Tracking Cells in DIC Images

I am working on under agarose assays, in which I take time lapse videos of cells moving on a plastic surface underneath agarose gel. While I have attempted to use fluorescent markers to label the cells and make them easier to identify, it never worked out well and the best images I have obtained are through DIC live cell microscopy. I know from reading the forum that Cell Profiler does not handle DIC images too well and I have tried to incorporate the ideas you have provided on dealing with these images, but I am still having trouble.

The pipeline I have currently is as follows. The goal is to identify the cells, track them through the entire set of time lapse images and obtain a video of the tracked cells moving as well as measurements on their total distance moved:

LoadImages:
File type to be loaded:individual images
File selection method:Text-Exact match

Crop:
Select the input image:Orig
Name the output image:Cropped
Select the cropping shape:Rectangle
Select the cropping method:Mouse
Apply which cycle’s cropping pattern?:First

EnhanceEdges:
Select the input image:Cropped
Name the output image:Edged
Automatically calculate the threshold?:Yes
Absolute threshold:0.2
Threshold adjustment factor:1
Select an edge-finding method:Canny
Select edge direction to enhance:All
Calculate Gaussian’s sigma automatically?:Yes
Gaussian’s sigma value:10
Calculate value for low threshold automatically?:Yes
Low threshold value:0.1

IdentifyPrimaryObjects:
Select the input image:Edged
Name the primary objects to be identified:Cells
Typical diameter of objects, in pixel units (Min,Max):30,100
Discard objects outside the diameter range?:Yes
Try to merge too small objects with nearby larger objects?:Yes
Discard objects touching the border of the image?:No
Select the thresholding method:Otsu Adaptive
Threshold correction factor:0.1
Lower and upper bounds on threshold:0,1.0
Approximate fraction of image covered by objects?:0.01
Method to distinguish clumped objects:Intensity
Method to draw dividing lines between clumped objects:Intensity
Size of smoothing filter:10
Suppress local maxima that are closer than this minimum allowed distance:5
Speed up by using lower-resolution image to find local maxima?:Yes
Name the outline image:CellOutlines
Fill holes in identified objects?:Yes
Automatically calculate size of smoothing filter?:Yes
Automatically calculate minimum allowed distance between local maxima?:Yes

TrackObjects:
Choose a tracking method:Distance
Select the objects to track:Cells
Select object measurement to use for tracking:Intensity_Area
Maximum pixel distance to consider matches:50
Select display option:Color and Number
Save color-coded image?:Yes
Name the output image:TrackedCells

ExportToSpreadsheet:

SaveImages:
Select the type of image to save:Movie
Select the image to save:TrackedCells

While the pipeline seems to work fairly well, it is not identifying the cells reliably and doesn’t always do a good job tracking them from image to image. I have attached an example picture of the cells I am trying to identify and track. Any suggestions you have for me would be much appreciated, whether on the pipeline itself or on any ways I could format the images prior to loading them into Cell Profiler.

Thanks,
Marshall

Hi Marshall,

Based on the image you have uploaded, I have difficulty seeing how CellProfiler (or any other biological image processing software) would be able handle this image. Unfortunately, there is almost no cell-to-cell or cell-to-background contrast to speak, and I would have a hard time distinguishing cells by eye. If a human cannot make out the features of interest, it is most likely not possible for a computer to automatically do the same either.

Regards,
-Mark

Hi Mark,

I realize that the image I provided is not the best resolution and I am working on getting better ones through further experiments. Do you have any suggestions on capturing more clear and distinct DIC images or any thoughts on how I could improve the pipeline I already have?

Would you mind posting the pipeline itself to the forum? I know that you placed the text in-line in your initial post, but it’s best to have the original just so nothing gets lost in the translation.
-Mark

Sure. Here it is…

Marshall
Cell Tracker.cp (8.32 KB)

After looking at the pipeline (and also a colleague who is an expert in ImageJ), I don’t think these images are tractable for analysis without either using a fluorescence-based solution, or teaming up with a computer scientist who wouldn’t mind taking this on as a project.
-Mark