Smart AI-Powered Detection in Digital Microfluidics: Active Matrix Electrowetting-on-Diel

Published Date::27/09/2024

Researchers develop an AI-powered system for multipurpose smart detection in active-matrix electrowetting-on-dielectric digital microfluidics.

To generate and manipulate submicron-litre biosamples, powerful tools that are easy to operate, accurate, and multifunctional are needed. To date, different technology platforms have been developed, including flow cytometry, microwell microfluidics, microdroplet microfluidics, optical tweezers, and digital microfluidics (DMF).


The advantages of DMF systems over other platforms are that they can realize sample separation, real-time manipulation, and parallel in situ analyses, all while enabling the simultaneous manipulation of biosamples on a two-dimensional surface. The high-throughput sample generation process demands a large number of electrodes driving droplets on a DMF chip.


However, passive-matrix (PM) EWOD systems typically accommodate fewer than 200 electrodes, as each PM electrode is physically connected to a peripheral connector. The large number of associated connection lines limits the scalability of electrodes, posing a challenge in DMF research and development work.


To address this issue, researchers have adopted active-matrix (AM) addressing, in which each pixel contains active transistors that act as switches and can be independently addressed by row and column driver lines. Several studies have been conducted using AM-DMF technology for molecular diagnosis, proteomics analysis, high-resolution concentration gradient preparation, and parallel single-cell manipulation tasks.


AM-EWOD technology can generate and analyze thousands of discrete droplets surrounded by filler oil in parallel. However, determining and screening samples pose significant challenges for researchers, where manually selecting samples is an inefficient and inconsistent strategy.


Traditional image processing techniques are mostly tailored for specific scenarios. They greatly rely on predefined distributions or manually designed features, which results in limited generalizability. In recent decades, deep learning (DL) has made significant technological progress and has shown great potential for use as a powerful tool in an AM-EWOD system for multipurpose smart detection.


The applications of DL in droplet-based microfluidics are becoming increasingly widespread. However, a limited number of studies have reported applications of DL technology in the AM-DMF field.


Therefore, in this work, we addressed the automated biosample selection and determination problems on an AM-EWOD platform by using a range of DL models for different tasks.


The phenomenon of “necking” occurs during droplet splitting, and the length of the droplet neck affects the homogeneity of the volumes of the split droplets. To achieve uniformity during droplet splitting, it is essential to ensure uniform control of the electrodes on both sides.


However, it is challenging for an AM-DMF system to achieve highly uniform control of large-scale pixel electrodes because of the utilization of thin-film transistors (TFTs) as driving switches. Therefore, we employed an image recognition method to assess droplet uniformity, allowing us to select droplets according to our needs under appropriate conditions, thereby compensating for the limitations imposed by the inherent challenges of the utilized system.


High-throughput droplet generation provides extensive parallel data for biological analyses. However, a trade-off exists between the success rate of droplet splitting and the number of droplets that can be generated and controlled per unit area in parallel.


Therefore, we propose a DL-based high-throughput droplet recognition method to iteratively design droplet-splitting paths and swiftly find an optimal solution. Furthermore, the method enables real-time monitoring and control of the droplet generation process, allowing for improved efficiency and accuracy in biosample analysis.


In this work, we demonstrated the potential of AI-enabled smart detection in active-matrix electrowetting-on-dielectric digital microfluidics for multipurpose biosample analysis. The proposed DL-based method enables real-time monitoring and control of the droplet generation process, allowing for improved efficiency and accuracy in biosample analysis.


Active-matrix electrowetting-on-dielectric (AM-EWOD) is a technology that enables the manipulation of droplets on a two-dimensional surface using electrodes. The technology has been widely used in various fields, including biosample analysis, drug discovery, and materials science.


FAQS:

Q: What is active-matrix electrowetting-on-dielectric (AM-EWOD) technology?

A: AM-EWOD is a technology that enables the manipulation of droplets on a two-dimensional surface using electrodes.


Q: What are the advantages of DMF systems over other platforms?

A: The advantages of DMF systems over other platforms are that they can realize sample separation, real-time manipulation, and parallel in situ analyses, all while enabling the simultaneous manipulation of biosamples on a two-dimensional surface.


Q: What is the challenge in achieving uniform control of large-scale pixel electrodes in AM-DMF systems?

A: The challenge in achieving uniform control of large-scale pixel electrodes in AM-DMF systems is due to the utilization of thin-film transistors (TFTs) as driving switches.


Q: What is the proposed DL-based method for high-throughput droplet recognition?

A: The proposed DL-based method enables real-time monitoring and control of the droplet generation process, allowing for improved efficiency and accuracy in biosample analysis.


Q: What is the potential of AI-enabled smart detection in active-matrix electrowetting-on-dielectric digital microfluidics?

A: The potential of AI-enabled smart detection in active-matrix electrowetting-on-dielectric digital microfluidics is to enable real-time monitoring and control of the droplet generation process, allowing for improved efficiency and accuracy in biosample analysis.


More Topics :