University of Utah              The JOHN A. MORAN EYE CENTER

HOME TOOLS SCIENCE PUBLICATIONS CYTOSOMICS & METABOLOMICS


Navigation: Top     Overview     Basic Functions    Basic Operations    Specific Processes    Legal   

CellKit Downloads & Links
Cell Kit v2.2 for Win2K/XP IDL *.sav file for IDL 6.0 or greater VM (right-click: save-as)
Cell Kit v2.3 for OS X (X11) IDL *.sav file for IDL 6.1 or greater VM (right-click: save-as)
IDL 6.0 VM from RSI
CellKit Description and Full Help Files
CellKit (CK) is an application suite to explore classified or segmented images and visualize multivariate signatures of pre-aligned images such as multichannel datasets. CK was planned as a light-weight laboratory tool for our own biological research but is based on capabilities found in more powerful (and expensive) remote sensing application suites. The basic goals are to provide quick histogramming and classification tools for multichannel datasets. CK is early in its development (the interface and functions will undoubtedly change), but may be useful to those beginning their explorations of multichannel worlds.
CK was developed under IDL Versions 6.0 Win32 (x86) and 6.1, Mac OS X (X11 darwin ppc m32). This version is pre-compiled ' *.sav ' form that runs in the IDL Virtual Machine™ environment (© 2003 Research Systems, Inc). The IDL VM is a free application/environment for Windows, Unix and Mac OS X (X11) platforms. Link: IDL VM from RSI.

Overview

CellKit displays, filters, histograms, scatterplots, N-plots and classifies data arrays from registered tiff files.

CellKit has five major fields (view the 1024 pixel interface):

1. The image view window (IW, left) where scaled images are displayed

2. The image view controls (lower left and center):
    Controls to show R, G, B, RGB images for display
    Controls to show indexed theme maps, binary theme classes or object maps
    Controls to filter displayed images
    Controls buttons to select lut type
    Controls to show the last (previous) image
    Controls to save the image
    Droplists (right bottom row) to select red, green, blue, and theme channels

3. The data view window (DW, right) where multivariate data are visualized.
    A status line at lower left reports current selection options and run-time information

4. The data view controls (lower right):
    Controls to select visualization type (e.g. 1d hist) with options
    Options to select image sampling percentages
    Mask option to select classes (applies to all except 2N and parallel)
    Equal option to equalize 2d hist and 3d hist data
    Slicer options to apply a slicing lut to 2d hist and 3d hist data
    Filter options to smooth 1d hist, 2d hist, 3d hist data
    An editable multichannel selection droplist
    An editable multiclass droplist
    An editable channel pairs droplist for 2N plotting
    An editable class pairs droplist
    Controls to run plots, save ascii plot data and save plot visualizations

5. The menu
    Files - to access imagery
    BatchFilter - to perform banks of filtering on multiple image sets
    Image View - zooms images around cursor (not implemented)
    Image Morph - morphological operators on images (not implemented)
    Image Classify - kmeans, Isodata and watershed-based clustering (partially implemented)
    Statistics - classifications separabilities, statistics dump, spatial statistics (not implemented)
    Plot Prefs - preferences for data view controls (partially implemented)
    Help - help files

Navigation: Top     Overview     Basic Functions    Basic Operations    Specific Processes    Legal   

Basic Functions
Menu / Button / Droplist Function
File > Preview Image loads an image preview into the IW
File > Load Images loads tiffs into the red, green and blue channel droplists
File > Load Theme Map loads indexed tiff image into the class droplist
File > Load Object Mask loads binary object mask tiff
File > Record Data Links records all loaded image sets
File > Load Data Links loads all recorded image sets
File > Exit exits CK
Batch Filter > Filter Sequence sets a sequence of up to 8 filters to apply to a file batch
Batch Filter > FilterSeq>Red applies the filter sequence to the selected red [x] channel
Batch Filter > FilterSeq>All and Save applies the filter to all loaded images and saves with new filenames
Batch Filter > FilterSeq>All, Save, Load applies the filter to all loaded images, saves and reloads
Image Classify > Cluster object-based quick classifier
Image Classify > Set Cluster Parameters sets initial cluster and iteration numbers
Image Classify > Save Cluster Data saves ASCII cluster data
Image Classify > Save Theme Map saves classification theme map
Image Classify > Explore Theme Map explore classification theme map
Plot Prefs> 1DH background sets plot background to black or white
Plot Prefs> 1DH background sets plot background to black or white
Plot Prefs> 2DS background sets plot background to black or white
Plot Prefs> 3DS background sets plot background to black or white
Plot Prefs> 2N background sets plot background to black or white
Help > Help help
Plot Prefs> Help > About CellKit About CellKit
 
