Error in TrackMate spot detection depending on previous file conversation

imagej
trackmate

#1

I am attempting to track cells within a time lapse image using TrackMate and am getting strange results when detecting my spots.

For background, my files are originally .nd2 files that I convert to .tif format using ImageJ. There seems to be an error in the files in which I use the Bio-Formats plug-in to do the conversion as opposed to using the Nikon ND2 Reader plug-in.

The examples below are from the same .nd2 file. The first example was converted using Bio-Formats and the second with the Nikon ND2 reader.

For this first example, the .nd2 file was converted into a .tif using the Bio-Formats plug-in. The image was cropped, thresholded, segmented, and saved as a .tif.

I used the LoG detector to detect my spots with the parameters as below:
ESTIMATED BLOB DIAMETER: 6.45 microns
THRESHOLD: 1.0
CHECKED: Do sub-pixel localization

Using these settings, I receive the following spot detection results, in which the spots are detected around my cells as opposed to the cell itself:


For my second example, the .nd2 file was converted into an .avi using the Nikon ND2 Reader plug-in. The image was then cropped, thresholded, segmented, and saved as a .tif.

I used the LoG detector to detect my spots with the parameters as below:
ESTIMATED BLOB DIAMETER: 10.0 pixels
THRESHOLD: 1.0
CHECKED: Do sub-pixel localization

Using these settings, I receive the following spot detection results, in which the spots are correctly detected:


This issue happens with all detectors but for the purposes of this example, I used the LoG detector. I have also tried different diameter and threshold values but this only varied the size and number of spots surrounded the cells. When it comes to actually tracking the cell movements, there seems to be no issue other than the fact that the first example has many almost identical tracks due to the way it is detecting the spots. I have attempted to re-threshold the first image to determine if there were small particles within the image that were being detected but this did not correct the issue. This error has occurred on both Windows 10 and Fedora 26.

Is there a way to correct the spot detection in the first example? I convert my files using Bio-Formats as opposed to the Nikon ND2 reader as Bio-Formats includes the metadata from the original experiments (image resolution, micron to pixel size, channels, etc.), but am now running into these issues with TrackMate.

Any help with this issue would be greatly appreciated.


#2

You are using an inverted LUT in the second image. The spots appear black but they are actually white, which is what TrackMate is looking for (bright blobs over dark background).


#3

Hi @tinevez,

Thank you for the reply! I inverted my images and TrackMate is now able to correctly detect the spots. Thanks again for the help!