Reviewed by Lexie CornerFeb 12 2025
Researchers from Monash University have developed an artificial intelligence program to assist scientists in analyzing environmental microplastics. The study was published in the Journal of Hazardous Materials.
Despite growing awareness of microplastics, many scientists and policymakers still lack precise data on the types of microplastics present and their environmental distribution. The Monash AI program addresses this by analyzing thousands of samples in seconds using machine learning algorithms, a process that would typically take months using conventional methods.
Identifying microplastics is complex, as natural materials like seashell fragments can resemble plastic particles. Instead of relying on visual analysis, the new algorithm processes data from Fourier transform infrared spectroscopy (FTIR), identifying unique chemical “signatures” that distinguish different microplastic types.
This program is the first to analyze a library of microplastic signatures, providing a critical tool for researchers addressing microplastic pollution. The research was led by Ph.D. candidate Frithjof Herb and senior lecturer Dr. Khay Fong from the Monash School of Chemistry.
We are addressing a significant bottleneck for progress in tackling the microplastics problem. Not only is the process of analyzing samples arduous and time-consuming, but until now, we have been unable to do it on a large enough scale to gain a comprehensive understanding of exactly what microplastics we are dealing with, where they are, and where they end up.
Frithjof Herb, Study Lead Researcher, Monash University
Herb said, “This is a very important first step in finding ways that we can clean up these damaging microplastics, and find ways to prevent them from entering environmental waterways in the first place.”
In addition to seashells, natural fibers such as algae, animal fur, and crustacean shells are often mistaken for microplastics.
According to Herb, the identification process is further complicated by the continuous evolution of synthetic materials, leading to variations in the chemical composition of microplastics.
Plastics are constantly changing, both in how they are made and how they break down in the environment. Traditional tools struggle to keep up with these changes. But our tool offers a crucial advantage to scientists who need something that can quickly adapt, which is important for analyzing data that continues to evolve.
Frithjof Herb, Study Lead Researcher, Monash University
Herb said, “We are really proud of what we have achieved here; it runs nicely on conventional laptops, reflecting our focus on sustainability and accessibility, which we sought through small and efficient models.”
Journal Reference:
Herb, F., et al. (2025) Machine learning outperforms humans in microplastic characterization and reveals human labeling errors in FTIR data. Journal of Hazardous Materials. doi.org/10.1016/j.jhazmat.2024.136989.