ImageJ Introduction OCCM

Transcription

ImageJ Introduction
•
•
•
•
•
•
•
•
•
spatial calibration
measurements
thresholding
regions of interest
particle counting
confocal files
stereology
plugins
ImageJ links
Microscopy Otago
occm.otago.ac.nz
October 2011
ImageJ Basics and a Bit Beyond Workshop
Otago Centre for Confocal Microscopy
ImageJ Overview
The program doesn’t have a pretty interface, but if you substance over appearance then ImageJ is
the Swiss Army knife of image quantification. The base program, as downloaded from the ImageJ
website (http://rsbweb.nih.gov/ij/) will provide you with a comprehensive range of analytical tools,
ranging from measurements of length, area, circumference to thresholding, particle counting and
density. The program will accept a wide range of data types, covering most of the major sources of
images used in research. Brightfield, fluorescence, confocal, electron microscopy, !CT and
electrophoresis gels for example. Versions are available for Macintosh, Linux and Windows.
Many specialised functions are included with ImageJ in the form or both plugins and macros, it’s a
case of dredging through them to find something to suit your needs, or searching the web with the
appropriate key words to find one.
Most users find they want to customise their copy of ImageJ, this can be done by downloading and
installing plugins from the comprehensive (and ever increasing) list found at http://rsbweb.nih.gov/
ij/plugins/index.html. Generally plugins are easy to install and are produced by users who wish to
expand ImageJ’s capabilities.
Another way to customise ImageJ to better suit your needs is to write, or record, marcos from
within ImageJ. This can be done as easily as recording the steps of a repetitive task as you do them,
the resulting macro is then played back and reproduces the steps. Alternatively they can be written
in a text program from scratch, or modifications can be made to recorded macros. The limits of
macros and plugins are determined by your personal level of nerdiness.
Main Panel
The main panel is your starting point for selecting regions of interest (ROI), adding text and
generally deciding which features in your image you’d like to investigate further.
Menus
Like an over ambitious restaurant, ImageJ’s menus can be a bit intimidating, particularly as they
often contain the primary functions you will want to use. The trick is to browse about and see what
things do, after a short time you will become familiar with the layout and gradually expand your
repertoire. Much like at the intimidating restaurant - but a great deal cheaper.
-1-
© Andrew McNaughton 2010, Microscopy Otago
Opening and Calibrating an Image
One of the commonest things done with ImageJ is measuring lengths, areas or perimeters. To do
this the program must be calibrated, either from a known length in the image itself, such as a scale
bar, or from a reference image taken under the same conditions. The calibration process in both is
much the same.
Step 1 Open the Image
Open your image from the File menu, either choosing the Open option for common file types, or
from the Import option which lists many other less common file types.
Some file types, such as Zeiss
confocal microscope images will
be opened with the correct
calibration already set as ImageJ
can read the scaling information
associated with the image.
Most will not be calibrated and this
will need to be set manually.
If your data is a series of images
they can all be opened
simultaneously using the Image
Sequence menu option. In most
cases calibrations will be applied
across the entire series.
-2-
Step 2 Image Calibration
An uncalibrated image will only show the pixel
dimensions of the image, a calibrated image will
also show the dimensions in the appropriate units.
To calibrate an image from a scale bar, or other feature of known length in the image, the line tool is
selected from the main control panel and a line drawn along the calibration bar or feature. To force
the line to draw horizontally or vertically hold down the shift bar while dragging the line tool.
Try to choose a feature of reasonable length to
use as your calibration source. Inevitably you
will be a few pixels out when drawing over it so
the error will be proportionally less across a
longer distance.
If you know you’ll be using a scale bar for this
purpose, make a big one right across the image.
-3-
From the Analyse
menu choose the Set
Scale option. The line
you selected should
still be active along the
feature you chose.
The value to change in the window which appears is the
Known Distance, this is the length of the feature you
measured. In this example the length of the scale bar was
50mm so 50 is entered in the box. The unit of length is also
entered in the appropriate box, although this is just a label
and has no effect on the numerical values of measurements.
To apply the same calibration to other open images, or ones
you will open, tick the Global option.
The scale at the bottom indicates how many pixels per unit
of measurement.
Once the calibration values are known they can be noted and entered directly into this window on
future occasions, without using the line tool step - if they were recorded under the same conditions.
