Real-time fluorescence imaging flow cytometry enabled by motion deblurring and deep learning algorithms.

Area of Science:

• Biomedical Engineering
• Cell Biology
• Machine Learning

Background:

• Fluorescence imaging flow cytometry (IFC) is vital for analyzing cell subpopulations.
• Motion blur in high-speed IFC images hinders accurate cell identification.
• Developing methods to overcome motion blur is crucial for advancing IFC applications.

Purpose of the Study:

• To develop a real-time single-cell imaging and classification system for fluorescence IFC.
• To enable direct cell type identification from motion-blurred images using deep learning.
• To address the challenge of motion blur in high-speed cell analysis.

Main Methods:

• A fluorescence microscope integrated with a deep learning algorithm was employed.
• A motion deblurring algorithm was developed for reconstructing blur-free images.
• A ResNet model was trained on deblurred HeLa cell images for classification.

Main Results:

• The system successfully identified cell types directly from motion-blurred images.
• Deblurred images of HeLa cells across different cell cycle stages were acquired.
• The ResNet model achieved 96.6% accuracy in classifying HeLa cells in three mitotic stages within 2 ms.

Conclusions:

• The developed system offers a novel approach to real-time single-cell fluorescence IFC and classification.
• This technology effectively overcomes motion blur limitations in IFC.
• The system holds significant potential for diverse biological and medical applications requiring rapid cell analysis.