AUTOMATIC 3-D MODEL-BASED NEUROANATOMICAL SEGMENTATION

Explicit segmentation is required for many forms of quantitative neuro anatomic analysis. However, manual methods are time-consuming and subj ect to errors in both accuracy and reproducibility (precision). A 3-D model-based segmentation method is presented in this paper for the com pletely automatic identification and delineation of gross anatomical s tructures of the human brain based on their appearance in magnetic res onance images (MRI). The approach depends on a general, iterative, hie rarchical non-linear registration procedure and a 3-D digital model of human brain anatomy that contains both volumetric intensity-based dat a and a geometric atlas. Here, the traditional segmentation strategy i s inverted: instead of matching geometric contours from an idealized a tlas directly to the MRI data, segmentation is achieved by identifying the non-linear spatial transformation that best maps corresponding in tensity-based features between a model image and a new MRI brain volum e. When completed, atlas contours defined on the model image are mappe d through the same transformation to segment and label individual stru ctures in the new data set. Using manually segmented structure boundar ies for comparison, measures of volumetric difference and volumetric o verlap were less than 2% and better than 97% for realistic brain phant om data, and less than 10% and better than 85%, respectively, for huma n MRI data. This compares favorably to intra-observer variability esti mates of 4.9% and 87%, respectively. The procedure performs well, is o bjective and its implementation robust. The procedure requires no manu al intervention, and is thus applicable to studies of large numbers of subjects. The general method for non-linear image matching is also us eful for non-linear mapping of brain data sets into stereotaxic space if the target volume is already in stereotaxic space. (C) 1995 Wiley-L iss, Inc.