DeepDDM: A Compact Deep-Learning Assisted Platform for Micro-Rheological Assessment of Micro-Volume Fluids

Abstract

An emerging differential dynamic microscopy technique has been successfully used for quantitative dynamic investigation of micro-particle suspension, leading to a rheological assessment of the solution. This technique exploits an optical microscope equipped with a digital camera for the assessment. However, the accessible measurement ranges at high frequencies are limited by the video frame rate of the camera, resulting in a limitation in investigating distinguish responses at the high-frequency region. With advanced deep learning technology, image-synthesizing deep learning-based algorithms can significantly increase the video frame rate, producing additional in-between frames. As a result, the rheological responses at the high-frequency region can be obtained. To address this problem, a video frame interpolation integrated differential dynamic microscopy-based device (DeepDDM platform) was developed. Our DeepDDM platform interpolates video frames to extend the maximum measuring angular frequency up to quadruple from 30.1 rad/s to 123.0 rad/s, resulting in a more comprehensive rheological assessment without hardware modification. Unlike reducing the camera exposure time approach, our approach requires only a single camera and works without brightness reduction. Furthermore, the device is compact and portable. It comprises a few main components, and requires only $8 \mu \text{L}$ sample volumes for the rheological assessment. Thus, it is easy to relocate to measure biological samples which are often do not retain their natural properties in a storage allowing for in situ studies of the fluids. In comparison, the obtained responses agreed with the reference mechanical rheometer, although the employed partially coherent source and out-of-focus image acquisition bring difficulties to our system.

Keywords

• Differential dynamic microscopy
• rheology
• deep learning
• frame interpolation