function & selection radio buttons Function
show + R,G,B radio loads grey images from the red, green or blue droplist
show + RGB radio loads rgb image from all three droplists
show + Theme Map radio loads an indexed image of the theme map
show + Theme Class radio loads an indexed image of the theme class
show + Object Mask radio loads an binary image of the object mask
filter + type & size radios hi, lo, median or edge filters (3x3, 5x5, 7x7) of the displayed grey image
lut + type radios stretch, equalization, inversion or slicer indexing of displayed grey image
save image saves greyscale IW image at true size
show last recovers last image (toggle)
percent radios percent of the image to sample
mask checkbox turn on class masking for all plots except 2N
equal checkbox turn on histogram equalization for 2d and 3d hist
slicer checkbox turn on slicer lut for 2d hist
filter radios select 'no', 'box'car or 'lee' (gaussian) filterting for 1d, 2d, 3d hist
run plot runs data visualization with selected options
save plot save ascii data
save image save plot image as true color tiffy
 
plot type radios Function
1d hist 1d hist of the red channel +/- theme class mask
1d chan 1d hist from 'multichannel list' channels +/- theme class mask
1d class 1d hist from 'multiclass list' & red channel droplist
2d hist 2d hist from x & y channels +/- theme class mask
2d scatter 2d scatterplot from x & y channels +/- theme class mask
3d hist 3dh of x, y, z channels +/- theme class mask (not implemented)
3d scatter 3d scatterplot from x, y & z channels +/- theme class mask
2N plot multi 2d scatterplots from 'channel pairs' +/- theme class mask
2N op1 multi 2d means + orthog 2sd channel pair ellipses +/-theme class mask
2N op2 multi 2d means + orthog 2sd channel pair ellipses +/-theme class mask
parallel parallel plots from 'multichannel' and 'multiclass' list (not implemented)
 
droplists Function
red channel [x] droplist the image used for the x dimension in 1d, 2d and 3d plots
green channel [y] droplist the image used for the y dimension in 2d and 3d plots
blue channel [z] droplist the image used for the z dimension in 3d plots
theme class droplist the class bitmask from the theme map
multichannel droplist the list of channels for 1d chan plots
channel pairs droplist the list of classes for 1d class plots
class pairs droplist a list of pairwise classes (not implemented)
 
Navigation: Top     Overview     Basic Functions    Basic Operations    Specific Processes    Legal   

Basic Operations
CK initial assumptions
      all data tiffs are are monochrome (1 channel)
      all tiffs are the same size
      all tiffs are registered
      theme map tiffs are properly indexed: 1 pixel value/class, starting at pixel value 0
     
Loading and Display
      1. Only monochrome, single channel tiffs are accepted as input
      2. Up to 96 tiffs may be selected, but only the first 16 are currently used
      3. The display is rescaled but true image sizes are used in all operations
     
Shortcuts to loading batches of files: using DataLinks
      To record a batch of files for future use...
      1. Load any combination of images, theme maps and object masks
      2. Select File > Record DataLinks and save as a *.dlk file
      To load a batch of recorded files ...
      1. Select File > Load DataLinks and select the desired *.dlk file
      As long as you do not move the files or rename directories, the DataLink will load them - even if you move the *.dlk.
      If some of the files have been moved, deleted, etc., DataLink will try to load the remainders
     
Showing loaded images
      1. Select the desired file in the red [x], green [y], blue [z] or theme class droplists
      2. Select the desired radio (R,G,B, RGB, Theme Map, Theme Class or Object Mask)
      3. Select 'show'
      NOTE: R,G,B are displayed as monochrome, RGB as rgb, all others as indexed
     
Filtering displayed images
      1. Select & show the desired monochrome (R,G,B) images
      2. Select the desired filter and size radios
      3. Select 'filter'
      4. Select save image to save the displayed image as a new file at the original resolution
      NOTE: 'filter' operates on the displayed true-size image in memory, not disk files shown in the channel droplists
      NOTE: Filtering an image and re-selecting 'show' for that channel will undo all filters
      NOTE: Selecting 'show last' will undo the last filter.
     
