The Research Of Mammogram Image Segmentation Method Based On Microcalcification Detection

Computer-Aided Diagnosis (CAD) techniques provide potential necessaries for the early detection and recognition of breast cancers. One of the early signs of breast cancer is the presence of micro-calcification clusters at the mammogram of symptomatic women. However, a number of such findings especially the micro-calcifications that have small size and low contrast could be missed or misinterpreted by doctors. For this reason, a reliable automated computer-aided diagnosis system (CAD) could be very useful. On the one hand, it can provide a valuable "second opinion" to a radiologist or doctor. On the other hand, it can provide more objective information for physicians to make further diagnosis.A CAD system proposed in this article is presented for the identification of micro-calcification clusters and the classification of benign or malignant in digital mammograms. The proposed method is based on a five-step procedure: preprocessing and segmentation, regions of interest (ROI) specification, and feature extraction and classification, the identification of micro-calcification clusters, the classification of micro-calcification clusters. This article mainly completes the first two modules, which are the foundation of the overall system. The key is to discover the micro-calcifications and regions of interest (ROI) existed in the mammograms, preparing for the corresponding processing module.First, this article completes the function of the first module through following several steps: at the beginning of preprocessing we carries on image normalization to transform a 256 levels-grays image, the second step we applies an improved median filtering algorithm to smoothing & clearing noise, then uses contrast enhancement algorithm for mammograms based on Gray homogeneity to enhance the image. After preprocessing, in the segmentation module we mainly use the threshold segmentation, image difference, edge detection and region growth ways to obtain the integrity of the micro-calcification information. The experiment shows that micro-calcifications detection rate could reach 94.8% with FP 2.41.Second, the second module is presented for an auto region of interest extract technique based on morphological operators. The Concrete mentality is as follows: first, it is necessary to identify neighboring pixels with connectivity of eight in the segmented image obtained in the previous stage. Second, eliminates these individual objects achieved through the use of morphological erosion operators. Next, apply the dilation operators in order to produce a ROI with sufficient area around the object. At last, identify groups of objects that are candidates for micro-calcification clusters and specify the ROIs. The experiment shows that ROI detection rate could reach 93%.The complement of two modules studied in this article may help reduce the following module operand, improve the sensitivity level, and accelerate the intellectualized advancement of the computer assistance diagnosis system.