[BioC] NIH image processing program

[Vincent Carey 525-2265]

> >>>>> "vincent" == Vincent J Carey, <stvjc@channing.harvard.edu> writes:
>
>     vincent> Is anyone doing anything with "NIH image" a MAC-based
>     vincent> utility for dealing with tiff images of e.g., gels?
>     vincent> I am getting some data from this system which translates
>     vincent> darkness to numbers.  Investigators then use the numbers
>     vincent> as density estimates.  There is a windows-based tool
>     vincent> from Scion with similar functionality.
>
> Is this different from the NIH ImageJ java-based imaging
> software/toolkit (http://rsb.info.nih.gov/ij/)?  It looks like it.

Thanks for the tip.
I did not download the PC version of Image from "Scion".  ImageJ
seems to run OK on windows and linux with jdk1.3.  There are plenty
of sample images to work with.

The problem I am facing is to assist in the determination of
the composition of a sample of fragments that have lengths that
are all multiples of a common number.  We want to estimate the
distribution of lengths or to test for different length distributions
in different experiments.  The image is a series
of equispaced blobs of different intensities and
of different sizes.  ImageJ doc seems to suggest manually segmenting
the profile (profile = density sketch with heights proportional to darkness),
and it provides a tool to calculate area under the curve in each
segment.  There are no references in the top level doc to
considerations of variability of the profile or to principles of
determining extents of segments.

The image processing interface could presumably be hooked into
R handily via SJava.  Where we want to keep the image data is
an open question.  The example tiff file for a 6 lane gel is only
about 100K.

But this interoperation is of minor concern at the moment.
Low throughput is OK for now.  I can use ImageJ offline to obtain
numerical profiles of the gel data.  I've done some work on
an R package for automatic segmentation and area calculations
based on local variability of image profile data produced by
Image or ImageJ and will release that once I have a better
understanding of where these profile numbers come from (if what
I did turns out to make sense).  It would be nice to be able to
precisely juxtapose gel image tiff data with the derived
profile and segmenting results in an R graph.  Do I want to
convert the tiff data to a matrix for plotting by image, or
is there a more streamlined approach?

> I've been meaning for years (2, actually) to integrate this with Orca.
> A student here evaluated it for image analysis of ELISPOT (ELISA
> spotting) assays.
>
> best,
> -tony
>
> --
> A.J. Rossini				Rsrch. Asst. Prof. of Biostatistics
> U. of Washington Biostatistics		rossini@u.washington.edu
> FHCRC/SCHARP/HIV Vaccine Trials Net	rossini@scharp.org
> -------------- http://software.biostat.washington.edu/ ----------------
> FHCRC: M: 206-667-7025 (fax=4812)|Voicemail is pretty sketchy/use Email
> UW:   Th: 206-543-1044 (fax=3286)|Change last 4 digits of phone to FAX
> (my tuesday/wednesday/friday locations are completely unpredictable.)
>
>
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor@stat.math.ethz.ch
> http://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
>