Digital Imaging for Hospital and Community Based Mammography

Institution: University of California, Los Angeles
Investigator(s): Daniel Valentino, Ph.D. -
Award Cycle: 1995 (Cycle I) Grant #: 1KB-0272 Award: $221,716
Award Type: New Investigator Awards
Research Priorities
Detection, Prognosis and Treatment>Imaging, Biomarkers, and Molecular Pathology: improving detection and diagnosis

Initial Award Abstract (1995)
Early detection is crucial in reducing the economic and human costs of breast cancer. Digital mammography, which generates electronic images of the breast instead of the traditional photographic films, offers a significant advance in the early detection and diagnosis of breast cancer. In addition to providing earlier detection, digital mammography will facilitate the use of advanced computer programs to help all California women obtain more prognostic breast cancer examinations. The goal of this project is to develop an innovative computer information system to provide cost-effective digital mammography in a hospital or community clinic.

The proposed system includes the following technological advances, many of which have been developed in our laboratory and will be evaluated for use with digital mammography: direct digital acquisition and storage of whole-breast images; loss/less image compression (reducing the size of images without losing any information); 'intelligent' image management including the integration of images, radiology reports, and medical records; and an innovative computer workstation. We will develop an innovative workstation that we call the "virtual view box." The virtual view box allows a radiologist to view pairs of images, and quickly switch to other images for additional comparisons; this is ideally suited to the rapid comparisons that mammographers currently perform using film. We will develop new electronic tools to optimize the brightness and contrast of the images depending upon the density of the breast and type of lesion, as well as to rapidly view reports and image databases. We will use a rapid prototyping technique to develop and implement the mammography workstation so that it accommodates the actual tasks the mammographer is required to perform. Finally, the proposed system will facilitate the use of other advanced digital technologies, such as teleradiology (providing radiological services remotely) and computer-aided diagnosis (programs that provide additional diagnostic information to radiologists), by providing a digital environment in which computerized tools can help doctors screen more patients at lower cost.

The primary benefits of this system will be the ability to quickly and accurately screen young women under the age of 50; the decreased cost of operating digital systems versus film-based systems; and a unique potential to provide improved access for underserved populations via the use of a mobile digital mammography unit and teleradiology to provide subspecialist diagnosis to poor and minority women. In addition, using computerized tools that are being developed by institutions across the nation, digital mammography can improve the detection of suspicious lesions during screening. The intended result of this research is a test-bed for whole-breast digital mammography that, working with medical industry leaders, will be used to ensure that digital mammography products will be widely available to benefit California women.

Final Report (1998)
Digital Mammography represents a significant improvement over film mammography in the early detection and diagnosis of breast cancer; early detection is crucial for saving women's lives and for reducing the economic costs of breast cancer. The focus of this research is to develop computer systems that enable cost-effective digital mammography in a hospital or community clinic.

During the third year of this grant, we began to research the clinical acceptance of the image compression algorithms that we developed in the previous grant years. We demonstrated that Lossy compression ratios greater than 25:1 were unacceptable. However, by eliminating the background of the image beyond the nipple, we are able to reduce the original image size by 3/5 to 1/3. This enabled us to compress the remaining portion of the image to achieve acceptable image compression at significantly higher overall ratios.

The clinical images obtained during the previous grant years were found unsuitable for mass and calcification detection trials. We obtained only 83 abnormal mammograms out of 22, and the abnormalities were too obvious to obtain statistically significant results from a comparison of workstation accuracy to film accuracy. We therefore constructed a database of digital mammograms with simulated lesions by imagining cadaveric breasts with superimposed lesions of known size, and will use the simulated images for testing the diagnostic accuracy of hard copy versus soft copy.

Finally, we have continued the development of an infrastructure to support direct digital mammography. We investigated high-performance interfaces, networks and servers that enable direct digital mammography units to send images quickly to display workstations, and to store images permanently in archival devices.

To summarize our work over the past three years: First, we implemented an infrastructure that provides for the acquisition, archival, and distribution of full-field digital mammography images, and we have integrated it with our hospital-wide imaging and information system. Second, we acquired digital mammography research data sets from both volunteer patients and cadavers, and established an image database for evaluating systems for processing and display of digital mammography images. Third, we developed compression software to reduce the size of the original images. Using "lossless" methods we can restore a compressed image back to its original size without any loss of data. Using "lossy" methods we can restore the compressed image back to its original size, but the restored image differs slightly from the original image. We evaluated the ability of radiologists to detect breast masses in images compressed using lossy compression methods, and we demonstrated that the lossy methods were unacceptable when images were compressed by more than a factor of 25. Fourth, we developed workstation software for the rapid display and screening of digital mammograms. The display software developed under this project is available in a software toolkit called the UCLA Digital Viewbox/TkTm. The Digital Viewbox/TkTm is an application programmers toolkit that can be used to research and develop new processing and display algorithms which can accelerate the development of commercial research products, and thereby benefit California citizens.

Efficient review and retrieval of images from large-scale PACS. Medical Physics, Vol. 22, No. 6, June 1995.
Index Medicus: Radiology
Authors: Chou H, Chen MS, Vogel E, Huang L, Valentino DJ, Villasenor J
Yr: 1995 Vol: 197 Nbr: P Abs: Pg:258

Preclinical visual tests of whole breast digital mammography units. Medical Physics, Vol. 22, No. 6, June 1995.
Periodical:Medical Physics
Index Medicus: Med Phys
Authors: Kimme-Smith C, Terwilliger R, Bassett LW
Yr: 1995 Vol: 22 Nbr: Abs: Pg:918

Integrated High-Speed Network for REgional PACS. Medical Physics, Vol. 22, No. 6, June 1995.
Periodical:Medical Physics
Index Medicus: Med Phys
Authors: Valentino DJ, Duerinckx AJ, Hagan G, et al.
Yr: 1995 Vol: 2 Nbr: 6 Abs: Pg: