Fractal Methods Research For Microcalcifications Detection In Mammograms

With the prevalence of breast cancer all over the world, it has been a research focus in theprevention and treatment for breast cancer. The survival rate is low for the terminal breast cancerpatients. Therefore,“early detection, early diagnosis and early treatment” is the most effectivemeans for preventing and curing breast cancer. Existence of microcalcifications in mammogramscan be considered as the early sign of breast cancer. So, it has very important meaning in preventingand curing breast cancer. However, microcalcifications are usually tiny, and changeful in form, itwill be a hard and heavy work to search for them only by human eyes. Furthermore, somemicrocalcifications are likely not to be detected by negligence. Now, it has been a hot spot ofresearch and the developing trend of this field to take the advantage of the powerful calculatingperformance of computers for microcalcifications detection and breast cancer diagnosis asassistants of radiologists.In this paper, a set of integrated methods is presented for the detection of microcalcificationsin mammograms using computers. Firstly, extract the mammary area of mammograms using C-mean clustering algorithm and histogram. Secondly, extract ROIs（Regions Of Interest, mean theregions quite possible existing microcalcifications）in grid blocks of mammary area using box-counting dimension. Finally, detect microcalcifications in ROIs mainly using the proposed rapidfractal image coding method. Before extracting ROIs, an image enhancement method thatimplementing histogram equalization to the low frequency part of an image is presented tostrengthen the performance of box-counting dimension method in extracting ROIs. In this paper,an improved rapidly fractal image coding method is presented, where domain blocks are classifiedto minish the codebook, the coding speed is increased without loss of detection performance, andtwo adaptive thresholds set for the binaryzation processing, with high detection rate ofmicrocalcifications.In this paper, about100mammograms from the mini-MIAS database are used in theexperiments, and a lot of analysis and comparative work is done. It can come to a conclusion fromthe result of experiments, comparison and analysis that the methods presented in this paper havehigh detection rate and low false-negative rate, with certain practical application value.