State-of-the-art in Medical Image Analysis

The human visual system is remarkably powerful in the tasks of recognition, segmentation and image retrieval. Much can be learned from recent neurophysiologic findings, such as with voltage sensitive dyes, showing neuronal firing activity over a large cortical area in experimental animals. The early stage of the visual system seems to measure mathematical derivatives in the image up to at least fourth order, has a multi-scale hierarchical structure, and has an elaborate mechanism for contextual analysis of orientations. We discuss possible physiological mechanisms for complex feature detection, for the detection of motion fields, edge preserving smoothing, and image segmentation, and image-based retrieval from PACS image databases. The methods turn out to be extremely robust, and match mathematically optimal schemes. Many illustrative examples will be discussed.