Viewer Appearance

Back to Contents

First, make sure you have:

(as of June 2020)

Convert the slices to a common image format

I mostly assume you have a way of doing this already. However, there may be 2 steps which differ from what you are used to:
  • If you intend on using windowing, export the image with the spectrum: HU level: 0, width 2000.
  • Try to export the image into something lossless like PNG because you will have to tile them later.

If you don't have a pipeline for converting dicoms to png, here are 2 free ways:

Windows Tutorial Video:

Linux Tutorial Video:

  • Option 1: Using my scripts
    • First download them here
    • Install python3 (if you don't already have it)
      • (Linux) $ sudo apt-get install python3
      • (Windows) use the Microsoft Store.
    • Go to the folder with your dicom (.dcm) files
    • Extract the scripts here (so that the script files are directly in the folder with the dicoms)
    • Open a terminal here
    • Convert the dicoms to ascii files
      • If using python, you will need the pydicom module installed
        • maybe: $ python -m pip install pydicom
        • or just: $ pip install pydicom
      • Otherwise if using xmedcon, you'll need that
        • Linux only $ sudo apt-get install xmedcon
      • Run $ ./ (Linux with xmedcon) or python3
    • When it finishes, look at the final values:
      • Ensure the slope and intercept are consistent
      • Ensure that the slope and intercept are 1 and -1024 respectively
        • If not, edit the slope and intercept values in to match
      • Ensure the values are in HU, if not, these scripts won't work. You will have to use a proper dicom viewer (you can probably still use the sort script mentioned above though).
    • Convert the ascii files to png
      • Open and set the desired window high and low bounds (if not using windowing in the web ct viewer).
        • These are in HU, but 0 is air, so add 1000
        • The scripts come set to a window from -1000 to 1000
      • run $ python3
      • This will take a while
    • The pngs will now be in the png8 folder
  • Option 2: Using Microdicom
    • Make sure you are in Windows
    • Optional (but only if you like menial labor): Use to presort your dicoms.
    • Install Microdicom
    • Open Microdicom
    • Go to File > Scan for DICOM files ... (Ctrl + Shift + O)
    • Navigate to the folder with your dicom (.dcm) files
    • Apply your desired window (create a custom one with Center Level at 0, and Width at 2000 if using windowing in the web ct viewer)
    • Scroll through every slice (otherwise, windowing isn't applied to every slice when you export)
    • Go to File > Export > To a picture file
    • Select a folder for your destination
    • As mentioned above, select PNG
    • Make sure there isn't any meta data set to burn into your images.
    • If you didn't presort your dicoms, you may have to order them by hand now.

Tile your images

Again, the exact process will depend on what software you use. I am sure there are many options. I will cover 2 free ones here.

Windows Tutorial Video:

Be aware the the gimp solution in this video is very slow, if you have a large slice count (depends on your computer, but probably greater than 150), set this to run overnight.

Linux Tutorial Video:

  • First, know that the end products should be as square as possible eg:
    • 7 x 7 if there are between 43 - 49 slices
    • 7 x 8 if there are between 50 - 56 slices
    • 8 x 8 if there are between 57 - 64 slices
    • etc.
    • It doesn't have to be perfect, but load time is shortened if we can chunk the whole image into the GPU and square collages will help with that.
  • Software for combining images:
    • Option 1: Imagemagick
      1. Install Imagemagick (on Ubuntu, just $ sudo apt-get install imagemagick), other platforms might require you to build this from scratch.
      2. Go to the folder with your pngs
      3. Open a terminal there and run: $ montage -mode concatenate -tile 8x *.png out.png
        • The 8x means: Tile the new image 8 by whatever, so the second dimension will be whatever it needs to be to fit all of the slices. Change the 8 to be the floor of the square root of your image count.
          Input Total Slices: 8x
      4. Note that for even slightly thin slices you will have to increase the memory Imagemagick allows itself in policy.xml. On Ubuntu, this file is found at /etc/imagemagick-6/policy.xml (the -6 will change as this page gets dated)
    • Option 2: Gimp (SLOW if slice count is > 150, consider running the Fuse layer step, 16, overnight).
      1. Install Gimp (2.8 or above)
      2. Download this plugin for turning layers into a sprite sheet here
        • This plugin was created by Eduardo Hernàndez, I am only hosting it here because Gimp doesn't have the best repo for plugins.
      3. Find the folder for adding scripts to Gimp.
        • Open Gimp
        • Go to: Edit
          • Click Preferences
            • Expand the Folders section
              • Click Scripts (Not Plugins)
                • This shows you the directories where Gimp looks for scripts when it starts up, if there is more than 1, choose the one with your user name in it and go to that folder.
      4. Once you get to your scripts folder (it may be empty) close Gimp.
      5. Extract the plugin you downloaded in step 2 here.
      6. Reopen Gimp (Gimp only looks for new plugins when it starts up)
      7. Make sure the plugin is available, go to:
        • Filters
          • Combine
            • Make sure there is an option: Fuse Layers, but don't use it yet.
              If this option isn't here, make sure the plugin was extracted into teh correct folder,
              and make sure Gimp was restarted.
      8. Now go to File > Open as Layers (Ctrl + Alt + O)
      9. An open dialog will appear, Navigate to the folder with your pngs
      10. There are sort options at the top, click Name until your images are listed in descending order.
      11. Now import all the images
        • click an image once
        • press Ctrl + A to Select All
        • Click Open
      12. The images should no be listed in the layers outline in ascending order
      13. Tile them with your plugin by again going to:
        • Filters
          • Combine
            • Click: Fuse Layers
      14. A dialog will appear asking you How many Layers on the X axis?
      15. Set this to the floor of the square root of your slice count
        Input Total Slices: - Layers on the X axis
      16. Click OK, this will take a LONG time, some images have taken over an hour.
        • The feed back isn't helpful in gaging how close the image is to completion.
        • Gimp will appear to be not responding from time to time, don't panic, this is fine.
        • If you do feel something has gone wrong, you will have to force close Gimp to stop the script.
           But honestly, I have never had the script not finish an image (it just takes forever).
      17. A new image will appear with all the slices in one
  • Again, this might not be the final image, so lossless formats (like PNG) are preferred.

