WAVELET TRANSFORMS FOR DETECTING MICROCALCIFICATIONS IN MAMMOGRAMS

Clusters of fine, granular microcalcifications in mammograms may be an early sign of disease. Individual grains are difficult to detect and segment due to size and shape variability and because the background m ammogram texture is typically inhomogeneous. We develop a two-stage me thod based on wavelet transforms for detecting and segmenting calcific ations. The first stage is based on an undecimated wavelet transform, which is simply the conventional filter bank implementation without do wnsampling, so that the low-low (LL), low-high (LH), high-low (HL), an d high-high (HH) sub-bands remain at full size. Detection takes place in HH and the combination LH+HL. Four octaves are computed with two in ter-octave voices for finer scale resolution. By appropriate selection of the wavelet basis the detection of microcalcifications in the rele vant size range can be nearly optimized. In fact, the filters which tr ansform the input image into HH and LH+HL are closely related to prewh itening matched filters for detecting Gaussian objects (idealized micr ocalcifications) in two common forms of Markov (background) noise. The second stage is designed to overcome the limitations of the simplisti c Gaussian assumption and provides an accurate segmentation of calcifi cation boundaries. Detected pixel sites in HH and LH+HL are dilated th en weighted before computing the inverse wavelet transform, Individual microcalcifications are greatly enhanced in the output image, to the point where straightforward thresholding can be applied to segment the m, FROG curves are computed from tests using a freely distributed data base of digitized mammograms.