Boundary Point Detection for Ultrasound Image Segmentation using Gumbel Distributions

Due to high noise, low contrast, and other imaging artifacts, region boundaries in ultrasound images often do not conform to the assumptions of many image processing algorithms. Specifically, the beliefs that region boundaries have a high gradient magnitude or a high intensity can break down in this context. In this paper, we present an alternative way of detecting likely boundary points in ultrasound images by decomposing the image into one-dimensional intensity scans. These intensity scans, mimicking traditional A-Mode ultrasound, are modeled using Gumbel distributions. Results show that the relationship between the modes of these distributions and regions boundaries is relatively strong.