Digital Breast X Ray Image Enhancement And Calcification Extraction Algorithm

Breast cancer remains a leading cause of cancer deaths among women in many parts of the world. There are many of tools to examine the breast cancer in clinical diagnosis. Mammography is an effective technology to checking out the early breast cancer among those. Clusters of microcalcifications in a mamrnogram have much meaning to find an early indication of breast cancer. Calcifications look brighter due to its high density in mammogram. However, there are small in size and there are small differences in X-ray attenuation between normal glandular and malignant tissue, so that calcifications is hard to be detected for presenting low local contrast. In this paper several algorithms are studied and compared to enhance the contrast of mammogram. Two methods are investigated here to segment these calcifications from enhanced mammogram. Then one pre-procession algorithm adapted to mammogram was introduced to reduce the number of pixels calculated. To display gray picture deeper than eight bits, the pseudo-color enhancement algorithm were further investigated.1. Mammographic enhancement is conducted in spatial domain, frequency domain and discrete wavelet transform domain. Conclusions are reached as follows:a) Piecewise linear transformation is superior to others for mammographicenhancement in spatial domain comparatively. This transformation stretches contrast between micro-calcification and surrounding normal tissue, which increase the dynamic range of the gray levels in mammogram. Other methods, for example histogram equalization, stretch contrast by losing the total number of possible levels, so that those algorithms aren't suitable to mammographic enhancement.b) Enhancement algorithm in the frequency domain work through filtering by meansof the characteristic of image's Fourier spectrum. But the size of microcalcification is similar to that of noises, and the contrast between microcalcification and surrounding normal tissue is very low. So it is very hard to select suitable filtering function for mammogram enhancement in frequency domain.c) Enhancement algorithm by wavelet transform is well done in the part ofmammogram. But the desired effect couldn't be obtained when it is applied in whole mammogram for the gray value at edge of breast image changed greatly. To solve the problem, this paper put forward a method that divides the mammogram into several part and processes different part with different enhancing coefficient, which proved to have solved the problem well.2. Two kinds of algorithm are introduced for segmenting the microcalcification effectively.The Fuzzy C-mean (FCM) is employed to distinguish the brighter area in mammogram.To get the high contrast area, two algorithms can be employed. One is that the standard deviation are worked-out for each pixel in local area, and then all these value are segmented by one threshold. Another is segmenting the mammogram in wavelet transform domain.By combining FCM with standard deviation segmenting, or with segmenting in wavelet transform domain, these possible microcalcification can be got, which show bright and high contrast in mammogram.3. The pre-procession algorithm suitable to mammogram is put forward in this paper, which can decrease the number of pixel calculated later. It also can avoid the infection of noise and artifact in background.4. Aim at solving the problem of displaying gray image deeper than eight bits on PC, pseudo-color enhancement is further studied and the corresponding enhancement function is improved. The algorithm can be applied in mammography effectively to stretch the range of gray value that can be showed.