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.


In this section we highlight the promise and the problems associated with building a robust real-time Internet microscope system. The promise of Internet microscopy is in some respects obvious: such systems can put large and unique collections of tissue safely into the hands of many investigators without the standard limits of time and space. A system of this type is needed for the following purposes.

Creating digital slide collections.

 Investigators accumulate vast and often irreplaceable histological collections over the course of their careers (e.g., the Scheibels collection of Golgi material, the Altmans series of thymidine-labeled material, the primate brain collections of Rakic and Goldman-Rakic, Paul Yakovlevs extensive collection of adult and fetal human brains, the Brain Biodiversity Bank started by Woolsey and continued by Wally Welker). The list could go on for pages. Almost every one of us accumulates some tissue that is unique. In my own case, I have spent over a decade studying a line of mutant Belgian sheepdogs that completely lack an optic chiasm. The anatomical repercussions of this mutation are without precedents (Williams et al., 1994; Hogan et al., 1999). What is the long-term fate of this sort of collection? Often a dumpster. The Yakovlev collection narrowly escaped such a fate and was fortunately be placed in a permanent collection maintained by the Armed Forces Institute of Pathology.
Key parts of these collections can now be saved by being digitized. The problem of curating digital archives can ultimately be reduced to finding the right digital medium and the most effective database structure and means of distribution. Only a few years ago, the idea of digitizing entire collections would have been totally impractical, but in the past year the cost of digital storage has dropped precipitously. One terabyte1000 gigabytes (GB) of hard drive space is enough capacity to accommodate over 300,000 extremely high-resolution images that closely match the quality of 35-mm ASA 64 slide film (3000 x 2000 x 24-bit images). Today this amount of digital storage space can be purchased at a retail price of about $8000, and the cost is likely to drop further. The storage cost of a single archived high-resolution image should drop to a quarter of a cent.
Even at current hardware costs, the important bottleneck is not storage but limitations in the technology used to acquire and process digital images. If a technician making $1015 an hour takes about 2 minutes to capture and process an image, then labor costs 100 times as much as storage. We need to develop methods for rapidly and automatically acquiring batches of high-resolution images from existing collections. A major motivation of the iScope project has therefore been to develop methods to convert collections of glass slides to their digital counterparts quickly and with minimal human intervention (Aim 4).
We have made an excellent start in archiving, distributing and exploiting low-power images from the MBL. These images are extremely useful. For example, two colleagues of R. Williams are now in the middle of mapping QTLs that modulate the size of the hippocampus (Lu et al., 1999) and the cerebellum (Airey et al., 1999). Their computer workstations are located less than 20 feet from where the MBL slides are stored in a room full of microscopes. But rather than using the slides and microscopes, Drs. Lu and Airey find it far easier to download the 25- and 4.5-µm/pixel images from the web site. They load these images into NIH Image, outline boundaries, and transfer the data into Excel worksheets.
Low-power images, however, are not sufficient. We would like to be able to zoom into any part of a slide or section, from the equivalent of a 1x objective to the equivalent of a 100x objective. Ideally all the images, or virtual tissue, would be available immediately from a digital collection. This is feasible for low and intermediate scales, from 25 µm to.5 µm per pixel. But high-magnification images with pixel dimensions of under .5 µm are more challenging to deliver. Aim 4 of this project describes how we intend to use the iScopes to automatically and systematically acquire high-resolution image stacks for several hundreds of sites per brain.

2. Collaborative research.  

The Internet, email, and courier services have transformed the way scientists collaborate. These innovations are fostering new trans-institutional collaborations. It is now possible to assemble the best team to carry out a complex multidisciplinary research project from among universities and corporations scattered widely across the globe. The iScope is just one more tool among many that will increase the ability of groups to collaborate effectively across institutional boundaries. This resource will be accessible to neuroscientists around the world without restriction or tracking. The only limitation will be that we may have more users than microscopes or bandwidth will be able to accommodate. We have already set up a simple but effective arbitration scheme that schedules microscope users by category on a first come, first served basis.

3. Efficient use of animal resources.  

A collection such as the MBL, which already contains samples from ~100 genetically defined strains of mice, increases the efficiency with which neurogenetic problems can be addressed. For example, a scientist investigating the effects of age on the volume of the mouse neocortex can explore and exploit the MBL collection rather than ordering and then aging a set of mice, and then processing all of their brains. In a matter of a few days, this investigator can mine the MBL and obtain excellent data for animals ranging in age from one month to two years. Information about many other factors, such as sex differences in the volume of the amygdala or hippocampus of mice, is likely to be embedded in the MBLs superb collection of celloidin sections. The answers to scientific questions can be obtained quickly without sacrificing a single additional mouse, and without the huge cost of processing hundreds of brains.

4. Teaching neuroscience and neurogenetics.  

Teaching neuroscience will be transformed over the next decade by Internet resources. Not only will publications be available on the Internet, but the datasets and analytic tools that are critical in carrying out research will become more accessible. For example, tutorials and resources already available on the <> site could be used as part of a course in quantitative neuroanatomy and neurogenetics. The iScope and MBL in particular would be useful adjuncts in many neuroanatomy courses.

Next Topic

  Status of the Current Internet Microscopy Project