Speeding up microparticle analysis with photonics

(08-09-2021) Can microparticles like blood cells or microbes be studied faster and cheaper thanks to artificial intelligence? Alessio Lugnan figured it out in his PhD.

Liquids can host huge numbers and a great variety of microparticles that are interesting to study, such as: cells in blood, microbes in water and food, pollutants (e.g. microplastics) in water, plankton in the ocean, etc...

“In order to ensure statistical validity in scientific studies or in biomedical applications, or to detect rare microparticles, a large number of particles have to be measured/analysed in sufficient detail”, Alessio Lugnan explains.

“Flow cytometry is a widely used technique that allows to study microparticles one by one, while flowing in a liquid at a high speed”, Alessio continues.

 Some applications are:

  • detection of circulating tumor cells in blood
  • blood analysis to monitor immune status
  • monitoring of waterborne microbes for water treatment and reuse
  • biological analysis of heterogeneous cell populations
  • cell sorting, to automatically isolate specific cell types
  • bacteria viability in probiotic products

“A goal of my dissertation is to investigate the use of a specific machine learning method (a branch of artificial intelligence), whose employment would make flow cytometry cheaper, more compact and easier to use, in order to enable versatile and in-situ implementations”, tells Alessio.

In particular, the proposed method consists in illuminating flowing microparticles with a laser (this is common), letting the transmitted light pass through an optical medium which diffracts the light in unpredictable and diverse ways, and taking a picture of the outcome.

A simple, versatile and very fast machine learning algorithm (a linear classifier) then learns how to classify different microparticles, using the obtained images.

“The strong point of the method is that the linear classifier, which is alone not very powerful, becomes more powerful by accessing the very diverse and multidimensional information generated by light diffraction. Still, it retains its simplicity, speed and versatility”, Alessio clarifies.

Light diffraction does not require computation, but happens naturally at the speed of light. Therefore, powerful microparticle classification can be carried out at high speed (low computation required), allowing to analyse large numbers of microparticles with simple and cheap instrumentation.


PhD Title: Photonics-Based Machine Learning to Speed up and Simplify Label-Free Flow Cytometry


Read the entire PhD


ContactAlessio Lugnan, Peter Bienstman, Joni Dambre


Editor: Jeroen Ongenae - Illustrator: Roger Van Hecke