[BioC] Image analysis for single channel macro arrays

[Nicholas Lewin-Koh]

Hi,
Thanks, your insights are pretty much the conclusions I was coming to. 
Spot doesn't use histogram methods for segementing, it uses
"seeded region growing" and morphological opening and closing.
If my understanding is correct it grows the pixels outward from the seed 
points with a decision rule based on the nearest neighbors.
Unfortunately, the way it is implemented is not very flexible and
only seems to work for 2 channel arrays. 

I found some work done by Adele Cutler using ImageJ
http://www.math.usu.edu/~adele/spotfinder/index.htm, But I can't get
their 
plugin to work. My java is very rusty, and I think the version of
Minpack_f77.java
they use is older than the one on the web, and the api has changed. Ugh

I will take a look at Arraypro, but I have not
been impressed with anything commercial yet.

Thanks Again,
Nicholas






 
On Fri, 30 Jul 2004 02:01:41 -0400, "Peter Wilkinson"
<pwilkinson at videotron.ca> said:
> Well, I have looked deeply into this and I am not impressed with any 
> software out there. Genepix is rather good but is painfully log because
> its 
> mostly manual.  Quantarray is poor at segmenting, and its adaptive 
> algorithm is poor. The histogram method is only good for well defined 
> spots, or high quality runs. Its automated feature does not work well at 
> all. I thought Spot uses a histogram .... but I don't remember, and its
> not 
> automated.
> 
> The problem with ALL the software I have seen is that if you put the 
> software in different technician's hands you will get different results.
> 
> I do have a demo of Arraypro analyser that I have not tried it yet, but 
> this one boats reproduceability between users. You can find a demo of
> that 
> one. However after talking with their scientists about the segmenting
> .... 
> I was not that impressed.
> 
> I don't think any commercial software has a chance if you have diffusion.
> I 
> can not imagine anything being good enough for dealing with that the way 
> one might expect
> 
> As far as background, unless you are dealing with radioisotopes,
> background 
> subtraction can be over-rated, and should not be done.
> 
> I am always on the hunt for good segmenting software .... soon I hope to 
> find the time to write my own.
> 
> Have a go with Array Pro  and good luck .... we all need it.
> 
> Peter
> 
> 
> At 11:26 PM 7/29/2004, Nicholas Lewin-Koh wrote:
> >Hi,
> >Does anyone have any insight into what works well for spot
> >identification
> >and quantification for membrane arrays? I tried Spot but it only seems
> >to work on dual channel
> >arrays. Most of the other commercial programs don't seem to be very
> >good, the features
> >of interest to me are automatic gridding, good treatment of overlapping
> >spots (they diffuse more on membrane
> >arrays), and good backgound estimation.
> >
> >Thanks for any insight/help.
> >
> >Nicholas
> >
> >_______________________________________________
> >Bioconductor mailing list
> >Bioconductor at stat.math.ethz.ch
> >https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
> 
>