Aggelos K. Katsaggelos - US grants

This grant provides partial support for an IEEE-sponsored database comprised of sampled images and image sequences that has been established at the National Center for Supercomputing Applications (NCSA) in Urbana, Illinois, with the Signal Processing Society serving as a gatekeeper. This database is accessible via the InterNet at no charge to the users. Data is being solicited from university, industrial and military sources. The data will provide a needed testbed for evaluation of image processing algorithms.

ABSTRACT
0311838
Pappas, Thrasyvoulos
Northwestern U

There has been much interest in supporting video communication over wireless networks. This is
not a simple task due to the stringent Quality of Service (QoS) required by video applications and
the many impairments of wireless channels. Two important QoS characteristics for video are the
degree of signal distortion and the transmission delay. Wireless channels are lossy, have a timevarying
response, and are subject to multi-user interference. Furthermore, since users of a wireless
network are mobile, e.ciently utilizing the available energy is a key consideration.
The di.culties inherent in wireless video can be addressed at both the physical layer and the
source coding layer. We jointly consider a combination of these approaches in a cross-layer framework.
Speci.cally, we consider jointly adapting source coding parameters, such as the quantization
step-size and prediction mode, along with physical layer resources, such as the transmission rate and
power. Our goal is to provide acceptable QoS while taking into account system constraints such as
the available energy. We propose a general framework that allows a number of resource/distortion
optimal formulations for balancing the requirements of di.erent applications.

ABSTRACT
0308197
Thrasyvoulos Pappas
Northwestern Univ

The use of distributed sensor networks for data gathering and analysis is one of the key emerging technologies at the beginning of the 21st century. Such networks consist of a possibly large number of devices (nodes) equipped with multiple sensors (acoustic, imaging, video, seismic, magnetic, infrared, micro-radar, etc.), storage and processing capability, and wireless communication links to neighboring nodes and/or a central location.

Distributed sensor networks offer a lot of exciting possibilities for a number of applications, ranging from environmental monitoring, to industrial process control, to military and security applications, to health monitoring. They also pose a lot of challenging problems for research that derive from the distributed nature of the network, the multimodal nature of the information collected by the sensors, the reliability of the individual nodes, the bandwidth and impairments of the communication links, the power consumption constraints, and the dynamic nature of the overall system.

Recent advances in several domains, including sensor technologies, computational speed, storage capacity, networking, and miniaturization, make the project possible and set the stage for significant breakthroughs. A substantial research effort on various aspects of the problem is already under way.

A recent issue of the IEEE Signal Processing magazine summarizes some of the ongoing research in the area of signal processing. A substantial effort is coordinated by the DARPA Sensor Information Technology (SensIT) Program, whose mission is the development of all necessary software for networked micro-sensors. SensIT was founded on the concept of a networked system of cheap, pervasive platforms that combine multiple sensor types, embedded processors, positioning ability and wireless communication.

The workshop will take place at the Allen Center on Northwestern University's Evanston campus and its duration will be two days. The goal of the proposed workshop will be to identify fundamental research directions in the signal processing and communications areas, the advancement of which will lead to major breakthroughs in the specific focus areas as well as important scientific progress, in general.

Based on the discussions of the workshop, a report will be generated that will provide advice to NSF on promising directions for near and long term research in the area of distributed communications and signal processing for sensor networks. The report will also be used to clarify the soon to be announced call for proposals under the title "Sensors and Sensor Networks for Information, Decision and Action."

