Label-free 3-D quantitative phase imaging cytometry with deep learning: identifying naive, memory, and senescent T cells

Abstract

The current prevailing methods for identifying immune cell subsets exploit a group of differentiation markers (CDs) targeted by fluorochrome or metal conjugated antibodies. However, such labeling methods, requiring a staining process and specific reagents, prevent rapid and cost-effective identification of immune cell subsets. Therefore we developed a label-free imaging cytometry platform that synergistically used refractive index (RI) tomography and three-dimensional (3-D) deep learning. We constructed and trained a deep learning classifier that learns unique representations from the 3-D RI map of each cell obtained using RI tomography without labeling. In this study, we were able to classify human naïve, memory, and senescent T cells according to the expression of CD4, CD8, CD45RA, CCR7 and CD57 using the label-free classifier within milliseconds, with high precision (>95%) even though the morphological and biochemical characteristics extracted from the RI tomograms of the T cells are almost homogeneous. This cannot be achieved by conventional machine learning approaches that only exploit the set of manually extracted features. Our label-free cell sorting platform will facilitate rapid and cost-effective immunological and biomedical studies by eliminating the laborious labeling process.

Noninvasive morpho-molecular imaging reveals early therapy-induced senescence in human cancer cells

“…QPI allows the quantification of a number of cellular and subcellular characteristics, such as dry mass, volume, surface, and thickness, in an entirely noninvasive manner, by exploiting the fact that the optical phase shift of a laser beam through a specimen contains information about its RI variations due to structural features ( 23 ). In the context of cell senescence research, it has been exploited to distinguish subsets of human senescent T cells through a cytometry platform for tomographic imaging ( 24 ).…”