Breast Biopsy Microscopic Image Segmentation And System Implementation

Breast cancer is a common gynecological disease, and it is the malignancies that impact women’s lives and health. Early detection is one effective way to improve the cure rate, but the current detections are mostly dependent on the microscope and the personal experience from pathologists, which will make the test result more subjective and differences. With the development of image processing technology, the computer-aided detection system has been proposed, breast biopsy microscopic image segmentation is one of the key questions need to be solved firstly. A good segmentation algorithm can provide objective and accurate "second idea" for pathologists. Therefore, in this paper, I will take the real breast biopsy image as research object; do research on digital image processing technology deeply, and meanwhile, combined with the specific experiments to do the segmentation and feature extraction on breast biopsy image. The main works are:1. In the stage of the image pre-processing, the thesis proposes the method of composite pre-processing. It includes gray scale of the image, image enhancement, image filtering, and the image from the color image into a gray image of the strong contrast.2. In the stage of the image segmentation, the thesis proposes the method of combination of segmentation, Firstly, compared several threshold method for image and choose the otsu, then completed the first division after combination method including morphological hole filling, decreasing of noise and blob marker.3. In the stage of clump splitting, the thesis improves the method of clump splitting based on the bottleneck rules, and proposes the method of automatic recognition and clump splitting. The paper uses the number of concave points to classify clump splitting and chooses different methods according to different types of segmentation on clump splitting.4. Cell feature extraction and realize the system.Experiments show that the thesis’s method of segmentation of microscopic image of pathological of breast cancer is effective, and segmentation of the clump splitting is automatic, this method can play a role in pathological diagnosis, and lay a good foundation for the research of computer-aided diagnosis.