Quantcast

Results table overwrite - Issues with Find Maxima?

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Results table overwrite - Issues with Find Maxima?

Feriel P
Hello everyone,

I have been struggling with a Results table overwrite issue that I have
been unable to debug. The attached piece of code came from a much larger
macro that worked fine until a month ago. The code in Test.ijm consists of
a for loop iterating 3 times. During each iteration, it opens one z-stack,
puts it through a few plugins (including 3D Spot Segmentation from Thomas
Boudier, and Find Connected Regions from Mark Longair) and prints spot
counts to the Results table. In the end, there should be 3 z-stacks
processed by the macro, with their spot counts displayed in individual rows
in the Results table. However, every time the loop carries out an
iteration, all results for that z-stack are printed to the first line (i.e.
Row 0) of the Results table. This means that the first z-stack has its
results displayed on Row 0, then the second z-stack overwrites its results
to the same row, and this gets overwritten by the third.

According to the workflow in the code, the first piece of data sent to the
Results table comes from the command *run("Find Maxima...",
"noise="+noiseTol+" output=Count");* where noiseTol is the variable for the
intensity threshold above which to count maxima/spots. Once the Count
output is sent to the Results table, a few more pieces of data (e.g. spot
count from 3D Spot Segmentation, Image name and number labels, etc) get
sent to new columns in that same row, with the help of setResult()
functions. To my knowledge, the *run("Find Maxima...", "noise="+noiseTol+"
output=Count");* function should create a new row in the Results table each
time, but within this loop structure, it only seems to write to row 0. This
may be due to bad code on my part that I have been unable to find and
correct, but I wonder if it has something to do with some change in Find
Maxima's behavior?

If anyone can read through/run the code with the attached Example.tiff
z-stack, and help me debug this Results line overwrite issue, I would be
most grateful. Currently running v1.50i of ImageJ and have the latest
versions/releases of Bioformats, 3D Image Suite, etc plugins. Note: I have
modified the code so that the one example z-stack has its intensity
increased in the second and third iterations of the loop, so that the macro
will be able to generate new higher spot counts from the second and third
iterations.

Thanks for the help!

--

Feriel Presswalla
Kellogg Eye Center

University of Michigan

------------------------------------------------------------------------------------------------------------------------------

Example.tiff z-stack:
https://www.dropbox.com/s/v5gtx4z5hemp4my/Example.tif?dl=0
Code for Test.ijm: https://www.dropbox.com/s/s4y6k39g6cfya4q/Test.ijm?dl=0

//

example1=1;

example2=2;

example3=3;

filterSize = 4;

threshold = 5000;

noiseTol = 5000;

gaussianRadius = 5;

titles = "Title";

//setBatchMode(true);

for(i=1; i<=3; i++)

{

  if (i==1)

  {

  //Dialog box allows you to open the Example file

  run("Bio-Formats Macro Extensions");

file = File.openDialog("Choose a .tif file");

Ext.setId(file);

print(file);

run("Bio-Formats Importer", "open=[" +file+"] autoscale color_mode=Default
view=Hyperstack stack_order=XYCZT");

  example=example1;

  }

  if (i==2)

  {

//Dialog box to open the same Example file

run("Bio-Formats Macro Extensions");

file = File.openDialog("Choose a .tif file");

Ext.setId(file);

print(file);

run("Bio-Formats Importer", "open=[" +file+"] autoscale color_mode=Default
view=Hyperstack stack_order=XYCZT");

//Increases the intensity of the Example file so that the macro will treat
it as a different image

//and generate a higher ROI count above the same fixed threshold

run("Add...", "value=1500 stack");

  example=example2;

  }

  if (i==3)

  {

//Dialog box to open the same Example file

run("Bio-Formats Macro Extensions");

file = File.openDialog("Choose a .tif file");

Ext.setId(file);

print(file);

run("Bio-Formats Importer", "open=[" +file+"] autoscale color_mode=Default
view=Hyperstack stack_order=XYCZT");

//Increases the intensity of the Example file so that the macro will treat
it as a different image

//and generate a higher ROI count above the same fixed threshold

run("Add...", "value=2500 stack");

  example=example3;

  }

  rename("rawBackSubtracted"+example);

  run("Z Project...", "projection=[Max Intensity]");

//Collects greatest number of local maxima without collecting noise. Can be
increased to 8^3 if stack contains bright artifacts even after background
subtraction.

selectWindow("rawBackSubtracted"+example);

run("3D Fast Filters","filter=MaximumLocal radius_x_pix="+filterSize+"
radius_y_pix="+filterSize+" radius_z_pix="+filterSize+" Nb_cpus=4");

print("Segmenting loop in Complete; filter size ="+ filterSize);

selectWindow("3D_MaximumLocal");

rename("3D_MaximumLocal"+example);

run("Z Project...", "projection=[Max Intensity]");

rename("MAX_seeds"+example);

