Multimedia Stream Modeling, Relationships and Querying

Project Award Number: #IIS 0140384

Principal Investigator

Alfonso F. Cardenas
Computer Science Department
University of California, Los Angeles
3731 Boelter Hall
Los Angeles, Ca 90024
Tel. (310) 825-7550
Fax. (310) 825-2578
cardenas@cs.ucla.edu

http://www.cs.ucla.edu/csd/people/faculty_pages/cardenas.html

Co-PI

Denise R. Aberle
School
of Medicine
University
of California, Los Angeles
B2-165 Center for Health Sciences
Los Angeles, Ca 90024
(310) 825-9704
daberle@mednet.ucla.edu

http://www.radsci.ucla.edu:8000/professors/aberle.html

Collaborator

Robert B. Cameron
School
of Medicine
University
of California
62-215 Center for Health Sciences
Los Angeles, Ca 90024
(310)794-7333
rcameron@mednet.ucla.edu

http://www.surgery.medsch.ucla.edu/asp/Doctors.asp?emp_name=Cameron, Robert B. M.D.

Keywords

Multimedia systems, Stream modeling, Stream querying, Timeline

Project Summary

We investigate and develop stream-based data model constructs and the corresponding querying facilities using a timeline paradigm, in response to the growing requirements of advanced multimedia database applications. Using the basic concept of a stream, we develop substreams, aggregated streams, and derived streams.  The various types of relationships between and within streams, and with standard database entities, are explored and developed as part of the data model. These stream constructs add to existing SQL database facilities and object-oriented database facilities. We focus on the challenges and requirements of medical applications and of geophysical satellite remote sensing, requiring similar underlying stream data management and timeline access advances that we put forth. A patient data and physician tested and an environmental satellite data test bed will be used to validate and evaluate the effectiveness of the proposed technologies.

Publications and Products

Cárdenas, A. F., Pon, R. K., Michael, P. A., Hsiao, J. T., "Image Stack Viewing and Access," Journal of Visual Language and Computing, 2003, in press.

Cárdenas, A. F., Pon, R. K., and Cameron, R. B., "Management of Streaming Body Sensor Data for Medical Information Systems," The 2003 International Conference on Mathematics and Engineering Techniques in Medicine and Biological Sciences (METMBS '03), June 23-26, 2003, Las Vegas, Nevada.

Cárdenas, A. F. and Michael, P. A., "Image Stack Stream Model of Multimedia Data," Proceedings of the 2002 Conference on Distributed Multimedia Systems, San Francisco, California., September 26-28, 2002.

Cárdenas, A. F., Michael, P. A. and Islam, B. S. , "Stack Database Model/View of Multimedia Data," Proceedings of the 2002 International Conference on Information and Knowledge Engineering, Las Vegas, Nevada, June 24-27, 2002.

Project Impact

There is an explosive growth in the amount and variety of data streams, generated by the large numbers of sensors being deployed in many domains, and becoming more readily available through the internet as well as from stored traditional databases. Our efforts are aimed at enhancing the management and access to multiple related data streams from different sources through a timeline paradigm. We focus first on the medical domain and the opportunities of enhancing the access to patient data by not only doctors and the traditional health care providers but also by patients. We will test our advances and hypothesis at the UCLA medical center. We also investigate the data management challenges of streams of medical data coming on a possible real time basis from new generation mobile on-body medical sensors. In particular we are using the LifeShirt, with on board medical sensors, as part of a collaborative effort with its inventor and vendor VivoMetrics, Inc. We focus second on the geophysical domain and the challenges and opportunities for enhancing the management and access to a variety of data streams collected by real-time sensors on new NASA satellites. We will test our advances using NASA satellite data needed by colleagues dealing with environmental challenges. Thus, we will gauge the impact of our R&D advances.

Goals, Objectives and Targeted Activities

We have been developing the substream, aggregated stream and generic stream derivation constructs and explicit relationships constructs between streams beyond the base stream model that we developed in prior research needed by medical applications and geographic information systems (GIS) applications. We are pursuing the image stack data model for querying and viewing data coming from different types of data streams, which map to the same location of interest. For example, in the medical domain for superimposing MRI and CT images for the same patient body location, and in the GIS domain for superimposing co-registered pollution level, population density, and street location data. We are devoting efforts to the challenge of co-registering images needed for the image stack. We are developing the database model extensions needed for the SQL, ODMG, and JDO database standards to handle the multimedia stream database constructs and the image stack model. We have also initiated the extension to the SQL language and the native object query language OQL for both ODMG and JDO, to deal with a number of sample queries on multimedia streams corresponding to these stream model developments.

 

We are furthering the timeline paradigm as the high-level user-interface to the variety of data streams. We are developing proof-of-concept prototypes for medical patient data and geophysical data, including NASA satellite data. Within in this paradigm, we are researching different visualization techniques for a variety of data in both domains.

 

We have addressed the various physical database challenges and prototype implementation alternatives, focusing on a “quick and dirty” approach for fast proof of concept and a subsequent one for large and varied real-time data streams. We have initiated the development of the image stack model and viewing, and are working on the querying strategies to access data. Several key queries in the medical domain and the GIS domain are the focus of current work.

Area Background

Stream data management, sensor data management, multimedia image access, timelines

Potential Related Projects

NSF IDM Funded: Enabling the Creation and Use of GeoGrids for Next Generation Geospatial Information, Management and Processing of Data Streams, Management of Immersive Sensor Data Streams

Project Websites

http://www.mmss.cs.ucla.edu

Provides an overview and details of our overall project.

Illustrations

http://www.mmss.cs.ucla.edu/TimeLine.htm

Provides a screen shot tour or glimpses of the anticipated eventual prototype we are developing.

Online Data

http://goliath.cs.ucla.edu/ImageArchive/

This data set provides web access to several hundred real life medical images (mostly CR, CT, and MR) of patients, mostly afflicted by cancer. Actual patient identity and demographic identity is deleted as required by confidentiality regulations.  This dataset content was gathered during a prior NSF funded project, the KMeD project.