Filtering batches of images:  will perform up to 8 serial filters on up to 16 files
      ...set up the filter sequence
      1. Select Batch Filter > Filter Sequence: select a string of 1-8 sequential filters to apply
      2. Select 'done'
      ...to test the filter sequence on an image
      1. Select the desired file in the red channel [x] droplist
      2. Select Batch Filter > Filter Seq>Red: filters and displays the result; save with 'save image' if desired
      ...to filter all loaded images
      1. Select Batch Filter > Filter Seq>All & Save: filters, appends '_batch.tif' to the image name and saves
      WARNING: overwrites files already named ' *_batch.tif'. Use Filter Seq>All,Save,Load to avoid this.
      ...to filter all loaded images and reload them for analysis
      1. Select Batch Filter > Filter Seq>All, Save, Load: filters, appends '_batch.tif', saves and reloads into droplists
      WARNING: click away willy-nilly and fill up a drive with ' *_batch_batch_batch.tif ' files!
     
Viewing the display lut
      1. Select & show the desired monochrome (R,G,B) images
      2. Select the desired lut
      3. Select 'lut'
      4. Select save image to save the displayed image as a new file at the original resolution
      NOTE: 'lut' operates on the displayed true-size image in memory, not disk files shown in the channel droplists
      NOTE: Remapping an image and re-selecting 'show' for that channel will undo all filters
      NOTE: Selecting 'show last' will undo the last lut.
     
GENERAL process to visualize signatures
      1. Select plot type radio
      2. Select the required droplist channels (red, green, blue) or multichannel etc
      3. Select the class droplist (optional)
      4. Select percent to sample the image (most operations require little data for accuracy, default =10%)
      5. Select mask invoke the class droplist (optional, default=OFF)
      6. Select equal and/or slicer to enhance 2d or 3d hist displays (optional, default=OFF)
      7. Select filtering radios to operate on 1d, 2d, or 3d hist displays (default=no)
      8. Select plot
      9. Select save image to save the displayed plot as an rgb tiff
      10. Select save plot to save plot data as an ascii file (not imlemented for all plot types)
     
Navigation: Top     Overview     Basic Functions    Basic Operations    Specific Processes    Legal   

SPECIFIC processes to visualize signatures
1d histogram signatures
      1. Select the '1d hist' radio
      2. Select a file from the red channel [x] droplist
      3. Select a class from the theme class droplist (optional)
      4. Select percent to sample the image
      5. Select mask invoke the class droplist (optional)
      6. Select filtering radios
      7. Select plot
      8. Select 'save image' to save the displayed plot as an rgb tiff
      9. Select 'save plot' to save plot data as an ascii file
     
multichannel 1d histogram signatures
      1. Select the '1d chan' radio
      2. Select a channel list from (or enter one into) the multichannel list
      3. Select a class from the theme class droplist (optional)
      4. Select percent to sample the image
      5. Select 'mask' invoke the class droplist (optional)
      6. Select filtering radios
      7. Select plot
      8. Select 'save image' to save the displayed plot as an rgb tiff
      9. Select 'save plot' to save plot data as an ascii file
1d chan notes: select from existing options or enter your own values: 0-9, 10=a 11=b 12=c 13=d 14=e 15=f. Acceptable channel numbers are shown next to each file name in any channel droplist. Enter values as a list without spaces, commas, etc.: e.g. '123b5c'. If plot does not run, check the status line for an "illegal entry" error message. This suggests a text error in entering your list.
     
multiclass 1d histogram signatures
      1. Select the '1d class' radio
      2. Select a class list from (or enter one into) the multiclass list
      3. Select a file from the red channel [x] droplist
      4. Select percent to sample the image
      5. Select filtering radios
      6. Select plot
      7. Select 'save image' to save the displayed plot as an rgb tiff
      8. Select 'save plot' to save plot data as an ascii file
1d class list notes: select from existing options or enter your own values: 1-9, 10=a 11=b 12=c 13=d 14=e 15=f 16=g. Enter values as a list without spaces, commas, etc.: e.g. '123b5c'. If plot does not run, check the status line for an "illegal entry" error message. This suggests a text error in entering your list.
     
2d histogram signatures
      1. Select the '2d hist' radio
      2. Select files from the red channel [x] and green channel [y] droplists
      3. Select a class from the theme class droplist (optional)
      4. Select percent to sample the image
      5. Select 'mask' invoke the class droplist (optional)
      6. Select 'equal' to equalize the histogram (optional)
      7. Select 'slicer' to lut the histogram (optional)
      8. Select filtering radios
      9. Select plot
      10. Select 'save image' to save the displayed plot as an rgb tiff
      11. Select 'save plot' to save plot data as an ascii file
2d hist notes: At low sampling percents, histograms visualized without equal or slicer may yield small noisy spots.
     