Once calibrated the image window will show its
dimensions in the appropriate units in addition to
the pixel dimensions.
-4-
Measuring Image Parameters
Once an image is calibrated any measurements made using ImageJ’s measurement tools will have
the appropriate dimensions. There are many choices in the default set of measurement options,
those of interest can be activated and those not required can be deactivated. Results are recorded in
a separate window and can be exported in Excel format or imported into other analysis programs.
ImageJ is excellent for obtaining quantitative data, but in many cases it may be best analysed
elsewhere, unless you install the appropriate plugins to perform statistical analysis on data.
Setting Measurement Options
ImageJ can measure all manner of
parameters but to avoid confusion it’s
best to disable those you don’t require.
Choose Set Measurements from the
Analyze menu to open the options
window where you enable, or disable,
measurement options.
In this example only Area and Perimeter
(doubles as Length for the line tool) are
enabled for recording. Other options can
be activated as required.
Always tick the Display Label option as
this adds the appropriate column label to
the exported data - very confusing
without them.
This example has Limit to Threshold
ticked, which means measurements such
as Area are limited to those parts of the
image fitting within a predetermined
threshold range. This can be very useful
and save a lot of time - more later.
-5-
Measuring Length
The next step is usually to measure something, in this first example it will involve measuring the
distance from one point to another using the line tool. The process is identical to that used to
calibrate the scale of the image by tracing a line along a feature of known dimensions. In this case
you simply don’t know the feature’s dimensions in advance.
Draw the line tool across the feature of interest.
Choose Measure from
the Analyze menu (or
press ‘apple m’). This
will cause the Results
window to open and
show the values of the
parameters you chose to
record. For the line tool
the Angle option is
always recorded, so try
and find a use for it.
By default the line tool will measure a straight line (hold Shift key to force it horizontal or vertical)
but segmented lines, freehand lines and arrows can be chosen by holding down the Option key
when selecting the line tool option.
If you repeat the steps above the Results window will include multiple recordings. These can be
accumulated and then saved as an Excel compatible file (or into other spreadsheets such as
Numbers). If you change measurement tools you will be prompted to save the existing results as
changing tools overwrites the previous tool’s results.
Results can be cleared manually by selecting the Clear Results option form the Analyse menu (see
above).
-6-
Measuring Area Using Selection Tools
Ensure Area is ticked in the Set Measurements window (Analyze / Set Measurements menu) and
then choose one of the the area selection tools from the main menu. Note that some of the tools
have small red triangles below them, this indicates there are additional functions available. To
access them press the Control key as you select the tool, this will cause an options menu to appear
from which you can choose.
The next step is to measure something, in this example it will involve manually tracing around a
region of interest (ROI) and recording the area and perimeter of the ROI. Choose the freehand
tracing tool, shown above, and begin tracing around the ROI.
This tool will always close the tracing, if
you don’t the two endpoints will be linked
with a straight line. Be sure to close the
trace where you want it to close, rather
than have it do it for you in an unexpected
way. As this tool measures area the ROI
must be enclosed.
For free hand tracing of unclosed
perimeters, hold down the Control key and
choose from the Line Tool’s additional
menu options.
To view the measurement results choose
Analyze / Measure from the main menu or press the shortcut keys (Command / m). The Results
window appears showing whichever parameters you chose to measure. Choose File / Export when
the Results window is active and they can be exported for further analysis.
The selection tools are useful for well defined ROIs, essentially features which are easy to trace
around by hand. Once features become more complex it can be next to impossible to do this by
hand unless the image is zoomed up high enough to show individual pixels. Such a task would soon
become prohibitively time consuming. Cue thresholding.
-7-
Measuring Area Using Thresholds
Threshold Introduction
In 8 bit greyscale images (and colour images, but more of that later) there are 256 (28) intensity
graduations which can be assigned to a pixel. A pixel with an intensity of 0 is black, a pixel with a
value of 255 is white, everything in between is a shade of grey. Images with 16 bits have 65,536
(216) intensity graduations, with 0 still representing black but 65,536 representing white.
Thresholding works by separating pixels which fall within a desired range of intensity values from
