Contrast enhancement procedure for digital images of dense breasts by using tissue attenuation coefficients

Computer-aided diagnosis (CAD) schemes are developed all over the world to aid radiologists in order to increase the probabilities of early detection of breast cancer. Among the challenges to overcome by these schemes, one is very important: the contrast enhancement in dense breast images, which frequently present poor contrast among the structures of clinical interest and the background due to predominance of fibroglandular tissues. In a previous work, we have investigated the application to dense breast images of some generic image processing techniques intended to enhance contrast, as well as the results from a CAD scheme aimed to detect clustered microcalcifications. Nevertheless, the results were not satisfactory. Thus, in this work we are discussing the proposal of a new technique to enhance the contrast specifically designed for digitized dense breast mammograms. This procedure uses the linear attenuation coefficients (/spl mu/) relative to the tissues commonly involved in dense breasts in order to determine the levels of absorption and, from those, to enhance the levels corresponding to microcalcifications. The resultant images were submitted to a previously developed processing for clustered microcalcifications detection. The results indicate that more structures could be detected in the images submitted to this new technique without increase in the number of false-positive.