A procedure for automatic identification of sulci of cerebral hemispheres in magnetic resonance images is presented. The algorithm departs from information of the spatial location of a given feature (e.g., sulcal contour) in one hemisphere and automatically defines its course in the other hemisphere. The automatic tracing of sulci was successfully performed in the mesial aspect of both hemispheres in 18 normal individuals. The sulcal location in the target hemisphere was first approximated by assuming an affine transformation between hemispheres, and then refined by local edge analysis. The method produced reliable results in comparing the intersulcal areas (the cuneus), sulcal length, their complexity, and the angle between the pariteooccipital and retrocalcarine fissures.