|
Semi-automated Boundry Tracing of Medical Images for Three-Dimensional Model Development |
|
|
Principal Investigator: |
Charles Anderson, Associate Professor |
|
Collaborating Companies: |
|
|
Company Representative: |
Thomas McCracken, President |
|
Segmentation is a critical step in many image processing applications. It is usually performed manually, since automatic techniques yield results which are less than satisfactory. We are experimenting with neural networks that learn a boundary discrimination given an initial hand-traced boundary segment, then automatically continue the tracing. Visible Productions produces three-dimensional human models that are recognized as some of the most accurate models in the world. Their current procedure requires many person-hours of hand-tracing regions in cross-section images. Visible Productions has identified our semi-automated technique as a technology that would greatly reduce their production time and cost. |
|
|
Modeling Student Pilots for Intelligent Training |
|
|
Principal Investigator: |
Charles Anderson, Associate Professor |
|
Collaborating Companies: |
|
|
Company Representative: |
Thomas Donohue, Vice President Business Management |
|
CTA Simulation Systems is developing pilot training systems for NASA. Based on their experience, student pilots not only learn at different rates, but in different manners. Some students, for example, tend to overcompensate early in their training while others experience problems with take-off and landing. The challenges presented to each student must be tailored to their unique learning experiences. This requires an intelligent training regime that exploits a model of each student that predicts where the student's performance will be deficient. This project will develop machine learning algorithms for predicting student performance as a basis for future intelligent training systems. |
|
|
Project Title: |
Automated Velocity Picking: A Computer Vision and Optimization Approach |
|
Principal Investigator: |
Darrell Whitley, Associate Professor |
|
Collaborating Companies: |
|
|
Company Representative: |
Barry Fish, Staff Geophysicist |
|
"Velocity Picking" is the problem of picking velocity-time pairs based on a coherence metric between multiple seismic signals. Coherence as a function of velocity and time can be expressed as a 2-D color image representing the "Semblance Velocity." Currently, humans pick velocities by looking at the Semblance Velocity image; picking velocities for a seismic survey can take days or even weeks. Automating the process as pure optimization without exploiting the Semblance Velocity image yields an essentially intractable problem. The problem can also be posed as a geometric feature matching problem similar to those which we currently solve in computer vision. A feature extraction algorithm can recognize islands (peaks) of maximal power corresponding to velocities in the Semblance Velocity image: a heuristic combinatorial matching process can then be used to find a subset of peaks which maximizes the coherence metric. Several heuristic search techniques will be utilized and compared, including random starts local search and CHC, as well as a new method, "Path Relinking". |
|
|
Project Title: |
Bandwidth Measurement and Tracking of Narrow Band Signal Components |
|
Principal Investigator: |
Tamal Bose, Associate Professor |
|
Collaborating Company: |
Data Fusion Corporation |
|
Company Representative: |
John Thomas, Vice President |
|
The goal of this project is to develop a suitable scheme for measuring the bandwidth of a very narrow band signal component in a spectrum which also contains several other sinusoidal modes corrupted with noise. The algorithm should be capable of continuously tracking the changes in this bandwidth. The proposed system consists of several filters, decimators and adaptive algorithms. The system is expected to enhance the narrow band components and perform very accurate measurements of bandwidth and frequency. One of the adaptive algorithms used is called the Fast Conjugate Gradient (FCG) algorithm. It was developed by the PI as part of a 95 (% CASI project). This algorithm is used along with other algorithms and filters for developing the overall system of this project. The developed scheme will also be studied for implementation in a fixed point processor. Fixed point processors offer several advantages over floating point in terms of speed, compactness, cost and power consumption. The algorithms and filters of this scheme will be theoretically analyzed for quantization noise. Several conditions will be derived on the filter coefficients and algorithm parameters so as to minimize quantization noise and ensure stability of the filters. This will in turn maximize the accuracy in the frequency and bandwidth measurements. |
|
|
Project Title: |
Meaurement of Implant Component Position and Orientation From X-Ray Images |
|
Principal Investigator: |
William A. Hoff, Assistant Professor |
|
Collaborating Company: |
Rose Musculoskeletal Research Laboratory |
|
Company Representative: |
Richard Komistek, Director of Development |
|
