Optimal texture feature extraction in digital mammograms
In the present work we propose novel techniques for preprocessing and texture feature extraction in digital mammograms: The preprocessing scheme is based on dividing the input mammograms into blocks and compute an estimate of the normal background tissue contained in each block. Texture feature extraction is performed using an single filter optimized with respect to the relative distance between the average feature values. The methods are applied to a database of 43 mammograms — 23 containing one cluster of microcalcifications, and 20 containing no clusters. The results show that our CAD system performs very well in the detection of clustered microcalcifications. In particular, at a rate of about 1.5 false positive clusters per image the method reaches a true positive rate of 100%.
Unable to display preview. Download preview PDF.
- 2.K. Fukunaga, Statistical Pattern Recognition. San Diego: Academic Press, 2nd ed., 1990.Google Scholar
- 4.J. Suckling, J. Parker, D.R. Dance, S. Astley, I. Hutt, C.R.M. Boggis, I. Ricketss, E. Stamatakis, N. Cerneaz, S.-L. Kok, P. Taylor, D. Betal, and J. Savage, “The Mammographic Image Analysis Society digital mammogram database,” in Proceedings of the 2nd International Workshop on Digital Mammography, (York, England), pp. 375–378, Jul 1994.Google Scholar