Measuring image texture to separate "difficult" from "easy" mammograms.

We are investigating computerized techniques for sorting mammograms according to whether the breast tissue is fatty or dense. The hypothesis is that areas of dense tissue are a major factor in making certain mammograms harder for both radiologists and computers to interpret. Being able to identify dense mammograms automatically could permit better use of the time and skills of expert radiologists by allowing the difficult mammograms to be examined by the most experienced readers. In addition, the scope for computer-aided detection of abnormalities might be increased by concentrating on the easier, fatty mammograms. The mammograms used in the experiment were classified independently by two radiologists, who agreed in almost all cases. A number of local statistical and texture measures were then computed for patches from digitizations of these mammograms. One of the measures (local skewness in tiles) gives a good separation between fatty and dense patches. This measure has been incorporated into an automated procedure that separates off approximately two thirds of the fatty mammograms. This finding has been replicated on mammograms taken from a UK screening programme. The relationship between the fatty/dense distinction and the classification proposed by Wolfe is discussed.