A Methodology for Detecting Vessels in X-Ray Mammogram Images

Identification of the blood vessels and ~ilk ducts within an X-ray mammogram image al~ws intelligent analysis of calcification clusters and lass borders detected by other means. These vessel mtures are however generally buried within noise ’ith variance of the order of the vessel signal it,ll. We present a scale-matched feature detector esigned to extract nominally 1-dimensional ridges ̄ ore an image with a signal-to-noise ratio approxllatlng unity. The algorithm uses first and second ifference gradient measures of the original non~aoothed images to drive a search exploiting the priori knowledge of the expected vessel (ridge) fea~res. The algorithm is illustrated with an ex~mle.