Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.
Non-WSU users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Access Type
WSU Access
Date of Award
January 2023
Degree Type
Dissertation
Degree Name
Ph.D.
Department
Electrical and Computer Engineering
First Advisor
Amar Basu
Abstract
Microfluidics is a technology that enables the precise control and manipulation of liquid at the nanoliter to picoliter scale. “Lab-on-a-chip” systems enabled by microfluidics have tremendous potential for applications in biology, chemistry, and medicine. One key function of microfluidics is the analysis of particles, cells, or droplets in a liquid sample, such as: cells in a blood sample, microorganisms in environmental samples, or droplet microreactors in single cell assays. Protocols requiring the analysis of a large number of droplets, cells, or particles (104-106 events) typically use laser-induced fluorescence (LIF) flow cytometry, which provides high throughput, but lacks spatially-resolved measurements. In contrast, microscopic imaging of samples can provide spatial information about their contents, but traditional manual microscopy approaches have low throughput. By automating image inspection and classification using computer vision (CV), high-resolution data can be obtained while maintaining the throughputs needed to analyze large populations. Past efforts to develop computer vision-based analysis have included MATLAB-based image processing systems, C++-based image processing using only CPUs, machine learning-based analysis using GPUs, systems using time delayed integration (TDI) of images, and other novel image acquisition techniques that deliver high-throughput. Thus far, these methods can provide either high speeds with low resolution or high resolution with low speeds. High speeds enable real-time analysis and sorting, while high resolution allows for better distinction between populations. This research develops a high-throughput droplet, cell, and particle analysis algorithm implemented on the GPU, achieving speeds of approximately 3000 frames per second or 7500 events per second at about 1 MP, making real-time analysis possible at high120 resolution. Real-time analysis can be applied to measurements for digital assays, quantifying cell encapsulation, shape-based chemical detection, analyses of aquatic samples for organic and inorganic objects, liquid biopsy, and physical sorting of samples. The algorithm was combined with a machine-vision camera to provide real-time analysis at ~300 frames per second or 1200 events per second at 700 X 500 resolution, limited by camera throughput. We addressed software rate limiting step of a serial contour finding algorithm by replacing with a modified GPU-based contour tracing algorithm which improved our throughput from 3000 to 3200 frames per second or 8000 events per second at 1 MP resolution. We also addressed a hardware-based rate limiting step of transferring data to and from GPU by using a modular GPU board with a unified memory. This brought the transfer time from 0.7 ms to 0.1 ms, and the overall system latency down to 8ms. Combined with piezoelectric actuators for sorting, the system was able to perform brightfield image-activated sorting of samples at 56 events per second at a resolution of 0.5 MP. At a total cost of <$1,000, our prototype system shows the potential to democratize access to image-activated sorting, making its versatile and programmable sorting capabilities available for a wide range of applications in medicine, environmental applications, and manufacturing.
Recommended Citation
Vedhanayagam, Arpith, "Gpu Accelerated Imaging Flow Cytometry And Real Time Sorting" (2023). Wayne State University Dissertations. 3936.
https://digitalcommons.wayne.edu/oa_dissertations/3936