Knife-Edge Scanning Microscopy: High-Throughput Imaging and Analysis of Massive Volumes of Biological Microstructures - High-Throughput Image Reconstruction and Analysis - page 11

CHAPTER 2
Knife-Edge Scanning Microscopy:
High-Throughput Imaging and
Analysis of Massive Volumes of
Biological Microstructures
Yoonsuck Choe, Louise C. Abbott, Donhyeop Han, Pei-San Huang, John Keyser,
Jaerock Kwon, David Mayerich, Zeki Melek, and Bruce H. McCormick

Recent advances in physical-sectioning microscopy have enabled high-throughput
imaging of massive volumes of biological microstructure at a very high resolution.
The Knife-Edge Scanning Microscope (KESM) we have developed is one of the few
that combines serial sectioning and imaging in an integrated process. The KESM is
capable of imaging biological tissue (about 1 cm 3 ) at 300 nm 500 nm

resolution within 100 hours, generating data at a rate of 180 Mbps. The resulting
data per organ (e.g., a mouse brain) can easily exceed tens of terabytes. High-
performance computing methods are required at every stage in the lifetime of the
generated data set: (1) distributed storage and retrieval, (2) image processing to
remove noise and cutting artifacts, (3) image and texture segmentation, (4) three-
dimensional tracking and reconstruction of microstructures, and (5) interactive
visualization. In this chapter, we will review the capabilities and latest results from
the KESM (Section 2.2) and discuss the computational challenges arising from the
massive amounts of data, along with a survey of our ongoing efforts to address
these challenges (Sections 2.3 to 2.5). We expect the integration of high-throughput
imaging and high-performance computing to lead to major breakthroughs in sci-
entific discovery in biological sciences.

× 300 nm ×

2.1 Background

In this section, we will provide a brief review of high-throughput imaging and
analysis methods, to provide a proper context for the work we will discuss in the
remainder of the chapter.

2.1.1 High-Throughput, Physical-Sectioning Imaging
Currently, the standard approach for microscopic imaging of a volume of tissue
is confocal microscopy [1]. The basic idea is to change the depth of focus (focal
plane), and use a pinhole aperture to detect photons originating only from the
target depth. This is called optical sectioning where virtual, not actual, sections
are obtained. In conventional optical sectioning, the main limiting factor is not

11