Model Extraction from Magnetic Resonance Volume Data Using the Deformable Pyramid

Jyrki Lötjönen,Pierre-Jean Reissman,Isabelle E. Magnin,Toivo Katila
DOI: https://doi.org/10.1016/s1361-8415(99)80031-0
IF: 10.9
1999-01-01
Medical Image Analysis
Abstract:A general framework for automatic model extraction from magnetic resonance (MR) images is described. The framework is based on a two-stage algorithm. In the first stage, a geometrical and topological multiresolution prior model is constructed. It is based on a pyramid of graphs. In the second stage, a matching algorithm is described. This algorithm is used to deform the prior pyramid in a constrained manner. The topological and the main geometrical properties of the model are preserved, and at the same time, the model adapts itself to the input data. We show that it performs a fast and robust model extraction from image data containing unstructured information and noise. The efficiency of the deformable pyramid is illustrated on a synthetic image. Several examples of the method applied to MR volumes are also represented.
What problem does this paper attempt to address?