SUPERPOSITIONING OF 3-DIMENSIONAL NEUROANATOMICAL DATA SETS

Superpositioning of multiple 3-dimensional digital volumes is essential to extending neuroimaging capabilities from single-subject analysis to group evaluation. In this study, 6 rat heads were cryoplaned and digital images of the specimen blockface were organized into data volumes. The midline plane of each brain guided coronal and horizontal rotations and medial-lateral positioning. Four different methods of determining sagittal rotation, dorsoventral position; and anteroposterior position were tested. The first technique uses lambda and bregma as fiducial points; the second method employs bony sutures in the base of the skull; and the third strategy fits both sets of alignments points. The fourth method differs from the previous landmark-based approaches by employing cross-correlation of image densities. A composite of midsagittal cutplanes through the volumes aligned by the first method revealed significant misalignment. Use of ventral fiducials resulted in improved correlation statistics for orthogonal cutplanes through the volumes and significantly reduced positional variability of anteroposterior landmarks. Combining dorsal and ventral fiducials produced a slight improvement in midsagittal image correlation but had mixed effects on landmark variability. Following the fourth method of registration, correlation statistics and the alignment of reference points were similar to those obtained with the 4-point approach.