Image Analysis Study

If you choose to participate, you will be asked to segment fluorescence z-stack images of nuclei in 3D cell culture or FISH staining in C. elegans and then complete a survey to report your results and describe your analysis strategy. The goal of this study is not to identify novel segmentation strategies, but to compare the ways in which diverse individuals approach a common image analysis problem. You do not need to be an image analysis expert to participate! We seek a broad range of participants.

The links, below, will guide you through the steps in participating in this study. Thank you in advance!

If you have already completed the study and/or are interested in the ground truth data and information for how we produced the images, please see here: https://github.com/ABRFLMRG/image-analysis-study. The page contains links to the ground-truth images corresponding to the study images, as well as the program and configuration files that were used to produce the datasets.

-The ABRF Light Microscopy Research Group (LMRG)

Note: Regarding the issue with the mismatch between the dimensions reported and apparent PSF for the nuclei images, you can find updates here as well as on the github page, above. We have identified the source of the problem and determined that, in addition to making deconvolution problematic, it reduced the blur of the nuclei making the images easier to segment. We have also determined that the issue did not affect the FISH images.


Your task is to segment the objects in a fluorescence z-stack image and provide us with 1) specific measurements of the objects, and 2) a detailed description of how you conducted your analysis.

There are two different image sets: "Nuclei" contain images of nuclei in 3D cell culture; "FISH" contain Fluorescence In Situ Hybridization staining in C. elegans. Each set has four different z-stack images that vary in their signal-to-noise ratio and the clustering of their objects. You may analyze as many or as few of the images as you like (but the more the better!). [Note: the images are not real; they were generated using the CytoPacq software developed by the Centre for Biomedical Image Analysis at Masaryk University, Czech Republic. More information available upon request.]

You may segment the images using any method and software you like and you may use different strategies for each image. However, you need to provide a full description of your strategy (so take notes and save your protocols/code!). You must also allow us to discuss your strategy (anonymously) in public presentations and publications (see the Data and Metadata Submission section for more details).

1. Image download

You may download and analyze as many of the images that you like. If you aren't sure how many you want to analyze, please prioritize image #1 in each set (the easiest ones) and then image #4 (the most difficult). The more times an image is analyzed by different people, the more useful the data are for our study, so please analyze as many as you can. Thank you, in advance!

Calibration Image: When you have completed your analysis of an image, please run the calibration image through the same protocol. Your analysis of the image will provide a reference for the coordinate system (location of the origin) as defined by your analysis program, and possibly altered by your analysis (e.g. image is flipped/rotated). The image is a simple, z-stack containing four spheres. Although the image only needs the most minimal segmentation (it is essentially binary, objects are 255, background 0), it should be processed through any analysis steps that affect the orientation or location of the origin. If your analysis protocol is essentially the same for all images in a set (e.g. all nuclei images), you only need to analyze the calibration image once for the set.

Note for ImageJ/FIJI users: We recommend opening the images through the Bio-Formats Importer Plugin. Dragging-and-dropping the file on the ImageJ control window to open it may prevent some of the metatdata from loading (specifically the voxel dimensions).

Nuclei in 3D cell culture

Widefield fluorescence, 20x/0.75NA

Voxel dimensions: 0.124x0.124x0.200 um

Emission peak wavelength: 500 nm

FISH in C. elegans

Spinning disk confocal fluorescence, 100x/1.3NA

Voxel dimensions: 0.162x0.162x0.200 um

Emission peak wavelength: 500 nm

Calibration Image

(see note, above)

Voxel dimensions:

1x1x1 um

Note: If you want to download all the images in a set, hover your pointer over the file list, and a gray box with an arrow will appear in the upper-right corner. Click on the box to open the files in Google Drive.

2. What we need from you


Please provide the specified metrics (with correct units) for each segmented object in the image. Format your data as a CSV file with one object listed per row and using the column headings specified below. Use a separate CSV file for each image file. Label your file: LastName_FirstName_ImageName.csv. Use just the first portion of each file name (e.g. nuceli1, fish1, etc.).

If possible, please also provide a 3D volume image of your segmented image.

Nuclei in 3D cell culture images (and associated calibration image[s])

  • Centroid location in microns (um) for each nucleus: x, y, and z coordinates listed in separate columns. Labels x, y, and z.

  • Integrated Intensity. Label intensity.

  • Volume in cubic microns (um^3). Label volume.

  • Download CSV template here.

FISH in C. elegans (and associated calibration image[s])

  • Centroid location in microns (um) for each spot: x, y, and z coordinates listed in separate columns. Labels x, y, and z.

  • Integrated Intensity. Label intensity.

  • Download CSV template here.

Calibration Images: Use the same metrics as the corresponding image (e.g. if you ran the calibration image through your nuclei analysis protocol, provide the nuclei metrics). Provide these data as a separate CSV file. Use the same naming convention, but add _calibration to the end of the file name.


When you complete the submission survey (see next section), you will be asked to provide details about your analysis strategy including:

  • The software package and any plugins/libraries/packages used

  • The steps used in your analysis, including details such as the names of filters/algorithms/etc. used

  • Why you chose this particular strategy

  • How long it took you to segment the image

  • How confident you are in your result

  • Upload a code/macro/pipeline file (if applicable)

3. Data and Metadata Submission

This Qualtrics survey will lead you through the submission process, including uploading your data, reporting your analysis strategy, and collecting demographic information. Please complete all your analyses before starting the survey; this will allow you to enter all your data at once and prevent you from having to reenter your demographic information. Please do NOT change your analyses after viewing the survey.

By completing the survey, you are granting us permission to use all of the information you provide for our study. Identifiable information will only be used to match records (i.e. if you complete the survey more than once). All data will be de-identified prior to downloading and analysis. We may quote short sections of your written responses anonymously in public presentations and publications, but will not reproduce your response in its entirety. Your name or identifiable information will not be associated with any public presentation or publication of this study unless you wish to have your name included in a list of volunteers in the acknowledgements.

Contact information

Questions about this study may be addressed to Kristopher Kubow (LMRG chair) at kubowke@jmu.edu.