High Shear Arterial Thrombosis: Microfluidic Diagnostics and Nanotherapeutics

(Georgia Institute of Technology, 2019-12-20) Griffin, Michael T.

Ischemic cardiovascular events remain the leading causes of death in the world, largely due to ineffective preventative therapies and diagnostic tools. This work investigated the development of a physiologically relevant, low-variability microfluidic thrombosis assay (MTA) capable of screening therapy efficacy. First, an experimental design was implemented to assess the effects of geometry, collagen surface coverage, and anticoagulant selection on MTA occlusion time (OT) variability. It was found that better control of shear rates through novel grayscale lithography techniques decreased OT variability. Fibrillar collagens was also found to have a significant impact. The MTA was then implemented to study the effects of current antiplatelet therapies, aspirin and Plavix, as compared to the endpoints of other platelet function tests (PFTs). It was found that aspirin use significantly increased MTA OT but did not prevent occlusion in the MTA. Results from Plavix use found a stronger response, where 20% of patients had complete OT inhibition. Comparison with other PFTs found that the MTA more closely matched the rates of ischemic events from larger clinical studies. Finally, the MTA was utilized to assess a nanoparticle therapy hypothesized to function through biophysical mechanisms. It was found that negatively charged nanoparticles were more effective than neutral or positively charged nanoparticles. The antithrombotic effect of charged nanoparticles persisted even with different base materials, but the effects of nanoparticle size were not consistent between materials. A mouse bleeding model was also used to show that hemostasis was maintained with the nanoparticle therapy. The implications of all results for clinical diagnostic and future antithrombotic therapy research are discussed.
Utilizing switched linear dynamics of interconnected state transition devices for approximating certain global functions

(Georgia Institute of Technology, 2019-12-17) Parihar, Abhinav

The objective of the proposed research is to create alternative computing models and architectures, unlike (discrete) sequential Turing machine/Von Neumann style models, which utilize the network dynamics of interconnected IMT (insulator-metal transition) devices. This work focusses on circuits (mainly coupled oscillators) and the resulting switched linear dynamical systems that arise in networks of IMT devices. Electrical characteristics of the devices and their stochasticity are modeled mathematically and used to explain experimentally observed behavior. For certain kinds of connectivity patterns, the steady state limit cycles of these systems encode approximate solutions to global functions like dominant eigenvector of the connectivity matrix and graph coloring of the connectivity graph.
Optimization-based methods for deterministic and stochastic control: Algorithmic development, analysis and applications on mechanical systems & fields

(Georgia Institute of Technology, 2019-12-17) Boutselis, Georgios

Developing efficient control algorithms for practical scenarios remains a key challenge for the scientific community. Towards this goal, optimal control theory has been widely employed over the past decades, with applications both in simulated and real environments. Unfortunately, standard model-based approaches become highly ineffective when modeling accuracy degrades. This may stem from erroneous estimates of physical parameters (e.g., friction coefficients, moments of inertia), or dynamics components which are inherently hard to model. System uncertainty should therefore be properly handled within control methodologies for both theoretical and practical purposes. Of equal importance are state and control constraints, which must be effectively handled for safety critical systems. To proceed, the majority of works in controls and reinforcement learning literature deals with systems lying in finite-dimensional Euclidean spaces. For many interesting applications in aerospace engineering, robotics and physics, however, we must often consider dynamics with more challenging configuration spaces. These include systems evolving on differentiable manifolds, as well as systems described by stochastic partial differential equations. Some problem examples of the former case are spacecraft attitude control, modeling of elastic beams and control of quantum spin systems. Regarding the latter, we have control of thermal/fluid flows, chemical reactors and advanced batteries. This work attempts to address the challenges mentioned above. We will develop numerical optimal control methods that explicitly incorporate modeling uncertainty, as well as deterministic and probabilistic constraints into prediction and decision making. Our iterative schemes provide scalability by relying on dynamic programming principles as well as sampling-based techniques. Depending upon different problem setups, we will handle uncertainty by employing suitable concepts from machine learning and uncertainty quantification theory. Moreover, we will show that well-known numerical control methods can be extended for mechanical systems evolving on manifolds, and dynamics described by stochastic partial differential equations. Our algorithmic derivations utilize key concepts from optimal control and optimization theory, and in some cases, theoretical results will be provided on the convergence properties of the proposed methods. The effectiveness and applicability of our approach are highlighted by substantial numerical results on simulated test cases.
Advances in online convex optimization, games, and problems with bandit feedback

