Plenaries

All plenaries take place in the Marquis Ballroom 3/4 (New).

Pattern Recognition in Video

Rama Chellappa, University of Maryland, College Park
Monday, October 9, 08:00 - 09:25

Efficient Representation and Distribution of Video and Related Media

David Taubman, The University of New South Wales (UNSW), Australia
Tuesday, October 10, 08:30 - 09:25

Download the plenary notes: PPT File (4.0 MB) or Zipped PPT file (2.1 MB).

PET Molecular Imaging Biomarkers of Disease

Michael E. Phelps, UCLA School of Medicine
Wednesday, October 11, 08:30 - 09:25

Pattern Recognition in Video

Presented by

Rama Chellappa
Minta Martin Professor of Engineering
Department of Electrical and Computer Engineering and UMIACS
University of Maryland, College Park, MD.

Date and Time

Monday, October 9, 08:15 - 09:25, Marquis Ballroom 3/4 (New)

Abstract

With the ubiquitous presence of inexpensive video cameras, new challenges to video-based pattern recognition problems are emerging. Video-based pattern recognition problems have applications in homeland security, healthcare, battlefield awareness, and video indexing and anomaly detection. The single most important feature that distinguishes video-based pattern recognition problems from still-image based recognition problems is the dynamical nature of patterns in videos. This creates new intellectual challenges and provides opportunities for novel approaches.

In this talk, I will first discuss some of the general principles for designing robust video-based pattern recognition systems. Classifiers must be designed based on the principles of invariance to illumination, pose, articulation and sensor parameters. They also must be able to account for the dynamics of patterns. To accomplish this, we use a combination of tools from statistical pattern recognition, computer vision, geometry and control theory. Characterization of class-conditional densities using pattern's appearance, shape, and motion and scene illumination conditions are discussed. Statistical classifiers for a multitude of video-based recognition problems such as human identification/verification using face and gait features, vehicle identification across non-overlapping cameras and human activity recognition are presented with examples. A method for compensating for the variations in the rate at which patterns evolve and the role of quasi-invariants in activity recognition are then discussed. Two non-parametric methods based on our novel "human gait DNA" signature and a Geometric Transform that generalizes the Radon transform are then described for recognizing human motion patterns. The problem of anomaly detection in video is posed and solved as a one class recognition problem. Finally, we discuss many theoretical issues and practical problems that remain to be addressed in this area.

Speaker Biography

Prof. Chellappa received the M.S.E.E. and Ph.D. Degrees in Electrical Engineering from Purdue University, West Lafayette, IN, in 1978 and 1981 respectively. Since 1991, he has been a Professor of Electrical and Computer Engineering and an Affiliate Professor of Computer Science at University of Maryland, College Park. He is also affiliated with the Center for Automation Research (Director) and the Institute for Advanced Computer Studies (Permanent member). Recently, he was named a Minta Martin Professor of Engineering. Over the past 25 years, he has published numerous book chapters, peer-reviewed journal and conference papers. He has co-authored and edited many books in visual surveillance, biometrics, MRFs and image processing. His current research interests are in face and gait analysis, 3D modeling from video, surveillance and monitoring, hyper spectral processing, and computer vision. Prof. Chellappa served as the associate editor of many IEEE Transactions and as the Editor-in-Chief of IEEE Transactions on Pattern Analysis and Machine. Intelligence. He served as a member of the IEEE Signal Processing Society’s Board of Governors and as its Vice President of Awards and Membership. He has received several awards, including an NSF Presidential Young Investigator Award, two IBM Faculty Development Awards, an Excellence in Teaching Award and a Technical Achievement Award from IEEE Signal Processing Society. He was elected as a Distinguished Faculty Research Fellow and as a Distinguished Scholar-Teacher. He is a Fellow of IEEE and the International Association for Pattern Recognition. He has served as a General the Technical Program Chair for several IEEE international and national conferences and workshops. He is a Golden Core Member of IEEE Computer Society.

