Zhimin Huo PhD, Presenter: Nothing to Disclose

Fan Xu, Abstract Co-Author: Nothing to Disclose

Avice M. O'Connell MD, Abstract Co-Author: Nothing to Disclose

Comparison of prior and current mammograms is a common practice in detecting suspicious changes over time in screening mammography. Our enhanced digital mammography technology is aimed at reducing differences in mammographic appearance of the same breasts from one exam to another as a result of variations in imaging techniques and/or post-acquisition image processing.

Our technology consists of the detection of dense breast tissue in each image and tone-scale rendering to normalize the image based on the amount of dense tissue detected. In the dense tissue detection step, k-mean clustering was first applied to segment the breast region into three sub-regions. The size and shape (e.g., the compactness) of each region were calculated and compared to determine the likelihood of each region having dense tissue and if further classification was required for lower-likelihood regions. The normalization for the fatty and dense tissue was performed separately according to the amount of dense tissue detected in each image. We collected 42 cases, each with two screening exams within three years, all acquired and previously processed from the same or two different digital systems. We selected 15 cases with two exams each acquired from a different system. A mammographer was presented with each of the 15 pairs before and after our normalization processing was applied. The reader was asked to: 1) rank the similarity of the two exams in each pair in terms of brightness and contrast in the dense, fatty, and entire breast regions, respectively, on a 5-point scale (1-significantly different to 5-identical); and 2) give the preference as to which one of the two pairs (before and after) was preferred for comparison purpose.

The reader ranked the similarity in brightness and contrast consistently higher for the pairs after normalization was applied for 13 cases. Two cases were ranked the same before and after. Overall, the pairs after normalization were preferred for comparison purposes.

Our enhanced digital mammography technology can improve the consistency in brightness and contrast for mammographic images acquired across modalities. The improved consistency is preferred for temporal comparison.

Our enhanced digital mammography would help radiologists compare prior and current mammograms more effectively and efficiently, thus may potentially allow improving the overall workflow.

Huo, Z, Xu, F, O'Connell, A, Enhanced Digital Mammography: Dense-Tissue-based Image Normalization for Improved Consistent Appearance of Mammographic Images.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11016670.html