Intensity measurements by Gaussian peak fitting

Hi all,

I’m trying to find the best way to estimate the total intensity of (Gaussian-shaped) spots in 2D and 3D images in ImageJ.

  • There is GaussFit OnSpot, but it is 2D only (still it might be interesting how it compares to the other options).

  • TrackMate has a SpotIntensityAnalyzer, but it simply sums up the pixels in a SpotNeighborhood, which might be too inaccurate for our use case, and I’d favor accuracy over performance in this case (if someone already did some benchmarking on this, I’d actually be happy to see it).

  • I noticed there is GaussianPeakFitterND in legacy-imglib (that appears to be used in @StephanPreibisch’s RadialSymmetryLocalization plugin) and I will check it out. Is there a reason why this is not yet in imglib2, and are there plans to migrate or replace it??

Or is there some implementation in Image Ops that I missed?

/cc @tinevez @dietzc

Is this wide-field or confocal imaging?

For wide-field, the intensity of a point object in the image is independent of the z-plane in which you acquire the data. If the spot moves out of focus, the light spreads over more pixels (i.e. the width of the Gaussian gets wider), but the sum stays the same (although it becomes harder to measure accurately, since the signal per pixel approaches background).

For confocal, this is no longer the case, and intensity distribution in z will become a function of the point spread function of the microscope. If all you want is intensity of the given spot, I think the most straight forward way is to deconvolute the image, and then measure the intensity (by summing the background corrected pixels values). For 2D, I would certainly trust the sum of background-corrected pixel values of the raw image better than the integral of the Gaussian fit as the first is much less dependent on z-height (see: Franke, C., Sauer, M. & van de Linde, S. Photometry unlocks 3D information from 2D-localization microscopy data. Nat. Methods 14, 41–44 (2017)).

1 Like