Iconic modelling for the progressive transmission of neurological images: segmentation

摘要

We present a knowledge-based segmentation scheme for use in the transmission of high resolution medical images. Segmentation is used to generate a compact iconic model which can be transmitted rapidly to provide an early indication of image structure. The boundaries of the iconic image are modelled using a novel superelliptic shape-tree. Each part of the iconic image is progressively updated, using a set of rules that take into account viewing requirements, to provide all informative image build-up, in a timely manner. We show that a simple knowledge base is adequate to describe a wide range of variation in MR and CT images, and achieve a segmentation that can be modelled to provide the iconic image.