The use of image texture analysis in the detection of micro-calcifications within the mammogram

An algorithm is reported for computer aided diagnosis of micro-calcifications in digital mammography. The procedure includes: (1) image texture analysis in X and Y directions; (2) spatial feature extraction; (3) density analysis to extend the sensitivity of the method to include subtle micro-calcifications grounded on a dense background; (4) morphology to group the responses; and (5) simple image enhancement to eliminate false small individual stimuli. The algorithm is straightforward and adapts well to variation in micro-calcification shape. When applied to complete mammogram images, the main focus was consistently within the region of interest (ROI) provided by an expert radiologist. Although the initial rate of false positive (FP) response is 2.25 events per image, these are always smaller in area than the main true positive (TP) response. Research is continuing to further optimise of the algorithm to reduce the number of false stimuli and establish the method´s overall clinical value