Lesion Areas Detection In X/CT Image

In recent years, with the improvement of living standard, people pay more and more attention to their health, which makes the doctor confront more and more heavily burden, especially for the radiologist abundant inspection not only increase their workload but also reduce the accuracy of inspection. Moreover, the rapid development of medical imaging technology make all parts of the body inspect more directly and clearly, which provides more favourable conditions for the computer-aided diagnosis of the medical imaging. This paper does the work on the basis of previous studies, which carry on research on mammograms and stomach CT images as following:Breast cancer is the top killer of women’s health, and mammography is the most frequently-used inspection method. Mammogram has the features that personalized greatly, more interference information and imbalanced properties. To these problems, this paper presents a personalization partition method for the detection of suspicious region in mammograms. By the personalization partition method, the problems that whole image with too many noise region and nonuniformity distribution on the region of interest are resolved. Furthermore, the v-SVM classification is used to deal with the imbalanced problem. Ultimately this method forecast a case which is ill or not and further to forecast which part the suspicious region belongs to, facilitated the following procedure.Mammograms is gray image, and the mass in mammogram is usually high concentrative in centre and arc-shaped outstanding in edge mass. According to the characteristics of breast masses above, this paper introduces visual attention model to the detection of breast cancer, the information entropy is applied to obtain the saliency map to reflect the masses arisen in normal region. Then, morphological features are used to remove the redundant regions in the binary image obtained by segmenting saliency map, support vector machine with over-sampling imbalance classification method is used to reduce false positive rate.Gastric cancer is one of the most common cancers, and takes the first incidence in all the malignant tumors in our country. Stomach CT image is a frequent measure for the detection of gastric cancer. And in the stomach CT images, the size, position and amount of lymph nodes play an important role in the diagnosis of the gastric cancer. A method detecting ROIs (region of interesting) in stomach image is proposed in this paper. Based on the detecting result, doctors can pay attention on only a local region than the whole; furthermore, the proposed algorithm can provide a useful method for other researcher studying on the detection of lymph nodes.