Microcalcifications detection using multiresolution methods

The detection of clustered microcalcifications can aid a radiologist to detect early breast cancer. Microcalcifications exhibit some important characteristics, like small size and high luminosity. So, a CAD method can be useful to avoid them being over-looked. In the paper a multiresolution analysis is proposed by decomposing the image through a band-pass filter bank, so that in each sub-band image only the details at the given scale become visible. Thereafter all the images will be combined in a final one in order to obtain an image that contains all the details of interest at the scale where microcalcifications tend to appear. Once the image is obtained, it is necessary to determine which details correspond with microcalcifications. Statistical analysis of the histogram permits us to classify the susceptible zones containing microcalcifications. Applying these statistics techniques over the whole image and representing the results in a 2-D map, clustered microcalcifications regions appear clearly distinguishable