ABSTRACT
NSF-0515929
Pappas, Thrasyvoulos

Embedded systems technology has reached the level at which it is practical to deploy numerous low-power wireless sensing devices with the capacity to execute sophisticated algorithms and to selectively allocate resources to the most critical situations. The plan is to develop a robust platform for distributed, pervasive, small sensing and computing devices to implement a cognitive information processing system that understands its surroundings and responds to changes in its environment by constantly evaluating and adapting its behavior. The novelty of the approach lies in the use of (1) multiple sensor fusion based on a dynamic Bayesian network (DBN) which provides a powerful, intelligent, and cognitive learning framework, and (2) collaborative and distributed Monte Carlo (CDMC) tracking methods for establishing the collaboration among multiple computational units. Once the DBN model is learned, solutions to
the problems under consideration are provided by the probabilistic inference of the model, which provides a self-evaluation of the tracker. The scientific objective will be to develop a framework and algorithms for multi-perspective, multi-modal fusion of sensor information subject to bandwidth, energy, reliability,and location-uncertainty constraints associated with low-cost distributed sensors. The proposed research covers topics from sensor calibration and feature extraction, to data fusion and collaborative processing, to resource/performance optimizations at the local and system level. While many of the individual problems have been studied on their own, the challenge is to bring them all together in a new context under tight constraints on power consumption for computing and sensor communication, as well as latency constraints and sensor failure considerations.

Distributed, collaborative sensing will play an increasingly critical role ina variety of applications involving security and surveillance, environmental monitoring, industrial
automation, and remote exploration. The techniques can also be usedin a number of other applications, such as human tracking and behavior analysis for medical
applications such as telemonitoring of elderly people. The broader impacts of the proposed activity also include improvements and integration of the signal processing and computer vision curricula at Northwestern and the opportunity for undergraduates to perform research on related topics.

ABSTRACT
NSF 0536994, Choudhary
NSF 0536947, Fox

Problems in managing, automatically discovering, and disseminating information are of critical importance to national defense, homeland security, and emergency preparedness and response. Much of this data originates from on-line sensors that act as streaming data sources, providing a continuous flow of information. As sensor sources proliferate, the flow of data becomes a deluge, and the extraction and delivery of important features in a timely and comprehensible manner becomes an ever increasingly difficult problem. More specifically, developing data mining and assimilation tools for data deluged applications faces three fundamental challenges. The amount of distributed real time streaming data is so large that even current extreme scale computing cannot effectively process it. Second, today's broadly deployable network protocols and web services do not provide the low latency and high bandwidth required by high volume real time data streams and distributed computing resources connected
over networks with high bandwidth delay products. Finally, the vast majority of today's statistical and data mining algorithms assume that all the data is co-located and at rest in files. Here, the real time data streams are distributed and the applications that consume them must be optimized to process multiple high volume real time streams. The goal is to develop novel algorithms and hardware acceleration schemes to allow real-time statistical modeling and change detection on such large-scale streaming data sets. By using Service Oriented Architecture principles, a framework for integrating high -performance change detection software services, including accelerations of commonly used kernels in statistical modeling, into a Grid messaging substrate will be developed and tested. Geographical Information System (GIS) services will be supported using
Open Geospatial Consortium standards to enable geo-referencing.

This project has the potential to have near-term and long-term impact in several important areas. In the near-term, the implementation of kernels and modules of statistical modeling and change detection algorithms will allow the end-user applications (e.g., homeland security, defense) to achieve one to two orders of magnitude improvement in performance for data driven decision support. In the longer term, the availability of toolkits and kernels for the change detection and data mining algorithms will facilitate the development of applications in many areas including defense, security, science and others. Furthermore, this research will bring the use of reconfigurable
architectural acceleration of functions on streaming data including change detection and data mining, thereby opening new avenues of research and enabling newer data-driven applications on complex datasets. Both graduate and undergraduate students (through undergraduate fellowships) are engaged in the research. In addition, team members actively engage with minority serving institutions using audio/video and distance education tools.

This National Science Foundation Research Traineeship (NRT) award prepares master's and doctoral students at Northwestern University with the data-analysis skills to advance the research frontiers in astronomy, physics, and Earth science. While providing students with training in data-enabled science and engineering, the program promotes collaborations with research and education that will lead to the development of new data tools with broad applicability. Through internships, evidence-based curricular approaches, and capstone citizen science projects, trainees will develop the core competencies in demand by a wide range of employers. Trainees will learn how to effectively engage the public in discoveries on the solar system, stellar explosions, star clusters and galaxies, gravitational waves, and seismic waves. The citizen science projects will be used for innovative recruiting, strengthening the participation of the public and students from underrepresented groups. By diversifying the graduate student population and improving instruction and mentoring, the research will be contributing to a diverse, inclusive scientific workforce in academia and industry.

