Deep-Learning Terahertz Single-Cell Metabolic Viability Study

Abstract

Cell viability assessment is critical, yet existing assessments are not accurate enough. We report a cell viability evaluation method based on the metabolic ability of a single cell. Without culture medium, we measured the absorption of cells to terahertz laser beams, which could target a single cell. The cell viability was assessed with a convolution neural classification network based on cell morphology. We established a cell viability assessment model based on the THz-AS (terahertz-absorption spectrum) results as <i>y</i> = <i>a =</i> (<i>x - b</i>)^<i>c</i>, where <i>x</i> is the terahertz absorbance and <i>y</i> is the cell viability, and <i>a</i>, <i>b</i>, and <i>c</i> are the fitting parameters of the model. Under water stress the changes in terahertz absorbance of cells corresponded one-to-one with the apoptosis process, and we propose a cell 0 viability definition as terahertz absorbance remains unchanged based on the cell metabolic mechanism. Compared with typical methods, our method is accurate, label-free, contact-free, and almost interference-free and could help visualize the cell apoptosis process for broad applications including drug screening.