Final compression

Again, the process depends on your image editing program.
  • Load the complete collage into an editor (if you went the Gimp route above, you are already here)
  • You can usually save a little space by makeing sure the image mode is in grayscale.
      In Gimp, go to Image > Mode > Grayscale
  • Typically each slice in my images are 512 x 512 (so the full 8 x 8 example would be 4096 x 4096)
  • Scale the image to whatever you want, but make sure the final resolution is evenly divisible by the tile resolution
    • For example, 4096 / 8 = 512 exactly.
    • If a dimension wasn't evenly divisible (eg 4095 / 8 = 511.875) the viewer will round off a pixel and the image will migrate slightly as it is scrolled.
    • Before you scale the image, make sure empty portions of the image are a color similar to your slice's background
      • Imagemagick makes the empty parts white, when you scale the image the smoothing algorithm will blend this white into neighboring slices (creating a white border along the bottom of some slices).
  • Export the image. This is hopefully the final product and you will typically want to use lossy compression. (JPG).
    • Because these images are low on color you can typically get away with a lot of compression.
    • If you are using windowing with very narrow widths you will want less compression.

Add your image to the viewer

Windows Tutorial Video:

Linux Tutorial Video:

  • Let's replace the simple trial's image with yours
  • Go to the build01 folder where you have your index.html
  • Go to the assets folder and change 3 things:
    1. Edit assets.txt
      1. Find the size of your new image in bytes, let's pretend it was 1,234,756 bytes large.
        • Maybe right clicking will tell you, otherwise, search the Internet for how.
      2. Open assets.txt and find a line that looks like i:slides/file01.jpg:1489715:image/jpeg
      3. This line contains the file name, size, and mime type (separated by colons):
        • If your image is a .jpg, you only have to change its file size, so the new line is: i:slides/file01.jpg:1234567:image/jpeg
        • If your image is a .png, the new line is: i:slides/file01.png:1234567:image/png
    2. Edit slidedim.txt
      1. Open slidedim.txt
      2. This file tells the viewer how to parse your image, more at Viewer Configuration.
        1. The first number is the index for the set of slides (think: a specific trial), no other rows can have this same index. This stays 1
        2. The next number is the file number, we used file01 in assets.txt so we'll keep this 1. Trials can use the same file if desired, this decreases download and image processing time (currently the reward trial is at index 4, and uses the same image as trial 3. This gives us a separate place to use click states for the fireworks, while leaving the user's clicks untouched).
        3. The next number is the width, in slides, our image is, let's say this was 8, so change this number.
        4. The next number is the height, in slides, our image is, let's say this was 7, so change this number.
        5. The next number is the total slides, our image is, let's say this was 53, so change this number (Note this would mean we would have some empty spaces in our image since 7*8 is 56, this is fine though, make sure to enter the actual number of slides 53, and NOT total possible slices 56).
        6. The next item is a boolean value indicating whether you wish to record clicks on that trial, change this to t and we can now see clicks show up on our new image.
        7. Lastly, we have the file extension, if the image was jpg, leave it as .jpg, if it was png, change it to .png. Case matters.
    3. Add your actual image
      1. Go into the slides folder
      2. Rename or remove file01.jpg
      3. Paste in your new image
      4. Name it file01.jpg, case probably matters (even if you are using Windows now, your final server will probably be Linux)
      5. You can have a total of 20 images with this naming scheme

Everything below is cosmetic, if you wish
you may proceed to learn how block out your experiment.

Other Appearance Options

The first half of this video also covers adding an image to the CT viewer, however the above tutorials are profoundly better. The last half covers changing colors/etc of the viewer itself.

Change the look of the click and highlight

This is explained in the config file section

Change the look of the ruler click

This is explained in the config file section

Change the look of the scroll bar

This is explained in the config file section

Change the look of the initial loading bar

  • Don't spend too much effort here. The viewer, and its loading bar, should be hidden while the user reads consent / instructions, and fills out demographics info. If everything goes correctly, the viewer should be finished loading before it has a chance to be seen.
  • Basic Changes
    • In the root of the build folder there is a file called styles.css
      • In this file, there is a section called Loading Bar
      • There are five groups of values
        1. …outer… is the outline of the loading bar
        2. …inner… is the loading bar itself
        3. …panel is the entire surrounding panel the loading bar and image are in.
        4. …panel > … > img is the format of the image.
        5. waitSpinner is the spinner that appears when the user input blocks while the viewer loads as well as when the user's data is submitted.
      • Add whatever values / change whatever colors you want here.
    • The CT viewer image is specified in html/logo.png You can change this to be whatever you want.
  • Advanced Changes (if you're familiar with html/JavaScript)
    • You can completely overwrite the loading screen by implementing two functions (check scripts/viewerEvents.js for skeleton code):
      1. window.viewerCustomLoadingBar(panel) is called when the loading bar is first created.
        • The parameter panel is the root panel to build everything inside of.
        • If you do not create an element here with the id viewer-meter-inner-bar you must create the function mentioned below as well. Otherwise, the viewer tries to advance a loading bar which doesn't exist (and crashes the loader, so no ct viewer will appear ever).
      2. window.viewerLoadingBarAdvance(percent) is called when the loading bar advances.
        • The parameter percent is a number between 0 and 1 indicating how much data is loaded.

Proceed to learn how block out your experiment.