A method for real-time mechanical characterisation of microcapsules.

Area of Science:

• Biophysics
• Materials Science
• Computational Science

Background:

• Characterizing microcapsule mechanical properties is crucial for fundamental understanding and applications.
• Existing methods may lack the speed and accuracy required for real-time analysis.

Purpose of the Study:

• To develop a novel machine learning approach for simultaneous, real-time prediction of microcapsule mechanical properties.
• To achieve high prediction accuracy with unprecedentedly low latency.

Main Methods:

• A multilayer perceptron (MLP)-based machine learning model was developed.
• The model predicts membrane mechanical law type, shear, and area-dilatation moduli from steady profiles in tube flow.
• The approach was validated using both simulation and experimental data.

Main Results:

• The MLP-based method achieved high prediction accuracy.
• Prediction latency was less than 1 millisecond on a personal computer.
• The approach is two orders of magnitude faster than convolutional neural network methods by analyzing 1D boundaries.

Conclusions:

• The developed ML approach enables rapid, real-time mechanical characterization of microcapsules.
• This method offers a foundation for advanced tools in active sorting of deformable microcapsules and biological cells in microfluidics.