Archiving+Digitized+Brain+Images

Development of a Brain Image Database requires a number of considerations, including the quality of the fixation, tissue preparation, the method of digitization of the sections, and storage/retrieval/display of the resultant digital files. The first several steps, from tissue fixation to digitization, are now readily defined and accomplished. The major problems associated with establishing a uniform file format, though still undergoing discussion, will largely be resolved as a result of the parallel development of the field of Digital Pathology in Medicine. However, the storage and retrieval of these images, has proven far more difficult than anticipated.This is consequent to the large size of the resulting files of Whole Slide Imaging of complete sets of serial sections of brains.

The most common level of resolution used in contemporary scanners is that provided by a 20X objective with an N.A. (Numerical Aperture) of 0.75. The nominal resolution is ca. 0.50 micrometers/pixel. A 25 x 75 mm slide is typically segmented into two zones: 1) the region of label of 25 x 25 mm, and 2) the region of tissue within the remaining 25 x 50 mm. Due to mechanical constraints of the slide carrier, only about 20 x 45 mm of that area is available for imaging. This limitation must be noted when preparing histological material. Thus the maximal area to be imaged on a 25 x 75 mm slide is 900 mm^2. Scanning this at ca. 0.5 um/pixel results in an image composed of 40,000 x 90,000 pixels = 3.6 x 10^9 pixels. Each pixel is typically composed of a Red, Green and Blue plane with a dynamic range of 0-255 for each color. This is stored as 8 bits per color plane, for a total of 24 bits per pixel. Thus, each pixel requires storage space of 3 Bytes. The resulting image file is 3 x 3.6 x 10^9 = ca 11 GB (GigaBytes).If the complete set of serial sections consists of sections cut at 30 um, and every 3rd section is stained and mounted on slide, that will result in ca. 50 to 100 slides (e.g., small rat brain), the complete set of images of the resulting scan will require between 0.5 to 1 Terabyte (TB).

Larger brains, such as those of Macaque Monkeys, require use of 50 x 75 mm slides, with tissue areas of 50 x 50 mm. The mechanical constraints limit the effective area to 45 x 45 mm for a total file size of ca. 25-30 GB for each slide.

Even at the rapidly falling cost of storage media, this poses major problems in storage, retrieval, display and analysis.

In order to reduce the burden of this large file size and speed up zooming and panning, various image compression algorithms have been developed. Each manufacturer of scanning devices may use their own file format. One of the more common formats, and that described in the recent DICOM Supplement 145 for Whole Slide Imaging (WSI) is JPEG200. This is a "pyramid" format, and uses "lossy compression". The compression algorithm reduces file size by approximately 12:1; i.e., a 12 GB file is compressed for storage as a 1 GB file on the hard drive.

The use of compression algorithms is deemed essential in order to store these very large collections of files. A typical set of serial sections of a mouse brain requires ca. 6-12 GB in compressed format, depending upon the number of sections you choose to digitize.

HJK: October 9, 2010