Chip-Sized Microscopy for Continuous Monitoring: Application in White Wine Fermentation and Yeast Cell Counting via Deep Learning

Abstract

Nowadays, continuous monitoring is a difficult issue in microscopy. A chip-sized microscope was developed, composed only of microelectronic components, with high optical resolution and a wide field of view. Due to its miniaturized size, it can be placed on or attached to the sample for continuous monitoring in the sample environment. An example of an application of this microscope for the food and beverage industry is described, referring to the study of the fermentation process of white wine. The comparison of the images acquired with conventional optical microscopy reveals similar results. To automatically count yeast cells, the traditional image postprocessing is compared with deep learning. Neural networks achieve similar cell recognition characteristics but with an ~100× speed improvement, by directly processing the obtained holograms.