Automatic boundary extraction of ultrasonic breast lesions

Breast cancer is the second leading cause of mortality in women. Automatically detecting tumors and extracting lesion boundaries in sonograms is difficult due to the specular and noisy nature of these images. Algorithms that claim to automatically classify ultrasonic breast lesions, rely on manual delineation of the tumor boundaries. This is however a slow and cumbersome task. In this work, we combine texture, intensity, gradient and a deformable model along with empirically determined domain specific knowledge to automatically find lesion margins in ultrasound images with no human intervention. The system has successfully extracted tumor boundaries in over 30 breast sonograms.