Automated Image Processing Workflow for Morphological Analysis of Fluorescence Microscopy Cell Images

Abstract

Computerized image analysis of biological cells and tissues is a necessary complement to high-throughput microscopy, allowing researchers to effectively analyze large volumes of cellular data. It has the potential to dramatically improve the throughput and accuracy of measurements and related downstream analyses that may be obtained from images. This study presents a novel workflow for automated analysis of fluorescence microscopy images, which benefits from running multiple segmentation workflows and combining them to produce the best final segmentation. It is tested using a dataset of 42 fluorescence microscopy cells, evaluated against a hand segmented dataset using the F1 score, and critically compared to a single segmentation workflow, which served as a control. The accuracy and reliability of the novel workflow are demonstrated to be superior to the control workflow, which achieved F1 scores of 0.845 and 0.608, respectively. The workflow and example code are available through an open-source software platform.