MRC Institute of Genetics & Molecular Medicine at the University of Edinburgh

www.ed.ac.uk/igmm

Image analysis basics using ImageJ/FIJI

Laura Murphy10th May 2017

What we’ll cover today

1st bit of talk - Basics of digital images

Pixels, bit depth, histograms and LUTs

2nd bit of talk - Introduction to ImageJ and FIJI

Tour of GUI

Opening your images

Changing display of your image

3rd bit of talk – Simple image analysis tasks

Thresholding

Measuring/ analyse particles

Batch processing simple tasks (intro to macro writing)

Using scripts provided by the facility

Any FIJI commands used on a slide will be listed at the bottom of said slide with “>” separating each menu option

Your pretty picture is really just a set of numbers.

0 0 0 0 0 0 0

0 0 0 120 0 0 0

0 0 100 130 120 0 0

0 100 255 255 255 255 0

0 90 255 255 255 255 0

0 88 255 255 255 255 0

0 0 110 255 255 0 0

0 0 0 90 0 0 0

0 0 0 0 0 0 0

The beauty is IN the numbers!

Where do the numbers come from?

Light given off by your sample is picked up each grid point of the camera and an intensity value is stored for each position in your field of view

This is dependent on how much light there is hitting each point of the camera (how strong your signal is there)

Saturated

Light from sample

Pixels?

Pretty images are patterns of numbers representing light intensities

Think of your images as data(pixel values) and display(the coloured squares that make up your image)

Each channel (or time point or Z-position) you captured with is a separate collection of pixel values

Bit Depth?

Pixels can’t have ANY numerical value. It is constrained by “bit-depth”.

This is essentially sensitivity of the camera to different levels of light – how many gray scales!

Each pixel can carry from 1 to 32-bits

For a 8-bit (monochrome) camera each pixel has a possible 256 greyscale values

Pure black = 0

Pure white = 255 (saturation)

For a normal 14-bit (most of the epifluorescence microscopes here)

Pure black = 0

Pure white = 16,384 (saturation)

8-bit 14-bit

Histograms?

It is often useful to consider the range of intensity values within your image which we can see on a histogram

X-axis is the intensity values from 0 -> 255 (8-bit)

0 = black, 255= white and the rest is the grayscale values in-between.

Y-axis the number of pixels in each of those tones

Big peak is the dark background of the image

Don’t lose your data

In general, more bit depth is better. You want all that information when you are analyzing.

Additionally some file formats (e.g. JPEG), squish these numbers to make a smaller file size. This is irreversible so don’t use these file formats.

These are referred to as “lossy” – we want “lossless”.

TIFFs are a lossless file format.

Take away message while analysing leave bit depth alone and keep things as TIFFs

Numbers to colourful images

But I don’t like looking at grey images!

Warning: before you start your experiment!

When you are planning your experiment, you should already have planned how to analyse it.

For true comparison, all settings at the time of imaging need to be the same (i.e. your exposure times between slides).

Additionally, sample preparation needs to be the same!

Need appropriate controls – positive AND negative.

Are you using the correct microscope to answer your scientific question?

If in doubt -> speak to the imaging staff (igmm-imaginghelpdesk@igmm.ed.ac.uk)

Why ImageJ/FIJI?

• Image analysis software that understand most image formats.

• Provides a Graphical User Interface with robust functionalities.

• Platform independent.• Open-source so free

Extensible via plugins.• Active community (can

find a solution for most scenarios).

Tour

Fiji Is Just ImageJ

Menu bar

Tool bar

Status bar

<Tip: hover over image and the status bar will return X,Y co-ordinates and pixel value>

Additional windows will pop for other things – images, histograms, line profiles.

Opening a file in FIJI

File > Open

Or Drag and drop the file on to the tool bar.

Different dimensions such as channels or time points or Z positions can be accessed by moving the slider bar at the bottom of the open image

Any commands you do will be performed on the “active” window or slice

Open “S-Shaped nephron.tif” by drag and drop or by finding it in the dialogue box when you go to File > OpenGo to Image > Show Info to see the metadata FIJI reads about your imageGo to Image > Type to see the bit-type

Brightness and contrast

Contrast and brightness can be used to ease the viewing of your image. In FIJI if you use the sliders to adjust the contrast and brightness you will see the minimum and maximum settings change.

