Colocalization Analysis Widget Help
This widget aids in the visualization of colocalized voxels (i.e., those
voxels with "roughly equal" amounts of label in the images being displayed,
either 2 or 3 images). It does this by first dividing the multiply-labelled
data set into mutually exclusive sets of voxels: those that are colocalized,
those that are not colocalized and have more of image1 (the "green" image) in
them than image2 or image3, and similarly for image2 (the "red" image) and
image3 (if present). It then displays (or doesn't display) each of these
subsets in one of several ways. By default it only displays the
colocalized subset, and they are displayed in a "grayscale".
Thus voxels with low colocalization can be removed from view. This is obviously
only used when two or more images have been read in by DAVE (e.g., dave -z 2 -I
image1.i2i -I image2.i2i).
The colocalization sliders work best if you are displaying volume data
in "cubes" mode ("planes" is the default). At times "planes" mode produces
streaks or other artifacts. Go to the menu in the Edit Volume widget, pick
Render Cubes from its menu.
[Actually, this difficulty with "planes" mode may have been removed.]
After performing colocalization, the Count Widget (in the Goodies menu
of the DAVE Master Widget) can be used to get statistics on the number of
colocalized regions (the Colocalization Widget only gives voxel based
statistics). The Count Widget can also be used to eliminate some unwanted
colocalized regions from consideration. See it for more details.
NOTE: This widget is not "active" until "Show Colocalized Voxels" is pressed.
When active this button is highlighted in yellow. Don't forget to
deactivate the widget before you remove (HIDE) it (assuming you don't
want it to affect what you see anymore).
Definition of Which Pixel Intensities are Compared
Many operations get applied to your initial data during the course of
displaying it. First each image gets rescaled into one byte (0-255). These
scaling factors (scale and black level) are usually different for each image
(you can control these scale factors via the -S command line option). Then
the RS slider in the Edit Volume widget gets applied (by default it does
nothing). After these the Opacity and Brightness sliders rescale intensity
values. Finally, at each voxel these resulting values for the two or three
images are combined to produce a color at that location.
For colocalization analysis, all intensities are compared after the
initial rescaling performed when reading in the data (i.e., rescaling each
image 0-255). The comparisons also occur after the RS slider on the Edit
Volume Data widget is applied to the images (by default this slider changes
nothing).
If "Apply Before O+B Sliders" is set in the menu, then intensities are
compared prior to doing the opacity and brightness mappings. In other words,
what you see is not necessarily what you are analyzing.
The philosophy behind this is that those sliders are being used merely to
enhance the visualization of the data and should not affect the colocalization
statistics.
If it is desired to have the opacity and brightness sliders affect
the colocalization analysis (e.g., the sliders are being used to threshold
the image) then the "Apply After O+B Sliders" menu option should be chosen.
Intensities are compared, for display purposes only, at the current
resolution level used for displaying the image. Since lower resolution levels
are created by taking the maximum (by default) of neighboring voxels, voxels
may appear more colocalized at lower levels of resolution than they will at
higher levels. Resolution level will NOT affect the colocalization statistics
(see below).
Volume Subset
Only voxels which are not clipped out by the volume clipping widget
will contribute to colocalization statistics.
REL Slider
This slider specifies the maximum relative difference in intensity which
the same voxel in pairs of images can have and still be considered colocalized.
Relative intensity is calculated as:
absolute_value_of(image1-image2)/(min_of(image1,image2)).
So it is the difference in intensities over the minimum intensity.
This method of calculation will yield the same number if image1 and image2 are
exchanged. If either intensity is zero the voxel will not be considered
colocalized. If a voxel is totally transparent it will not be considered
colocalized.
ABS Slider
This slider specifies the maximum (absolute value) difference in intensity
two voxels can have and still be considered colocalized. Since the default is
255 (and all data values have been rescaled 0-255), no voxels are eliminated
from the display based upon the default setting, except if either intensity is
zero the voxel will not be considered colocalized. If a voxel is totally
transparent it will not be considered colocalized.
