GUIDE
TO THE PROGRAMS (last updated
INTRODUCTION
Computed Corpuscle Sectioning is a general method to determine
the volume, profile area, and perimeter of any corpuscle at any
orientation sectioned at any position and thickness, and also models the
concentration and distribution of the material filling the corpuscle
(1). Here, it has been applied to the modeling of nuclei in
histologic sections, to provide insight into the measurement of
quantitative nuclear DNA content (ploidy) by image cytometry
(1,5). The Reference Curve Method (1,5) is a method for analysis
and correction of ploidy measurements in tissue sections. I have
provided the programs and data on this web page for those who
want to gain experience with these methods, and to augment my
published articles in Analytical and Quantitative Cytology and
Histology. This web site became fully functional on
uploaded the "u" data sets: if you have not downloaded these yet, the ALG.COM
and the THK.COM programs will not work. The contents of this web page can be
best understood as intended only with reference to the cited papers; reprints
may be requested by e-mail. The Computed Corpuscle Sectioning Program is not
provided on this web page, but was used to create all of the synthetic data
files posted here.
INSTRUCTIONS FOR DOWNLOADING
First create and name a folder (e.g., CORPUS) using Windows
Explorer (File - New - Folder); the equivalent process in DOS is
to type MD CORPUS at C:\>. On the Web page, click on each of
the hyperlinks (you will need to download all of them into the
folder you created). When the download is complete (it may take a
while-each data file is almost 1 megabyte), save the file to the
folder you created using File - Save. Save all of the files to the
same folder. To unzip each file, do: Start - Run (finding and
running each file in the folder you created). Choose to automatically
overwrite files. This will unzip the .EXE files to produce approximately
5,000 data files and the four programs mentioned below.
PROGRAMS
Four DOS-executable programs are provided on this web page,
together with data files specifically designed for use with them.
These programs can be run in Windows 95 or later (Start - Run), from Windows
3.1 (using File Manager), or from the DOS-prompt by typing the program's name
(once you are in the correct directory). To get to the appropriate directory,
the DOS command is CD\ CORPUS if 'corpus' is the name of the directory you
created. To make matters very easy, you can find the .COM files in Windows
Explorer. Click on it using the right mouse button to show a drop down menu--
then click on CREATE SHORTCUT. A shortcut icon will be added. Now drag the
icon onto your desktop and you can run the program from the icon (recommended).
To enable direct printing of the screen display, you will need to type GRAPHICS
LASERJET or some similar command at the DOS-prompt. Then, when the Enter and
Print Screen keys are pressed simultaneously, the screen display should print.
Alternatively, you can press ALT and PRINT SCREEN simultaneously, which copies
the image to your clipboard. This image can then be pasted (EDIT - PASTE) into
a Word document or can be pasted into Paint (a program bundled into Windows:
do START-PROGRAMS-ACCESSORIES-PAINT). If you paste it into Paint, you may wish
to invert colors (IMAGE-INVERT COLORS). Then save the file if you wish.
All of the data here were generated by a computer simulation of nuclear
sectioning (the Computed Corpuscle Sectioning Program).(1)
ALG.COM provides a comparison of various algorithms (those of
Bins and Takens (3), McCready and Papadimitriou (2), and
Haroske et al (4)) for correcting ploidy measurements in tissue
sections. THK.COM compares the histograms produced by
sectioning a nucleus of your choice at different section thicknesses;
the Very Thick and Relatively Thick sections are uncorrected,
while the Standard Thickness and Ultra-Thin sections are corrected
by the method of McCready and Papadimitriou (2). The
histograms are all vertically scaled to fit in the available space, so
there is no absolute vertical scale. The relative number of nuclei is
plotted on the vertical axis, and the DNA Index (DI) on the
horizontal axis, with the vertical lines representing, from left to
right, a DI of 0, 1, 2, and 3. Each histogram bar encompasses a
range of 0.1 and is centered on a multiple of 0.1. For all of these
programs, if you just press the Enter key when asked to input a
value, the default value in parentheses will be entered
automatically. Once the histograms are displayed, pressing the
Enter key once more will terminate the program and return you to
the DOS-prompt. If a program terminates unexpectedly, your text
at the DOS-prompt may be altered in size and color; if this
happens, just rerun any of my programs and let it terminate
properly, and the text properties will be restored.
WRITEDAT.COM will create a Pascal binary data file with your
data (you key it in); if I have the time, and depending on the
demand, I may do one or two interpretations of your data for your
interest (not for clinical use or research) if you e-mail the file to
me. Until I announce a fee for this service, I am offering it (with
the understanding that I may not have the time or inclination to do
it) free of charge. I might also be interested in collaborating on
research. Certain choices of section thickness and of nuclear sections
for analysis are strongly suggested by my computer models, and if you
would like my advice please make your request to me by e-mail and I shall
be happy to give it.
