Extracting regions of lipid droplets from confocal microscopy images utilizing optical properties of oleaginous yeast.

Non-invasive methods for observing the morphology of living oleaginous yeast are ideal for optimizing the production of various oils, such as food oils, oleochemicals, and biodiesel, from oleaginous yeast. However, existing methods have been developed to target budding yeast without high oil production ability and extract regions of entire cells. This study is the first to target oleaginous yeast, namely, Lipomyces starkeyi, demonstrating a method for extracting regions of L. starkeyi directly influencing oil production through the unique optical properties of L. starkeyi. Specifically, we exploited changes in the brightness along the z-stack depth of multiple z-stack images obtained using confocal microscopy. Because this brightness change was unique to lipid droplets, pixels corresponding to lipid droplets were easily identified, allowing calculations of the parameters of visual features. The obtained parameters of visual features were then used as input for a semantic segmentation algorithm to accurately distinguish lipid droplets from other organelles, including organelles similar to lipid droplets in shape. Experimental results showed that our method successfully estimated the growth status of L. starkeyi, which, to date, is obtainable through invasive biochemical methods only. Moreover, our method non-invasively determined the shape of each yeast cell over the cultivation period to enable single-cell analysis, which has not been achieved with conventional biochemical methods. KEY POINTS: We propose a method to extract regions of L. starkeyi influencing oil production. Our method requires only confocal microscopy images and is completely non-invasive. Our method estimated the growth status of L. starkeyi to enable single-cell analysis.