On-chip label-free cell classification based directly on off-axis holograms and spatial-frequency-invariant deep...

Area of Science:

• Biophotonics
• Computational Biology
• Cellular Imaging

Background:

• Imaging flow cytometry is crucial for cell analysis.
• Current methods often require time-consuming cell profile reconstruction for classification.
• Label-free techniques are desirable for preserving cell viability and simplifying workflows.

Purpose of the Study:

• To develop a rapid, label-free cell classification method using raw digital holograms.
• To bypass the computationally intensive phase profile reconstruction step in holographic imaging flow cytometry.
• To enhance the speed and robustness of cell classification in high-throughput screening.

Main Methods:

• Acquisition of raw off-axis digital holograms of cells in flow.
• Development of a convolutional neural network (CNN) trained on raw holographic images.
• CNN designed to be invariant to spatial frequencies and fringe directions in holograms.
• Validation using four distinct cancer cell types.

Main Results:

• Demonstrated successful label-free classification of individual cells directly from raw holograms.
• Achieved classification without the need for quantitative phase profile reconstruction.
• Showcased the CNN's ability to handle variations in holographic fringe patterns.
• Validated the approach on multiple cancer cell lines.

Conclusions:

• The presented method significantly accelerates label-free cell classification in imaging flow cytometry.
• Direct analysis of raw digital holograms offers a faster and potentially more robust alternative to traditional reconstruction-based methods.
• This approach holds promise for real-time, high-throughput cellular analysis in various biological and clinical applications.