A Fully Automated End-to-End Process for Fluorescence Microscopy Images of Yeast Cells:From Segmentation to Detection and Classification

In recent years, an enormous amount of fluorescencemicroscopy images were collected in high-throughput lab settings. Ana-lyzing and extracting relevant information from all images in a shorttime is almost impossible. Detecting tiny individual cell compartmentsis one of many challenges faced by biologists. This paper aims at solvingthis problem by building an end-to-end process that employs methodsfrom the deep learning field to automatically segment, detect and classifycell compartments of fluorescence microscopy images of yeast cells. Withthis intention we used Mask R-CNN to automatically segment and labela large amount of yeast cell data, and YOLOv4 to automatically detectand classify individual yeast cell compartments from these images. Thisfully automated end-to-end process is intended to be integrated into aninteractive e-Science server in the PerICo (https://itn-perico.eu/home/)project, which can be used by biologists with minimized human effortin training and operation to complete their various classification tasks.In addition, we evaluated the detection and classification performance ofstate-of-the-art YOLOv4 on data from the NOP1pr-GFP-SWAT yeast-cell data library. Experimental results show that by dividing originalimages into 4 quadrants YOLOv4 outputs good detection and classifi-cation results with an F1-score of 98% in terms of accuracy and speed,which is optimally suited for the native resolution of the microscope andcurrent GPU memory sizes. Although the application domain is opticalmicroscopy in yeast cells, the method is also applicable to multiple-cellimages in medical applications.