This program will bridge computer science, electrical engineering, applied math, and statistics to physics, astronomy, and Earth sciences in order to develop students with the skills required to analyze datasets of unprecedented size and complexity. Trainees will be prepared for the technical challenges of extensive data generated at major research equipment and facilities, including Large Syntopic Survey Telescope, Advanced Laser Interferometer Gravitational-wave Observatory, and EarthScope. Trainees will learn data analytics and management, statistical methods, and image processing skills to ask new questions and choose the appropriate tools from the method disciplines and adapt them to solve problems in physics, astronomy, and Earth sciences. The program will also prepare students to work in a collaborative scientific community, advancing their leadership, communication, mentoring, and management skills. Internships at major research facilities, national laboratories, and the private sector will help students gain transferrable professional skills to pursue a range of career paths.

This grant will allow the purchase of a high-performance-computing cluster, which is required by researchers at Northwestern University to investigate sources of gravitational-wave emission. Modeling and understanding of gravitational-wave sources is becoming extremely important as scientists are utilizing data from the Laser Interferometer Gravitational-wave Observatory to detect black holes and other exotic astrophysical objects (such as "neutron stars"). On one level, astronomers require large computers to develop programs that can help us understand the large amounts of data from gravitational-wave detectors, and pick out the real, astronomical sources from a wide variety of noise sources. But in addition, as detections of gravitational-wave sources occur more frequently, astronomers need to understand what those detections can tell us about the population of exotic objects throughout the universe. For instance, given the detections of gravitational-wave sources already, researchers ask how many black holes there might be, in our galaxy, and in other galaxies, and how massive are they? Also, researchers ask: how do those black holes form, and how do systems of pairs of black holes form? These are important questions in understanding objects at the extreme of the known laws of physics, and to answer these questions, astronomers need to understand not only the lives of individual stars and the formation of black holes and neutron stars, but they need to simulate populations of millions of stars, and the wide variety of events and processes that can affect the development of those stars. For that work, astrophysicists require large collections of very efficient computers, such as the computer cluster that will be purchased with these funds, to run massive simulations that help us explore these questions. In addition to the pure research that will be done with this cluster, the PI's group is also well known for generating award-winning visualizations of their simulations, which will be used at a variety of Science, Technology, Engineering, and Mathematics (STEM) education and public-outreach events. The entire team also has extensive experience in attracting a diverse group of researchers to STEM research and fields; this cluster will help that group to train a new generation of diverse researchers in data-science methods and scientific computing, contributing to the technical workforce of the nation.

This grant will allow the acquisition of a high-performance computing (HPC) cluster that will enable research in the emerging area of gravitational physics. The cluster will be essential to the development and optimization of codes needed for both gravitational-wave (GW) data analysis as part of the Laser-Interferometer Gravitational-wave Observatory (LIGO) Scientific Collaboration, and GW source modeling for the physical interpretation of the detections. The equipment consists of a large computer cluster (56 nodes) with InfiniBand networking and 70 TB of usable storage. This cluster will incorporate three innovative Graphics-Programming-Unit nodes (using GPUs appropriate for scientific computing) which will be used as accelerators for special-purpose massively parallel computations. The cluster will be housed at a top-of-the-line HPC data center on the Northwestern University campus and will be operated and managed by an experienced team of HPC professionals led by one of the co-PIs. In more detail, the cluster will be used for GW research focused on binaries with two compact objects (neutron stars and/or black holes) in interdisciplinary collaborations between GW data analysts (members of the LIGO Scientific Collaboration), astrophysicists, and computer scientists. The goals are to optimize data searches for GW signals (through effective detector characterization), to extract as promptly as possible and accurately the physical properties of the signal sources (through continuous improvements of our parameter-estimation algorithms), and to advance the astrophysical interpretation of the discoveries so we can better understand the sources' origin and constrain theoretical models using our GW observations (through the development of state-of-the-art formation models in different environments and comparing predictions to data). The team is led by PI Kalogera and co-PI Rasio, who are well recognized for their significant impact in these research areas and for their innovative development of new computational tools for GW data analysis and astrophysical modeling of GW sources.

