Electrical Impedance Tomography (EIT) seeks to provide a new modality by which to image portions of the body and tissues which provide differences in conductivity, depending on their state. Unlike bulky, expensive, and hard-to-access traditional medical imaging equipment, such as those used in magnetic resonance imaging and computed tomography, EIT is potentially capable of being contained to small, portable, inexpensive hardware. Here, a Red Pitaya (RP) device is used to provide and process signals that can be generated and multiplexed into 16 channels, all while data is gathered from a set of electrodes embedded into a tank containing electrically conductive simulated tissue. A voltage-controlled current source provides a known value of current to be introduced with each new set of measured signals, which are then amplified and filtered to preprocess the signals. The data is then gathered together by the RP and communicated to a MATLAB program running on a nearby PC through a Standard Commands for Programmable Instrumentation interface, where the data is reconstructed using regularization with EIDORS, a MATLAB toolkit. Precision and accuracy were measured by evaluating the signal-to-noise-ratio (SNR) and a 2D Pearson correlation coefficient to the ideal images of the tank. The first tank design used electrodes of a small, pointed design, while the second used 1 cm2 copper plates for electrodes. The first had a mean system SNR of 31.616 dB (11.347) before filtering and 34.176dB (11.9803) after, and a correlation coefficient of r = 0.702 (0.055), while the second tank showed negative results.

Over the past century, society has been riding a wave of technology revolution empowered by Moore’s Law that has enabled trillion devices proliferating into every aspect of our daily lives. Only through innovation in material science of silicon-based transistors and manufacturing capability of advanced deposition and lithography process has made this possible. The next century is beginning to see a new emergence of “more than Moore” with devices attributing new properties such as conformability, transient/edible, self-healing, actuating motion and so much more. Similarly, these devices will be formed from union of new and old technologies, where they can be fabricated using high-throughput printing technologies for make them affordable and ubiquitous. This thesis focuses on the composition of various innovations of particle-polymer composites with additive manufacturing process such as screen-printing to design and fabricate multiple technologies such as self-propelling motors, energy harvesting, energy storage, stretchable electronics, and wearable devices. The thesis provides strategy to develop new composites for their respective technologies by sourcing the composite materials necessary to optimize the composite characteristics, which would be studied through characterization and device evaluation by varying the composite ratios. The development of new composite materials for emerging technologies is enabling a new vision of how technologies can continue to weave into every fabric of our lives.

Neuronal tracing methods are essential tools to understand the fundamental architecture of the neural circuits and their connection to the overall functional behavior of the brain. Various viral vectors have been developed to achieve cell-type specific and directional specific labeling of the neuronal connections. The incorporation of the viral tracers with optogenetics techniques can further provide precise control over the viral propagation. Herein, a novel approach to guide the infectious pattern of the Rabies Virus (RV) retrograde tracer with light is described. The key concept is to use Baculovirus (BV) as a helper virus to deliver all the functional components prior to the rabies infection. These functional components include the optogenetic tool parts, a fluorescent CRE reporter to mark RV infected cells, and other genes necessary for RV spread. GoldenGate DNA assembly method was used to efficiently construct three large BV expression cassettes in bacteria plasmids that can be transposed into the baculovirus genome in a single TN7 transposon recombination step. The actual imaging of BV and RV infected cells and stimulation of optogenetics tools for the spread of RV can be achieved with CMOS sensor and OLED display panels. By integrating these commercially available electronics with proper microcontrollers, a synchronized communication protocol for optogenetic stimulation, image capturing, and file transferring was established in a lensless imaging setup. The OLED and CMOS sensor’s large field of view and low cost allows this tracing method to scale many orders of magnitude greater than using microscopy-based methods.

Glaucoma, the leading cause of irreversible blindness, affects more than 70 million people worldwide. Lowering intraocular pressure, the most common method of delaying both the development and progression of glaucoma, is typically done with daily topical administration of therapeutic eye drops, which has notoriously high non-adherence rates ranging from 30% to 80%. The advent of smart phone enabled technologies creates an opportunity to address the non-adherence problem. However, previous eyedrop electronic monitoring solutions had awkward medication bottle adjuncts and crude software design for monitoring the administration of a drop which adversely affected their ability to foster significant, sustainable improvements in adherence. Here we present our prototype, “smart drop” bottle, that is capable to detect each eyedrop medication administered while maintaining the shape or size of the eyedrop bottle. This is achieved by developing a smart-phone application that interact with small rigid-flex electronics beneath the bottle and bottle label respectively. We have shown we can achieve wireless communication close to 100 feet with 0% of false positive results thus providing a potential solution for adherence monitoring of glaucoma patients.