(Georgia Institute of Technology, 2019-12-16) Rivera Cardoso, Adrian

In this thesis we study sequential decision making through the lens of Online Learning. Online Learning is a very powerful and general framework for multi-period decision making. Due to its simple formulation and effectiveness it has become a tool of daily use in multibillion companies. Moreover, due to its beautiful theory and its tight connections with other fields, Online Learning has caught the attention of academics all over the world and driven first-class research. In the first chapter of this thesis, joint work with Huan Xu, we study a problem called: Risk-Averse Convex Bandit. Risk-aversion makes reference to the fact that humans prefer consistent sequences of good rewards instead of highly variable sequences with slightly better rewards. The Risk-Averse Convex Bandit addresses the fact that, while human decision makers are risk-averse, most algorithms for Online Learning are not. In this thesis we provide the first efficient algorithms with strong theoretical guarantees for the Risk-Averse Convex Bandit problem. In the second chapter, joint work with Rachel Cummings, we study the problem of preserving privacy in the setting of online submodular minimization. Submodular functions have multiple applications in machine learning and economics, which usually involve sensitive data from individuals. Using tools from Online Convex Optimization, we provide the first $\epsilon$-differentially private algorithms for this problem which are almost as good as the non-private versions for this problem. In the third chapter, joint work with Jacob Abernethy, He Wang, and Huan Xu, we study a dynamic version of two player zero-sum games. Zero-sum games are ubiquitous in economics, and central to understanding Linear Programming Duality, Convex and Robust Optimization, and Statistics. For many decades it was thought that one could solve this kind of games using sublinear regret algorithms for Online Convex Optimization. We show that while the previous is true when the game does not change with time, a naive application of these algorithms can be fatal if the game changes and the players are trying to compete with the Nash Equilibrium of the sum of the games in hindsight. In the fourth chapter, joint work with He Wang and Huan Xu, we revisit the decade old problem of Markov Decision Processes (MDPs) with Adversarial Rewards. MDPs provide a general mathematical framework for sequential decision making under uncertainty when there is a notion of `state', moreover they are the backbone of all Reinforcement Learning. We provide an elegant algorithm for this problem using tools from Online Convex Optimization. The algorithm's performance is comparable with current state of the art. We also consider the problem under the large state-space regime, and provide the first algorithm with strong theoretical guarantees.
Development of ultrasound and photoacoustic imaging tools for tracking of cells and particles

(Georgia Institute of Technology, 2019-12-16) Kubelick, Kelsey P.

Development of novel therapies relies heavily on histology to evaluate outcomes. While histology provides detailed information at the molecular level, drawbacks include highly invasive, destructive sample preparation. To this end, clinical translation of novel therapies would be expedited by further developing minimally invasive, longitudinal imaging methods to inform therapy design or provide clinical feedback. Combined ultrasound (US) and photoacoustic (PA) imaging augmented with contrast agents is an excellent option to address this need. This research describes development of a US/PA imaging toolbox, consisting of contrast agents, imaging protocols, imaging hardware, and detection algorithms, that can be tailored for a variety of applications where longitudinal, in vivo imaging of specific cells or particles is desired. To demonstrate versatility, these US/PA imaging tools were developed and combined in different ways for implementation in three distinct applications: 1) stem cell monitoring in ophthalmology to aid development of glaucoma therapies; 2) intra- and post-operative monitoring to guide stem cell therapies of the spinal cord; and 3) monitoring particle trafficking to the lymph node to inform vaccine design. Although the applications investigated here were extremely different, common themes were identified, highlighting broad relevance of the US/PA imaging toolbox and common opportunities for later development. Overall, the tools developed here lay the foundation for design of custom US/PA imaging platforms in the future.
ORGANIC CONTAMINANTS DESTRUCTION USING THE UV/FREE CHLORINE PROCESS: MECHANISMS AND MODELING