PI: Kenneth Shull (Northwestern University)
co-PIs: Francesca Casadio (Art Institute of Chicago)
Oliver Cossairt (Northwestern University)
Aggelos Katsaggelos (Northwestern University)
Marc Walton (Northwestern University)

Non-Technical Abstract:
Historic art objects provide a collection of materials that have been naturally aged for decades or even centuries. In addition to the intrinsic archival value of these materials, they are also models for understanding property degradation over long periods of time. This project aims to develop computational and experimental tools needed to understand how these changes take place. To accomplish this task a research network has been established between Northwestern University and leaders in cultural heritage science from the Rijksmuseum and the University of Amsterdam in the Netherlands, the National Research Council in Italy, and the Synchrotron Soleil in France. This new infrastructure promises to deliver a significant enhancement of research and education resources (networks, partnership and increased access to facilities and instrumentation) to a diverse group of users. The art objects central to the project provide a series of well-defined case studies for investigating complex materials systems that are both applicable to materials education and push the limits of the existing analytical tools, thus inspiring instrumental innovations across broad sectors of the physical sciences. Further development of these tools will enable art conservators to more effectively make informed decisions about treatments of works of art, and to understand long-term materials degradation more generally. The project will also deliver a significant enhancement of research and education infrastructure by a diverse group of users and will provide meaningful, international research experience to 50 participants, with a strong emphasis on scientists at the beginning of their careers. In addition, the connections between science and art will illustrate the creative aspects of both disciplines to a very broad audience, attracting a more representative cross section of people into science.

Technical Abstract:
The purpose of the proposed program is to probe the properties of heterogeneous material composites at multiple length scales, with a focus on the materials used in creating works of art. The grand challenge is to understand the coupling of material structures from nano- to macro- length scales to the visual appearance, and to use this coupling as a probe of material properties. This coupling will be addressed by incorporating light/matter interactions into computational chemistry approaches, which will also be developed to understand the physical-chemical changes that occur in materials over long periods of time. This information can be used to reconstruct the appearance of an object as originally created, and project the appearance into the future. This methodology is of primary importance to the art conservation community, which has developed advanced research infrastructures in Europe that are rare or nonexistent in the U.S. These European resources are essential for the completion of the project goals, which are to provide a greater understanding of the way in which chemical and physical changes within a material gradually distort its visual perception, and to develop a mechanistic understanding of these alteration pathways. The proposed PIRE project integrates teams from leading cultural heritage science institutes in France, the Netherlands and Italy with their American Counterparts. While the tools will be applicable to modern engineered materials as well, examples from art provide a much broader educational impact. A combination of individual and cohort visits to the three primary international sites will provide students with an international perspective on science and research, while building skills in communicating the role of science in society.

This project is designed to advance research on problematic eating behavior. The project investigates wearable sensors to measure eating behavior and developing models of behavior that comprise multiple observable behaviors such as eating alone or with friends, or chewing speed. These data can help scientists improve upon current traditional methods such as self-reported eating diaries, which tend to be inconsistent, sparse, and rarely timely. We capture human behavior using a custom wearable augmented camera. Wearable cameras provide rich data, but raise privacy concerns. The project will address these concerns by building a framework using machine learning and information theory while including human-reported privacy concerns. The framework will address wearers' concerns that may limit recording authentic behavior in real-world settings and will optimize algorithms to enhance the detection and classification of human behavior.

The project explores the acceptability of obfuscation techniques on varied activities and their requisite tasks. The proposed research will design a suite of computationally efficient task-specific algorithms that use raw images in computationally restrictive (in situ) and obfuscated images in unrestrictive environments (offline) to build information-performance curves for the scalable development of personalized ground truth wearable cameras. The project also will develop a modular, plug-and-play, low-complexity and efficient obfuscation computing hardware device to facilitate and accelerate the use of the proposed methods and algorithms. This work will validate an overeating behavior model in a real-world setting using the design framework and device, providing visual confirmation of eating behaviors, showing how it can be used to test existing models. This project is likely to be useful to other domains in the social sciences, fundamentally changing the way researchers build and validate behavioral models in real-world settings. There are potential applications in health (especially preventive medicine), social, and economic sciences: energy balance, infant development, medication adherence, consumer behavior, and human-environment interaction.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

