Analysis of Cell Viability in Microfluidic Spheroid Arrays by Image Analysis and Neural Networks

Abstract

The outcomes of clinical studies are strongly influenced by their preceding preclinical studies. Due to higher physiological relevance, 3D spheroid arrays can potentially improve drug screening outcomes in preclinical studies. Analytical accessibility is still limited. To further increase usability and simplify subsequent analyses of spheroid arrays, the authors present an automated method for analyzing spheroid viability. The developed easy-to-use workflow provides viability analysis using fluorescence images of cell aggregates. It allows the automated analyses in an early development stage of microarrays with a low amount of available experiment data. The proposed workflow provides an accurate spheroid extraction (with a segmentation accuracy evaluated by Dice score with a score of 0.89) which is compared to the performance of the workflow with the segmentation of a Unet. By reducing human intervention in the analysis task, information extraction and the evaluation process are simplified leading to an overall shorter analysis time, while objectivity and comparability are increased.