Particle sorting and automatic particle identification for advanced medical diagnostics

Rodriguez Luna, Juan Carlos (2018) Particle sorting and automatic particle identification for advanced medical diagnostics. PhD thesis, University of Glasgow.

Abstract

The physical separation of micro-particles is very important in many research field as diverse as chemistry and medicine. The main goal of the current separation techniques is to extract micro-particles such as cells at a high processing rates and purity. Chromatography, for instance, is commonly applied for the detection and enrichment of pathogens, which is useful for the medical diagnostics of parasitic infections. Many separation techniques have been developed over the years, applying physical phenomena of different kinds and/or taking advantage of unique physical properties of the particles themselves.
From all of these techniques, one that has remained popular over the years is Dielectrophoresys(DEP). One of the main reasons for its popularity is that it does not require markers of any kind; it takes advantage of differences in the particle’s polarizability, size and shape. Another distinctive characteristic of dielectrophoresis is its selectivity due to its capacity to be controlled using frequency and voltage amplitude and its suitability for small microfluidic systems. In very general terms the work I have done during my PhD studies was oriented towards the development of novel and robust technology for aiding in the micro-particle sorting and bio-particle recognition by using computer tools. The ideas and concepts I will be introducing throughout this document were allowed total freedom to evolve and change to better fulfill the main goals of the project and also to better adapt to the many technical challenges I had to face during my research. As well as developing a new dielectrophoresis method I have also tried to maximize the impact of this work by doing it in a truly accessible way for anyone, regardless if they are interested in basic research, a possible application or just looking to adapt this concepts and tools for a different purpose.

The central work in this PhD thesis focuses on two main topics:

- Computerized bio-particle tracking and identification using a machine-learning algorithm that incorporates a number of predictors, including colour histogram comparison.
- A portable dielectrophoresis(DEP) electronic device able to tailor the potential across a microfluidic channel for particle separation.

The first project is about computerized vision system designed to track and identify micro-particles of interest through the use of video microscopy, machine learning and other video processing tools. This system uses a novel particle recognition algorithm to improve specificity and speed during the tracking and identification process. We show the detection and classification of different types of cells in a diluted blood sample using a machine-learning algorithm that makes use of a number of predictors, including shape and color histogram comparison. This software can be considered as a stand alone piece
of work. Its open source nature makes it ideal for scientific purposes or as a starting point for a different application. In the context of this PhD thesis, however, it is an invaluable tool for validating and quantifying experimental results obtained from the micro-particle separator experiments presented in Chapter 4.
The central piece of work in this PhD thesis is introduced in Chapter 4. This project is about the development of a all-in-one continuous flow DEP based microparticle separator which uses a system of individually addressable electrodes to shape and control the particle’s potential energy profile across the entirety of a microfluidic channel. These tailored potential landscapes are created by averaging the electric field generated by 64 individual electrodes, where the electronic device has complete control over each electrode’s on/off
state, frequency, AC voltage amplitude and pulse duration. All the characteristics of the potential landscapes are controlled wirelessly through a mobile phone application. These specially designed potential landscapes allow us to make lateral sorting and/or concentration of a binary mixture of particles at the same time they move through a microfluidic channel; all this without the need for buffer flows or additional external forces. One of the outstanding characteristics of this new sorting technique is that it relays exclusively on negative DEP. Most previous techniques require a combination of positive and negative DEP and possibly and external force of different nature to achieve particle sorting; all of which requires the use of a crossover frequency and hence a careful control of the conductivity of the suspending medium. Here by using only negative DEP we eschew the careful control over the conductivity of the suspending medium and the use of any other external force; all this contributes to make our device small and robust. In addition to this, our electronic device was designed to include all the supporting electronics it needs in a small and robust printed circuit board that can also be operated by batteries. We present simulation results to illustrate the physics behind this new technique along with experimental results demonstrating the separation of polystyrene beads.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: This thesis was supported by CONACYT (Consejo Nacional de Ciencia y Tecnologia).
• Keywords: Machine learning, support vector machine, image recognition, particle classification, openCV, DEP, dielectrophoresis, particle sorting, portable device, tailoring the potential, microfluidics.
• Subjects: Q Science > Q Science (General); T Technology > T Technology (General)
• Colleges/Schools: College of Science and Engineering > School of Engineering > Biomedical Engineering
• Supervisor's Name: Neale, Dr. Steven and Cooper, Professor Jonathan
• Date of Award: 2018
• Depositing User: Mr. Juan Carlos Rodriguez Luna
• Unique ID: glathesis:2018-9072
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 08 May 2018 09:04
• Last Modified: 13 Jul 2018 07:32
• URI: https://theses.gla.ac.uk/id/eprint/9072