This project will build our understanding of how to enable non-expert volunteers in a citizen-science project to contribute to analyses of large volumes of data by searching for potentially causal relations. The increasing use of automated scientific-data-collection instruments has led to an explosion in the amount of scientific data collected, challenging the ability of scientists to analyze them. Volunteers have less background knowledge than experts about the purpose, context, content, provenance and processes associated with the data. A system that provides such background knowledge will enable non-experts to make sense of the data. The research plan also includes building system support to augment the capabilities of the volunteers, for example by searching for related data and by performing causal inference in conjunction with volunteers. Citizen-science projects provide a vehicle to disseminate scientific practice, knowledge and findings to the general public to increase awareness and understanding of the practices and techniques of data-intensive science. Findings should be directly applicable to the target context of involving citizen-science volunteers in navigating and analyzing large quantities of science data and generalize to other settings with big data.

In this research, volunteers classify noise events (glitches) produced by the Laser Interferometer Gravitational-wave Observatory (LIGO). Along with glitches observed in the main Gravitational Wave channel, the detectors record around 400,000 auxiliary channels of data that may provide information about the origins of the glitch. The research will test hypotheses about the kind of additional information needed to enable non-experts to productively navigate this large dynamic dataset to find related information and will develop processes, techniques and tools to allow the volunteers to manage and efficiently process the data. It will develop our understanding of how and when to introduce which different types of background knowledge about the data to enable non-experts to work on a task, such as by providing maps and visualizations of particular data and relationships at the time they are most needed in the volunteers' work process. The gravitational physics and astronomy communities will directly benefit from advances in LIGO detector characterization, data quality vetoes and hence signal searches.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Project Summary/Abstract: Despite the accuracy and versatility of cardiovascular MRI, its footprint is only 1% among cardiac imaging tests (SPECT, echocardiography, CT, MRI) in the US. While there are several factors such as referral patterns favoring SPECT and echocardiography among cardiologists that account for low utilization, the two addressable obstacles that preclude widespread adoption are lengthy scan time (imaging facility operational cost) and reading (physician cost). These obstacles must be addressed for community hospitals with limited resources to adopt cardiovascular MRI into clinical routine practice. While compressed sensing (CS), since its introduction into the MRI world in 2007, has led to highly-accelerated cardiovascular MRI acquisitions, the subsequent image reconstruction remains too slow (> 5 min for 2D time series, > 1 hour for 3D time series) for clinical translation (unmet need 1). Downstream, image analysis for cardiovascular MRI is notoriously labor intensive (e.g. 30- to 60-min) and limited (?circles? at two cardiac phases for cine MRI, whereas perfusion and late gadolinium-enhanced (LGE) images are evaluated visually), for what is essentially a basic computer vision task (unmet need 2). In direct response, we will address these two unmet needs and unlock the enormous potential of CMR using deep learning (DL). DL applications have exploded since advancements in optimization and GPU hardware. While several recent studies have applied neural networks such as convolutional neural networks (CNNs), U-Nets, and Generative Adversarial Nets (GANs) for reconstruction and segmentation, no study has implemented an inline end-to-end pipeline that receives raw k-space from the MRI scanner and delivers both reconstructed images and fully processed images automatically with high speed (< 1 min). The objectives of this study are: a) developing a network for image reconstruction with maximal acceleration (aim 1), (b) developing a network for image processing tasks (aim 2), and c) developing an integrated, end-to-end network that does both (aim 3). By developing an architecture that can simultaneously learn maximal acceleration, fine tune end-to-end performance, and perform reconstruction/inference using feed-forward networks, we anticipate a disruptive technology that will lead to a paradigm shift in cardiovascular MRI and increase its footprint in community hospitals. This 2-year study is doable because of the requisite database of raw k-space (not derived from DICOM) data (N = 617) and annotated cardiac MR images (N=3,021) from over 3,000 patients existing at our institution. Success of this proposal will deliver a disruptive technology that has potential to cause a paradigm shift in cardiovascular MRI and enable widespread adoption of cardiovascular MRI into clinical routine practice.