Detection of microcalcifications in digitized mammogram film images using wavelet enhancement and local adaptive false positive suppression

Microcalcification clusters are the primary radiological indicator of early breast cancers. Microcalcifications appear as small, bright specks in mammogram films. Detection of these abnormalities is often hampered by the presence of prominent background generated by glandular and connective tissue. There exists a great need for computer-assisted techniques to prompt radiologists to examine potential abnormalities that might otherwise be missed. The authors have developed a method for the detection of microcalcification clusters in digitized mammogram film images. This is a multi-step process consisting of wavelet enhancement of objects occurring at scales characteristic of microcalcifications, local adaptive thresholding of the enhanced image and false-positive object suppression. Results from the application of these methods to test images resulted in a true-positive microcalcification cluster detection rate of approximately 94%. The total false-positive rate was 0.21 false-positives per true-positive microcalcification cluster detection.