(Georgia Institute of Technology, 2019-12-16) Zhang, Weiqiu

Advanced oxidation processes (AOPs) are effective technologies to oxidize recalcitrant organic contaminants in the aqueous phase. The UV/free chlorine process has gained attention as a promising AOP technology, and it generates various reactive radicals (i.e. HO∙, Cl∙, Cl2-∙ and ClO∙) at room temperature and pressure. These electrophilic radicals eventually mineralize refractory organic contaminants into CO2 and H2O. Compared with other common AOPs (e.g. UV/H2O2 and UV/Persulfate processes), the UV/free chlorine process has many advantages, for example (1) it has much lower chemical reagent costs; (2) it has higher energy efficiency; (3) it is only slightly impacted by chloride ions (Cl-) (We found Cl- significantly inhibits the effectiveness of the UV/Persulfate process). For large scale applications, understanding the degradation mechanisms is critical to the design of the UV/free chlorine process that has the lowest energy consumption and greatest toxicity reduction. A number of related studies have shed light on the degradation of some selected organic compounds (e.g., atrazine, naproxen, etc.). However, these previous studies of the UV/free chlorine process have not comprehensively examined the mechanistically complex radicals-initiated chain reactions. Many researches have conducted experiments to determine the degradation mechanisms. However, these experimental studies are very time consuming and expensive. With respect to developing kinetic models that can simulate the reaction pathways in the UV/free chlorine process, most studies have used simplified lumped reactions or invoked the simplified pseudo steady state assumption because the rate constants between reactive radicals and organic compounds are unknown. Accordingly, conducting experiments and developing simplified kinetic models would be impossible to fully elucidate the oxidation mechanisms of all organic contaminants that may be found in the aqueous phase (Chemical Abstracts Service lists about more than 147 million compounds). To overcome the above-mentioned challenges, we developed a first principles-based kinetic model to predict the oxidation of organic compounds in the UV/free chlorine process. First, we collected photolysis and chemical reactions that describe the oxidation of target organic compounds from literature. Second, we developed a rate constants estimator to predict the rarely reported second-order rate constants between reactive radicals and organic compounds (i.e. kHO∙/R, kCl∙/R, kCl2-∙/R and kClO∙/R). kHO∙/R was estimated by the group contribution method (GCM). kCl∙/R, kCl2-∙/R and kClO∙/R were estimated by using the genetic algorithm that was fit to our experimental data (i.e. experimental observed time-dependent concentration profiles of target organic compounds). Third, we developed a stiff ordinary differential equations solver using Gear’s method to predict the time-dependent concentration profiles of target organic compounds, and our prediction results agreed with our experimental data for various operational conditions. Accordingly, our first principles-based kinetic model was successfully verified using our experimental data. Based on our UV/free chlorine kinetic model, we developed four quantitative structure activity relationships using Hammett constants of organic compounds and our predicted rate constants. We then determined relative contribution of these reactive radicals and photolysis, and, we found ClO∙ was the dominant radicals for organic contaminants oxidation. We also optimized the operational conditions (i.e. UV intensity and free chlorine dosage) that has the lowest energy consumption. Furthermore, we successfully implemented graph theory to develop a computerized pathway generator, which was built based on the predefined reaction mechanisms from experimental observations. The pathway generator can automatically predict all possible reactions and byproducts/intermediates that are involved in the degradation of target organic contaminants during the UV/free chlorine process (e.g. the degradation of TCE involves more than 200 byproducts /intermediates and more than 1,000 reactions). Therefore, the pathway generator significantly advances our understanding about the degradation pathways. However, we have noticed that it is difficult to estimate the rate constants of all possible involved reactions at current stage, because we only have very limited amount of experimental data (e.g., we do not have data on peroxyl radicals reactions) to develop a GCM. Consequently, future work will mainly focus on developing new methods (e.g. quantum chemistry) to estimate the rate constants of all possible involved reactions, and then predicting the time-dependent concentration profiles of byproducts. Finally, we investigated the disinfection byproducts (DBPs) and disinfection byproducts formation potentials (DBPFPs) in the UV/free chlorine process. In practical applications, natural organic matter can react with residual free chlorine to produce toxic DBPs. As a result, both the micropollutants and the DBPFPs must be decreased. Therefore, we need determine the controlling factor (i.e., organic contaminant destruction or DBPFPs reduction) in the design of the UV/free chlorine system. Overall, our study can be used to design the most cost-effective UV/free chlorine process.
Transformation elasticity and anelasticity