AND/OR Buttons
If OR is chosen (the default) a pair of voxels will be displayed if the
difference in their intensities meets the criterion specified by the REL slider
(i.e., relative difference less than the specified amount) OR the criterion
specified by the ABS slider (i.e., absolute difference less than the specified
amount). If AND is chosen a voxel pair will be displayed only if both
criteria are met.
If three images are being displayed, a voxel will only be considered
colocalized if all 3 pairs of images (1 vs. 2, 2 vs. 3, and 1 vs. 3)
qualify as colocalized.
Typical Usage
Typically you might only want to display voxels which are within 2535520453640f each
other, so REL would be set to .25. But for dim voxels you might want more
leeway (since a difference of, say, 5 when the voxels are only 10 is already a
50-314419296ifference), thus you would set ABS to perhaps 5. If you don't care about
dim voxels (perhaps you are not seeing them anyway because you have set the
opacity slider in the Edit Volume Widget so they are transparent) you can set
the ABS slider to 0.
Alternatively, you may not really want to worry about the relative amounts
of the two labels since their intensities depended upon the amount of
fluorescent label loaded into the cell, exposure time, etc. All you really
want to do is threshold the images, and any voxel with both image values above
threshold should count as a colocalized voxel. In this case, pick the menu
option "Apply After O+B Sliders". Now use the brightness slider (on the Edit
Volume widget) to threshold the image (you may want to pick "Map - Threshold"
from the Edit Volume widget's menu). Set the ABS slider on the colocalization
widget to 255 and leave the OR button active. Thus any voxel with the two
intensities within 255 of each other (which is all of them since the rescaled
numbers only range from 0 to 255) will be colocalized. The definition of
"colocalized" excludes any voxel from being colocalized if either of its two
values (or its opacity) is zero. Thus all voxels with two nonzero values
after the threshold will be counted as colocalized.
A note of caution: visually unnoticed voxels still may count
Visually you may not notice voxels which are almost totally transparent,
yet they count towards colocalization as long as they are not totally
transparent. You can make these voxels more visually apparent (if you desire)
by choosing "Opaque Data" from the Edit Volume Options Widget's menu (this
widget is gotten from the Edit Volume Widget's menu). Thus all
voxels will either be totally transparent or totally opaque. Similarly, if you
set the right brightness slider to the same level as the left brightness slider
all voxels will be either totally black or totally white. Rather than do this
you might also just change the Mapping function from Rescale to Threshold, this
will permit "dim" objects to be somewhat more visible. You might also change
the background color (Edit Bkgd Color from the DAVE Master Widget) to something
other than black. The HELP button in the
Edit Volume widget explains the function of its sliders and menu options in
more detail.
STATISTICS SHOWN ON THE WIDGET
The window shown to the right of the sliders is used to display
colocalization statistics. The top line is the number of voxels colocalized.
Line 4 is the fraction of colocalized voxels in the region being displayed
(i.e., "total" is the number of voxels in the image at the current level of
resolution unless the Volume Clipping widget is being used to display a subset
of the image). Line 2 is the number of colocalized voxels divided by the number
of nonzero (and nontransparent) voxels that are in the green image (image 1) in
the region being displayed. Line 3 is the number of colocalized voxels divided
by the number of nonzero (and nontransparent) voxels that are in the red image
(image 2) in the region being displayed.
Line 4 is the number of colocalized voxels divided
by the number of nonzero (and nontransparent) voxels that are in the blue image
(image 3) in the region being displayed.
So lines 2 and 3 are the colocalization percentage by volume.
The statistics are not calculated (or recalculated when the sliders
change) until the "Recalc Co-Localization" button is pressed. The statistics
are all based upon colocalization at the initial (highest) image resolution
(even if you are viewing the data at lower resolution). Thus the statistics
may disagree slightly from the rendered view.
Statistics will be out of date if the left brightness slider, the
volume sliders, or the colocalization sliders are moved. When the
statistics are out of date, the "Recalc Co-Localization" button becomes
highlighted in blue and an asterisk is added to the label to remind
the user that the button needs to be pressed to update the displayed
statistics.
