Single-detector multiplex imaging flow cytometry for cancer cell classification with deep learning.

Area of Science:

• Biomedical Engineering
• Cell Biology
• Optical Imaging

Background:

• Imaging flow cytometry integrates flow cytometry and microscopy for advanced cell analysis.
• Existing methods face limitations in simultaneous multi-channel imaging and automated data processing.
• Ovarian cancer detection requires sensitive and accurate single-cell analysis techniques.

Purpose of the Study:

• To develop a novel multiplex imaging flow cytometry (mIFC) system using spatial wavelength division multiplexing.
• To implement a deep learning framework for automated processing and analysis of mIFC data.
• To evaluate the performance of mIFC in distinguishing between normal and cancerous ovarian cell lines.

Main Methods:

• Developed a single-detector mIFC system excited by a metal halide lamp.
• Utilized spatial wavelength division multiplexing for simultaneous brightfield and multicolor fluorescence imaging.
• Designed a deep learning model comprising U-net, VDSR, and VGG19 networks for image processing and cell classification.

Main Results:

• Validated mIFC performance with resolution and magnification test lenses, demonstrating micron-scale differentiation capabilities.
• The cluster of differentiation 24 (CD24) channel showed higher sensitivity in classifying ovarian cell lines compared to brightfield, nucleus, or cancer antigen 125 (CA125) channels.
• Achieved an average classification accuracy of 97.1% for three ovarian cell line types using deep learning analysis across all four imaging channels.

Conclusions:

• The developed single-detector mIFC system offers consistent imaging channels and high differentiation capability.
• Deep learning analysis significantly enhances the accuracy of single-cell classification from mIFC data.
• This mIFC technology holds promise for advancing automated single-cell analysis in cancer detection and other biomedical applications.