Study On Methods Of Micro-Calcifications Detection In Mammograms

Breast cancer has become the leading cause of death among women in the city, which is a serious threat to the physical and mental health of modern women.It is known that the living pattern and environment contributed to the Breast cancer in some way, but it is impossible to find out the exact cause. The keys for decreasing the mortality are early discovery, early diagnosis and early treatment. Because of the limitation of radiologists’work experience and consistent work, computer-aided technology plays a very important role in the early diagnosis of mammograms. Micro-calcifications present an early sign of breast cancer. Because of the low contrast with issues and other factors, finding an effective way to detect the micro-calcifications in mammograms using computer aided technologies is one of the most significant objects on the research of breast cancer.This paper is focused on the image processing and analysis of the micro-calcifications auto detection based on the difficulties of micro-calcification detection and the characteristics analysis of biomedical images and mammograms in three modules. The main research work and innovations are as follows:1. The Mammogram Preprocessing Module. Based on the present situation that most extraction methods of breast region are highly complex, this paper presents an extraction method of breast region based on morphology. The experimental results show the proposed method has obtained the highly performance on the speed of extraction and the universality based on the complete and effective extraction of breast region.2. The Region of Interest Extraction Module. Based on the fact that medical images usually consist of weighted independent components and the strength of independent component analysis on high data statistical correlation, this paper proposed a method to extract the region of interest based on independent component analysis. Using FastICA based on maximization of negative entropy to ensure the speed of ICA convergence. We separated the false signals by means of artificial neural networks and support vector machine classifiers in order to extract of the region of interest. The experimental results show that support vector machine classifier does a better performance than artificial neural networks in both accuracy and simplicity, and the highest accurate rate and true positive rate are91.67%and93.33%. The compared results to the methods proposed by other papers showed the high performance on characters extraction of the mammograms based on independent component analysis3. The Micro-calcifications Detection Module. In the technologies of detection of micro-calcification, the key issue is improving the contrast between micro-calcifications and background. Based on the character performance in the frequency domain of mammograms and the time-frequency of wavelets, this paper proposed a method combing the wavelets analysis and wavelets domain denoising. Compared results to Top-Hat and partly weighted wavelet coefficients show that the proposed method has obtained the high removal of background and high contrast of micro-calcification by overcoming the weakness of high noise, which achieved the effective detection of micro-calcification.