Журнал наномедицины и нанотехнологий

Абстрактный

Emergence of Data-Driven Microfluidics.

Takasaki Yumi

Microfluidic devices currently rely on researchers to select the initial operating conditions empirically, then monitor and modify parameters throughout studies to achieve and maintain stable conditions that produce repeatable data. Multiple factors can cause inconsistent experimental conditions in polydimethylsiloxane-based microfluidic devices, including fabrication flaws, clogging, bubble formation, chemical impurities, or long-term effects like temperature and pressure fluctuations, surface fouling, and substrate deformation. Because of the adsorption of hydrophobic small molecules, these differences can alter chemical synthesis by changing solution concentrations or biological analyses. In this section, we discuss new research on the application of machine intelligence in microfluidic chips in conjunction with optical microscopy. This method has been used to forecast droplet size and stability, as well as to assess droplet chemical composition and fluid characteristics, as well as to monitor and correct flow rates and droplet sizes to avoid undesirable consequences such as bubble formation. Multiple inputs are converted into low-dimensional feature representations (codes) using autoencoders, which can then be processed and decoded to reconstruct the inputs.