A fast way to visualize the brain surface with volume rendering of MRI data

The preprocessing of 3-dimensional (3D) MRI data constitutes a bottleneck i n the process of visualizing the brain surface with volume rendering. As a fast way to achieve this preprocessing, the authors propose a simple pipeli ne based on an algorithm of seed-growing type, for approximate segmentation of the intradural space in T1-weighted 3D MRI data. Except for the setting of a seed and four parameters, this pipeline proceeds in an unsupervised m anner; no interactive intermediate step is involved. It was tested with 15 datasets from normal adults. The result was reproducible in that as long as the seed was located within the cerebral white matter, identical segmentat ion was achieved for each dataset. Although the pipeline ran with gross seg mentation error along the floor of the cranial cavity, it performed well al ong the cranial vault so that subsequent volume rendering permitted the obs ervation of the sulco-gyral pattern over cerebral convexities. Use of this pipeline followed by volume rendering is a handy approach to the visualizat ion of the brain surface from 3D MRI data. Copyright (C) 1999 by W.B. Saund ers Company.