Research On Edge Extraction Of Breast Tumors In Ultrasonic Images

Breast tumor is one of familiar female diseases, and breast cancer is also the most prevalent malignant tumor in breast tumor, its disease incidence rate and the mortality rate occupy head of the feminine malignant tumor. The curative effect of breast cancer lies on the early diagnosis and treatment. So it is more imminently needed to find out the breast cancer in the very beginning. Ultrasonic inspection is adopted widely in early diagnosis of breast tumor for its specialty of harmlessness, high resolution and sensitivity, and it usual uses tumor boundary, the shape and the echo distribution and other ultrasonic picture characteristic to distinguish the benign tumor and the malignant tumor. The edge extraction of tumor is an important precondition of obtaining these characteristics. Therefore, accurate extraction of breast tumor edge has the vital significance.A novel method for edge extraction of breast tumors in ultrasonic images based on the region growing and multi-scale morphology is proposed. The central point of the picture is chosen as the initial seed spot, which can restrain texture noise of adipose and gland tissue around the tumor. The average gray level of target region and homomorphic optimal threshold are taken as the decision criterion of region growing to get the original tumor region. Multi-scale morphology filtering and connected component labeling are used to extract the exact outline of tumor.Experiments indicate that this method can extract tumor edge automatically, which provides important reference for computer-aided diagnosis of breast tumors in ultrasonic images.