those which do not, (also known as ‘segmentation’). In the ideal world the pixel intensity values of
the feature(s) you’re interested in will be a unique subset of the image, and thresholding them from
the other features in the image is easy. In practice it seldom is.
Thresholding can be a very effective method of measuring complex or disjointed features in an
image. A few keystrokes can measure what would have been a near-impossible task by hand. But
there is a catch - potentially quite a big one. When we look at an image we are extremely good at
separating out what we are interested in, irrespective of how complex the image may be. Image
analysis software has no intuitive moves up its sleeve, it will take all the information in an image
and treat it literally. If a stray pixel falls within a threshold range it will be counted, even if to us it’s
obviously a false positive.
Example
In this example an image has been chosen that ‘behaves’ quite well, in many cases additional steps
will be required to produce reliable thresholding results.
Activate thresholding (Image / Adjust /Threshold ... ) and the window above appears. The
histogram represents the distribution of pixel intensities in the image, black towards the left (0) and
white at the right (255). Dragging the sliders selects different regions within the greyscale, pixels in
the image which fall in the selected range are highlighted red. In this case, all the pixels between 76
(dark grey) and 182 (mid grey). Discontinuous regions can be selected and those with complex
boundaries delineated from the background.
The selected area can be measured using the usual measurement option (Analyze / Measure menu
option or apple / m keystrokes) but first measurements must be limited to only the selected pixels.
Open the Analyze / Set Measurements menu option and tick the limit to Threshold box. Press Ok to
save the changes and then make your measurement.
-8© Andrew McNaughton 2010, Microscopy Otago
Without the Limit to Threshold option
ticked the entire area of the image (44,697
mm2) is measured.
Ticking the
Limit to Threshold option
measures only the selected thresholded
areas of of the image (1,595 mm2).s
The Results window will appear and can be saved as usual for future use.
Combining Thresholding and ROI
Although thresholding is very useful, as can be seen in this example if you were trying to measure
the area shown in the previous example you would be including areas outside your ROI.
Eliminating these from measurements can be relatively easy, or next to impossible. With image
analysis there is a great diversity in images and what people want from them, consequently there are
often no standard answers. Imagination and
perseverance are often your best options.
To remove the unwanted areas at the bottom
of the image, use a selection tool to mark
your ROI, either before or after you
threshold it.
Measurements will now be limited to pixels
which (1) fall within the selected area and
(2) are within the selected threshold
intensity range.
Both selection area and threshold ranges can
be adjusted ‘live’ and new measurements
taken by pressing apple / m. Each new
measurement is added to the Results
window.
-9-
Combining Thresholding and Multiple ROIs
If you want to measure multiple ROIs then ImageJ’s ROI Manager is a powerful tool for marking
and measuring many separate regions within an image, or a stack of images.
Activate the ROI Manager via the Analyse / Tools / ROI Manager ... menu path.
Select a ROI as described earlier
and add it to the ROI Manager’s
window by pressing ‘t’, or
clicking the button in the
window.
Select a new ROI (with any of
the selection tools) and add them
to the ROI Manager’s window.
Repeat as required.
Using the ROI Manager’s option
buttons ROIs can be renamed,
their boundary colours changed
and so forth.
A ROI will only be visible in the
image if it is highlighted on the
ROI Manager’s list.
If multiple ROIs are marked on the
list (Command key to make noncontiguous selections and Shift for
contiguous selections.)
As with any part of ImageJ, (or other image analysis programs), explore the menus of the ROI
Manager as it contains other features which may be of use to you. One of the reasons instructional
documentation can be sparse is it’s impossible to anticipate all the possible uses it will be put to.
Your images and information you wish to extract from them may be unique so a customised method
may be the only option.
- 10 -
Particle Analysis
Another common task in image analysis is determining ‘how many?’. Counting discrete particle
requires a different approach to measuring continuous variables such as length or area. Unless your
raw image shows your particle conveniently contrasted against the background, it’s likely the first
step will be to separate them using thresholding. This time the image will be turned into a binary
image, that is, consisting of only black pixels or white pixels, with no intermediate grey values.
Counting Binary Particles
In most image analysis there is a point where a subjective assessment has to be made to decide
‘what’s in and what’s not’. This is when a good knowledge of the sample and its preparation
technique are vital to ensure sensible decisions are made. Quantitative measurements will give you
