Automatic segmentation and diagnosis of breast lesions using morphology method based on ultrasound

The main objective of this paper is to use the auto segmentation with morphological technique to find out predictable region of interest (ROI), especially the center and margin area of the tumor. The proposed method has employed moving average method for detecting edge of tumor after estimating the corresponding center using the aid of medical domain knowledge. In our re-search, after computing distance between center and edge of tumor we get factual and numerical data of tumor to calculate multi-deviation and circularity test. It is useful to construct tumor profiling by splitting up the lesion into 4 divisions with the mean of multi-standard deviation (benign: 13.7, malignancies: 38.32) and 8 divisions with the mean of multi-standard deviation (benign: 3.36, malignancies: 15.29) with equal segments. We used K-means algorithm to make classification between benign and malignance tumor. This technique has been fully validated by using more than 100 ultrasound images of the patients and found to be accurate with 90% degree of confidence. This study will help the physicians and radiologist to improve the efficiency in accurate detection of the image and appropriate diagnosis of the cancer tumor.