Project 1: The Mouse Brain Library Project 2: Internet Microscopy (iScope) Project 3: Neurocartographer and Segmentation of the MBL Project 4: The Neurogenetics Tool Box


















Principal Investigator/Program Director Williams,Robert W.


Problems and issues addressed in this project

A spatial database, like any other database, consists of content and infrastructure to support access and analysis. This project is largely concerned with infrastructure, focusing on the analytic tools needed as part of a neuroanatomical 

database. To transform a collection of images such as the MBL into a powerful research database, we need a suite of special tools that have some capacity to detect, define, and extract numerous compartments, nuclei, ventricles, and fiber tracts that make up the mouse brain.

Typically, a user of the MBL will want to compare a region of interest (ROI) across strains. The investigator may wish to visually inspect histological sections, perhaps sorted over the range of strains according to the size of the ROI. Alternatively, the investigator can issue a command to the iScope system (Project 2) to automatically sample  a small volume of tissue (typically 50 x 50 x 30 m) within an ROI across a set of brains. If manual delineation of each brain were a prerequisite for data extraction, the utility of the NeuroCartographer project would be diminished for the simple reason that our throughput would be severely limited.

Automatic segmentation is crucial. In addition to saving time, automated segmentation avoids the subjective judgement involved in manual delineation that is the major source of error in data analysis (Eilbert et al. 1990). Our objective is to develop a reliable computerized system to expedite analysis. Algorithms will be designed to segment each brain in the MBL into a set of standard anatomical structures like those defined in the rat atlas produced by Nissanov and Bertrand (1998a). An investigator who prefers to define a nonstandard division will be able to modify neuroanatomical templates residing in the atlases and have these modifications automatically propagated throughout the database. Besides providing the means to navigate through the MBL, segmentation of the database content will have another benefit: positional variability and volume estimates of the anatomical structures will be analyzed in the Neurogenetics Tool Box (NTB; see Project 4), then used to map QTLs.

Automatic brain parcellation is a familiar theme in neuroinformatics (Toga and Thompson 1999). We believe that this project and this neurogenetics research setting lends itself particularly well to automation: the data set is highly homogeneous, consisting of cresyl violet and Loyez fiberstained tissue fixed in precisely the same manner, and embedded and cut in a single laboratory. These celloidin-embedded sections are of the highest quality. It is our hope that the technological innovations proposed here not only will serve well in the context of MBL but will provide insight to the general problem of brain segmentation.



Next Topic