Cell segmentation methods for time-series images?

Hi, I am currently a student in a lab where we use fluorescence microscopy. For our current data analysis process, we measure the fluorescence of cells by using HCImage and manually selecting cells in the time lapse video (scrolling through the video and selecting cells that are lighting up). There are generally 50-100 cells per sample in each time lapse video so this process takes 5+ minutes and is tedious. Is there any way to automatically select these cells/ROI using a specific program? I am hoping to use a program that is suitable or specific for time lapse data and can measure the fluctuating fluorescence for each automatically selected ROI.

I do not have much machine learning or cell segmentation experience so specific recommendations and steps are very appreciated. I do not have a time lapse video example to share, but I have shared two images; one at time 0 seconds and another at time 6 seconds.

*Images removed at request of author.

1 Like

Hey @tshast2,

maybe you want to give Fiji a try? You can download it here:

And you may find this video on image segmentation interesting

And this one on automating the procedure with ImageJ macro in Fiji:

Let us know how it goes and if you need further pointers for specific tasks :slight_smile:


Thank you, Robert! Do you recommend using Fiji as well for time-series images that are not in video format? For one sample, I have an image at time 0 seconds, time 3 seconds, time 6 seconds, time 9 seconds, and so on for about 3 minutes. The cells remain static, however the fluorescence of each cell changes.

That is how most time series are analyzed. You would create a Stack out of the individual images (look up stacks).

If the cells remain static, that reminds me of an MITx course where they created a Maximum Intensity Projection of a time series, segmented the MIP using Analyze Particles… and then exported the data from those ROIs as a CSV file.

Not sure the time frame would work for you for the course, but what was presented is almost exactly what you are describing, but for neurons in a time series of the brain.
1 Like