Efficient Representation and Distribution of Video and Related Media

Presented by

David Taubman
The University of New South Wales (UNSW), Australia

Date and Time

Tuesday, October 10, 08:30 - 09:25, Marquis Ballroom 3/4 (New)

Abstract

Recent years have seen a number of significant advances in source compression for images, video and other high dimensional media. These include content-adaptive oriented bases to follow motion or geometric flow; efficient schemes for representing motion and other complex structural properties; and distributed video coding. While compression efficiency and visual fidelty are important driving forces behind some of these innovations, an arguably even more important motivator is the need for flexible representations and systems. The goal is to allow decisions regarding quality, bandwidth, display resolution, region of interest or computational requirements to be moved away from the source compressor itself and placed in more dynamic components of the system such as servers, interactive clients, network transcoders and the like. Tightly connected with this trend is a rapidly expanding body of work on rate-distortion optimized distribution of compressed media. The first part of this talk provides a review of some of these topics, highlighting approaches which have proven effective to date and areas where we can expect to see further gains. The second part of the talk is devoted to a high level discussion of some of the challenges currently facing flexible and efficient compression of video and other media. One of these is the challenge of region accessibility; another is the challenge of resolution accessibility; in fact, they can be seen as related challenges. We will look briefly at various attempts to overcome these challenges. In the third and final part of the talk, we suggest future directions for multimedia compression and distribution. We consider the roles played by source compression, intelligent servers and intelligent clients in addressing the needs of diverse, interactive clients, browsing large, potentially open ended multimedia content sources.

Speaker Biography

David Taubman is with the School of Electrical Engineering and Telecommunications, at the University of New South Wales, where he heads the Telecommunications Research Group. Before joining UNSW at the end of 1998, he spent 4 years at Hewlett-Packard’s research laboratories in Palo Alto, California. He received the B.S. and B.E. (Electrical) degrees in 1986 and 1988 from the University of Sydney, Australia, and the M.S. and Ph.D. degrees in 1992 and 1994 from the University of California at Berkeley. He contributed extensively to the JPEG2000 standard for image compression and the JPIP standard for interactive image communication. He is author, with Michael Marcellin, of the book “JPEG2000: Image compression fundamentals, standards and practice” and author of the popular Kakadu software for JPEG2000 developers. He is recipient of two IEEE Best Paper awards: for the 1996 paper, "A Common Framework for Rate and Distortion Based Scaling of Highly Scalable Compressed Video;" and for the 2000 paper, "High Performance Scalable Image Compression with EBCOT". He is also co-author with J. Thie of a 2004 ICIP best student paper award, for work on hybrid ARQ with LR-PET. His research interests include scalable image and video compression, joint source/channel coding, multimedia distribution and perceptual modeling.

PET Molecular Imaging Biomarkers of Disease

Presented by

Dr. Michael E. Phelps
Norton Simon Professor
Chair, Department of Molecular and Medical Pharmacology
Director, Institute for Molecular Medicine
Director, Crump Institute for Molecular Imaging
UCLA School of Medicine

Date and Time

Wednesday, October 11, 08:30 - 09:25, Marquis Ballroom 3/4 (New)

Abstract

Various in vivo molecular, functional and structure imaging technologies are being developed and applied to the drug discovery and development processes, from preclinical to clinical, from PET to optical, MRI and CT. The information that the molecular imaging technologies are seeking to provide are to image:

1. the biology & biochemistry of disease for molecular diagnostics.
2. the drug target to stratify subjects into those who have the target and those who don't.
3. biological processes involved with the drug target or biological processes associated with the disease, as surrogate markers of disease to identify the presence of disease and its response to the drug.
4. the pharmacokinetics and pharmacodynamics of drugs - Did the drug hit the target? Did the drug occupy enough of the drug target to induce the desired pharmacologic effect? Where else did the drug go in the body over time?

It is critical that these biomarkers, surrogate markers and pharmaco - kinetic and – dynamic approaches develop as a measurement science and that measures developed and applied in preclinical settings be translatable to the patients for clinical research, guiding therapeutic trials and molecular diagnostics in patient care. This presentation will show some of the technological approaches and applications of PET and PET/CT to the questions listed above, using cancer and dementia as the model disease for the approaches.

The analytical measurement capability of PET for imaging and measuring the tissue concentration of labeled molecules over time allows measurement of rates of transport and reactions for molecules in vivo, from mice to patients. While having limits in spatial resolution of today’s scanners of about 1 mm in animals and 3 to 5 mm in patients, the sensitivity is excellent with detection of labeled molecular probes typically in the range of pico to femtomoles/g tissue from injections of nanomoles of the labeled probe. Thus, measures can be typically performed without mass effects on the biological systems or safety concerns regarding pharmacologic / toxic risk to the patient.

