Files and links (1)
CC BY V4.0, Open Access
Thesis
Deterministic Lateral Displacement Particle Detection Tool Utilizing Machine Vision and Machine Learning
Washington State University
Master of Science (MS), Washington State University
01/2021
DOI:
https://doi.org/10.7273/000003354
Handle:
https://hdl.handle.net/2376/122214
Abstract
Deterministic Lateral Displacement (DLD) is a microfluidic method of separatingparticles in a solution, with one goal being the separation of circulating tumor cells
from blood. With a properly designed DLD for high-throughput flow, it is possible for
the DLD to be effective for particle separation while optimizing the time it requires
to process a sample. DLD is in its research and development stage, with much of the
analysis simulated or completed manually. There was a need to reduce the time of
analysis from hours with manual techniques, to less than a minute using automated
software analysis.
The Particle Detection Automation Tool provided initial developmental efforts for
a program that will reduce the analysis time of DLD videos in an easy-to-use manner.
For the basis of the tool, a reliable method of particle detection had to be created.
This was accomplished through Python and its available packages in machine vision
techniques such as probabilistic Hough transforms, Canny edge detection, and existing
tools such as OpenCV’s SimpleBlobDetector. The program development began with
detection methods such as simple thresholding, but added a GUI and was able to
detect particles consistently and with a high accuracy. Version 1 of the program had
an overall detection accuracy of 78.45%. After improvements to the program, the
final version of program, Version 4, the overall particle detection accuracy improved
to 96.84%. The particle distribution determination still needs to be improved.
Three machine learning techniques, Complement Naive Bayes, K-Nearest Neighbors,
and Support Vector Machines Radial Basis Function kernel, were implemented
and compared in the determination of the DLD mode. More data needs to be acquired,
analyzed, and labeled to have a larger dataset when re-training the machine
learning models to avoid over-fit machine learning models. There is still much work to
be done with improving the current version of the program and expanding the machine
learning analysis, while developing additional functions that enhance the capabilities
of the particle detection automation tool. However, it can be said that the program
presented here provides a solid basis for future development and implementation in
automated analysis.
Metrics
11 File views/ downloads
29 Record Views
Details
- Title
- Deterministic Lateral Displacement Particle Detection Tool Utilizing Machine Vision and Machine Learning
- Creators
- Eric Anthony Gioe
- Contributors
- Linda (Xiaolin) Chen (Advisor)Mandana Bozorgi (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Engineering and Computer Science (VANC), School of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University
- Number of pages
- 370
- Identifiers
- 99900652204701842
- Language
- English
- Resource Type
- Thesis