(Georgia Institute of Technology, 2019-12-12) Golgoon, Ashkan

We present a theoretical framework for studying a large class of elastic and anelastic problems in nonlinear solids. We specifically use the transformation properties of nonlinear and linearized elasticity in this theory. Given an anelastic deformation, a non-vanishing strain does not correspond to a non-vanishing stress. That part of strain that is related to the corresponding stress is called elastic strain. The remaining part of strain is called eigenstrain or pre-strain. Eigenstrains (or anelastic sources) such as inclusions, defects, growth, phase transformations, and nonuniform temperature changes can cause residual stresses. The relaxed (natural) configuration of a residually-stressed body is a non-Euclidean manifold that cannot be isometrically embedded in the Euclidean ambient space. Using transformation anelasticity, one can construct the Riemannian material manifold of the body. In particular, the material metric explicitly depends on the distribution of eigenstrains. In this PhD thesis we utilize transformation anelasticity to study the induced elastic fields of a circumferentially-symmetric distribution of finite eigenstrains in nonlinear elastic wedges; the stress field of a nonlinear elastic solid torus with a toroidal inclusion; nonlinear elastic inclusions in anisotropic solids as well as distributed line and point defects in nonlinear anisotropic solids. The goal in transformation elasticity is to transform the nonlinear or linearized boundary-value (or initial-boundary-value) problem of an elastic body to that of another elastic body using a diffeomorphism (or a smooth mapping). The diffeomorphism, in turn, explicitly determines how the different elastic fields (and elastic parameters) of the two bodies are related. In particular, it is noted that the two boundary-value problems are not related by push-forward or pull-back under the diffeomorphism. We apply this theory to formulate the nonlinear and linearized elastodynamic transformation cloaking problem in the context of the classical elasticity, the small-on-large theory of elasticity, i.e., linearized elasticity with respect to an initially stressed configuration, and in solids with microstructure, namely gradient and (generalized) Cosserat solids. In particular, we note that a cloaking transformation is neither a spatial nor a referential change of coordinates (frame). Rather, a cloaking map transforms the boundary-value problem of an isotropic and homogeneous elastic body (virtual problem) to that of an anisotropic and inhomogeneous elastic body with a finite hole covered by a cloak that is to be designed (physical problem). The virtual body has a desired mechanical (wave-guiding) response, whereas the physical body is designed such that the same response is mimicked outside the cloak using a cloaking transformation. Finally, starting from nonlinear shell theory, we utilize transformation elasticity to formulate the transformation cloaking problem for Kirchhoff-Love plates and for elastic plates with both the in-plane and the out-of-plane displacements.
"Assessment of oxidative potential of ambient water-soluble and insoluble of PM2.5