Because of the relatively short half-lives of the most common radioisotopes used in PET, F-18 of ~2hrs and C-11 of ~20 min., new automated chemical synthesis technologies are being developed that employ conventional chemistry and in situ click chemistry in integrated microfluidics chips operated by a PC to accelerate, simplify and diversify the types of molecular imaging probes produced. There are commercial PET biomarker production facilities within 100 miles of about 98% of the hospital beds of America and within this range of the major universities and pharmaceutical companies. The new microfluidics technology provides for a new approach to ship chips to users to allow them to make whatever biomarkers and labeled drugs they choose.

Technologies are being developed to allow for the convenient translation of microPET and human PET scanners into measurement devices with outputs of information desired in biological and pharmaceutical studies. As part of this effort, a Kinetic Imaging Software (KIS) system has been developed that has three domains: 1) “Library” to define the terminology and principles of tracer and pharmacokinetics, 2) “Simulation” that contains compartmental models of biological systems with their mathematical engines along with blood and tissue time activity curves with easy ways to change transport and reaction rate constants and display results in graphs and simulated whole body images as a function of time and 3) “Analysis” to be used with real data to extract tracer- and pharmaco-kinetic parameters. KIS has been designed to be easy and fun to use. Examples of PET pharmacokinetic studies in mice and patients will be shown with labeled drugs. KIS is part of an overall image processing environment through which biological, pharmacological and medical information is provided.

PET studies in cancer patients will be used to illustrate the use of biomarkers and surrogate markers to:

1. Accurately diagnosis, stage and evaluate drug responses - responders vs non-responders – using FDG to image alterations in glycolosis.
2. Access the drug response of patients in different stages of cancer – drugs matched to stage of disease.
3. Pre-selection of patients for drug therapy by the presence or absence of the drug target using imaging probes specific for the drug target.

While many of the goals above can be realized today, there are others with technical and scientific hurdles to overcome to provide their benefits to biological and pharmacological sciences and the drug discovery and development processes, as well as the use of these developments in expanding in vivo molecular imaging diagnostics in the care of patients. It is reasonable and appropriate to integrate the goals and resources of molecular diagnostics and therapeutics to each other’s benefit. In addition, it is important to integrate in vitro and in vivo molecular diagnostics into a systems biology view of disease.

Speaker Biography

Dr. Phelps earned B.S. degrees in chemistry and mathematics at Western Washington State University in 1965, and a Ph.D. in chemistry, at Washington University, St. Louis, in 1970. Subsequently, he was on the medical school faculty of Washington University (1970-75), University of Pennsylvania (1976) and UCLA (1976-present). Dr. Phelps is the inventor of the Positron Emission Tomography (PET) scanner.

In addition to developing several generations of PET scanners with his colleague Dr. Edward Hoffman and CTI, Dr. Phelps and his other Ucla colleagues and students have used PET to study both the biological basis of normal organ functions, as well as numerous disorders of the brain and heart, as well as cancer.

In 1981, he established the use of imaging technologies for what today is called “Brain Mapping” for imaging how the brain performs various functions such as seeing, hearing, thinking, working and remembering with PET. Ten years later other imaging technologies such as fMRI were developed for brain mapping.

With his colleague Dr. Chugani, Dr. Phelps performed seminal studies with PET on the biological basis of how a child’s brain develops its behavioral repertoire, specialized learning in the formative years and unique means by which the child’s brain can reorganize to compensate for a lesion or surgical resection. Using PET, he and his colleagues in pediatric neurology and neurosurgery identified epileptic tissues, defined the surgical resection criteria and established a clinical service for the surgical treatment of childhood seizure disorders, from focal to hemispheric resections.

Along with his colleagues Gary Small, Dan Silverman, Pete Engel, Jeff Cummings and John Mazziotta, Dr. Phelps developed criteria for the use of PET in differentiating various types of dementia (e.g., Alzheimer’s, frontotemporal, vascular, etc.) early in the degenerative process, as well as the biological alterations in early stages of Parkinson’s, epilepsy and Huntington’s diseases. In Huntington’s and familial Alzheimer’s, metabolic abnormalities were identified with PET, 7 and 5 years before symptoms, respectively. It was also shown that PET provided a diagnosis of Alzheimer’s with a 93% accuracy 3 years before the clinical diagnosis of “probable” Alzheimer’s.

