Re: Area or height??

From: Howard Shapiro (hms@shapirolab.com)
Date: Tue Feb 27 2001 - 09:22:55 EST


Bob Zucker wrote-

>At a local flow cytometry users group ( RTCA)  on Feb 22 2001, we heard a
>about the  BD DIVA digital electronics. It was impressive and like all new
>developments in flow cytometry it is a welcomed addition.
>BD is now measuring area on all pulses instead of height that is normally
>used on previously versions of BD  equipment.  For you BD users,  there is
>one board in the machines currently that can yield area measurements which
>are required for DNA measurements using Doublet discrimination. Now I
>remember that our old Ortho 50H yielded area measurements on all signals
>and this is the way we always ran our machine. We reluctantly  gave up this
>measurement to use the User friendly BD FacsCalibur  with their CellQuest
>program.
>
>My questions are the following:
>How is our Flow data going to change by  using area instead of peak height
>measurements
>It should be more accurate with area but how bad is the height measurement
>that we have obtained for many many years..
>This Question  was raised at the meeting but it was not adequately answered
>by anyone in the audience. How bad is height measurements?. Is it only bad
>( inaccurate) for linear measurements like DNA??? Let's discuss this. I
>think it is important.

The area or integral of a pulse is always an accurate representation of the
total signal, i.e., the intensity of fluorescence or scattering.  When the
height of the illuminating laser beam (i.e., the axis dimension in the
direction along the flow stream) is substantially greater than the diameter
of the cell or particle being analyzed, so that the whole particle is
within the illuminated region during transit, the pulse height or peak is
proportional to the area.  When the illuminating beam height is close to or
smaller than the cell or particle diameter, passage through the beam
produces a "slit-scan"; the pulse height is now proportional to the density
of fluorescent or scattering material in the cell, but not to the
area.  The height is thus sensitive to shape; the area is not, and that is
why plots of peak vs. integral can be used to remove data from many (but
usually not all) doublets in DNA analysis in instruments with small beam
heights.

The Ortho 50H had a very tightly focused beam, with a height supposedly
around 5 um; the pulse integral provided accurate measurements of
fluorescence, while the peak height did not.  In B-D's earliest
stream-in-air instruments, the focused laser spots were originally round;
while later models and the FACScan, Calibur, etc. use elliptical beams, I
believe that the heights are larger than what Ortho used.  If the beam
height is 20 um or more, pulse height will probably be proportional to
integral and thus provide an equally accurate measurement for particles 10
um or less in diameter.

In instruments which do not use high-resolution analog-to-digital
conversion, both pulse height and integral are derived using analog
electronic circuits.  Peak height is detected by peak detector circuits,
but there are two ways to get an integral. One can use an integrator
circuit, or one can pass the pulse through a low-pass filter (electronic,
not optical), which also effectively integrates it, and use a peak detector
to get the peak height of the filtered pulse, which is proportional to the
integral of the original pulse.  When the pulses are log amplified, the
second method should be used, because the desired quantity is the log of
the integral, not the integral of the log.  This problem goes away when you
use digital electronics.

B-D's DIVA electronics compute the pulse integral or area from 32 14-bit
samples or "slices" of a pulse, and this is almost certainly more accurate
than analog circuits would be; since log amps are dispensed with, their
associated inaccuracies disappear from the measurement. Digital pulse
processing is potentially less accurate than peak detection for capturing
peak height, because the highest point of a pulse may come between
samples.  Thus, while one can get an integral accurate to better than 1 per
cent from only 8 samples of a pulse (as Luminex does), more samples are
needed to get an accurate peak height.  I would imagine that accuracy on
peak height measurements with 32 slices as used in the DIVA electronics is
within a few per cent.  For what it's worth, pulse width measurements are
also less accurate than integral measurements with digital processing,
because the pulse may rise above and/or fall below threshold between slices.

So, I wouldn't worry about the accuracy of the digital measurements in the
DIVA; the integral (analog or digital, provided it is taken properly) is
always a good measurement.  As far as your old data goes, one would not
expect manufacturers to use heights instead of integrals if the heights
were not accurate.  This shows up in linearity; look at a fluorescence
distribution from beads or from nucleic stained for DNA.  If the modal
value of the peak [the peak of the distribution, not the pulse] from bead
doublets isn't within a channel or so of twice the value for the peak from
singlets (and/or if the peak from G2/M cells and doublets isn't within a
channel or so of twice the value for the peak from G0/G1 cells), there's a
problem.  And, if you can use the peak-vs.-integral method for doublet
discrimination, you can assume that the integral is an accurate measurement
of total intensity and that the peak is not.

B-D is to be congratulated for implementing digital pulse processing at a
high enough speed to be compatible with sorter operation.  But I should
point out that the high speed processing is essentially happening in a
black box; I have only seen the software briefly (at Montpellier), and I
can't comment on what it's like for the user (I'll try to get more hands-on
experience with the commercial instruments in time for the 4th Edition).

-Howard



This archive was generated by hypermail 2b29 : Sun Jan 05 2003 - 19:01:08 EST