Thor Ole Gulsrud

Associate Professor, Stavanger University College - Signal Processing Group

Texture analysis of digital mammograms

Breast cancer is a leading cause of cancer deaths among women. For women in Norway and in the other developed countries, it is also the most frequently diagnosed cancer. In 1992, about 2100 new cases of breast cancer were detected in Norway, and in the same year 17\% of all cancer deaths among women were caused by this type of cancer disease. Early detection is the most effective way to reduce mortality, and mammographic screening is currently the best method for early detection. An increasing number of countries have started mass screening programs which have resulted in a large increase in the numbers of mammograms requiring interpretation. In the interpretation process radiologists carefully search each image for any visual sign of abnormality.

Taber and Dean have classified the visual signs for which radiologists search during mammographic screening into three basic categories; circumscribed masses, stellate lesions and calcifications. From a texture point of view the ideal case would be that each category represents a well defined texture class, making the interpretation an easy task. However, a wide variability in tumor appearance exists. In addition, there is often a visual similarity of some tumors to normal structures in dense and fatty-glandular breasts. As a consequence, radiologists often find the task of interpretation difficult. It is not uncommon that in many centers films are studied by two radiologists in an attempt to reduce error rates. To improve the accuracy of interpretation, a variety of computer based aids have been proposed. The aim of these aids may be to detect and prompt specific signs of abnormality or to distinguish between benign and malignant tumors.

Ideally, we would like to construct a system in which all forms of abnormality are automatically correctly detected and classified. However, the current state of the art in mammographic image analysis renders this a long-term goal. Thus, our main goal is to design computer-based aids which will be useful in their own right, but will also eventually contribute to a completely automated system.

Conference paper abstracts

• Thor Ole Gulsrud and John Håkon Husøy, Optimal filter for detection of clustered microcalcifications. In Proc. 15th International Conference on Pattern Recognition, Barcelona, Spain, 3-8 September 2000, pp. 508-511.
• Thor Ole Gulsrud and John Håkon Husøy, Optimal filter for texture feature extraction in digital mammograms. In Proc. NOBIM konferansen 2000, Trondheim (Norway), 6-7 June 2000, pp. 61-66.
• Thor Ole Gulsrud and Eirik Løland, Multichannel filtering for texture feature extraction in digital mammograms. In 18th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Amsterdam, The Netherlands, 31 October - 3 November 1996.
• Thor Ole Gulsrud, Automated detection of clusters of microcalcifications using a multichannel filtering approach. In Proc. International Symposium on Digital Signal Processing 1996, London, UK, 22-25 July 1996, pp. 16-21.
• Thor Ole Gulsrud and Sissel Kjøde, Optimal filter for detection of stellate lesions and circumscribed masses in mammograms. In SPIE's Visual Communications and Image Processing 1996, Orlando, FL, March 1996, pp. 430-440.
• Thor Ole Gulsrud and Stein-Ole Gabrielsen, Classification of microcalcification using a multichannel filtering approach. In Proc. 17th Annual International Conference of the IEEE Engineering in Medicine and Biology Society and 21st Canadian Medical and Biological Engineering Conference, Montreal, September 20-23, 1995, pp. 889-890.
• Thor Ole Gulsrud and John Håkon Husøy, Image texture classification using Quadrature Mirror Filter bank in combination with Co-occurrence matrices. In Proc. SPIE International symposium on optical instrumentation and applied science: Applications of Digital Image Processing XVII, San Diego, July 1994, pp. 497-506.
• Thor Ole Gulsrud and John Håkon Husøy, Application of Quadrature Mirror Filter bank in combination with Co-occurrence matrices for classification of texture images. In Proc. NOBIM-konferansen-94, Asker (Norway), June 1994, pp. 53-59.
• John Håkon Husøy, Trygve Randen and Thor Ole Gulsrud, Image texture classification with digital filter banks and transforms. In Proc. SPIE International symposium on optical instrumentation and applied science: Applications of Digital Image Processing XVI, San Diego, July 1993, pp. 260-271.

Links to other WWW pages

• Stavanger University College, Signal Processing Group
• Stavanger University College WWW homepage
• Stavanger University College gopher-server

Address

• Office address:
  Stavanger University College
  Department of Electrical and Computer Engineering
  P.O. Box 2557 Ullandhaug
  4091 Stavanger
  Norway
  Tel.: +47-51 83 20 58 (direct)
  Tel.: +47-51 83 10 00 (switchboard)
  Fax.: +47-51 83 17 23
• Private address:
  Rektor Oldensgt. 52
  4022 Stavanger
  Norway
  Tel.: +47-51 89 69 75
  
• E-mail address:
  thor.gulsrud@tn.his.no