Abstract:
Globally, cancer is becoming a major health issue as advances in modern medicine continue to extend the human life span. In the U.S., cancer is the second most-common cause of death, exceeded only by heart disease and accounting for nearly one of every four deaths. Breast cancer ranks second as a cause of cancer death in women (after lung cancer).
Thus, early detection and treatment are critical in reducing breast cancer related mortality. Working with Breast ultrasound (BUS) data or image is regarded as a challenging task due to the inherent nature of ultrasound imaging. Ultrasound imaging is characterized by speckle patterns, anisotropy and signal drop-out. Moreover, proper image acquisition techniques by the clinicians and their level of expertise also play a dominant role in determining the image quality. The fuzziness in the shape and boundaries of the breast lesions make it very difficult to automate the segmentation of BUS images. In order to improve the issues prevalent in the existing approaches, a complete qualitative and quantitative analysis of Breast ultrasound (BUS) images is proposed in this thesis. The method involves three steps – (a) Finding out strain image by means of strain estimation, (b) Final Segmentation of the detected lesion and (c) Quantitative analysis of the lesion.
