Single-cell cytometry via multiplexed fluorescence prediction by label-free reflectance microscopy

Area of Science:

• Biomedical Imaging
• Computational Biology
• Cell Biology

Background:

• Traditional imaging cytometry relies on physical staining, which limits throughput and efficiency.
• Existing methods struggle with high-content screening due to labeling bottlenecks.

Purpose of the Study:

• To develop a label-free imaging cytometry technique using deep learning for enhanced throughput and multiplexed readouts.
• To demonstrate accurate prediction of subcellular features and cell cycle phenotypes without physical staining.

Main Methods:

• Leveraging reflectance microscopy for rich structural information and superior sensitivity.
• Employing a deep learning-augmented digital labeling method trained on immunofluorescence images.
• Validating predictions against established imaging cytometry techniques and cell cycle markers.

Main Results:

• Achieved up to a three-fold improvement in prediction accuracy compared to state-of-the-art methods.
• Successfully reproduced single-cell structural phenotypes of cell cycles, including Golgi twins, Golgi haze, and DNA synthesis.
• Enabled accurate multiparametric single-cell profiling across large cell populations through multiplexed digital readouts.

Conclusions:

• The developed label-free digital labeling method significantly enhances imaging cytometry throughput.
• This approach offers a powerful alternative for phenotyping, pathology, and high-content screening.
• Digital multiplexing provides accurate and efficient single-cell analysis without physical staining.