SOLICITATION NOTICE
R -- Developing Feature Indexing and Revelence FeedBack Techniques for Improved CBIR of Spine X-ray Images
- Notice Date
- 7/7/2005
- Notice Type
- Solicitation Notice
- NAICS
- 519190
— All Other Information Services
- Contracting Office
- Department of Health and Human Services, National Institutes of Health, National Library of Medicine, 8600 Rockville Pike, Bethesda, MD, 20894
- ZIP Code
- 20894
- Solicitation Number
- 05-143-CYC
- Response Due
- 7/22/2005
- Archive Date
- 8/6/2005
- Description
- It is the intent of the National Library of Medicine (NLM) to procure professional services from Brigham Young University, Provo, Utah (D.J. Lee) on a sole source basis. Period of Performance: November 1, 2005 - October 30, 2006. Professional services over a 12-month period to support the ongoing research in image processing for biomedical informatics research and development. The investigator will support data collection, algorithm R&D for relevance feedback and image indexing, and testing to meet the needs for the ongoing research at the National Library of Medicine in content-based image retrieval (CBIR) of digitized spine x-ray images. The investigator shall also provide medical validation for segmented vertebral shapes from approximately 1000 images of the digitized spine from the Second National Health and Nutrition Examination Survey (NHANES II). Task Description: Task 1: Shape-based relevance feedback for retrieval. Subtask A: Relevance Feedback: The investigator will focus on (1) improving existing relevance feedback technique to allow users to indicate feature characteristics of interest, (2) allow query expansion and (3) investigating and adopt pertinent features from other relevance feedback techniques that lend themselves to working with indexed features. Subtask B: Evaluation and Final Testing: A comprehensive evaluation and final testing of the whole system will be performed on expert marked data. A quantitative method for evaluating retrieval relevance and efficiency will be developed. Relevance will indicate quality of retrieved results while efficiency will depend on number of iterations and amount of user feedback required to achieve desired results. Task 2: Shape indexing for fast retrieval. Subtask A: Study of Existing Indexing Structures: Investigate the possibility of using an existing high-dimensional indexing structure for relevance feedback. The Government shall provide source code for hierarchical shape indexing. The investigator shall explore possibilities for modifying code to meet desired results. Investigate the possibility of adapting shape matching methods for relevance feedback technique developed from previous work. Subtask B: A New Hybrid Indexing Structure: Develop a new hybrid indexing structure that is suitable for PSM based relevance feedback. This includes establishing partial shape categories and low level shape matching using relevance feedback. The structure shall also allow for shapes with different resolution to be indexed and searched simultaneously or independently. Task 3: Shape data collection and Medical Validation. Data collection: At least 600 images provided by the Government shall be segmented to provide a sufficiently large database for this work. Additional images may be segmented upon request by the Government based on agreement with the contractor and on progress of the research work. The contractor shall use existing segmented shapes for research in addition to the shapes generated from this activity. Medical Validation: The investigator shall, through the involvement of a board certified medical expert, build an extended reference set of segmented vertebrae from the digitized images. The Government shall provide the investigator with segmented boundary shapes. The investigator shall collect from the board certified medical expert reviews on segmentations of vertebrae in approximately 1000 images of the digitized spine according to specifications and methodology negotiated with the government. The reviews will include validation with corrections, if necessary, of existing segmented boundaries, 9-point landmark markup validation, and pathology detail (anterior osteophytes, disc space narrowing, subluxation and/or spondylolisthesis, and other observations, if any) for 5000 vertebrae from images be provided by the Government. Both cervical and lumbar spine images and segmentations will be included. The protocol for segmentation review and the software will be provided by the Government. It is expected that the medical experts will have available to them the necessary hardware (512MB RAM, late model PC running Windows XP SP2) for the task. The following criteria need to be met in order for competing bids to be considered for this sole source contract: (1) The contractor shall represent an institution that exhibits a sufficiently long history of successful R&D work in the area from the Computer Science and Electrical Engineering departments. (2) The contractor shall be an internationally known researcher who possess a Ph.D. in Computer Science, Electrical