Research On The Cascade Detection For Micro-calcifications In Mammograms And The Compression For Digital Mammography

Breast cancer is a common form of malignant tumors in women. Early diagnosis and early treatment are the key to reduce breast cancer mortality of patients. Microcalcifications are early sign of breast cancer appearing as an isolated or a cluster of bright spots in mammography. However, it is very difficult to accurately detect microcalcifications due to their characteristics, such as, small size, low contrast from their surroundings, various shapes and distributions. Along with the fact that the public pays close attention to breast cancer, the large-scale breast cancer general investigation job already begins to be put into effect, so the large amount of mammogram have brought about suitable pressure for detecting and diagnosing. Hence, it is our urgent need to improve the efficiency of detecting microcalcifications. Now, various methods can be used to detect microcalcifications, among them typical Gauss- Laplace operator (LOG) is one of effective methods. Although LOG can accurately detect microcalcifications, it is inefficient. In order to improve the speed of detecting, many scholars attempt to cascade many algorithms or muti-scale of a algorithm.. The existing algorithms can't satisfy the high efficiency need of large-scale investigation since there is no direction in cascade. In this article, we study the general theories of cascading many algorithms or muti-scale of a algorithm. First, we establish the model by reasonable assumption; Then, we solve the model and obtain the theories of cascade. Finally, we propose a quick and efficient detecting algorithm in the detecting stage and a real-time and efficient detecting algorithm in whole stage by theories. We do our experience on the breast cancer database of Nanjing Zhongda Hospital and achieve an average costing time of 5.0317s, so the real-time detecting can be realized.The mammography has the characteristics of high resolution and large pixel depth, such as, the experimental mammography in this paper is digitized at 2294×1914 pixels with 16 bits. Along with the implement of large-scale breast cancer general investigation job, patients'mammogram need to be preserved for return visit, so mammogram compression is one of fundamental means to solve large-scale data transmission and memory. In this article, we study ROI-based mammogram compression by local K-L transform and linear SPIHT coding algorithm and achieve high compression efficiency and quality.