Recently, total joint hip and knee implants have revolutionized orthopedic health care, enabling millions of people to walk again. However, implants can wear out prematurely, requiring additional surgery to replace the dysfunctional prosthetic joints. To solve this problem, there is an urgent need to measure the position and orientation of implant components in patients undergoing normal activities. Recently, the PI and his collaborators at the Rose Musculosketetal Research Laboratory (RMRL) have developed an algorithm and software system to provide this data using X-ray fluoroscopy images. This system has been used for over a year and although highly successful, is limited in accuracy and requires a great deal of operator effort. The objective of the proposed research is to develop and evaluate a new system that directly addresses these problems. |
|
|
Project Title: |
Predicting Program Behavior to Support Instruction Level Parallelism |
|
Principal Investigator: |
Dirk Grunwald, Assistant Professor, Department of Computer Science |
|
Collaborating Company: |
|
|
Company Representative: |
Thomas Christian, Engineer/Scientist |
|
The principal investigators will investigate the trade-offs between information gathered and the optimizations that are enabled by information collected. In particular, they will investigate the range of possible gathering times from compile-time (via source), to link time (via executable), to load time (incorporating shared library information) to execution time (using profiling or hardware). The research will investigate specific optimizations, such as branch prediction and load latency prediction, and look at the impact of having different information available on overall program performance. This is the second year of a two-year project. |
|
|
Project Title: |
|
|
Principal Investigator: |
Scott Leutenegger, Assistant Professor |
|
Collaborating Company: |
|
|
Company Representative: |
Jean-Claude Franchitti, President |
|
This project will concentrate on developing techniques for efficient loading, retrieval, and update of multi-dimensional data. The PIs expect to adapt techniques from computational geometry, a field that has matured independently of databases, as well as to develop new ones. The PIs intend to transfer their research into the extensible hybrid database system being developed by MetaComp. A database product that incorporates the PIs' proposed techniques for efficient support of multi-dimensional data will be competitive in the fast growing markets of scientific and geographic databases. This is the second year of a two-year project. During the coming year, the indexing structure will be extended to handle 10-20 dimensions. Support for general data types and arbitrary predicates will be developed. An interface to MetaComp's Object Query Language (OQL) compiler will be developed. Proximity queries and new insertion/deletion algorithms for dynamic data issues will be addressed. |
|
|
Project Title: |
Control of Sensor Information in Distributed Multisensor Systems |
|
Principal Investigators: |
Lucy Pao, Assistant Professor |
|
Collaborating Company: |
Data Fusion Corporation |
|
Company Representative: |
John Thomas, Vice President |
|
Because multiple sensors in many surveillance systems provide more information than can be processed with the available computational resources, this project proposes to develop a controller that manages the sensor information. Based upon our understanding of controllers for centralized processing architectures, which we are developing in a parallel CASI Unsolicited Technology Transfer Project, we propose to extend and adapt these control approaches for general distributed processing architectures. While the mathematics is more tractable for a centralized processing architecture, considerations such as survivability, computational resources, and communication bandwidth often make distributed processing architectures the only alternative. This distributed controller that we will develop will mange the rates and resolutions at which information from various sensors need to be processed, taken into account the quality of the sensor information as well as which quantities the sensors measure. |
|
|
Project Title: |
Multiuser Detection and Diversity Techniques for Indoor Wireless Communications |
|
Principal Investigator: |
Mahesh Varanasi, Associate Professor |
|
Collaborating Companies: |
|
|
Company Representative: |
Steven L. Maddy, Director, RF Engineering |
|
This project proposes noncoherent multiuser detection and diversity methods for indoor wireless communication systems in order to (a) achieve near-far resistance (b) support high user densities and (c) make highly efficient use of spectrum and power. Our task in the first year of this project would be carry out theoretical research aided by computation and software simulation to obtain noncoherent modulation-detection-diversity schemes that meet our objectives. The second year of this project will focus on developing new signal processing software modules that incorporate those schemes into standard communications software packages such as SPW or System View. The goal here is to provide quick validation, testing and performance evaluation of our theoretical results when our subsystems are embedded in the SpectraLink Pocket Communication System (PCS). |
|
Back to Current Projects home page.
© Colorado Advanced Software Institute 1997