[Histonet] Re: Image Analysis for Fat Tissue

[Nathanael Reveal]

Hi Ateret,

You've got a few choices, some of which are free and some of which are
quite expensive.


In many soft tissue staining protocols, adipocytes show up as white, oval,
void spaces. Which is good!

Software like ImageJ <http://rsbweb.nih.gov/ij/>, or its more integrated
variant FIJI <http://fiji.sc/wiki/index.php/Fiji> (which stands for Fiji is
Just Image J) is an excellent place to start. They have very good and easy
to use measurement tools for area and number.


Unfortunately, the topic takes a turn for the complex. Cell number and cell
size (i.e. cell volume) can be subject to systematic bias due to section
orientation, section thickness, and related parameters. This may cause
under or over estimation of the data. This brings us to
Stereology<http://en.wikipedia.org/wiki/Stereology>,
the science of estimating data from 2d cross-sections of 3d structures.

Say for example, you're sectioning at 5 microns. The 2d profiles (i.e.
cross-sections) of a large adiopcyte will appear on many sections, but the
2d profile of a small adiopcyte will show up only once or twice. In this
way, if you count all the adipocyte profiles you see, you could be
overestimating the number of adipocytes. You could be counting some
adiopocytes more than once.

There are standard techniques for estimating the total number of cells in a
region and the mean volume of a group of cells. The Optical Fractionator
estimates the total number of cells. The Rotator estimates mean cell
volume. These tools are found in most stereology systems, commercial or
free. The drawback to these methods are that they require specialized
commercial software and a microscope capable of monitoring real-time
position in X, Y, and Z. This is commonly done with either a motorized
stage or encoded stage.


Some practical alternatives you could implement with ImageJ may be:

1. Measure "Adipose Volume". That is the area of adipose tissue measured
over a set of serial sections. The Delesse
principle<http://en.wikipedia.org/wiki/Stereology>(again from
Stereology) says that area fractions are proportional to volume
fractions in a set of serial sections. So no special hardware or random
sampling is required.

2. Take care to only count nucleated adipocyte profiles. This helps
minimize the problem of overcounting in serial sections.

3. Count adipocytes in sections that are not adjacent. Meaning, cut 5
micron sections but make sure the sections you count are far enough apart
that you won't see the cross-sections from the same adipocyte.

4. As far as measuring the volume of individual adipocytes, you might be
able use the shortest diameter of the nucleated adipocyte profiles in you
see in your sections. This is a tricky one and very much subject to the
shape of your adipocyte population.


As always, you'll do well to mimic the methods of those that have performed
similar work in the past. This will make it easier for reviewers to
understand your methods and comment constructively on your paper. You might
find this Google Scholar
search<http://scholar.google.com/scholar?as_ylo=2008&q=adipocyte+volume+number&hl=en&as_sdt=0,43>helpful.
It covers references to adipocyte volume and adipocyte number
since 2008.


Best of Luck,
Nathanael


--
Nathanael Reveal, CEO
BIOQUANT Image Analysis Corporation

5611 Ohio Avenue, Nashville, TN 37209
phone: 615-350-7866, toll free: 800-221-0549, fax: 615-350-7282

email: nathanael <@t> bioquant.com, web: www.bioquant.com