and the calculated particle diameters were found to deviate by <% from actual size (Figure ). To minimize algorithm latency, a multiprocessing architecture was implemented, reducing the hologram reconstruction time by x.

Detection Models

All microplastic classification tests were performed using lake water containing natural particulates, with a distinct validation dataset consisting of particles, comprising both natural and microplastic classes. Ground truth images were obtained using optical microscopy to validate particle type. Various hyperparameters (model architecture, epoch number, and image size) of the YOLO neural network were refined to enhance overall detection reliability, yielding improvements to key performance metrics (e.g., x increase in recall, Figure ). For identifying PS spherical microbeads (nominal diameters of and m), the AI model exhibited % accuracy, with a recall and precision of % and %, respectively. For classifying the irregular-shaped PE, PS, and PET microplastics encompassing a wide range of shapes and sizes (– m), an accuracy of % was achieved, with % recall and % precision (Figure ). The total misclassification rate was low, primarily caused by the partial obstruction of particles from neighbouring objects (Figure ).

Robotic Integration

The prototype imaging system was reduced in size and installed on a biomimetic robotic platform to demonstrate feasibility of autonomous monitoring. CFD simulations were performed on CAD models with and without the holographic hardware, showing a minimal % change to hydrodynamic drag (Figure ). Tests conducted at varying depths underwater confirmed that the robot maintained fluidic swimming kinematics under realistic environmental conditions.

Field Tests

Collectively, in situ field trials of the holographic imaging system were performed at distinct waterways. The prototype accumulated > hours of underwater operation without water ingress. These tests yielded , holograms, containing million microparticles. To facilitate comparisons between test locations, statistics on microparticle characteristics were derived, including size distributions and microplastic concentrations (Figure ). Microplastics were detected in several lakes, with average concentrations of particles/L (Figure ). Additionally, numerous unique microorganisms were identified in high-resolution, highlighting the system’s potential for tracking bioindicators as a measure of ecosystem health (Figure ).

Microplastic contamination in aquatic environments has been recognized as a significant concern by leading global organizations, including the United Nations and the U.S. Environmental Protection Agency , . To better understand the risks it poses, there is a growing need for novel identification methods that overcome the drawbacks of existing laboratory-based techniques , . This project developed a low-cost, submersible holographic imaging system capable of rapidly analyzing microplastics directly in water, successfully achieving all development objectives.

By employing digital holography, high-resolution 3D imaging is attained using a portable, low-profile design measuring x cm. When combined with advanced numerical algorithms and a custom-trained neural network operated on a Raspberry Pi , microparticles ranging from – m can be detected with high accuracy at capture rates of Hz, leading to a high-throughput sampling rate of L/h. Field testing further demonstrated robust performance under dynamic natural conditions, including turbulent and turbid waters.

This system addresses key limitations of traditional microplastic detection methods. It eliminates the need for time-consuming laboratory procedures, supports direct field-deployment with continuous monitoring, offers a cost reduction of ~x, and reduces analysis time from days to minutes , . Integrating the device onto a robotic platform enables routine, autonomous monitoring across large geographic regions with the ability to perform depth-resolved sampling along the water column. Additionally, it can identify microorganisms in situ, supporting their use as bioindicators of ecosystem health. Finally, the system provides a scalable framework whereby multiple robots can cooperatively survey in tandem, enabling comprehensive monitoring with minimal human effort and low ecological disruption.

Future work will involve:

Developing AI models to classify microorganisms (copepods) in situ as bioindicators of ecosystem health, complementing the detection of microplastic contamination. Feasibility studies are already underway.

Extracting phase information from hologram reconstructions to enhance training data with potentially unique, material-specific signatures.

Applying the holographic system to the detection of additional aquatic contaminants, including chemicalbiological toxins (harmful algae).

Exploring opportunities to benchmark the performance of the holographic imaging solution against existing laboratory methods (FTIRRamanOptical), in collaboration with a research group.

Conducting field trials to map microparticle spatial distributions across waterbodies with the objective of sharing results with conservationists.

I gratefully acknowledge the individuals and organizations who supported me over the past year:

My grandparents – for allowing me to use their backyard pool, where initial device testing and evaluation were conducted.

Dr. Chelsea Rochman – for providing the irregular microplastic reference samples.

Halton Region Conservation Authority – for providing access to the waterways used during field testing.

BASEF Organization and Committee – for planning and facilitating the CWSF trip and offering this wonderful opportunity to young scientists. Thank you for also providing guidance and review assistance during the preparation of my project for CWSF.

My parents – for supporting me during the entirety of my project. Thank you for being there during the numerous hours of work it took to get this project to CWSF.

[1] H. K. Lotze, “Marine biodiversity conservation,” Curr. Biol., vol. 31, no. 19, pp. R1190–R1195, Oct. 2021, doi: 10.1016/j.cub.2021.06.084.

[2] United Nations Department of Economic and Social Affairs, The Sustainable Development Goals Report 2021. United Nations, 2021. [Online]. Available: https://unstats.un.org/sdgs/report/2021/The-Sustainable-Development-Goals-Report-2021.pdf.