Engineering or closely related field and at least 10 years experience in R&D in Computer Vision, Image Processing algorithms and techniques. (3) The contractor should have demonstrated a verifiable track record of past research on the specific topics covered in this statement of work and a successful completion of tasks on time. (4) The contractor should exhibit research background in the area exhibited by peer reviewed publications on the topic. (5) The contractor should have experience in addressing challenges in content-based image retrieval problems for biomedical images. Specifically, the contractor should have strengths in shape analysis, similarity techniques, and relevance feedback. (6) The contractor should have been exposed to challenges posed in the large collection of spine x-ray images, such as those in the NHANES II collection. (7) The contractor shall provide evidence of contact with a board certified medical expert in the pertinent field of research. The contractor should also be able to provide evidence of agreement with the said expert to obtain medical validation on the required number of medical images/shapes. Dr. D.J. Lee is an associate professor in the Department of Electrical and Computer Engineering at Brigham Young University. He received his B.S.E.E. from National Taiwan University of Science and Technology in 1984 and his master's and doctoral degrees in electrical engineering from Texas Tech University in 1987 and 1990, respectively. Prior to joining BYU in 2001, he served in the machine vision industry for over eleven years as system designer, researcher, and technical and project manager. His last employment prior to joining BYU was with Robotic Vision System Inc. (RVSI) where he served as the Director of Vision Technology and was responsible of designing the state-of-the-art high speed semiconductor chip and wafer inspection systems. He has designed over 40 real-time machine vision systems and products for various industries including automotive, medical, pharmaceutical, semiconductor, surveillance, and military, etc. His hands-on experience includes project costs and budget management, computer vision and image processing algorithms development, large-scale software system implementation, hardware design, and system integration. He has applied, received, and managed many research grants from federal government agencies as well as private sectors. His current research focus at BYU is in content-based medical image retrieval, shape-based pattern recognition, relevance feedback, medical image analysis, biomedical informatics, robot vision, motion analysis, and object tracking. Dr. Lee spent the summer of 2002 working with the researchers in the Communications Engineering Branch at National Library of Medicine on the content-based spine x-ray image retrieval project. He successfully evaluated and implemented several shape representation and matching methods and obtained and delivered valuable results. Since September 2003, Dr. Lee's research team has provided services in developing partial shape matching algorithms for content-based spine x-ray image retrieval project for the Communications Engineering Branch at National Library of Medicine. The recent project his team performed is on improving the retrieval system to allow the users to provide relevance feedback. The development project is near completion and Dr. Lee's team will be able to accomplish their tasks before the deadline. In collaboration with the researchers at NLM, he has since submitted and/or published eleven technical papers as author or co-author on content-based image retrieval of spine x-ray images and relevance feedback. Dr. Lee is uniquely qualified for this research work because of his understanding of the project, extensive experience in content-based image retrieval and relevance feedback, and thorough understanding of the characteristics of the image data. The proposed acquisition will be procured under FAR Part 13 - Acquisitions of Non- Commercial Items. This is not a Request for Quotations (RFQ), nor is an RFQ available. However, all responsive sources may submit a capabilities statement in a timely manner that will be considered by the Government. Sources interested in responding to this notice must be able to provide convincing evidence that they possess the requisite expertise and experience to successfully perform the services as specified above. Responses must be in writing and must be received in the office within fifteen (15) business days from the publication date of this notice. Proposals must include pricing information. NLM Synopsis No. NLM 05-143/CYC. Inquires regarding this procurement may be made to Cara Y. Calimano, Contract Specialist, NLM on (301) 496-6127.
- Place of Performance
- Address: 8600 Rockville Pike, Bethesda, Maryland
- Zip Code: 20894
- Country: USA
- Zip Code: 20894
- Record
- SN00842794-W 20050709/050707211827 (fbodaily.com)
- Source
-
FedBizOpps.gov Link to This Notice
(may not be valid after Archive Date)
| FSG Index | This Issue's Index | Today's FBO Daily Index Page |