lots of numbers - make sure they are worth having.
The first step is to threshold the image to eliminate regions of no interest. In this example the image
is converted into a binary image (Process/Binary/Make Binary) so pixels are only black or white.
Zooming in on the image shows it is not a clear case of
discrete particles, but a snow storm of pixels which may,
or may not, be part of a discrete particle. In this case
more subjective decisions need to be made before
measurements can be taken.
Filters can often modify the image to a form better suited
to measurement, the process is however something of a
‘dark art’ and there may be no ‘right’ answer. Above all,
be cautious and always compare your modified image to
the original as it’s easy to stray far from reality.
In this case a mean filter was applied (Process / Filter /
Mean ... ) with a two pixel radius. This means each pixel
is replaced with the mean value of the pixels within two
pixels of it. The result is black spaces between isolated
white pixels are filled with a shade of grey. The result is
an amalgamation of discrete pixels into larger groups of
particles.
- 11 -
The grey pixels are then replaced with white pixels by
making the image binary once more.
The image is now in a form which allows the number of
particles to be counted.
The next step is to analyse the particles, that is, count
them.
Analyse Particles
A convenient tool for this is under the Analyse / Analyse Particles menu option. This also provides
another point to decide what will be analysed and what will not be, this time based upon size. More
often than not there are a few, or many, small areas in
the image, usually the result of noise in the original
image.
As noise is usually small in area relative to the ROIs
it can be removed by choosing an appropriate size
filter. In this example only particles consisting of 50
pixels or more will be counted.
The Show / Masks option generates an additional
window showing which particles were counted. A
results window lists all the particles and indicates
their size. It’s worth having a visual record of which
parts of your image were counted otherwise you’re
generating a list of numbers and assuming they
correspond to what you think you’re measuring.
- 12 -
Confocal Files
Introduction
Confocal microscope images generally consist of up to three colour channels, red, green and blue
(RGB). Each channel is independent of the others and in some cases one or two channels may not
contain any information, ie. in a single label sample two channels will be entirely black. The
channels are independent because they are collected by separate detectors, so splitting a RGB image
into its components shows the information collected by the detector 1, 2 and 3.
The other important point about confocal images is they are greyscale images that have had a false
colour applied to them. Once the fluorescent signal is detected by the photo-multiplier and becomes
as stream of electrons all colour information is lost. The appropriate colours are added by the
software based upon the user’s choice(s). Treating confocal files as greyscale images does not
therefore alter the underlying information they contain with regards to intensity or distribution.
Many of the ImageJ plugins relating to quantifying co-localisation require split colour channels
(multiple greyscale images) for them to work.
The other common task when working with confocal files is viewing and manipulating z-series,
many of these options are built in to ImageJ’s standard menus.
ImageJ can read the metadata (embedded information) contained in confocal files and can therefore
calibrate them automatically, making area / length measurements correct as soon as they are
opened.
Viewing Z-Stacks
By default a z-series will appear in a single window with a scroll bar at the bottom to move through
the stack.
To view the stack in other ways it is necessary to use the
stacks menu and choose one of the options.
- 13 -
Z-Stack Menu Options
Add a slice at the current location in the stack.
Delete a slice at the current location in the stack.
Move forward through stack.
Move backwards through stack.
Move to a specified slice in the stack.
Merges a series of images into a coherent z-stack. Images
need to be numbered sequentially.
Splits a z-stack into a series of images.
Tiles the images in a z-stack into a montage.
Re-slices the z-stack at user defined plane (see later).
Orthogonal produces additional XZ and YZ slice planes.
Z-project flattens z-stack to a single extended focus image.
3-D Project keeps the z axis spacing to give a 3D image.
Plots ROI intensity statistics for each slice in the z-stack.
Label slices with user defined information.
Reslice [/]... Option
Most of the menu options are self explanatory but the Reslice [/]... option is worth describing in
more detail. It allows a user-defined line to be drawn anywhere in the image to re-slice the 3D
volume along that line. The width of the slice can also be defined. The effect is to cut out a block of
data and examine it as a new data set.
Slice count defines how wide the
block will be and can be changed
within this menu if not suitable
before proceeding.
