1 of 60

Getting Started with Fiji/ImageJ

Helen Wilson

hwilson23@wisc.edu

Adapted from work by Edward Evans, Michael Nelson, Ellen Dobson, and many others

http://tinyurl.com/yj332a6p

2 of 60

Who am I?

Helen Wilson
University of Wisconsin-Madison

3rd year BME Graduate student in Dr. Kevin Eliceiri’s Lab
Undergraduate in BME at the University of Delaware

Currently:

Fluorescence Lifetime Imaging Microscopy
Second Harmonic Generation Microscopy

http://tinyurl.com/yj332a6p

3 of 60

Outline:

Software introduction
Examining digital images

Different types of images and formats
2D and 3D display
Enter known scale information

Segmenting areas of interest
Making measurements
Segmentation (StarDist and Labkit)
Tracking (Trackmate)

http://tinyurl.com/yj332a6p

4 of 60

What is ImageJ/Fiji?

Open source tool for scientific image analysis
ImageJ 1.x maintained since 1997, eventually modernized into ImageJ2
FIJI is “batteries included” ImageJ2

Wayne Rasband

Curtis Rueden

5 of 60

Fiji is just ImageJ

Is an image processing package— "batteries-included"
For Users:

easy to install (Mac, Windows, Linux)
easy to update
bundles a lot of plugins (added functionality)
offers comprehensive documentation

https://fiji.sc

Use a search engine

6 of 60

What’s it for? What’s the point?

IMAGES ANSWER QUESTIONS (and maybe create more)

What size is it?

Smaller than x? Larger than y?

Did the shape change?

More round? More protrusions?

Does it move?

What direction? How far? How long?

Does it have the molecule?

Is it where it should be? “How much?”

7 of 60

How do we know what to look at?

Information in the images is driven by the experiment
What is important is driven by the biological question

8 of 60

Generally…

https://www.youtube.com/watch?v=kor9E3Qen28

Images

Regions of Interest (ROI)

Measurements

9 of 60

What is a digital image?

10 of 60

Generating scientific digital images

Output

Usually 8-bit, 12-bit, 16-bit gray scale images.
Can be n-dimensional (i.e. “XYCTZ”).
Images are usually .TIFF format or proprietary format (e.g. .nd2).

11 of 60

A closer look: Images are pixels, pixels are values

Notice inverted LUT (colorscale)

12 of 60

Bit Depth - Binary Representation

Pete Bankhead

Mike Oakely

4 bits = 4 places, 16 combinations

13 of 60

Visualizing image bit depth through gradients

32-bit

16-bit

8-bit

14 of 60

Don’t trust your eyes with image bit depth

Depending on your screen you generally will be unable to detect differences in the gradient after 8-10 bits.
Converting from a higher bit depth to a lower bit depth (even if the image looks the same) results in data loss.

16-bit -> 8-bit

Adapted from: https://gregbenzphotography.com/photography-tips/8-vs-16-bit-depth-photoshop/

15 of 60

Visualizing image bit depth through “resolution spheres”

8-bit

16-bit

32-bit

Courtesy of Edward Evans

16 of 60

File type (.tif, .asc, proprietary files, try to avoid .jpeg)
Dimensions

Not limited to 2D or 3D
X,Y… C? Z? T?

Pixel/voxel size
Coordinates - 0,0 is which corner?
Understanding your metadata is important

In some cases the XY pixel sizes are not the same!

What other metadata do you have?

Microscopy: Cameras and Digital Image Analysis (Nico Stuurman)

17 of 60

A great image is better than perfectly detailed analysis
Common artifacts and issues:

Scratches, debris
Poor staining/staining issues
Uneven illumination
Spectral crosstalk/bleed-through
Improper stitching/stage movement
Focus drift
Saturated signal
Inconsistent parameters

A Note on Artifacts

QUALITY IN = QUALITY OUT

18 of 60

Developing an Analysis Workflow

Define what information you need and research existing tools

Define your control
Segmentation, tracking
Start simple

Consider how each step might “skew” the original image or analysis process during pre or post processing

Filters, thresholding (avoid manual thresholds), processing masks
Record what you do so it is reproducible (consider automation/scripting)

Take measurements from the initial image

WORK ON A COPY OF THE ORIGINAL

https://imagej.net/imaging/principles

19 of 60

FIJI Basics

20 of 60

The Main Window

Tip: click on the status bar
Tip: right / double-click on Tools

Getting Started page of the ImageJ wiki

21 of 60

The Main Window

Right / double-click on Tools
Hold shift to draw horizontal or vertical lines
Hold alt and use arrow keys to change sizes
Hold spacebar to move

22 of 60

CTRL + L: Search Bar

Edit ▶ Options ▶ Search Bar... ▶ Pressing L focuses the search bar

23 of 60

Staying Up-To-Date

Not this one!

https://imagej.net/plugins/updater

24 of 60

Staying Up-To-Date

https://imagej.net/update-sites/following

25 of 60

What to do if you have analysis questions…

https://forum.image.sc/

https://imagej.net

26 of 60

How do we analyze images in FIJI?