He and his colleague, Dr. Heinrich Schelbert, developed techniques with PET for the early detection of coronary artery disease and cardiomyopathies. They established PET as a gold standard for determining if tissue in the heart tissue was metabolically viable to allow selection of patients who would benefit from revascularization by bypass or angioplasty and who would not or required heart transplant.

Dr. Phelps and his colleagues also developed a technique with PET for imaging the entire body for cancer; to detect tumors, differentiate benign from malignant lesions, determine extent of metastasis and therapeutic effectiveness. Analyzing publications involving over 24,000 patients, Drs. Gambhir, Phelps and Coleman at Duke, established that PET improved detection, staging, detecting recurrent disease and assessing therapeutic responses in 12 different cancers, with an accuracy 8 to 43% higher than conventional imaging and changed treatment selection in 15 to 60% of the patients, depending on the clinical question. Today, the literature has expanded to over 60, 000 patients further confirming the earlier results published by Gambhir et. al.

Dr. Phelps has also led the worldwide transition of PET from research to clinical service, establishing the first clinical PET service at UCLA, obtaining FDA approval and Medicare and private insurance reimbursement.

Drs. Simon Cherry, Arion Chatziioannou and Phelps developed a miniaturized PET scanner, microPET, for imaging mice in biological and pharmaceutical research. Along with his colleagues Drs. Sam Gambhir, Harvey Herschman, Jorge Barrio and Nagichettiar Satyamurthy, a novel technique to image gene expression in vivo with PET was developed.

In 2001 Lee Hood, Jim Heath and Phelps established the “Alliance for NanoSystems Biology” between the Institute for Systems Biology in Seattle, Caltech and UCLA to merge systems biology, nanotechnology and molecular imaging. The focus of Alliance is to create new in vitro and in vivo molecular diagnostics along with the technologies to achieve them, based on a systems biology view of disease.

Recently, with his colleagues Drs Heath, Quake, Kolb and Tseng, Dr. Phelps developed technologies for accelerating, diversifying and simplifying the synthesis of PET molecular imaging probes, biomarkers and drugs with integrated microfluidics chips and new classes of high affinity and specificity molecular imaging biomarkers using “Click Chemistry” and traditional chemistry on a chip.

With colleague Dr. Henry Huang, he developed a software based Kinetic Imaging System (KIS) for performing tracer- and pharmaco-kinetics and pharmacodynamics with PET. This system combines the educational aspects of learning systems with mathematical engines buried in KIS and presented to the user in a simple game like approach for analyzing experimental data from mice to patients to provide the results desired by biologists and pharmaceutical scientists.

Dr. Phelps built a new combined basic science and clinical Department of Molecular and Medical Pharmacology at UCLA that includes the clinical PET and nuclear medicine services to bring together molecular diagnostics and molecular therapeutics from cells and mice to the care of patients. He also founded the Institute for Molecular Medicine and the Crump Institute for Molecular Imaging at UCLA.

Dr. Phelps has:

• Published over 670 peer-reviewed scientific articles, books and book chapters.
• Over 410,000 citations to publications.
• Been principle or co-principle investigator of over $245 million in grants.
• Received over $21 million in private donations to support his research.
• Received international honors and awards such as the George von Hevesy Prize, 1978, 1982), von Hevesy Foundation in Zurich (von Hevesy won the Nobel Prize in chemistry); the S. Weir Mitchell Award, 1981, Academy of Neurology; chaired the 1983 Nobel Symposium; the Paul Aebersold Award, 1983, Society of Nuclear Medicine; The Ernest O. Lawrence Award, 1884 from DOE; Rosenthal Foundation Award, 1987, American College of Physicians; the Enrico Fermi Presidential Award awarded by President Clinton, 1998; Kettering Prize, 2001, General Motors Cancer Research Foundation; Benedict Cassen Memorial Prize, Society of Nuclear Medicine, 2002.
• Been elected to the Institute of Medicine of the National Research Council in 1985 and in 1999 to the National Academy of Sciences.
• Become Chairman of the Board, Norton Simon Foundation; Chairman of the Board, Norton Simon Research Foundation; Member of the Board of the Norton Simon Art Foundation. These foundations have over $2.5 billion in assets.