Phenotypic Analysis of Microalgae Populations Using Label-Free Imaging Flow Cytometry and Deep Learning

Environmental factors such as temperature, nutrients, and pollutants affect the growth rates and physical characteristics of microalgae populations. As algae play a vital role in marine ecosystems, the monitoring of algae is important to observe the state of an ecosystem. However, analyzing these microalgae populations using conventional light microscopy is time-consuming and requires experts to both identify and count the algal cells, which in turn considerably limits the volume of the samples that can be measured in each experiment. In this work we use a high-throughput and field-portable imaging flow cytometer to perform automated label-free phenotypic analysis of marine microalgae populations using image processing and deep learning. The imaging flow cytometer provides color intensity and phase images of microalgae contained in a liquid sample by capturing and reconstructing the lens-free color holograms of the continuously flowing liquid at a flow rate of 100 mL/h. We extracted the spatial and spectral features of each algal cell in a sample from these holographic images and performed automated algae identification using convolutional neural networks. These features, alongside the composition and growth rate of the algae within the samples, were analyzed to understand the interactions between different algae populations as well as the effects of toxin exposure. As proof of concept, we demonstrated the effectiveness of the system by analyzing the impact of various concentrations of copper on microalgae monocultures and mixtures.

Automated summary

Environmental factors such as temperature, nutrients, and pollutants influence the growth and characteristics of microalgae populations; monitoring algae is crucial for ecosystem observation. Conventional light microscopy for algae analysis is time-consuming and sample volume-restricting; researchers have developed a high-throughput, portable imaging flow cytometer for automated label-free phenotypic analysis.