27 of 60

Data Visualization

2D, 3D, Hyperstacks

28 of 60

Look up tables (LUTs)

See different types of LUTs
Examine change in perceived information based on LUT
Add calibration bar
See how the image looks for different types of colorblindness

Mlp-viridis can be a good option as perceptually uniform

29 of 60

Image color selection

Estimated 1/12 men and 1/200 women have some type of colorblindness
Grayscale when printing
Include colorbar

FIJI - Simulate Color Blindness

BioVoxxel Plugin

30 of 60

Profile Plots

Easily look at background noise level
Help estimate size of objects

31 of 60

Get to Know Your Data…

File ▶ Open Samples ▶ Boats
Analyze ▶ Histogram (Ctrl+H)

Can look at pixel values and bit depth changes

ImageJ User Guide: Analyze

What would cause this histogram?

32 of 60

A real example of bit depth loss

A look at image histograms
See how changing image type affects the pixel information

33 of 60

Thresholding Segmentation

ImageJ User Guide: Thresholding

Use histogram to threshold
Compare with other auto threshold methods
Create a mask or binary image

34 of 60

Masks from thresholds may need adjustments…

Images may be difficult to threshold and create usable masks
Considering preprocessing the image before thresholding
Masks can be altered with watershed, erode, or dilate

If you are doing this a lot, look into plugins like MorphoLibJ or SAMJ

Cell Image courtesy of Kasia Wiesh, Skala Lab UW-Madison

35 of 60

Analyze Particles

Select and identify individual blobs
Use size exclusion to eliminate smaller blobs
Add the selections to the ROI (region of interest) manager

36 of 60

Take Measurements

Decide which measurements to include
Measure individual blobs

37 of 60

Background Subtraction - Gaussian Filters

Designed to blur out objects, especially good for filtering noise
Width controlled with sigma value
Difference of Gaussians can highlight different objects within size range

Adapted from Pete Bankhead: https://petebankhead.gitbooks.io/imagej-intro/content/chapters/filters/filters.html#sec-filters_gaussian

Input

Gaussian blur (σ = 5)

Blur Image - Input Image

38 of 60

Warning: ImageJ Subtract Background may introduce artifacts

https://forum.image.sc/t/consensus-on-subtract-background-built-in-or-other/7061

39 of 60

2D Gaussian Subtraction Script

https://github.com/elevans/fiji-scripts/

40 of 60

Background suppression: Mean * Input image

Input

Mean filter

(radius = 5)

Input * Mean

Nuclei 1

Nuclei 3

41 of 60

(Mean*input) Threshold Comparison

42 of 60

StarDist - Object Segmentation

Uses deep learning to segment cells, which can be done with pretrained models

Can sometimes have better results if you have overlapping objects

Can be used to classify cells
Works for 2D in FIJI, supports 3D elsewhere (Napari)

Source: https://github.com/stardist/stardist

Additional info: https://stardist.net/index.html

43 of 60

Labkit - Pixel Classifier

Supervised pixel classifier - segmentation tool to identify objects of interest
Works to separate foreground and background using calculated pixel metrics
For 2D and 3D

Source: https://github.com/juglab/labkit-ui

Additional Info: https://imagej.net/plugins/labkit/

44 of 60

Scripting in FIJI

Why are scripts useful?

They facilitate reproducible science…

Document your work
Automate your analysis
Share with the world

Supported Languages:

Groovy, ImageJ Macro, Python(Jython), JavaScript, Ruby(JRuby), Lisp(Clojure), R(Renjin), Java, BeanShell, Scala

Using the Script Editor in ImageJ

45 of 60

Scripting in FIJI...

ImageJ Macro language

Less powerful
Easy to learn and use

Macro Recorder

https://imagej.net/Introduction_into_Macro_Programming#The_recorder

Built-in Macro Functions list

https://imagej.net/ij/developer/macro/functions.html

46 of 60

Batch Processing

Simple and quick way to run macro script on a folder of files
Can sometimes have odd functionality with certain plugins

https://imagej.net/scripting/batch

47 of 60

Batch Processing Within Code

Usually completed with a for loop - template provided with FIJI
Requires some basic knowledge of coding, especially if not using the ImageJ macro language
More versatile and easy to make changes to code