Uncertainty quantification and estimation in differential dynamic microscopy.

Area of Science:

• Physics
• Materials Science
• Biophysics

Background:

• Differential dynamic microscopy (DDM) combines microscopy and scattering for dynamical property analysis.
• Current DDM methods face challenges with computational cost and robustness, limiting routine adoption.
• Accurate characterization of spatiotemporally correlated systems is crucial in various scientific fields.

Purpose of the Study:

• To develop a more computationally efficient and robust DDM analysis method.
• To quantify noise and reduce computational complexity in DDM.
• To validate the new method against existing techniques like conventional DDM and multiple particle tracking.

Main Methods:

• Statistical analysis to quantify image noise and its impact on DDM estimators.
• Fourier analysis to study bias in model parameter estimation and detect negligible bias.
• Gaussian process regression (GPR) for efficient prediction of image structure functions, reducing computational load.

Main Results:

• DDM with uncertainty quantification (DDM-UQ) significantly reduces computational cost (0.5%-5% of Fourier transforms).
• The DDM-UQ method demonstrates enhanced robustness and accuracy compared to conventional DDM.
• Validation through simulations and experiments confirms the efficiency and reliability of DDM-UQ.

Conclusions:

• DDM-UQ offers a foundation for new DDM applications and high-throughput characterization.
• The developed statistical framework improves the reliability and accessibility of DDM analysis.
• This approach enables broader adoption of DDM as a routine characterization tool for complex systems.