[3] T. Backhaus et al., “Assessing the ecological impact of chemical pollution on aquatic ecosystems requires the systematic exploration and evaluation of four lines of evidence,” Environ. Sci. Eur., vol. 31, no. 1, p. 98, Dec. 2019, doi: 10.1186/s12302-019-0276-z.

[4] E. Jeong, J.-Y. Lee, and M. Redwan, “Animal exposure to microplastics and health effects: A review,” Emerg. Contam., vol. 10, no. 4, p. 100369, Dec. 2024, doi: 10.1016/j.emcon.2024.100369.

[5] A. R. Sunny et al., “Microplastics in Aquatic Ecosystems: A Global Review of Distribution, Ecotoxicological Impacts, and Human Health Risks,” Water, vol. 17, no. 12, p. 1741, Jun. 2025, doi: 10.3390/w17121741.

[6] N. Wu, S. W. D. Grieve, A. J. Manning, and K. L. Spencer, “Flocs as vectors for microplastics in the aquatic environment,” Nat. Water, vol. 2, no. 11, pp. 1082–1090, Nov. 2024, doi: 10.1038/s44221-024-00332-4.

[7] United Nations Environment Programme, From Pollution to Solution: A Global Assessment of Marine Litter and Plastic Pollution. 2021. Accessed: April 03, 2026. [Online]. Available: https://wedocs.unep.org/handle/20.500.11822/36963.

[8] J. Mutuku, M. Yanotti, M. Tocock, and D. Hatton MacDonald, “The Abundance of Microplastics in the World’s Oceans: A Systematic Review,” Oceans, vol. 5, no. 3, pp. 398–428, June 2024, doi: 10.3390/oceans5030024.

[9] D. Gola et al., “The impact of microplastics on marine environment: A review,” Environ. Nanotechnol. Monit. Manag., vol. 16, p. 100552, Dec. 2021, doi: 10.1016/j.enmm.2021.100552.

[10] G. Abbas, U. Ahmed, and M. A. Ahmad, “Impact of Microplastics on Human Health: Risks, Diseases, and Affected Body Systems,” Microplastics, vol. 4, no. 2, p. 23, May 2025, doi: 10.3390/microplastics4020023.

[11] N. Zolotova, A. Kosyreva, D. Dzhalilova, N. Fokichev, and O. Makarova, “Harmful effects of the microplastic pollution on animal health: a literature review,” PeerJ, vol. 10, p. e13503, June 2022, doi: 10.7717/peerj.13503.

[12] A. K. M. M. Hasan, M. Hamed, J. Hasan, C. J. Martyniuk, S. Niyogi, and D. P. Chivers, “A review of the neurobehavioural, physiological, and reproductive toxicity of microplastics in fishes,” Ecotoxicol. Environ. Saf., vol. 282, p. 116712, Sept. 2024, doi: 10.1016/j.ecoenv.2024.116712.

[13] S. Thanigaivel et al., “Microplastic pollution in marine environments: An in-depth analysis of advanced monitoring techniques, removal technologies, and future challenges,” Mar. Environ. Res., vol. 205, p. 106993, Mar. 2025, doi: 10.1016/j.marenvres.2025.106993.

[14] V. Hidalgo-Ruz, L. Gutow, R. C. Thompson, and M. Thiel, “Microplastics in the Marine Environment: A Review of the Methods Used for Identification and Quantification,” Environ. Sci. Technol., vol. 46, no. 6, pp. 3060–3075, Mar. 2012, doi: 10.1021/es2031505.

[15] A. Käppler et al., “Analysis of environmental microplastics by vibrational microspectroscopy: FTIR, Raman or both?,” Analytical and Bioanalytical Chemistry, vol. 408, no. 29, pp. 8377–8391, Nov. 2016, doi: 10.1007/s00216-016-9956-3.

[16] U. Schnars, C. Falldorf, J. Watson, and W. Jüptner, Digital Holography and Wavefront Sensing: Principles, Techniques and Applications. Berlin, Heidelberg: Springer Berlin Heidelberg, 2015. doi: 10.1007/978-3-662-44693-5.

[17] T.-C. Poon, Digital Holography and Three-Dimensional Display: Principles and Applications, 1st ed. New York, NY: Springer, 2006.

[18] J. W. Goodman, Introduction to Fourier optics, 2. ed., 7. [pr.]. in McGraw-Hill series in electrical and computer engineering Electromagnetics. New York: McGraw-Hill, 2003.

[19] S. Pertuz, D. Puig, and M. A. Garcia, “Analysis of focus measure operators for shape-from-focus,” Pattern Recognit., vol. 46, no. 5, pp. 1415–1432, May 2013, doi: 10.1016/j.patcog.2012.11.011.

[20] F. C. A. Groen, I. T. Young, and G. Ligthart, “A comparison of different focus functions for use in autofocus algorithms,” Cytometry, vol. 6, no. 2, pp. 81–91, Mar. 1985, doi: 10.1002/cyto.990060202.

