K-means clustering for high-resolution, realistic acoustic maps

Abstract

In this work, we describe a method for converting fat-water-separated magnetic resonance imaging (MRI) volumes to acoustic maps for ultrasound simulations. An acoustic map is a mapping of acoustic imaging parameters such as speed of sound and density to grid points in the ultrasound simulations. Tissues are segmented into five primary classes of tissue in the human abdominal wall (skin, fat, muscle, connective tissue, and non-tissue). This segmentation is achieved using an unsupervised machine learning algorithm, called soft k-means clustering, on a multi-scale feature representation of the MRI volumes. We describe an automated method for utilizing soft k-means weights to produce an acoustic map that achieves approximately 90% agreement with manual segmentation. Two-dimensional (2D) and three-dimensional (3D) nonlinear ultrasound simulations are conducted, demonstrating the utility of realistic 3D maps over previously-available 2D acoustic maps.