4 - Segmentation

4 - Segmentation#

4.1 - DAPI segmentation with thresholding#

  1. open the DAPI.tif image (drag & drop, or File > Open)

  2. change LUT to Grays

  3. Image > Duplicate (IJ:28.9) (shift + d)

  4. Image > Adjust > Threshold (IJ:28.2.4)

  5. understand the function of the Dark Background checkbox (inspect pixel values)

  6. try setting sliders manually. Can you find a good threshold range?

  7. try different algorithms by selecting them in the left dropdown menu. Can you find one that gives a good result?

    • NOTE: Image > Adjust > Auto Threshold, if you want to see all at the same time

  8. try different display options (Red, B&W, Over/Under) by selecting them in the right dropdown menu, do you understand what they show?

  9. when happy with result, click Apply

  10. save the resulting binary image: File > Save As > Tiff

  11. apply watershed to divide touching objects

    • select the binary image

    • Process > Binary > Watershed

  12. proceed with Analyze > Analyze Particles (IJ:30.2)

    • select Exclude on Edges and Add to Manager

    • click on OK

  13. bonus: repeat step 12 but use the Size and Circularity options to try to exclude some particles and the Show dropdown menu to visualize different outputs.

  14. set the parameters you want to measure:

    • Analyze > Set Measurement (IJ:30.2)

    • select Area, Mean gray value, Min & max gray value, Display label

    • click on OK

  15. select the original image (open it again as in step 1 if you do not have it)

  16. in the ROI Manager, click on Deselect and then on Measure

  17. save the Results table as .csv: select the table and click on File > Save As

4.2 - DAPI segmentation with filters and thresholding#

  1. open the DAPI_noise.tif image (drag & drop, or File > Open)

  2. change LUT to Grays

  3. Image > Duplicate (IJ:28.9) (shift + d)

  4. Image > Adjust > Threshold (IJ:28.2.4)

  5. understand the function of the Dark Background checkbox (inspect pixel values)

  6. try setting sliders manually. Can you find a good threshold range?

  7. try different algorithms by selecting them in the left dropdown menu. Can you find one that gives a good result?

    • NOTE: Image > Adjust > Auto Threshold, if you want to see all at the same time

  8. apply a filter of your choice (Mean, Gaussian Blur, Median, …)

    • Process > Filters

    • check the Preview checkbox

    • change the Radius / Sigma. What happens to the image?

    • when you are happy, click on OK

  9. repeat steps 3, 7 and 8 until happy with result, then click Apply

  10. save the resulting binary image: File > Save As > Tiff

  11. apply watershed to divide touching objects

    • select the binary image

    • Process > Binary > Watershed

  12. proceed with Analyze > Analyze Particles (IJ:30.2)

    • select Exclude on Edges and Add to Manager

    • click on OK

  13. bonus: repeat step 12 but use the Size and Circularity options to try to exclude some particles and the Show dropdown menu to visualize different outputs.

  14. set the parameters you want to measure:

    • Analyze > Set Measurement (IJ:30.2)

    • select Area, Mean gray value, Min & max gray value, Display label

    • click on OK

  15. select the original image (open it again as in step 1 if you do not have it)

  16. in the ROI Manager, click on Deselect and then on Measure

  17. save the Results table as .csv: select the table and click on File > Save As

4.3 - DAPI segmentation with Labkit#

  1. open the hela.tif image (drag & drop, or File > Open)

  2. change LUT to Grays

  3. Plugins > Labkit > Open Current Image With Labkit

  4. sidebar, under Segmentation: click Labkit Pixel Classification

  5. topbar: select the pencil tool

  6. sidebar: select foreground. Draw a line inside a nucleus

  7. sidebar: select background. Draw a line outside a nucleus

  8. sidebar: click the play button next to Labkit Pixelclassification

  9. repeat the last three steps until happy with result

  10. click the drop down menu next to Labkit Pixel Classifier. Select Show Probability Map in ImageJ

  11. inspect the probability maps, do you understand the meaning of the values of the pixels in the different channels?

  12. export the segmentation: click the drop down menu next to Labkit Pixel Classifier: Segmentation > Show Segmentation Results in ImageJ

  13. inspect results, do you understand the meaning of the pixel values?

    • you now have a binary image, but not the kind Fiji likes

    • to measure, proceed by thresholding (Image > Adjust > Threshold...): “set” both threshold values to 1, then Analyze Particles, etc

    • alternatively, multiply all values in the Labkit output image by 255, then apply Binarize, etc

  14. save the resulting image with name “myLabkitHeLa1.tif”: File > Save As > Tiff

4.4 - DAPI double-segmentation with Labkit#

  1. open the hela.tif image (drag & drop, or File > Open)

  2. change LUT to Grays

  3. Process > Enhance Contrast. check Equalize histogram. Then, OK

  4. Plugins > Labkit > Open Current Image With Labkit

  5. sidebar, under Segmentation: click Labkit Pixel Classification

  6. sidebar, under Labeling: click add label

  7. rename Label 1 by doublecklicking. For instance into cytoplasm. Optional: choose a different label color by clicking onto the color swatch.

  8. topbar: select the pencil tool

  9. sidebar: select foreground. Draw a line inside a nucleus

  10. sidebar: select cytoplasm. Draw a line inside the cytoplasm

  11. sidebar: select background. Draw a line where there is no cell

  12. sidebar: click the play button next to Labkit Pixelclassification

  13. repeat the last four steps until happy with result

  14. click the drop down menu next to Labkit Pixel Classifier. Select Show Probability Map in ImageJ

  15. inspect the probability maps, do you understand the meaning of the values of the pixels in the different channels?

  16. export the segmentation: click the drop down menu next to Labkit Pixel Classifier: Segmentation > Show Segmentation Results in ImageJ

  17. inspect results, do you understand the meaning of the pixel values?

    • you now have an image with three values

    • to measure, proceed by thresholding (Image > Adjust > Threshold...) at 0, 1, and 2, to extract each class (use Set and then set both thresholds to 0, 1, or 2)

    • then proceed with Analyze Particles, etc for each of the classes of interest (nuclei and cytoplasm)

  18. Bonus round: play with Settings