Object-surface-based extraction and manipulation of 3-D medical-imaging data

In acoustic, magnetic-resonance, or X-ray tomographic medical imaging, a number of slice cross-sections of elongated physiological structures are recorded, spaced relatively widely and nonuniformly along the object´s longitudinal axis. Traditionally interpolation between the observed slices is used to assign an appropriate density value to each cubical voxel in 3D space. In generating pictures from this set of voxels, curve fitting is then used to reduce the measurement and spatial-quantization noise. Such interpolation and curve fitting implies well-defined and well-behaved interfaces between regions of distinct physiological tissue. By formulating these assumptions explicitly, the author can drastically reduce the data, to a compact representation of the interface surfaces. This avoids processing-generated spatial-quantization noise, and opens the way for highly efficient algorithms for data extraction, manipulation, and display