Using machine learning to detect microplastics

Using machine learning to detect microplastics

A new approach combines 3D coherent imaging with machine learning to detect microscale microplastics in filtered water samples.

There is an increasing awareness that the largest percentage of marine litter consists of plastic waste which can take decades to biodegrade, and the ensuing concern over the negative effects of microplastics – plastic material with a diameter of less than 5mm – on fragile marine environments.,

Already found abundantly on the sea-bed, in the water column and at the sea surface, recent studies have also detected microplastics in freshwater and drinking water sources, causing concern about their potential health threat.

Microplastics are usually identified visually in environmental samples by using an optical microscope, so long as the particle size ranges between 1 and 5 mm. Anything smaller remains unidentified so an automated identification and counting method to perform effective ecological risk assessments and a means of identifying, isolating, and extracting the microplastics is a timely requirement.

Researchers from the Institute of Applied Sciences and Intelligent Systems (ISASI) and the National Research Council (CNR) in Italy have developed a process called holographic plastics identification (HPI), which combines 3D coherent imaging with machine learning to detect microplastics in filtered water.

Digital holography is used to image microplastic particles, providing a fast and effective means of identification with a potentially field-portable, real-time, low-cost system for environmental monitoring. Combining this technique with AI should improve its accuracy and capability.

“We used a machine learning paradigm relying on features extracted from holographic images,” said the researchers. “We demonstrate that it is possible to determine an optimal set of ‘holographic features’ extracted from the digital holograms, with the scope of identifying a distinctive marker for the [microplastic] class. Thus, these can be thought of as a specific ‘fingerprint’ for the whole MP population.”

Because other microscale organisms such as plankton are easily confused with microplastics, the researchers have created a library of holographic images of other populations of micro‐objects: nine single-celled plankton species and mix of microplastics sized from 1mm down to 20 μm.

The process was able to recognise microplastics in pre-filtered water samples with 99% accuracy and differentiate between sizes, shapes and plastic types – creating an automatic prescreening tool that could replace unaided microscope observation of pretreated water samples.