Binary logistic regression is often used in clinical applications to predict the occurrence of medical conditions that arise within a patient population. Point estimations are often made to approximate the unknown regression weights. In doing so, information about the underlying posterior distribution of the weights is lost. We propose a method that views the logistic regression model from a Bayesian perspective and takes into consideration the full posterior of the un- known regression coefficients when computing the probability of belonging to the positive class. This method will be referred to as the Fully Bayesian method. The Fully Bayesian method allows us to quantify the uncertainty in our probability calculations. The work in this paper builds on Kim and Ma’s previous work in which they demonstrated efficiently solvable fully Bayesian estimation techniques. By solving a (convex) Kullback-Leibler divergence problem they were able to obtain a mapping from any log-concave prior to its corresponding posterior distribution thus enabling one to draw independent samples from the posterior with ease. Hav- ing the full posterior is useful in revealing how credible a prediction is and can be utilized to define an abstain strategy. The data set was created from a subsample of de-identified patient data from Kaiser Permanente and consists of a highly im- balanced number of patients that have and have not been diagnosed with asthma. The results show that the overall performance of a Fully Bayesian scheme produces a higher measure of accuracy than the point estimate method.

Background & Objective: Although the reasons behind the biological necessity of sleep remain to be fully discovered, emerging data show that sleep is crucial for maintaining physical health and mental well-being. Despite its great importance, nearly 40% of US adults experience problems with sleep, ranging from sleep deprivation (decreased total sleep time) to sleep disorders (e.g. obstructive sleep apnea, OSA). Furthermore, many millions of these individuals with sleep disorders are severely underdiagnosed. The major limitation to improved understanding of sleep in health, and streamlined identification of at-risk individuals, is the inability to objectively measure sleep at scale. Gold standard sleep measurement by polysomnography (PSG) requires bulky, obtrusive measurements of multiple biological signals that interfere with sleep. Analysis of sleep measurements undergo a manual/visual process that is lengthy, expensive, and subjective. These impediments to sleep make testing beyond a single night in selected patients impractical, and make addressing growing public health issues in sleep seemingly impossible. Methods & Results: To address these shortcomings in the current sleep medicine paradigm, a systems approach to clinically-guided engineering spanning sensing and analytical aspects of sleep monitoring is described herein. Adhesive-integrated stretchable sensors were engineered to achieve scalable, peel-and-stick clinical sensing on ubiquitously used clinical adhesives, resulting in a cost-effective, facile, and familiar means for clinical implementation of unobtrusive electrophysiological sensing technology. Simultaneously and in collaboration with sleep medicine clinicians, small-data statistical analytics were engineered to dynamically assess clinical-grade sleep architecture from single-channel electrophysiology, the results of which are on par with in-lab PSG for both healthy/normal individuals and patients with OSA. Conclusion & Significance: The result of this applied clinical engineering research is a suite of tools that emphasizes miniaturization of sleep sensing methods, complemented by analytical endeavors to generate clinical-grade sleep measures from consequential small-data platforms. The gestalt is not only envisioned for objective assessment of sleep-based clinical outcomes, but to also stimulate basic science investigation of sleep dynamics and their context in other comorbidities. By performing engineering in medicine in a systemic manner, the principles of this work can be extrapolated for the continued advancement of general wearable health monitoring practice.

One of the primary components of GI motility is the gastric slow wave, which travels along the surface of the stomach and locally actuates smooth muscle contractions, known as peristalsis. Pathological slow wave patterns, detected through intrasurgical recordings, have been shown in high incidence with GI disease and symptoms. The ability to non-invasively detect pathological slow wave patterns is of high clinical interest, as this could significantly enhance diagnostics and lead to tailored treatment regimens. In four parts, we develop statistical analysis methods to infer spatial slow wave patterns from cutaneous high-resolution electrogastrogram (HR-EGG) waveforms.

We first establish a deep three-dimensional convolutional neural network (3D-CNN) framework to classify normal and abnormally initiated slow waves from in silico multi-electrode waveforms. We develop a linear discriminant classifier using spatial features of wave propagation to compare performance. We report the first successful classification study and show robustness across recording and anatomical variability.

Second, we develop a transfer learning framework, in silico, to build a 3D-CNN classifier robust to heterogeneity in both the location of the slow wave abnormal initiation as well as the recording start times with respect to slow wave cycles. We find that by building complexity in training sets, transfer learning one model to the next, the final network exhibits high classification accuracy on data with closer resemblance to human physiologic recordings, suggesting downstream potential for use as widely deployable clinical screening tools for GI disorders.

