High-Performance Computing Applications for Visualization of Large Microscopy Images - High-Throughput Image Reconstruction and Analysis - page 304

The mesoscale gap arises in part from the requirement to use multiple imaging
technologies to examine a specimen across scales. Each technology requires differ-
ent expertise, specimen preparation techniques, and contrast mechanisms, and also
requires a severe reduction in the amount of tissue. For example, if the pipeline be-
gins with an entire brain, the end results in one small block of tissue,

0.5 mm 3 .
These requirements make it difficult for individual researchers to bridge scales,
both because single researchers may not be familiar with a given technology and
because there is significant loss of context as the scope decreases with increasing
resolution of imaging technologies. Bridging techniques such as multiphoton mi-
croscopy and electron tomography, correlated microscopy is a key methodology
for acquiring the necessary multiscale data in order to fill in the resolution gaps
between gross structural imaging and protein structure: data which is central to
bridging the mesoscale gap and to the elucidation of the nervous system.
From a computer science perspective, the mesoscale presents a number of chal-
lenges. The two major ones are large data sizes and ultra-high resolution content.
The typical size of a dataset collected by a microscope, capable of acquiring ultra-
wide field mosaics, ranges from a couple of gigabytes to a few terabytes on disk.
The content resolution ranges from a few hundred megapixels to hundreds of gi-
gavoxels. The largest CCD sensors for electron microscopes are now approaching
64 megapixels. Figure 14.1 shows an image acquired from a 64 megapixel sensor
on an electronic microscope. With image formats of 8,000

<

8,000 pixels, these
systems can be used to acquire wide-field mosaics (2D, 3D, and 4D) that exceed

×

Figure 14.1 A section of the mouse heart as seen under National Center for Microscopy and
Imaging Research's 8K

× 8K CCD camera. The camera has a collective resolution of 64 megapixels
and the specimen stage can be moved in X and Y to collect tiles. These tiles can then be stitched
together form ultrawide field of view mosaic.