Development of a 3-D MRI Atlas of the Baboon Brain

By Ryan Collins Mentor: Dr. Kevin Black Department of Psychiatry, Washington University School of Medicine, St. Louis, MO

This summer, while working under the guidance of Kevin J. Black, M.D., I have taken steps to develop a 3-dimensional magnetic resonance image atlas of the olive baboon (Papio anubis). An atlas for neurological imaging provides a common means to communicate data, a means to increase weak signals by averaging data over subjects, and a way to account anatomic and functional variability across subjects.

A means of aligning PET images to atlases of the human brain has existed for almost 15 years (Fox et al. 1985) and newer methods of creating atlases have been developed. However, atlases of the human brain are only useful to studies involving humans as research subjects. Certain studies require nonhuman primates for reasons of feasibility and ethics, such as investigations of the effects of experimental psychiatric drugs (Black et al 1997, Perlmutter et al 1995). In 1997 Black et al. developed a Davis-Huffman or "dh68" atlas of the baboon brain using 6 points of reference to align MR images to atlas space. The creation of this atlas was labor-intensive, and it required the consultation of an expert in primate neuroanatomy. The final result was of high quality in subcortical regions, but of limited quality at the edge of the cortex. The purpose of this study was to create a new, high-quality 3-dimensional anatomic MRI atlas of the baboon brain from as many subjects as possible, to be used for automated image registration of future subjects.

MPRAGE (T1 weighted, 3D) magnetic resonance images of the brain were acquired from 11 baboons of various ages, both male and female. The images were then "segmented;" brain and non-brain voxels were identified and labeled using ANALYZE_AVW software and a semiautomated procedure developed by Robb et al. in 1989. Technical difficulties prevented the project from continuing beyond this point, due to an irreparable hard disk error. All work was repeated from backup copies of the MPRAGE images.

Future steps include the alignment of the 11 segmented images to a "bootstrapping" atlas image using Automated Image Registration software. These 11 images will then be averaged in atlas space to create a new, improved atlas image. The automated alignment will then be tested against the "dh68" alignment. An expert rater will test points within the brain, and an automated process will test the overlapping percentage. Once the atlas is completed, PET images may be aligned to MRI images, which can be aligned to this new atlas. This will allow the communication of functional data at a high degree of exactitude.

Acknowledgements: Kevin J. Black, M.D., Ed Miner, WUMS; Tom Videen, Ph.D., Joel Perlmutter, M.D., Howard Hughes Medical Institute.