Auto – FIJI optimises brightness and contrast for your image

Reset makes sure Minimum and Maximum match with the actual minimum and maximum pixel values in the image (or 0 and 255 for an 8-bit image)

Set allows you to input their values manually

NEVER PRESS APPLY – this DOES change your underlying numbers

Change the brightness and contrast of the image –Image> Adjust > Brightness/ContrastShortcut (on windows) = Ctrl+Shift+C

Other basic functions

Crop Use tool bar to make a

shape and then Image > Crop

Split channels Image > Colour > Split

channels

Merge channels Image > Colour > Merge

channels

Look up tables Image > Lookup tables

Duplicate Image > Duplicate..

Selection of the LUTs available

Other basic functions

Z- projection (e.g. Max) Image > Stacks > Z-

Project

Histogram! Analyze > Histogram

Plot profile Draw a line using this icon

on tool bar and then go to Analyze > Plot profile

Now to start analysing…

What is binarization and why do I want to do it?

As the word suggests, binarization divides your image into two parts usually referred to as foreground and background. This is the simplest method of image segmentation

The simplest method is to set one or two (upper and lower) cut-off value(s) separating specific pixel intensities from each other.

This is referred to as a threshold

Thresholding

Different thresholding algorithms in drop-down menu. Good option is Otsu – chooses the

threshold to minimize the intraclass variance of the thresholded black and white pixels.

Can manually adjust it using the sliders, BUT should you?

Manual vs. automatic thresholding

Manual

high user bias

time-consuming fiddling-around finding "appropriate" cut-off value

incompatibility with automatic processing

high intra- and inter-user variability

Automated

fully reproducible (on the same image they will always lead to the same binarization result)

no user bias during thresholding

fast (no fiddling) and can be automated (e.g., in macros, plugins, batch jobs – back to that later)

So automated thresholding is better but which one should I use?

16 different thresholding algorithms come as standard

Can try them all and see which one works best for your images.

Image> Adjust > Auto ThresholdSet drop down menu to “Try all” and press okay

None of the thresholding options work to segment my objects of interest!

Process> Filter > Gaussian Blur

Process> Binary >Watershed

Measuring what has been thresholded

File > Open samples > Blobs (set a threshold)Analyze > Set measurementsAnalyze > Measure

Analyse particles

File > Open samples > Blobs (set a threshold)Analyze > Analyze particles (also subject to set measurements options)

Batch processor

FIJI has an in-built batch processor

Allows you to select an input directory (make this a folder containing only your images) and an output directory (make this a different folder)

Process > Batch > Macro..

Macro recorder

Plugins > Macros > Record…

Records what you are doing in the macro language

Menu commands I used here:Image > Adjust > Threshold (Otsu)Analyse > Set measurementsAnalyse > Measure This was all recorded in the macro recorder

Macro commands I use the most (which the macro recorder won’t give me)

getTitle()

E.g. imgname=getTitle()

selectWindow()

Example use

Want to select one channel in your image

imgname=getTitle()

run (“Split Channels”)

selectWindow("C1-"+imgName);

Using already-made scripts from the facility

We store scripts written by the facility on our Githubhttps://github.com/IGMM-AIR/

You can get there via the “Bioimage Analysis section on the facility website: http://www.igmm-imaging.com

On the page you should see a link entitled “scripts and macros”

Okay, so I have a script. What do I do with it?

Drag and drop file (.ijm) on to FIJI window to open Macro tool dialogue

Plugins > new > macro

Copy and paste (or to try and write your own!)

Run button at the bottom of window.

Lines of commands preceded with “//” indicates comments or notes

Tips, tricks and oddities

Know the name of the command but not where it is in the menu? CTRL+L to see (and search) list of commands

Undo is very limited in FIJI so don’t rely on having it. (can sometimes use File>Revert but good to get into habit of Image>Duplicate before you start making changes)

Shortcuts (and create own in plugins-> shortcuts)

Space bar for grabber (little hand symbol)

There’s too many windows and I’ve lost the main window! – Press Enter key on keyboard

Can use the batch processor as a batch file format changer (e.g. ND2 files to TIFFs)

Finally…

The facility staff are here to help BUT the website has a lot of information that might answer your question.

If you have stuff that you think could be added to the website, let me know.

For Loops

You have a box with 10 toys in it, and you want to take all of them out of the box.

You start the count of toys removed from the box at 0, and you keep taking toys out of the box until you have 10 of them. That looks like this:

for (toys removed = 0; toys removed < 10; toys removed ++)

{

remove toy from box

}

The ++ after toys removed signifies that it increases by 1 after each time through the loop. The {} contain the loop (commands)

Batch processer as a macro