(Georgia Institute of Technology, 2019-12-11) Gao, Dong

Oxidative stress has been proposed as a major mechanism responsible for adverse health effects associated with particulate matter (PM) pollution. Various methods have been developed to measure PM oxidative potential (OP), the potential for particles to generate reactive oxygen species and elicit oxidative stress. But no consensus has been reached as to the best OP assay. Both water-soluble and insoluble PM components contribute to PM OP, but the water-insoluble OP fraction has been less studied. This dissertation aims to characterize water-soluble PM OP measured by different OP assays and water-insoluble OP in terms of temporal variability and chemical determinants. This dissertation provides a direct inter-comparison between two health-relevant acellular OP assays, the synthetic respiratory tract lining fluid (RTLF) assay and the dithiothreitol (DTT) assay. These assays were used to measure the water-soluble OP of ambient fine PM collected in urban Atlanta over a year-long period. The results showed that these assays were driven by different groups of aerosol species, ranging from organic species to transition metal ions. The OP responses in the RTLF assay were affected by the composition of synthetic lung fluid, which emphasizes the importance of developing a “standard” technique for OP assays. Multivariate regression models for these OP metrics capture interactions among species, expanding our understanding of the relationships among species in the OP assessment. To develop a method for quantifying total PM OP, we compared three commonly used extraction methods for total OP assessment, involving methanol extraction (1) with or (2) without filtering the extracts, followed by solvent removal and reconstitution with water, and (3) water extraction without removing the particle-laden filter. The results indicated that performing the OP assay directly on the water extracts that still contained the particle-laden filter was a more effective way to capture water-insoluble OP compared to organic solvent extraction. An automated system was developed based on the DTT assay to facilitate the total OP analysis. The water-soluble and total OP of ambient particles collected at two urban sites and one roadside site were analyzed, with water-insoluble OP determined by difference. The results clearly demonstrated a measurable OP contribution from water-insoluble PM, which accounted for 20–35 % of total OP. The spatial and temporal variations in OP measures suggested that the insoluble OP contributors were largely secondary and related to biomass burning emissions. Multivariate regression analyses indicated that water-insoluble OP was related to incomplete combustion products and surface properties of soot and water-insoluble metals. Overall, assessing water-insoluble or total PM OP may provide important information in elucidating the health risks related to PM exposure and ultimately in promulgating effective control strategies to protect public health.
Audio classification and event detection based on small-size weakly labeled data

(Georgia Institute of Technology, 2019-12-11) Cheng, Chieh-Feng

The objective of this research is to perform audio event detection and classification using small-size weakly labeled data. Although audio event detection has been studied for years, the research on this topic using weakly labeled data is limited. Many sources of multimedia data lack detailed annotation and rather have only high-level meta-data describing the main content of various long segments of the data. In this research, we illustrate a novel framework to perform audio classification when working with such weakly labeled data, especially when dealing with small-size datasets. Traditional approaches to this problem is to use techniques for strongly labeled data and then to deal with the weak nature of the labels via post-processing. In contrast, our approach directly addresses the weakly labeled aspect of the data by classifying longer windows of data based on the clustering behavior of the acoustic features over time. We evaluate our framework using both synthetic datasets and real data and demonstrate that our method works well under both situations. Also, it outperforms other existing methods when using small size datasets.
Engineering High-Efficiency Adsorption Contactors via 3D Printing of Microporous Polymers

(Georgia Institute of Technology, 2019-12-10) Zhang, Fengyi

Adsorption is a promising energy-efficient separation process, which selectively removes one or several components from a mixture by transporting a fluid through a mass transfer contactor. The most traditional mass transfer contactor design is a packed bed of adsorbent pellets, which suffers from high pressure drop, low mass transfer rate, difficulty in heat integration, etc. State-of-the-art structured mass transfer contactors have been developed to address these problems. For instance, hollow fiber sorbents can achieve rapid temperature manipulation by flowing heat-exchange media through the bore channels, and monoliths provide uniform fluid channels to minimize pressure drop. However, limited by manufacturing techniques, existing structured mass transfer contactors struggle to address all of the aforementioned problems with one structural design. 3D printing techniques can fabricate complex architectures without molding-based approaches, which is suitable for rapid prototyping of novel mass transfer contactor designs. The overarching goal of this thesis is to engineer high-efficiency adsorption contactors via 3D printing of microporous polymers. To achieve this goal, three objectives were established: (1) develop 3D printing techniques that can process adsorptive materials and generate hierarchical porosity, (2) prototype scalable mass transfer contactors with optimized energy efficiency, (3) perform proof-of-concept adsorption experiments to demonstrate the advantages of 3D printing in mass transfer contactor fabrication.