We Broke the Bottleneck of Microplastics Analysis

Nostra Culpa, we did it.

But why was there a bottleneck in the first place? And how did it happen? Well …

There are many methods to analyze microplastics, but some carry more of the workload than others. IR imaging, for instance, has been one of the powerhouses and driving forces of microplastics analysis for the last decade.

With IR imaging, microparticles are not only characterized by size and shape but also by their chemical makeup. This enables you to obtain valuable information regarding their origin, characteristics, and impact.

Countless publications have analyzed contamination in the air we breathe, the water we drink, and the soil we grow our food in using IR imaging. 

This research, besides requiring expert knowledge for sample preparation, shares one common ingredient: time.

Because analyzing whole filters by IR imaging requires a lot of time. One part is the measurement itself, but this can and has been automated. The remaining bottleneck came from the immensely large datasets produced.

We’re talking terabytes of data for a single publication.

The sheer size alone has a massive impact on computer hardware, and the algorithms used since 2015 are not equipped to reliably and quickly analyze this enormous data treasure.

However, developments over the last three years have brought a new approach to data analysis.

It might sound like a buzzword from meeting rooms or marketing documents, but AI, when applied by experts and trained by professionals, is an immensely powerful ally for data analysis and streamlining.

This year, Bruker has launched “MPID”, the MicroPlasticsIDentifier, a software tool for accurate, efficient, and regulation-aligned microplastic analysis using IR imaging and artificial intelligence.

How does it work? The exact mechanism is explained in detail in our product note (downloadable at the end of this article), but here’s the abridged version:

The MPID classification pipeline processes hyperspectral images through a series of automated steps and deep neural network classification.

Sounds complicated, but actually, it is indeed quite an easy concept.

Hyperspectral images are what we call the IR data collected during IR imaging. Each pixel in such an image is directly linked to an IR spectrum.

The algorithm examines every spectrum in every pixel and classifies it according to a defined ruleset:

• First, each spectrum undergoes spectral standardization.
• Then, a neural network determines the presence of relevant material by distinguishing foreground pixels from the background.
• Next, a second algorithm classifies each foreground pixel into a polymer class (e.g., ABS, PE, PET, you name it).
• Finally, each pixel is labeled based on surrounding spatial context, creating smooth and continuous chemical maps of the measured microplastics.

Let me be clear: This is not a black box.

This is a software tool that helps experts see clearer, faster.

Now IR imaging devices can run on auto, acquiring data 24/7, and users can finally keep up with the data, breaking it down into results in minutes and tackling the growing microplastics problem.

That’s how and why we broke the bottleneck.

By eliminating manual evaluation and external processing, MPID accelerates time-to-result and ensures consistency across large-scale studies.

Want to learn more? Download our Product Note.

Questions? Contact us