Matching a computerized brain atlas to multimodal medical images

A method for matching a digital brain anatomical atlas to multimodal medical images (MRI, PET, and SPET) was implemented. The digital atlas was derived from anatomical templates of the brain, cut according to the orbitomeatal orientation. The atlas consists of a set of contiguous slices schematically describing the brain as anatomical contours and of a set of regions of interest (ROIs) classifying the brain into functionally homogeneous areas. The matching procedure includes (a) an edge detection method for the extraction of anatomical contours and (b) a warping algorithm based on contour matching to fit the atlas to the individual brain anatomy, as described by MRI. Once the atlas is matched to MRI, the associated templates of ROI can be overlapped with functional PET/SPET studies, individually registered to MRI. The method was tested on MRI studies. The efficacy of the warping algorithm in overlapping atlas and MRI contours was assessed by calculating for each slice an index representing the extent of overlapping (I). Values of I in the range 0.8-0.9 were found (I = 1 complete overlapping). Local accuracy was also verified by comparing the position of correspondent anatomical ROI in the atlas and MRI images before and after warping. The atlas-matching procedure was applied to representative MRI/PET clinical images for an objective regional analysis of functional data.