Quantifying dynamics of soft and active matter with microscopy and machine learning

Understanding the dynamics of soft and active materials is crucial for the characterization and development of these materials. Differential dynamic microscopy (DDM) offers a powerful technique to quantify the dynamics, combining principles from optical microscopy and light scattering. By analyzing real-space image intensity fluctuations in a manner similar to dynamic light scattering, DDM helps in quantifying phenomena like thermal diffusion of particles or the dynamics of active gels. However, DDM typically requires analysis of 1000s of imaging frames, often taking 10s of seconds, to produce meaningful results. Analyzing fewer frames typically yields noisy, less accurate results. This can limit DDM analysis to studying systems that are in steady state or evolving slowly in time. We show how employing a machine learning approach to DDM analysis can effectively reduce the required number of frames for accurate analysis and facilitates more high-throughput material screening, monitoring gelation processes with fine temporal resolution, and precise quantification of rapidly evolving active systems.