Automated classification of cell morphology by coherence-controlled holographic microscopy.

Area of Science:

• Cell biology
• Biophysics
• Machine learning in biology

Background:

• Machine learning is increasingly used for cell classification in biology.
• Current methods often rely on morphometric (MO) features, lacking quantitative cell mass data and potentially limiting classification accuracy.

Purpose of the Study:

• To investigate the utility of coherence-controlled holographic microscopy for quantitative phase imaging in cell morphology classification.
• To compare the performance of quantitative phase features against traditional MO features for cell classification.

Main Methods:

• Utilized coherence-controlled holographic microscopy for quantitative phase imaging.
• Employed supervised machine learning algorithms to classify nutritionally deprived cancer tissue cells.
• Compared classification performance using quantitative phase features versus morphometric (MO) features.

Main Results:

• Most machine learning classifiers demonstrated higher performance when quantitative phase features were utilized.
• Quantitative phase features were found to be crucial for enhancing classification accuracy.
• The developed methodology shows potential for automated live cell monitoring.

Conclusions:

• Quantitative phase imaging via coherence-controlled holographic microscopy significantly improves cell classification performance.
• This technique offers a noninvasive, label-free method for accurate, automated live cell analysis.
• The approach provides high spatiotemporal phase sensitivity essential for detailed cell behavior studies.