Label-free imaging flow cytometry for cell classification based directly on multiple off-axis holographic projections

Area of Science:

• Biophotonics and Imaging
• Cellular Biology
• Machine Learning in Medicine

Background:

• Imaging flow cytometry (IFC) enables detailed cell analysis but often requires staining and complex processing.
• Off-axis holography offers label-free imaging but traditional methods involve computationally intensive preprocessing, limiting throughput.
• Real-time cell classification is crucial for high-throughput biological assays and clinical diagnostics.

Purpose of the Study:

• To develop an automatic cell classification scheme for stain-free IFC.
• To directly utilize off-axis holographic projections for classification without preprocessing.
• To enhance classification accuracy and throughput for white blood cells.

Main Methods:

• A dedicated off-axis holographic microscopy system was built for acquiring white blood cells in flow.
• Deep learning models were applied directly to off-axis holographic projections (hologram space).
• Multiple-viewpoint holographic projections were utilized to capture comprehensive cellular information.

Main Results:

• The proposed method achieved a 7.69% accuracy improvement using ten holographic projections compared to a single projection.
• Direct holographic projection analysis outperformed methods requiring quantitative phase profile preprocessing by 17.95%.
• The technique simplifies computational processes, enabling significant increases in cell classification throughput.

Conclusions:

• This label-free IFC approach facilitates high-throughput, high-content analysis of cellular datasets.
• The method shows great potential for real-time cell classification in clinical settings.
• Direct analysis of holographic projections offers a more efficient and informative alternative to traditional IFC methods.