Automated Methods for the Analysis of Skeletal Muscle Fiber Size and Metabolic Type - International Review of Cell and Molecular Biology - page 294

optimizing the ability of the algorithm to recognize the laminin labeling and
estimate “missing” segments of the fiber outlines; it should also be men-
tioned that the sample could be labeled for other extracellular proteins,
instead of laminin, such as collagen type IV, or proteins which are associated
with the sarcolemma of the muscle fibers such as dystrophin or dystrophin-
associated proteins. The algorithm was optimized to analyze images from
several species (e.g., mouse, pig, and monkey) and to identify fibers photo-
graphed over a variety of magnifications ( Fig. 7.5 ).
An interesting aspect of algorithm development concerned the thickness
of the fiber boundary labeling. Upon close inspection of images from skeletal
muscle, it became apparent that the thickness of the boundaries between the
fibers varied between preparations, and this can be of interest as remodeling
of the extracellular material is a feature of skeletal muscle development,
regeneration, and certain pathologies. Accordingly, methods were added
to the algorithm enabling the thick membrane mask area to be defined,
which reports the extracellular matrix area as a separate data parameter in
addition to the fiber CSA ( Fig. 7.6 ).
To test the accuracy of the CSA calculations by CyteSeer ® , images
were acquired from three species (mouse, pig, and monkey) and analyzed
by “manual” versus “automated” methods. For the manual method, the
images were imported into Image J and an experienced muscle researcher
(Dr. Kostrominova) outlined the fibers utilizing a digitizing pen interfaced
to the computer, a process that required > 1 hour per image. For the
automated method, images were imported into CyteSeer ® and analyzed
utilizing the Skeletal Muscle Algorithm, which typically requires < 10 s
per image.
For mice, nine different images were compared, which included gastroc-
nemius, plantaris, and soleus muscle, imaged at either 100

; the
number of fibers segmented by the manual method ranged from 56 to
311 fibers per image ( Table 7.2 ). Notably, the CSAs by the automated
method were within 4% of those calculated by the manual method. For
swine plantaris muscle, the two methods yielded results that were even closer
in agreement (CSAs by the automated method were within 2% of the man-
ual method). Very good agreement was also obtained with images obtained
from rhesus monkey vastus lateralis muscle (

or 200

5% difference).
In summary, CSA measurements obtained via automated analysis using
CyteSeer ® were rigorously tested, utilizing images of skeletal muscle from a
several species. The automated method, in general, yielded CSAs that were
very close to those obtained with manual segmentation.