Fast automated object detection by recursive casting of search rays

Abstract Anatomical object detection and coarse delineation is a frequent task in medical image processing. Tight time-constraints in diagnostic workflows and at the same time steadily increasing image matrices call for fast target–object and not image-matrix related procedures. In this context, we present a scheme for search ray based object detection in medical images. The scheme consists of a set of building blocks for generation, intensity sampling and analysis of search rays. Limitations caused by the one-dimensional sampling along the rays are overcome by a recursive scheme: the results of a search are used to set up a subsequent search in order to probe the image data in other directions to either confirm or discard the results of the previous search. The presented scheme covers a broad potential range of anatomical objects to be detected or coarsely delineated. As examples, we describe the implementation for finding the center of the heart and for coarsely delineating the myocardium in cardiac CTA scans.