Third, we establish an optimal transport theory-based philosophical framework for the detection of anomalous probability distributions. We demonstrate a rigorous mathematical proof and show an example algorithm within the framework of our methodology to detect an anomaly in a set of probability distributions that model slow wave direction estimates from HR-EGG recordings on human subjects.

Finally, we implement robust regression and clustering methods on optimal transport-based features of probability density estimates of slow wave directions from HR-EGG recordings on human subjects. We report highly accurate robust regression and suggest the ability to predict treatment response and underlying disease etiology. This methodology may provide clinicians with an opportunity to screen patients and optimize treatment regimens.

The field of optimal transportation is a broad area of theory pertaining to the computation of a mapping to transform one probability distribution into another. Many computational strategies solving the transport problem spanning decades of research have led to several interesting applications in areas such as statistics, economics, machine learning and computer science. However, there also exist many limitations to these algorithms, including the ever-present difficulty of scaling to larger datasets, both with respect to dimension and number of available samples from which we would like to extract useful information. Furthermore, although we have seen a dramatic increase in the availability of powerful distributed computational resources throughout the past few decades, such as publicly available CPU clusters and GPU compute nodes, many modern algorithmic approaches to the transport problem are not specifically designed with parallelism in mind to accommodate such frameworks. In a world where the continuous interaction between human users and distributed computational resources drives our technologically modern lives, this capability is critical, especially with applications that are expected to work in real-time.

Building upon previous research from our group, this work investigates a parallelized, computational problem to create optimal transport maps that is designed with the notion of remote scalability in mind; this work focuses on a CUDA-based implementation of the framework that utilizes GPU computational resources to accommodate analysis of data in higher dimensions not often seen in the field of computational optimal transportation, and that furthermore scale with the availability of additional hardware. We will also present applications using this framework in several different areas of machine learning with an emphasis on human-computer interfaces, including a novel multi-user brain-computer interface, a classification problem pertaining to automated sleep-staging based on electroencephalographical (EEG) data, and an example of generative modeling using the MNIST dataset. Finally, we'll discuss a future direction for application of this framework to preference-based deep reinforcement learning.

Characterizing and sampling from probability distributions is useful to reason about uncertainty in large, complex, and multi-modal datasets. One established and increasingly popular method to do so involves finding transformations or transport maps between one distribution to another. The computation of these transport maps is the subject of the field of Optimal Transportation, a rich area of mathematical theory that has led to many applications in machine learning, economics, and statistics. Finding these transport maps, however, usually comprises computational difficulties, particularly when datasets are large both in dimension and the number of samples to learn from.

Building upon previous work in our group, we introduce a formulation to find transport maps that is parallelizable and solvable with convex optimization methods. We show applications in the field of health analytics encompassing scalable Bayesian inference, density estimation, and generative models. We show how this formulation is scalable with the dimension of data and can be parallelized utilizing a sweep of architectures such as cloud computing services and Graphics Processing Units. Within the context of Bayesian inference, we present a distributed framework for finding the full posterior distribution associated with LASSO problems and show advantages compared to traditional methods of computing this posterior. Finally, we discuss novel methods to reduce the number of parameters necessary to approximate transport maps.

Light-responsive proteins enable control of biological processes with unprecedented precision, holding great promise for clinical and industrial applications. Introducing these proteins into cultured cells or tissues of live animals allows investigation and control of various cellular and organism functions, from neuronal activity, to intracellular signaling, gene expression and cell proliferation, for example. In this work, we take different approaches to the optimization of phytochromes and phytochrome-based optobiology tools. We also present a deep learning framework for protein biology with direct implications on identification of functionally relevant residues in phytochromes.

First, by co-expressing cyanobacterial enzymes, we show that it is possible to increase endogenous chromophore production. Chromophores are bilin molecules that covalently bind to phytochromes, enabling photoconversion. Endogenous production of chromophores is a key development for phytochrome its use in mammalian cells. We demonstrate the limiting factors in chromophore production are two of the required enzymes in the chromophore’s pathway, and not solely heme as previously reported. We show how stoichiometry and species-matching affect chromophore production, and how chromophore levels can impact the performance of phytochrome-based optogenetic systems. Next, we demonstrate the utility of coupling the endogenous chromophore pathway and a light-responsive module composed of cyanobacterial Phytochrome B (PhyB) and its interacting factor (PIF3) to control expression of reporter genes.

Finally, we present a deep learning framework to identify complex relationships inherent in multiple sequence alignments. We develop a Hierarchical Attention network (HAN) for protein sequence families (HANprot) and demonstrate its performance in terms of relevant residue matching. We also demonstrate its utility in finding relevant residues for PhyB, towards potential optimization of its photolabile properties. The residues identified by HANprot can be used as a starting point for further protein investigations when structural or database annotations are lacking.