Integration of light scattering with machine learning for label free cell detection

Area of Science:

• Biophotonics and optical sensing
• Cell biology and biophysics
• Machine learning in diagnostics

Background:

• Label-free cell detection methods are crucial for real-time biological analysis.
• Light scattering patterns offer unique cellular fingerprints based on morphology.
• Existing classification methods may lack efficiency for complex cellular analysis.

Purpose of the Study:

• To develop and validate a machine learning (ML) method for classifying cell light scattering patterns.
• To assess the efficacy of integrated light scattering and ML for analyzing cellular responses.
• To evaluate the potential for real-time, label-free cell detection and diagnosis.

Main Methods:

• Utilizing angular light scattering patterns for label-free cell characterization.
• Developing a machine learning algorithm to classify unique light scattering signatures.
• Applying the technique to staurosporine-treated versus non-treated SH-SY5Y neuroblastoma cells.

Main Results:

• The ML classification achieved over 90% accuracy in distinguishing treated from non-treated cells.
• Predicted percentages of treated cells in mixed solutions were within 5% of ground-truth values.
• Demonstrated the distinct light scattering profiles correlating with cellular changes.

Conclusions:

• The integrated light scattering and ML approach provides a highly accurate, label-free method for cell analysis.
• This technique holds significant potential for real-time cellular detection and diagnostic applications.
• Further development could enable rapid identification of cellular states and drug responses.