[21] United Nations Environment Programme, “Microplastics: The long legacy left behind by plastic pollution,” UNEP. Accessed: April 03, 2026. [Online]. Available: https://www.unep.org/news-and-stories/story/microplastics-long-legacy-left-behind-plastic-pollution.

[22] United States Environmental Protection Agency, “EPA, HHS Announce Historic Actions to Protect Americans from Microplastics and Safeguard Drinking Water,” US EPA. Accessed: April 03, 2026. [Online]. Available: https://www.epa.gov/newsreleases/epa-hhs-announce-historic-actions-protect-americans-microplastics-and-safeguard.

[23] S. Dutchen, “Microplastics Everywhere,” Harvard Medicine Magazine. Accessed: April 03, 2026. [Online]. Available: https://magazine.hms.harvard.edu/articles/microplastics-everywhere.

[24] E. R. Forgione, C. Ferry, B. Neumann, and K. D. Good, “Initiating environmental microplastics analysis: A review and planning guide with practical insights from laboratory implementation,” Sci. Total Environ., vol. 1014, p. 181220, Feb. 2026, doi: 10.1016/j.scitotenv.2025.181220.

[25] T. Vural, S. Çetinkaya, V. Yeğen, S. Şapcıoğlu, and S. Gündoğdu, “Protocol for extraction and analysis of microplastics in freshwater, sediment, and fish samples,” STAR Protoc., vol. 6, no. 3, p. 104057, Sep. 2025, doi: 10.1016/j.xpro.2025.104057.

[26] World Health Organization, “WHO calls for more research into microplastics and a crackdown on plastic pollution,” World Health Organization. Accessed: April 03, 2026. [Online]. Available: https://www.who.int/news/item/22-08-2019-who-calls-for-more-research-into-microplastics-and-a-crackdown-on-plastic-pollution.

[27] European Commission, “Microplastics,” European Commission. Accessed: April 03, 2026. [Online]. Available: https://environment.ec.europa.eu/topics/plastics/microplastics_en.

[28] W. Zheng, “FTIR Spectrometer Prices: Understanding Costs, Types, and Applicatio,” Sourcify China. Accessed: April 03, 2026 [Online]. Available: https://www.sourcifychina.com/ftir-spectrometer-price.

[29] Hench Technology, “A Practical 2025 Buyer’s Guide to Infrared Spectrometer Price: 7 Factors to Consider,” Hench Technology. Accessed: April 03, 2026. [Online]. Available: https://www.hcftir.com/a-practical-2025-buyers-guide-to-infrared-spectrometer-price-7-factors-to-consider-article/.

[30] H. Li et al., “Deep learning assisted ATR-FTIR and Raman spectroscopy fusion technology for microplastic identification,” Microchem. J., vol. 212, p. 113224, May 2025, doi: 10.1016/j.microc.2025.113224.

[31] S. Wang, S. M. Mintenig, J. Wu, and A. A. Koelmans, “Implications of method- and instrument-based size detection limits in μFTIR-based microplastic analysis,” Talanta, vol. 296, p. 128417, Jan. 2026, doi: 10.1016/j.talanta.2025.128417.

[32] A. M. Othman, A. A. Elsayed, Y. M. Sabry, D. Khalil, and T. Bourouina, “Detection of Sub-20 μm Microplastic Particles by Attenuated Total Reflection Fourier Transform Infrared Spectroscopy and Comparison with Raman Spectroscopy,” ACS Omega, vol. 8, no. 11, pp. 10335–10341, Mar. 2023, doi: 10.1021/acsomega.2c07998.

[33] Y. Chen et al., “Identification and quantification of microplastics using Fourier-transform infrared spectroscopy: Current status and future prospects,” Curr. Opin. Environ. Sci. Health, vol. 18, pp. 14–19, Dec. 2020, doi: 10.1016/j.coesh.2020.05.004.

Images

[34] haritonovstock, Artist, Colorful microplastic particles in sea water environmental pollution concept. [Online Image]. Freepik. Available: https://www.freepik.com/premium-ai-image/colorful-microplastic-particles-sea-water-environmental-pollution-concept_173422732.htm.

[35] VitalyEdush, Artist, Underwater world with corals and tropical fish. [Online Image]. iStock. Available: https://www.istockphoto.com/photo/underwater-world-with-corals-and-tropical-fish-gm472563478-63485799.

[36]  M. Eriksen et al., “Microplastic sampling with the AVANI trawl compared to two neuston trawls in the Bay of Bengal and South Pacific,” Environ. Pollut., vol. 232, pp. 430–439, Jan. 2018, doi: 10.1016/j.envpol.2017.09.058.

[37]  H.-U. Dahms et al., “Potential of the small cyclopoid copepod Paracyclopina nana as an invertebrate model for ecotoxicity testing,” Aquat. Toxicol., vol. 180, pp. 282–294, Nov. 2016, doi: 10.1016/j.aquatox.2016.10.013.

[38] Rayne Water, Artist, Microplastics: A Hidden Hazard. [Online Image]. Rayne Water. Available: https://www.raynewater.com/blog/unseen-dangers-the-truth-about-microplastics-in-our-water-and-raynes-solution/.