03 Nov 2022
University of Gothenburg technique combines microscopy with deep-learning.
A project at the University of Gothenburg has now developed a method combining holographic microscopy and deep-learning that can follow microplankton throughout their lifespan, monitoring their position and dry mass.
As described in eLife, the microscopy-based technique could provide insights into the feeding strategies of ocean plankton and new data for global ocean models.
"We have a good understanding of who eats who and where they go in the case of larger organisms such as animals and birds that we see every day," commented Gothenburg's Giovanni Volpe. "The method we have developed is the only one that works to study microscopic organisms at the individual level."
Holographic imaging has already found applications in microbial studies, especially for in situ measurements of particle size distributions and their identity, noted the project in its paper.
"However, its full potential has not yet been exploited, namely for the quantitative investigation of the growth and feeding patterns of individual plankton over prolonged times. Arguably, this is because the data acquisition and processing pipelines are very computationally expensive."
A marriage between microscopy and AI
The solution was to combine holographic microscopy with deep-learning algorithms, which can circumvent the long computational times and, once trained, allow rapid determination of three-dimensional position and dry mass of individual microplanktons over extended time periods.
In trials, the project's lensless holographic imaging technique used a narrow-band 632-nanometer LED light source to illuminate plankton located in circular wells, with a 1024 x 1280 pixel CMOS sensor positioned below the samples.
The diffraction patterns formed by the interference of the unscattered light and the light scattered by the plankton act as a unique fingerprint of their size, refractive index, lateral and axial position, commented the project. The dry mass can also be calculated from the scattering cross section.
Although the volume of water that can be monitored by holographic microscopy is limited by the coherence length of the light source, the project believes that simple modifications to its proof-of-concept platform could allow the same lens-less approach to be adapted for smaller organisms such as bacteria, and larger organisms including small crustaceans.
"The marriage between holographic microscopy and deep learning provides a strong complementary tool in microbial ecology," commented the project in its paper. "It allows the nondestructive and minimally invasive determination of the three-dimensional position and dry mass of individual microorganisms, outperforming traditional methods in terms of speed and individual resolution, and rivals the precision and accuracy of current methods."