Biologically derived companding algorithm for high dynamic range mammography images.

Abstract

The screening mammography is currently the best procedure available for early detection of the breast cancer. The acquired mammograms are high dynamic range (HDR) images having a 12 bit grayscale resolution. When viewed by a radiologist, a single image must be examined several times, each time focusing on a different intensity range. We have developed a biologically derived mammography companding (BDMC) algorithm for compression, expansion, and enhancement of mammograms, in a fully automatic way. The BDMC is comprised of two main processing stages: 1) preliminary processing operations which include standardization of the intensity range and expansion of the intensities which belong to the low intensity range. 2) Adaptively companding the HDR range by integrating multiscale contrast measures. The algorithm's performance has been preliminarily clinically tested on dozens of mammograms in collaboration with experienced radiologists. It appears that the suggested method succeeds in presenting all of the clinical information, including all the abnormalities, in a single low dynamic range companded image. This companded and enhanced image is not degraded more than the HDR image and can be analyzed without the need for professional workstation and its specific enhancement software.