//Error messages displayed in the Console Window come after Watershedding
and before 3D segmentation begins

run("3D Spot Segmentation", "seeds_threshold="+threshold+"
local_background=0 radius_0=2 radius_1=4 radius_2=6 weigth=0.50
radius_max="+gaussianRadius+" sd_value=1.17 local_threshold=[Gaussian fit]
seg_spot=Maximum watershed volume_min=1 volume_max=1000000
seeds=3D_MaximumLocal"+example+" spots=rawBackSubtracted"+example+"
radius_for_seeds=2 output=Both 32-bits");

print("Segmenting thresh ="+threshold+"; gaussian radius ="+gaussianRadius);

selectWindow("seg");

//These window names are generated with each run, so they must be renamed
to avoid using an old one in the next iteration. None of them will be
closed, allowing user to look at them all, compare, and evaluate settings

rename("seg"+example);

//"seg"+example is currently 32 bit. To be binarized by dividing by itself
it needs to be something other than 32bit, and we chose 8bit because the
Find Connected Regions plugin needs 8bit

run("8-bit");

//Divide the segmented stack by itself to create a binary stack

imageCalculator("Divide create stack", "seg"+example,"seg"+example);

  selectWindow("Result of seg"+example);

  //Invert the intensities so that the ROIs are highest intensity and the
rest is black and 0

run("Invert", "stack");

//Using the Find Connected regions plugin to eliminate stray 1-2 pix ROIs
so that they do not unnecessarily inflate the ROI count

run("Find Connected Regions", "allow_diagonal display_one_image
regions_for_values_over=1 minimum_number_of_points=3 stop_after=-1");

objCount=0;

for (s=1;s<=nSlices;s++)

{

setSlice(s);

List.setMeasurements;

max=List.getValue("Max");

if (max>objCount)

{

objCount=max;

}

if (max<objCount)

{

s=nSlices;

}

}

//Generated All Connected Regions stack is always 16bit. Since it is a
count mask we divide it by itself to make it binary

imageCalculator("Divide create stack", "All connected regions","All
connected regions");

selectWindow("Result of All connected regions");

rename("Result of All connected regions"+example);

run("16-bit");

run("Invert", "stack");

run("Subtract...", "value=65533 stack");

close("All connected regions");

//Segmented stack needs to be same bit depth as raw stack so that the max
projections can be merged

//7000 is a good noise tolerance; a good modification would be to stop and
let the user examine the point selection Complete to get a better sense of
Find Maxima...countrun("Z Project...", "projection=[Max Intensity]");

selectWindow("MAX_rawBackSubtracted"+example);

run("Select None");

run("Find Maxima...", "noise="+noiseTol+" output=Count");

print("NoiseTol after Count is done = "+noiseTol);

//To beginn with 0th row of the Results table and end with (seriesCount-1)
to get Find Maxima..., 3D ROI Count, and %Difference in Results window

//Store Find Maxima... count from Results Window in variable and use to
calculate error with objCount

selectWindow("Result of All connected regions"+example);

//run("3D Objects Counter", "threshold=1 slice=1 min.=1 max.=65024
summary");

updateResults();

j=nResults-1;

setResult("Series #",j,example);

//Title number here will show the row you intended it to be on. i-1 SHOULD
be the row number, but it keeps printing on row 0

t=titles+(i-1);

setResult("Series/Condition Name",j,t);

setResult("3D ROI Count", j, objCount);

updateResults();

//This works after the two setResult commands to refresh the Results window
with new values

objCount=0;

maxCount=0;

/*selectWindow("Connected seg"+example+"-3Dseg");

rename("Connected 3DROI"+example);

imageCalculator("Divide create stack", "Connected 3DROI"+example,"Connected
3DROI"+example);*/

//Create a max projection of the binary Connected 3DROI stack. We do it
this way because you cannot do a Max projection of the seg stack directly,
since the ROIs have pxel values of the object labells, not their intensities

selectWindow("Result of All connected regions"+example);

run("Z Project...", "projection=[Max Intensity]");

rename("MAX_Connected 3DROI"+example);

//Create a composite where the red channel is the max projection of seg,
and the green channel is the max projection of raw. This allows user to
compare colocalization, to see if the segmented ROIS match the raw spots,
or if there are false positives, or missed true spots

selectWindow("MAX_rawBackSubtracted"+example);

run("Enhance Contrast", "saturated=0.35");

run("Merge Channels...", "c1=[MAX_Connected 3DROI"+example+"]
c2=MAX_rawBackSubtracted"+example+" create keep ignore");

rename("Composite"+example);

//close("seg"+example);

close("Connected seg"+example);

close("Connected 3DROI"+example);

close("Result of seg"+example);

close("Result of All connected regions"+example);

close("3D_MaximumLocal"+example);

close("MAX_raw"+example);

close("rawBackSubtracted"+example);



}

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Loading...