Automated Methods for the Analysis of Skeletal Muscle Fiber Size and Metabolic Type - International Review of Cell and Molecular Biology - page 292

scanners) are similar but are adapted to work with histology slides and
emphasize visualization of colorimetric histochemical stains via bright-field
imaging ( Martina et al., 2011; Potts, 2009 ). For HCA and digital pathology,
thousands of individual photomicrographs are generated per sample (for
slides, the individual fields of view are “stitched” together into a large image
of the tissue for viewing). To analyze these ever-increasing data sets, image-
analysis algorithms are needed, and this is a current area of intense research
amongst biomedical researchers and software engineers.
6.1. CyteSeer ®
An example of a cell image-analysis program developed specifically for HCA
and digital pathology applications is CyteSeer ® (Vala Sciences Inc.), which is
Java encoded and compatible with most computers and digital microscopy
workstations. CyteSeer ® includes a menu-driven Graphical User Interface, a
“plug-in, pipe-line” architecture for rapid development and prototyping of
new image-analysis algorithms, and a standardized reporting structure for
algorithm-derived data parameters. One of the first algorithms derived for
CyteSeer ® was the Membrane Analysis Algorithm, which quantifies the
expression and cellular location of membrane-associated proteins; examples
include proteins which undergo palmitoylation ( Mikic et al., 2006 ), protein
kinase C, or cadherins, which are important regulators of cancer biology
( Prigozhina et al., 2007 ). Notably, when confluent cell monolayers are
labeled for cadherins, which are found at the junctions between cells, a
“cobble-stone” pattern is obtained, which is readily analyzed by the Mem-
brane Analysis Algorithm, provided that the cell nuclei are also covisualized
(e.g., by DAPI staining) in a separate fluorescence field (the algorithm ini-
tially identifies the centrally located nuclei to help identify the cellular
regions— Fig. 7.4 A and B).
It was subsequently noticed that the images of cadherin labeling from
confluent cultured cells resembled images of laminin labeling of skeletal
muscle fiber outlines obtained via fluorescence microscopy by muscle
researchers ( Fig. 7.4 C). In fact, CyteSeer ® 's Membrane Algorithm could
accurately identify laminin-labeled fiber outlines in images from skeletal
muscle, if “artificial nuclei” were added to the image, at central locations
within the fibers ( Fig. 7.4 D). These observations prompted formation of
a collaboration between researchers at Vala and the University of Indiana
(Drs. Kostrominova and Sturek) to further develop methods for the digital
analysis of skeletal muscle structure via HCA and digital pathology methods