A Pipeline of a Resilient Workforce that integrates Advanced Analytics to the Agriculture, Food and Energy Supply Chain

We propose a novel sensor system with accompanying data analytics to explore the capability of wearable multimodal sensors to address the short-comings of the traditional polysomnography systems. If successful, this project will lead to improved capacity to carry out sleep research and to detect and treat sleep disorders. The miniaturization and low power consumption will pave the way for rapid adoption and deployment of these systems for home-use in real-world settings.

Our main goals are to: (1) develop a statistically-sound data-driven framework for signal quality characterization of wearable devices in the real-world; and (2) use this framework for enhancing algorithmic developments and hardware design. Current approaches depend on rules or indicators derived from expert knowledge in controlled environments, so they do not generalize well to the use at-home. Our main application will be the early asthma exacerbation detection. We aim to employ the prototypes designed by the NSF-ERC ASSIST center, which aims to develop nano-enabled energy harvesting, energy storage, nanodevices and sensors to create innovative battery-free, body-powered, and wearable health monitoring systems.

The North Carolina State University Blueberry Breeding Program has long supported and collaborated with NC blueberry growers to breed for blueberry cultivars that stand up to our challenging climate and meet the needs of high yield, quality, and machine harvest-ability. New machine-harvest trials coupled with on-farm grower trials will inform the release of new NC State blueberry cultivars, while new technologies such as machine learning and in vitro micropropagation will allow us to grow and test larger numbers of selections in our berry evaluations. In addition, our cooperative breeding and partnership with NC blueberry growers help develop new varieties faster.

Inconsistent quality and aesthetics in agricultural crops can result in increased consumer and producer food waste, reduced industry resiliency and decreased farmers������������������ and growers������������������ profit, poor consumer satisfaction, and inefficiencies across the supply chain. Although there are opportunities to characterize and quantify sources of phenotypic variability across the agricultural supply chain - from cultural practices of growers and producers to storage and handling by distributors - the data available to allow for assessment of horticultural quality drivers are disparate and disconnected. The absence of data integration platforms that link heterogeneous datasets across the supply chain precludes the development of strategies and solutions to constrain variability in produce quality. This project������������������s central hypothesis is that multi-dimensional produce data can be securely integrated and used to optimize management practices in the field while simultaneously adding value across the entire food supply chain. We propose to develop multi-modal sensing platform along with a trust-based, data management, integration, and analytics framework for systematic organization and dynamic abstraction of heterogeneous data across the supply chain of agricultural crops. The projects short term goals are to (1) engage growers to refine research and extension priorities; (2) develop a first-of-its-kind modular imaging system that responds to grower needs by analyzing existing and novel multi-dimensional data; (3) establish the cyberinfrastructure, including analytics and blockchain, to make meaningful inference of the acquired data as related to management practices while ensuring data security; (4) deploy the sensing system at NCSU������������������s Horticultural Crops Research Station in Clinton, NC and on a large-scale system at a major commercial farm and distribution facility, and (5) extend findings to producers and regulators through NC Cooperative Extension. The proposed sensing and cyberinfrastructure platforms will be crop-agnostic and our findings will be transferable to other horticultural crops produced in NC and beyond.

The objective of the proposed research is to enable the rapid translation from aptamer selection to deployment on a potentiometric biosensor������������������s surface for highly selective detection. Optimization of the sensor surface will be accelerated through the use of advanced machine learning techniques to distinguish target-specific responses from non-specific binding events and electrode drift effects in complex, clinically-relevant fluids that most studies struggle to overcome. To demonstrate the effectiveness of the proposed approach, a biosensor platform for extracellular histone detection will be developed. The understanding that extracellular histones mediate tissue injury and propagate organ failure is relatively new, while the report of aptamer-based therapies is even more recent. Despite this, there have been no reports of electronic microsensors with targeted affinity for circulating histones. We therefore hypothesize that aptamer chemistry can be leveraged to functionalize the surface of potentiometric microsensors in order to perform early-stage, point of care (POC) detection of circulating histones, facilitate the identification of individuals at high risk for development of Multiple Organ Dysfunction Syndrome (MODS), and allow early treatment.

This project aims to develop a proper methodology for determining the quality of ECG signals and heart-rate estimation. We will compare and contrast: (1) algorithms for real-time denoising of ECG signals and heart-rate estimation on wearable platforms; and (2) the quality of the data from different system platforms including Holter monitors for ground truth, the off-the-shelf Shimmer platform, and dry and wet electrode ASSIST platforms. After this characterization, we will look at implementing machine learning approaches for motion artifact identification based on ECG and inertial sensing information. Determining the specific type of artifact will help select the type of denoising approach to be applied. There are a number of techniques developed for specific noise models that can be elaborated for this purpose. We will finally, perform data fusion from multiple sensor platforms in order to come up with more robust estimation algorithms. Finally, we will explore the usage of the artifact detection algorithms to identify faulty sensors or improper use. The methodology developed for this type of sensing will also be applicable to other sensing modalities (e.g., respiratory rate, blood pressure, etc.).

Paleoceanography, among other research fields, depends crucially on ubiquitous ocean dwelling single celled organisms called foraminifera. Undergraduate workers are often employed to pick several thousands of specimens from ocean sediments for each study. Depending on deposition rates and abundance of the species, such manual processing can become tediously repetitive with little intellectual motivation for the undergraduate workers, and time and cost-prohibitive for research scientists. The proposed project aims to develop a completely autonomous system for visual sorting of foraminifera, which is accessible to the scientific community. This system will be compatible with existing off-the-shelf microscopes, it will make use of microfluidics in order to facilitate the transport of the samples from a container to their sorted receptacles, and will utilize machine learning for recognition. These tools will be made available to the entire scientific community, and aim to keep the fabrication cost under three thousand dollars.

Recent developments in miniature, low-power wireless sensors has provided systems capable of long-term deployment and continuous operation. Plant sciences will be benefited by the application of such a tailored suite of sensors, in which the output data will inform real-time changes to growth conditions in order to minimize cost and maximize yield. This proposal develops a physical and computational framework that will determine the necessary suite of plant physiology sensors and uses the collected data to model the complex interactions between phenotypical response and growth conditions. The contributions of this award will facilitate the broad adoption of new plant physiology sensors and analytic platforms for plant/crop management.

The objective of this proposal is to develop a computational framework that integrates statistical and computational geometric data analysis techniques for the processing, analysis and representation of patterns in order to unleash the potential of physiological and environmental multi-modal wearable sensing health systems for continuous monitoring and tracking of human wellness and physiological state. To accomplish this objective, this proposal will: (1) develop algorithms for the concurrent modeling of physiological, kinematics and environmental states for inference purposes; (2) develop techniques to transform models between different sensing systems in order to make information sharing compatible across platforms; and (3) develop techniques to maximize the impact on the behavior of individuals by elaborating on schemes for data representation. These techniques will empower users and medical practitioners to understanding, analyze, and make decisions based on patterns present in the data.