oebele@idca.tds.PHILIPS.nl (Oebele Dijkstra) (05/09/89)
Hi, I'm going to work on a programming project for a department of Nuclear Medicine. The project is meant for images from a SPECT camera. The goal is to develop a program (programs) to calculate the volume of an organ or to calculate the volume of a part of an organ. Are there people on the net who can give me any info about this subject ? Maybe this kind of programs is already developed ? Any help/info would be appreciated. Thanks in advance! oebele@idca.tds.philips.nl
klg@dukeac.UUCP (Kim Greer) (05/15/89)
In article <414@ssp6.idca.tds.philips.nl++ oebele@idca.tds.PHILIPS.nl (Oebele Dijkstra) writes:
++Hi,
++
++I'm going to work on a programming project for a department of
++Nuclear Medicine. The project is meant for images from a SPECT camera.
++
++The goal is to develop a program (programs) to calculate the volume
++of an organ or to calculate the volume of a part of an organ.
++
++Are there people on the net who can give me any info about this subject ?
++Maybe this kind of programs is already developed ?
++Any help/info would be appreciated.
++
A lot of published work has been done by us and others in the field of
spect imaging in general and volume calculations in particular. Volumes
can be estimated by several methods: count-based or pixel-counting, to
name two.
Count-based methods will only work if you can equate counts in
the reconstructed images back to some known microcuries of activity. This
is generally just not very likely if you rely on commercial software. The
reason is that the reconstruction filters generally do not preserve the
dc value properly and that arbitrary scaling is performed to save the image
out to disk. You could write your own recon program, though this is
somewhat drastic step, in that you would have to probably re-write all the
other support software (display, roi, stat-package...). Other filter
problems we have seen with commercial software is that changing filters
will change the counts extracted from the roi. Also most systems do
not preserve the negative undershoots from the backprojection process.
(Why save them ? Do you simply ignore "undershoots" in your bank checking
account ? "Well, I just wrote a check for $200 but I have only $100
in my account. Let's see, that leaves me a balance of $0." This attitude
will get you in trouble on both accounts (no pun intended).) This method
also requires a calibration scan (with all imaging parameters the same !!)
at some time or other with which to relate patient scans to. Note that
attenuation compensation is a definite requirement as well.
Pixel-counting methods are a little less stringent, in that only
(generally speaking) the pixel size be known. This is done with a series
of test scans (spect) of point sources in air, in x,y & z planes. (The
pixel size in the direction parallel to the axis of rotation can be
determined with coronal/sagittal slices to determine slice location vs.
cm separation). Knowing the dimensions will lead to a volume-per-pixel
value. Thresholding the image somewhat will help to eliminate the effects
of resolution loss which make the organ appear larger than it truely is.
A literature search will turn up loads of papers on this subject.
The Journal of Nuclear Medicine is the best overall source of papers (see
the December issue of the last 6-7 years if you don't have acces to computer
search help.)
BIAS ALERT: Don't believe everything you read, particularly in "M.D." type
journals ... stick with JNM or IEEE type publications that emphasize
"physics" as opposed to "pretty picture" publications. Phone me if you
want to know which ones I mean.
Yes I know there are a lot of generalities here. Yes I know there
are lots of details left out. Hope this helps.
--
Kim L. Greer
Duke University Medical Center try: klg@orion.mc.duke.edu
Div. Nuclear Medicine POB 3949 ...!mcnc!ecsgate!dukeac!klg
Durham, NC 27710 919-681-2711x223 fax: 919-681-5636