Spatial statistics 2D/3D ImageJ plugin

Image Analysis
, ,
#1

Hello,

I’m trying to use the ImageJ/Fiji plugin Spatial statistics 2D/3D to measure spots distribution inside cells. I’m testing the plugin with two images, one where there is a cluster of spots, and another one where the spots are kind of randomly distributed. When I run the plugin I get reasonable and good SDI values for the G-function. However, the SDI value for the F function is always 1. I have fed the plugin with different images with different patterns, but the SDI for the F function is always 1. Any advice or suggestion?

Thanks a lot.
JJ

#2

Hi @juanje,

The algorithm works with an image of the structure where the spots can be located, are you providing the plugin a mask image of the cell ?

Best,

Thomas

#3

Hello Thomas,

Thanks a lot for your help. Yes, I’m providing a mask image. It is basically a binary mask of the cell. But I’m clearly doing something wrong because it gives me strange results. I’m working on figuring it out.

Best,

Juanje

#4

Hello @juanje were you able to figure out what was happening? I keep getting an SDI = 1 for F function and SDI = 0 for G function and I am hesitant to trust it entirely, because I have been unable to get a result using the random image provided. Any idea what should be entered for the first two parameters on the plugin (I entered the number of spots, but I am not sure this is correct).

Thank you!

#5

Hi @dmajo,

Getting SDI of 1 for F and 0 for G is quite normal. For the first two parameters :

  • Nb Points (F Function) : this is the number of points used to compute the F function; this has nothing to do with your number of spots. The more points you use the better the results, usually we use like 10, 000 points.

  • Samples : This is the number of randomized organization we will simulate to compare with the spatial organization of your spots. Usually 100 is quite enough.

Do not hesitate if you have more questions or comments.

Best,

Thomas

#6

Thank you so much @ThomasBoudier. I will try the analysis again with the parameters you mentioned. Based on the previous comments on this forum and the Andrey et al. paper, my understanding is the closer the SDI is to zero, the more clustered and the closer the SDI is to 1, the more regular the pattern. Does this apply to both functions? Is an SDI of 1 for F and 0 for G a contradictory result? Thank you again.

#7

Hi @dmajo,

A SDI between 0.05 and 0.95 will refer to random organization, below or above these values, it will refer to non-random organization. For a clustered organization, G should have a SDI close to 1 and F a SDI close to 0.

SpatialStatistics
SpatialStatistics1002×760 93.5 KB

Best,

Thomas

#8

Thank you so much for this @ThomasBoudier. Also, the parameters you mentioned (10000,100) are working well.

Much appreciated