Session: IMECE Undergraduate Research and Design Exposition
Paper Number: 119888
119888 - A Deep Learning Semantic Segmentation Approach to Investigate Organic Fouling on Thermal Bubble-Driven Micro-Pumps
Microfluidics has the potential to revolutionize healthcare by providing point-of-access care to regions lacking healthcare infrastructure. However, before any microfluidic technology can be applied to healthcare, the biostability and biocompatibility must first be assessed. Thermal bubble-driven micro-pumps are an emerging micro-pump technology that can be integrated directly into micro/mesofluidic channels, has no moving parts, and leverages existing semiconductor mass production infrastructure. Physically, thermal bubble-driven micropumps are thin film, high-power resistors that use rapid (< 5 us) electrical pulses to locally boil fluid over the resistor’s surface generating a vapor bubble that performs mechanical work. The ability to directly integrate these micro-pumps inside of micro/mesofluidic channels and leverage existing mass production infrastructure makes these micro-pumps well-suited to enable true “lab-on-a-chip” technologies. However, little is known about these micro-pump’s biostability nor biocompatibility with biofluids such as blood or protein-rich fluids like albumen. In this study, we assess the biostability of thermal bubble-driven micro-pumps by analyzing organic fouling through a deep learning semantic segmentation network.
Characterization of organic fouling on thermal bubble-driven micro-pumps is arduous due to the fact that imaging takes place in a particle-laden fluid of albumen or blood which obscures the vapor bubble. Additionally, it can be difficult to distinguish between the vapor bubble and regions of fouling making conventional image processing algorithms infeasible. Past work has utilized manual segmentation but such an approach is tedious and fails to make use of the large amount of data collected. As such, this work seeks to utilize and validate a deep learning semantic segmentation network based on transfer learning of RESNET-18 to identify vapor regions from stroboscopic high-speed images of the resistor surface. We trained the network based on manually labeled datasets to serve as a baseline. Once trained, we used the network to characterize fouling across a range of concentrations and biofluids; specifically, we utilized diluted bovine blood (10% - 100% blood by weight in PBS solution) and albumen (1-5g/100 mL dissolved in PBS solution). After characterizing the biostability of thermal bubble-driven micro-pumps, mitigation techniques were studied in order to prevent surface fouling. To mitigate fouling, we used various surface treatments such as hydrophobic coatings, plasma treating, and PVD sputtered barrier layers to study their effect on organic fouling of these high-power, thin film resistors. The inclusion of surface coatings or barrier layers that repel fouling agents would offer a solution to organic fouling on thermal bubble-driven micro-pumps in the presence of biofluids. Ultimately, this would allow thermal bubble-driven micro-pumps to be utilized in biological applications which is the first step in enabling lab-on-a-chip systems with this technology.
Presenting Author: Janeth Marquez Rubio University of Colorado Boulder
Presenting Author Biography: Janeth is an undergraduate student at the University of Colorado Boulder majoring in Biomedical Engineering. Her focus is on research in the Biomedical and Mechanical Engineering fields. Her current research project is focused on the characterization and mitigation of organic fouling in thermal bubble-driven micropumps. Her current interests are in reducing medical device rejection.
Authors:
Janeth Marquez Rubio University of Colorado BoulderBrandon Hayes University of Colorado Boulder
Robert Maccurdy University of Colorado Boulder
Cillian Murphy University of Colorado Boulder, University College Dublin
A Deep Learning Semantic Segmentation Approach to Investigate Organic Fouling on Thermal Bubble-Driven Micro-Pumps
Paper Type
Undergraduate Expo