COSMOS: a platform for real-time morphology-based, label-free cell sorting using deep learning

Area of Science:

• Cell biology
• Bioinformatics
• Artificial Intelligence

Background:

• Understanding cellular heterogeneity is vital for biological and medical research.
• Current methods for cell characterization and sorting often rely on specific labels or dyes, limiting real-time analysis.
• Morphological analysis provides a label-free way to assess cell characteristics.

Purpose of the Study:

• To develop an AI-powered platform for real-time characterization and sorting of single cells using brightfield microscopy.
• To enable label-free cell sorting based on morphological features.
• To address the technical gap in deep learning-based cell assessment and sorting.

Main Methods:

• Development of Computational Sorting and Mapping of Single Cells (COSMOS) platform integrating microfluidics and AI.
• Application of supervised deep learning models for morphological analysis of high-resolution brightfield images.
• Utilizing high-dimensional embedding vectors of morphology for cell characterization and sorting without biomarkers or stains.

Main Results:

• Demonstrated COSMOS capabilities on multiple human cell lines and tissue samples.
• Showcased the ability of neural network embedding space to capture deep visual characteristics of cells.
• Successfully purified unlabeled viable cells with desired morphological traits in real-time.

Conclusions:

• COSMOS platform effectively characterizes and sorts single cells based on morphology using AI and brightfield imaging.
• The approach enables efficient, label-free purification of cells with specific visual traits.
• This technology offers a novel solution for real-time deep learning-based cell analysis and sorting.