2d scatterplot signatures
      1. Select the '2d scatter' radio
      2. Select files from the red channel [x] and green channel [y] droplists
      3. Select a class from the theme class droplist (optional)
      4. Select percent to sample the image
      5. Select 'mask' invoke the class droplist (optional)
      6. Select plot
      7. Select 'save image' to save the displayed plot as an rgb tiff
      8. Select 'save plot' to save plot data as an ascii file
2d scatter notes: At high sampling % on large images, time to plot may be long.
     
3d histogram signatures  not implemented
     
3d scatterplot signatures
      1. Select the '3d scatter' radio
      2. Select files from the red channel [x], green channel [y] and blue channel [z] droplists
      3. Select a class from the theme class droplist (optional)
      4. Select percent to sample the image
      5. Select 'mask' invoke the class droplist (optional)
      6. Select plot
      7. Select 'save image' to save the displayed plot as an rgb tiff
      8. Select 'save plot' to save plot data as an ascii file
3d scatter notes: At high sampling % on large images, time to plot may be long. The dataset may be truncated and, if so, will be indicated on the status line
     
2N plot signatures as scatterplots
      1. Select the '2N plot' radio
      2. Select a list of channel pairs from (or enter one into) the channel pairs droplist
      3. Select a class from the theme class droplist (REQUIRED)
      4. Select percent to sample the image
      5. Select plot
      6. Select 'save image' to save the displayed plot as an rgb tiff
      9. Select 'save plot' to save plot data as an ascii file
2N plot notes: select from existing options or enter your own values: 1-9, 10=a 11=b 12=c 13=d 14=e 15=f 16=g. Enter values as a list without spaces, commas, etc.: e.g. '123b5c'. The list is interpreted as a sequence of 'x vs y' pairs. Channel numbers may be replicated as desired. If plot does not run, check the status line for an "illegal entry" error message. This suggests a text error in entering your list.
     
2N plot signatures as means with 2sd ellipses   - plotted as orthogonal sds
      1. Select the '2N op1' radio
      2. Select a list of channel pairs from (or enter one into) the channel pairs droplist
      3. Select a class from the theme class droplist (REQUIRED)
      4. Select percent to sample the image
      5. Select plot
      6. Select 'save image' to save the displayed plot as an rgb tiff
      9. Select 'save plot' to save plot data as an ascii file
2N op1 plot notes: select from existing options or enter your own values: 1-9, 10=a 11=b 12=c 13=d 14=e 15=f 16=g. Enter values as a list without spaces, commas, etc.: e.g. '123b5c'. The list is interpreted as a sequence of 'x vs y' pairs. Channel numbers may be replicated as desired. If plot does not run, check the status line for an "illegal entry" error message. This suggests a text error in entering your list.
     
2N plot signatures as means with 2sd ellipses   - plotted as tilted sds calculated from gaussian fits
      1. Select the '2N op2' radio
      2. Select a list of channel pairs from (or enter one into) the channel pairs droplist
      3. Select a class from the theme class droplist (REQUIRED)
      4. Select percent to sample the image
      5. Select plot
      6. Select 'save image' to save the displayed plot as an rgb tiff
      9. Select 'save plot' to save plot data as an ascii file
2N op1 plot notes: select from existing options or enter your own values: 1-9, 10=a 11=b 12=c 13=d 14=e 15=f 16=g. Enter values as a list without spaces, commas, etc.: e.g. '123b5c'. The list is interpreted as a sequence of 'x vs y' pairs. Channel numbers may be replicated as desired. If plot does not run, check the status line for an "illegal entry" error message. This suggests a text error in entering your list.
     
Parallel plot signatures   not implemented
     
Classifying Images  
      1. Load images
      2. Load an object mask
      3. Select File > Image Classify > Set Cluster Parameters to change if desired
      4. Select File > Image Classify > Cluster Image
      5. Select File > Image Classify > Save Cluster Data if desired
      6. Select File > Image Classify > Save Theme Map to save the displayed plot as an indexed tiff
      9. Select File > Image Classify > Explore Theme Map to browse the data
     
Navigation: Top     Overview     Basic Functions    Basic Operations    Specific Processes    Legal   

Legal
© 2003 Robert E Marc
This is a pre-release version for laboratory investigation only. It is not certified as a clinical, forensic, or policy tool. The code is not optimized for system performance. We aren't responsible for what you do with it or what happens when you use it. That may be obvious, but we have to say it these days. CK may be freely distributed as long as the original copyright is included. Bugs, requests for features, etc may be reported to robert.marc@hsc.utah.edu.
 
HOME TOOLS SCIENCE PUBLICATIONS CYTOSOMICS & METABOLOMICS

Moran Eye Center 75 N Medical Dr Salt Lake City, UT 84132
University of Utah Disclaimer   ·   MarcLab Website Terms of Use/Disclaimers/Contacts