First define a line where the data
set will be re-sliced
- 14 -
The new data set appears as a slice in
through the z axis following the userdefined line. In this case the cells can be
seen as they lie along the microscope slide,
viewed edge on.
The selected block is now shown on
the main image.
Moving the scroll bar moves horizontally
through the data set, this is the distance
defined in the dialogue box.
More Z-Stack Menu Options
Additional options are available from the ‘Tools’ menu item at the
bottom of the z-stack menu. Detailed explanations of these (and
all the other menu items) are available on the ImageJ website at:
http://rsbweb.nih.gov/ij/docs/index.html
Additional plugins can also be downloaded from the ImageJ
website which relate to z-stacks, check the plugins’ page
regularly as it is updated frequently.
http://rsbweb.nih.gov/ij/plugins/index.html
- 15 -
Adjusting Background and Merging Files
Adjusting Background Example
Red, green, blue (RGB) cameras produce a single colour image, in microscopy often only one of
those colours is required, but the image still contains three channels. When measuring background
levels in such images it’s worth checking to see the effect of the other ‘invisible’ channels before
adjusting for background.
Blue image no adjustments
Red image no adjustments
The above images have high backgrounds which can be removed
if it is first measured using the line tool and the profile tool
(Analyse / Plot Profile or apple/k). Draw a line over just a region
of background in the blue image, include a region of signal if you
want to measure the difference between them, and press apple/k.
The plot window shows there is a low intensity in the background
regions of the blue image, clicking the ‘list’ button shows the pixel
values along the line drawn in the image.
In this case the values suggest the background intensity is about
12, so changing the greyscale range of the image from 0-255 to 12
to 255 should make the background regions black. Splitting the
channels to show separate red, green and blue channels produces a
different result due to the way the plot profile reads the pixel values in an image.
Splitting the channels produces two black images for the green and red channels and the following
image for the blue. Note the profile plot for the background is
now noticeably higher than it
was for the RGB image, the
average pixel intensity being
around 36. This is because the
plot profile tool is measuring
the average pixel value across
all three channels. The blue
channel’s intensity is 36, but
averaged with red (0) and green
Blue channel uncorrected
(0) produces a value of 12.
background
- 16 -
To remove the background use the Image / Adjust / Brightness/Contrast (apple/shift/c) and change
the lower greyscale threshold to 36. This means any pixel with a value less than or equal to 36
appears as black. The adjusted images below show the results for the blue and red channels.
The red channel was adjusted for
a background of 136 and retains
only the brightest pixels, which are
likely to correspond to the
fluorescent marker.
Blue channel corrected
background
Red channel corrected
background
Merging Adjusted Images
If the original images are merged
(using Process / Image Calculator ...
and use ADD ) the result is as shown
on the left. The background
overwhelms the signal and little
useful information can be
determined.
Blue/red channels merged,
uncorrected background
By merging the two greyscale images, corrected for background, a much clearer indication of the
distribution of the fluorescent markers can be seen.
To merge the images use Image / Colour / Merge Channels and select the appropriate images for the
red, green and blue channels as shown in the dialogue box below.
The background adjusted and
merged image is shown on the left,
considerably different from the
original above. The image is
probably more indicative of where
the fluorescent markers occur in the
sample.
Blue/red channels merged,
corrected background
- 17 -
Co-localisation
Determining the degree, or location, of co-localisation is usually done by determining if pixel
intensities from one colour channel coincide with similar intensities in other channels in the same
xy position. Co-localisation plugins are needed as they are not part of the standard ImageJ tool
collection.
Use the plugin called Confocal Anaylsis
(Plugins’ menu) and choose the Colocalisation Finder option from the list in
the sub-menu. Assign the channels in the
dialogue box as shown.
To eliminate meaningless co-localisation of
say background signals a threshold can be
set to only count pixels within a defined
range.
The plugin produces a red/green image
with white highlights indicating where co-localised pixels occur.
A scattergram shows the co-localisation within the image and a re-sizeable green box can be
dragged to different areas of the scattergram to show where co-localisation occurs at different
intensity ranges. Live changes occur in both the image and the results window.
A results window shows the intensity range measured and the percentage of co-localisation.
Other options in this plugin’s sub-menu can produce correlation coefficients and other useful