CHOOSING THE DISPLAY MODE AND COLOR FOR EACH SUBSET
The popup menu/toggle buttons to the far right of the sliders
determine how each of the subsets - colocal, image1, image2, and image3
(see explanation at top) - are displayed. You can change this by
toggling through the options (by clicking with the left mouse button) or by
picking the desired option from the pop-up menu gotten by holding down the
right mouse button. The options are as follows.
Note: If two or three images are on (i.e, not set "off" in this widget)
then those components of a non-colocalized voxel will be displayed
in their normal colors (e.g., RED/Bi.Op.Green will be displayed as
RED/GREEN, but Bi.Op.Red/Off will be displayed as Bi.Op.Red).
This is because it is unclear how to otherwise combine multiple colors.
Note:Opacity is the maximum of the opacities of the displayed images, so
turning a subset off may change the opacity of the remaining colors.
For the colocalized voxels
Default:the voxels will tend to look grayish. The same algorithm is used to
color them as normally when both data sets are displayed. To wit:
Image1 is displayed in green, image2 in red. Voxels which have similar
intensities in the two images will normally appear gray (dark gray if
the intensities are low and brighter white if the intensities are
higher). The gray will have a reddish tint if image2 is brighter than
image1; similarly, a greenish tint implies that the two voxels are
similar in intensity but image 1 is slightly brighter.
Red: The red value is displayed.
Blue and green are set to zero. Opacity is based
upon the red level (brighter => more opaque).
Green: As for "Red" but the green channel is set.
Blue: As for "Red" but the blue channel is set.
Bin Red:Binary red. The red component is set to maximum brightness regardless
of the voxel value. Green and blue are zero. Opacity is based upon
voxel values as in the "Red" option above.
Bin Green:As for "Bin Red" above, but the green channel is set instead.
Bin Blue:As for "Bin Red" above, but the blue channel is set instead.
This seems to be buggy, looking the same as just "Blue".
Bin White:The voxel is set to maximum brightness (full white).
Bi.Op.Red:Binary Opaque Red. Same as Bin Red, but the voxels are also set to
be totally opaque. This is the option that will tend to make them
maximally visible.
Bi.Op.Green: As for "Bi.Op.Red" above, but the green channel is set instead.
Bi.Op.Blue: As for "Bi.Op.Red" above, but the blue channel is set instead.
This seems to be buggy, looking the same as just "Blue".
Bi.Op.Wht:The voxels are set to full white and are set totally opaque.
For the non-colocalized voxels in image 1:
Green:The voxel will be green. The larger the data value the brighter the
green (and the more opaque). This is similary to the way image1 data
is normally displayed (although it is not combined with image2 data).
Bin.Green: Binary Green. All nonzero (image1) voxel values (which aren't
colocalized) are shown full green; but opacity is still varied depending
upon the voxel value.
Bi.Op.Green: the same as Binary Green but all nonzero voxels are set to be
totally opaque.
Grey: The red, green, and blue channels are all set to the intensity of the
image1 ("green") voxel. Thus a gray voxel is produced. Opaciy is based
upon the voxel intensity.
OFF: Don't display the image1 data set.
For the non-colocalized voxels in image 2 or image 3:
The same as the items described above for image 1, with the obvious changes.
NOTE: Image1 and Image2 voxels (or any other pair of images) cannot be
simultaneously displayed other than at their default color values.
NOTE: At least 2 images must be set on in the Edit Volume widget
for colocalization to work. Otherwise only one data set is examined and
no colocalization is observed. These data sets are turned on by pressing
their buttons in the Edit Volume widget (ON is the default).
If any 2 images are on, colocalization will be done between them.
If 3 images are on, colocalization comparisons are done on
all 3 pairs of images and only if all 3 pairs colocalize is a voxel colocalized.
NOTE: All operations controlling display mode RENDERING are applied to the voxel
data AFTER it is rescaled by the Opacity and Brightness slider in the
Edit Volume widget (unlike dividing the data INTO the subsets, which is
applied before or after the sliders depending upon whether "Apply After
O+B Sliders" or "Apply Before O+B Sliders" is chosen from the menu).
Copyright 1995 by Lawrence M. Lifshitz and the University of
Massachusetts Medical School. All rights reserved.