[ome-users] FindSpots

[Andrew Patterson]

Forwarding to list.

Begin forwarded message:

> From: Ilya Goldberg <igg at nih.gov>
> Date: 11 May 2010 18:54:41 GMT+01:00
> To: jfaro at uvigo.es
> Cc: Andrew Patterson <ajpatterson at lifesci.dundee.ac.uk>
> Subject: FindSpots
> 
> Dear Jose
> 
> Hopefully these are useful answers:
>> A1) 1st table, Signals, has headings:
>> Feature, Mean, GeometricSigma, Background, GeometricMean, Integral, CentroidZ, Sigma, CentroidY, TheC, CentroidX
>> 
>> -what is the meaning of "Background", "Integral", "TheC"?
> Background is the intensity average of the pixels surrounding the spot that didn't make it into the spot.
> Integral is the sum of pixel intensities in the spot.
> TheC is the channel where the spot was found.
> The Centroids are the weighted mean of the X,Y and Z coordinates for the pixels in the spot - a "center of mass".
> The Gemoteric Mean and Sigma are computed from the log of the intensities
> 
>> A2) 2nd table, Extent, has headings:
>> Feature, Volume, MinX, MinZ, MinY, MaxZ, SigmaZ, FormFactor, MaxX, Perimeter, SurfaceArea, SigmaX, SigmaY
>> 
>> -what is the meaning of "Volume", "MinX", "MinZ", "MinY", "MaxZ", "FormFactor", "MaxX", "Perimeter", "SurfaceArea"?
>> -does Volume refer to the number of voxels in the corresponding spot? If so, does every voxel counted in the volume have neccessarily an intensity above the threshold? or it counts every voxel within the spot surface, irrespective of their intensity?
>> -does SurfaceArea refer to the number of voxels in the spot surface?
> 
> Volume is the number of pixels in the spot.  The "height" of 2D images is 1.
> The Mix and Max (X,Y,Z) define a minimum bounding box that contains the entire spot, where the box sides run along the X,Y,Z axes. They are the minimum and maximum X,Y,Z coordinates for the pixels in the spot.
> The FormFactor is the degree to which the spot differs from a circle, which has a form-factor of 1.0.  Its a dimensionless ratio of perimeter to surface area.
> Perimeter is only meaningful for 2D spots - it is the length around the spot.
> SurfaceArea though meaningful for 3D spots is not always accurately computed in the 3D case.  The algorithm for computing this accurately in all cases turned out to be quite complex to implement, and a simplified/faster alternative hasn't been found.
> 
> In all casses, only pixels in the spot are counted for all statistics with the following exceptions: background only counts the non-spot boundary pixels.  Perimiter, surface area and form factor count only the "outer edge" of the spot, ignoring any holes in it.  Other than these exceptions, the holes are not counted as part of the spot.
> 
>> A3) 3rd table, Location, has headings:
>> Feature, TheX, TheY, TheZ
>> 
>> -what is the meaning of "TheX", "TheY", "TheZ"? It seems redundant with CentroidX, CentroidY, CentroidZ.
> 
> Unlike centroids, these are un-weighted by the pixel intensities, so they are the geometric center of the spot, not its center of mass.
> 
>> B) for any particular 3D image, is it possible to visualize slice-by-slice the spot areas determined by FindSpots superimposed to the corresponding slice image? (This would be required to decide if the spots as determined by FindSpots capture or not the image features of interest, which is required in turn to adjust the threshold and minimum volume parameters...)
> 
> The specific coordinates of each boundary pixel are not retained, so the exact boundary of each spot cannot be displayed.
> 
> Because the algorithm is based on a global threshold, the exact boundaries of each spot can be determined by setting both the black level and white level to the same value - the threshold.  This generates a binary mask which will be an exact representation of what is counted in FindSpots. The only exceptions are small spots that will be ignored because their volume is below the minimum size.
> 
> Hope this helps,
> Ilya
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openmicroscopy.org.uk/pipermail/ome-users/attachments/20100512/1561a2e9/attachment.html>