information. To see what is, and what is not, useful to you try out the other options and see the
result. This is true of all the plugins and functions in ImageJ.
- 18 -
Stereology
Stereology methods can be used if the appropriate plugins are downloaded from the ImageJ
website. A good starting point is the Plugins/Analyse/Grid option which overlays points, crosses,
lines or a grid over the image. The starting point of the overlay can be randomised and the size of
the point spacing varied.
Counting of intersect points
can then be done as normal
for stereology or another
part of the plugin collection
can be used to easily keep
track of tallies.
Another stereology-like plugin in found in the same menu as the
previous one, Plugins/Analyse/Cell Counter. It allows the user to
assign numbers to different categories of cell, tissue, etc. and then
tally them automatically each time they click on the appropriate
feature. Up to eight different categories can be used.
If cycloids are your thing then Plugins/Grid Cycloid Arcs might be of use. It draws cycloids over
the image and can be customised to suit the user’s needs.
Many traditional stereology methods used to quantify images can probably be achieved using
ImageJ’s functions, in some cases by-passing stereology altogether.
- 19 -
Extending ImageJ’s Features With Plugins
The core program of ImageJ downloaded from its website can handle most of the tasks asked of it.
If you need to expand its capabilities this can be done using plugins. They can be accessed by
downloading and installing them from the comprehensive (and ever increasing) list found at http://
rsbweb.nih.gov/ij/plugins/index.html. Generally plugins are easy to install and are produced by
users who wish to expand ImageJ’s capabilities. Some are great, some are mediocre so be prepared
to win some and lose some (but they are all generally free).
Scan the long list of plugins before downloading anything and read the descriptions associated with
their links, don’t grab the first one that looks promising as there may well be a better one further
down the page. It’s also worth digging through the collections near the bottom of the page. They
tend to be collated by individuals with a particular interest, so if that coincides with your interest
then you may be in luck.
Installing a Plugin
The download link for a plugin is found on the page describing its features, that is, the page the link
takes you to from the plugins list page.
Once downloaded, move the plugin to the Plugins folder which is found inside the ImageJ folder in
the Applications folder. Sometimes the plugin’s instructions will tell you to place it in another folder
within ImageJ, it’s generally best to do as suggested.
To activate a new plugin it’s necessary to restart ImageJ. The
new plugin will then appear in the Plugins menu, or within a
subfolder within it, depending upon where it was placed.
Alternatively, use the Help/Refresh Menus option to update
changes and have them appear in ImageJ’s menus without
restarting.
To remove a plugin simply delete it from the Plugins folder
and restart ImageJ.
Plugin files will have an extension of .class or .jar, these should
run as they are. Sometimes a plugin will be have a .java
extension and will need to be compiled first.
This is done by using the Plugins / Compile and Run ... menu
option. After re-starting ImageJ the plugin is then available for use.
- 20 -
Summary
With any quantification software do not fall into the trap of thinking ‘there’s now has a number to
describe it, therefore it’s real’. Measuring nonsense is easy, numbers are only as good as the
underlying information in an image.
ImageJ’s main strength is its user base and the constant introduction of new plugins and marcos,
keep checking the Plugins’ page on the website, often somebody has already done the work for you.
Always keep a copy of your starting image in its original state, work on duplicates so you can
always refer back to the raw data. It’s very easy to drift a long way from where you started,
particularly if the numbers are looking good.
Finally, always back up your data! Keep the backup, ideally more than one, in different locations. If
they are all in the same location the fire/flood/earthquake/thief will take them all. Do not rely
entirely upon other people or automatic backup systems to care for your data as much as you do they don’t. If your research is going to change the world then you put the effort into ensuring it
survives long enough to make the world a better place.
Useful Links
Download ImageJ
http://rsbweb.nih.gov/ij/
Plugins
http://rsbweb.nih.gov/ij/plugins/index.html
Macros
http://rsbweb.nih.gov/ij/developer/index.html
Documentation
http://rsbweb.nih.gov/ij/docs/index.html
Other useful links
http://rsbweb.nih.gov/ij/links.html
- 21 -

ImageJ Introduction OCCM

Transcription

Similar documents

July, 1993 - Stumpf Welding Supplies

cromwell - Central Otago District Council

NZASI Newsletter JULY 2016 - Australasian Society for Immunology

OPTICAL MICROSCOPY UEF graduate school 2012

Newsletter 21 06 2013

Setting an Android Device to AutoFill Password Fields in

brightsign xd smart menu board

Alexandra - CENTRAL OTAGO PINOT NOIR LTD