Image-based evaluation of single-cell mechanics using deep learning.

Area of Science:

• Cellular mechanobiology
• Biophysics
• Artificial intelligence in life sciences

Background:

• Cell mechanical properties are crucial biophysical markers for cell phenotypes and functions.
• Existing methods for measuring single-cell mechanics are low-throughput, complex, and require specialized equipment, limiting large-scale analysis.
• There is a need for advanced, non-invasive techniques to assess cell mechanics efficiently.

Purpose of the Study:

• To develop and validate deep learning models for evaluating single-cell stiffness non-invasively.
• To apply these models for assessing the functional states of mesenchymal stem cells (MSCs) and macrophages.
• To explore the potential of image-based deep learning in mechanobiology research and clinical applications.

Main Methods:

• Development of image-based deep learning models to predict cell stiffness.
• In situ, non-invasive measurement of cell stiffness for mesenchymal stem cells (MSCs) and macrophages.
• Application of models to evaluate MSC functions (senescence, stemness, immunomodulatory capacity) and macrophage phenotypes/functions.

Main Results:

• Deep learning models accurately predicted cell stiffness ranges for MSCs and macrophages in situ.
• The models demonstrated high throughput and high sensitivity in stiffness evaluation.
• Successful application of the models to assess diverse cellular functions and phenotypes.

Conclusions:

• Image-based deep learning offers a powerful, non-invasive, and high-throughput method for single-cell mechanical property assessment.
• This approach enables detailed evaluation of cell functions and diversity, advancing mechanobiology.
• The developed models hold significant potential for future cell-based research and clinical translation.