The first letter in the name of each data file identifies the set to
which it belongs, and corresponds to the composition of the sample
analyzed. Thus, until you have downloaded the the "d" data set,
you will not be able to select data set "d", for example. (The first
question to be answered in any of these programs is which data set
you would like to analyze.) Also, ALG.COM and THK.COM will not run
without the three "u" data sets.
THE REFERENCE CURVE METHOD
The DOS-executable program is RCMNEW.EXE. It replaces RCM.COM and
runs when RCMNEW is typed at the DOS prompt. If you will want to
print the screen displays, the command GRAPHICS LASERJET or similar
command (depending on the printer) should be typed at the DOS
prompt before running the program. After you enter a series of
letter choices which define the corpuscle, selection, and section
parameters, you choose a section thickness. The actual section
thickness is the default value which will be entered if you simply
press the Enter key without entering a number. Next you are asked
to choose a screen expansion factor. This will only be necessary
on rare occasions where you find the screen dimensions are too
small; in that case choose a number larger than 1 (e.g., 1.2),
otherwise you may press the Enter key without entering a number.
The data points are then displayed with the Reference Curve and
the Reference Line. As described in my article (1), each data point
is represented in the data file as (Ds/Dd,As/At), but is horizontally
scaled ("prescaled") so that the point with the largest value of
Ds/Dd in the data set lies on the Reference Line. (Dd is the DNA
content of a whole diploid nucleus; Ds, the DNA content of each
sectioned nucleus in the sample, As, the profile area of each
sectioned nucleus in the sample; and At, the area of a circle the
diameter of which equals the section thickness.) At the bottom of
the screen the value of Vi/Vt and the DNA Index are displayed. These
values are meaningless until the appropriate Reference Curve has been
found that best fits each subpopulation of data points. You are
prompted at the bottom of the screen to enter Q to quit, or P0 to P7
to activate various kinds of perimeter correction. The Second Graphic
Screen is now displayed. Next you are prompted to enter the Factor
for Horizontal Scaling (enter a real number, usually one or very close
to one). This Factor for Horizontal Scaling is used to calculate H
(which moves the data points horizontally on the screen). H is
initially set equal to one. Each time you enter a value for the Factor
for Horizontal Scaling, H is set equal to the product of its latest
value and the value of the Factor for Horizontal Scaling you last
entered. Note that the product of H, the largest value of Ds, and the
"prescaling" factor equals F (the Horizontal Scaling Factor in my
article).(1)
After entering the Factor for Horizontal Scaling, you are prompted
to enter the Reference Curve Factor. The value of the Reference Curve
Factor is initially set such that the DNA Index equals one, and
thereafter the value of Vi/Vt of the Reference Curve is the product of
its current value and the last-entered Reference Curve Factor. A new
Reference Curve is calculated and redrawn each cycle. Only the smallest
Reference Curve is 'active' for the purposes of these calculations-
the larger reference curves correspond to a Vi/Vt twice, three times,
and four times that of the smallest.
The goal is to line up the terminus of each data curve with the Reference
Line (which is basically automatic but may require slight adjustment)
by finding the appropriate value of H, then to determine the Reference
Curve that coincides with each subpopulation of interest. (It should
be noted that the DNA Index displayed will not always be correct,
depending on the nuclear and section parameters selected. In these cases,
you must compare the subpopulation of interest to the internal diploid
standard. You will find the same limitations and remedy applies in the
histogram-based correction algorithms.)
The histogram of the data corrected by the Method of McCready
and Papadimitriou (MMP) (2) is displayed below the reference
curve.
Perimeter correction (5) involves replotting data points according
to how much the profile of the sectioned corpuscle deviates from a
circle. If P1 is entered, the perimeter-corrected points are moved
toward the origin (in an attempt to reposition the data curve so that its
terminus gives the correct DNA index); if P2 is entered, the
perimeter-corrected points are moved perpendicular to the reference line
(in an attempt to maximally discriminate each ploidy subpopulation from
the others). The perimeter-corrected points are shown in green, and all
the original points are shown in red. Perimeter correction is also
performed on the MMP histogram, in which all the perimeter-corrected
points are reassigned to a leftward bin depending on how much the profile
of the sectioned corpuscle deviates from a circle; those points from a
nucleus with a round section are assigned to their original bins.
Perimeter correction is in effect only on the cycle in which it is
requested; the next cycle reverts to no perimeter correction.
The power of the Reference Curve Method is due to the fact that
the data are displayed in a way that allows you to make inferences
about the nuclear and section parameters, and it also allows you, in
most cases, to identify distinct ploidy subpopulations with greater
ease and confidence than does any histogram-based method.
Identification of distinct subpopulations is the key to successful
ploidy analysis.
At any time the screen display may be printed by pressing the Shift
and Print Screen keys simultaneously, provided that the keys have
been enabled as described in the first paragraph, or you can copy the screen
display to the clipboard by pressing the ALT and PRINT SCREEN keys together.
WHAT COMPUTED CORPUSCLE SECTIONING TEACHES ABOUT HOW TO DO PLOIDY ANALYSIS
IN TISSUE SECTIONS
It is always necessary to select only center-containing sections of nuclei (5).
Ultrathick sections (at least as thick as the diameter of the largest nuclei)
will generally give the best results (5), for these reasons:
1) no correction and no diploid internal standard are needed, and
2) nuclear shape, inhomogeneous intranuclear DNA distribution, variable
internuclear DNA concentration, variable section thickness, and errors in
estimation of section thickness do not significantly affect the results.
However, the problem of nuclear overlap will be exacerbated, and if the nuclei
are large the required section thickness may be excessive. A choice of section
thickness slightly less than ultrathick may be used, but if the data are
displayed in a histogram and the nuclei are prolate ellipsoids, the height of the
peak of greatest ploidy may be falsely shortened if no correction is used, but
correction may restore the peak height.
If ultrathick sections are not appropriate for the reasons mentioned above, an
ultrathin section is the next-best choice (5), which is at most one-third (or
preferrably one-fourth or less) of the diameter of a diploid nucleus. A
diploid internal standard is required (if the nuclei are ellipsoidal), but
problems of nuclear overlap are minimized. If the nuclei are spherical, then
variable internuclear DNA concentration, inhomogeneous intranuclear DNA
distribution, and variable section thickness can be overcome, alone or even in
combination, if the RCM is used (histogram methods perform poorly except in the
case of variable internuclear DNA concentration). If the nuclei are prolate
ellipsoids, variable internuclear DNA concentration can be overcome by the RCM
or histogram methods, but not in combination with variable section thickness or
inhomogeneous intranuclear DNA distribution. The RCM alone is suitable for
oblate ellipsoids, and is problematic but more acceptable than histogram methods
for prolate ellipsoids with variable section thickness, inhomogeneous intranuclear
DNA distribution, or both. The RCM alone is suitable for documenting differences
in shape between different ploidy subpopulations which may have an effect on the
measured ploidy, and it alone provides a way to compensate for such differences
in shape. The RCM alone can document operator selection bias, thus avoiding
false shifting of peak position when histogram methods are used with ellipsoidal
nuclei. With ellipsoidal nuclei, a selection bias in favor of the most eccentric
nuclear section profiles is helpful in the RCM, as it allows perimeter correction
to be fully utilized; with histogram methods, such a selection bias may be
essential to ensure consistent peak position. Histogram methods underestimate
the height of the peak of greatest ploidy when the nuclei are prolate ellipsoids,
but this is not a problem with the RCM. You can replicate many of the above
observations reported by me in the cited articles (1,5) by playing with the
programs and data available on this web site.
You are cautioned that these data files represent a fairly small
number of nuclei, and that in some cases a sample size which is too
small may give the appearance of false aneuploidy. I intended the
file sizes and the number of nuclei in the various subpopulations to
reflect the number of nuclei which might be analyzed in an actual
analysis. Also bear in mind that these programs have been
designed not to work with unauthorized data files.
Comments are most welcome and may be addressed to
[email protected]. I hope you find this useful.
Jeffrey A. Freed, M.D.
REFERENCES
1. Freed JA. Possibility of correcting image cytometric nuclear
DNA (ploidy) measurements in tissue sections: insights from
computed corpuscle sectioning and the reference curve meth od.
Analyt Quant Cytol Histol 1997; 19(5): 376-386.
2.
cytomorphometry on tissue sections in a rat liver model. Analyt
Quant Cytol 1983; 5:117-123.
3. Bins M, Takens F. A method to estimate the DNA content of
whole nuclei from measurements made on thin tissue sections.
Cytometry 1985; 6:234-237.
4. Haroske G, Meyer W, Dimmer V, Kunze KD, Theissig F.
Feasibility and limitations of a cytometric DNA ploidy analysis
procedure in tissue sections. Zentralbl Pathol 1993; 139:407-417.
5. Freed JA. Improved correction of quantitative nuclear DNA
(ploidy) measurements in tissue sections. Analyt Quant Cytol Histol
1999; 21(2):103-112.
6. Freed JA. Conceptual comparison of two computer models of
corpuscle sectioning and of two algorithms for correction of
ploidy measurements in tissue sections. Analyt Quant Cytol Histol
2000; 22(1):17-25.
Return to CCS page: www.fortunecity.com/skyscraper/terabyte/562/jfccs.htm