Quick Start Tutorial

Back to Home.

First attempt at directions for use. This will be updated in the future with pictures and other such helpful things. For now, this is what you have. Questions can be directed to Russell Taylor at rjtaylor2@wisc.edu and programmatic inquiries can be made with Justin Carrington at carrington2@wisc.edu.

Before beginning use: This program was constructed by an undergraduate in the Dent Lab, Justin Carrington. It has been designed in cooperative process between Justin and Russell Taylor (Graduate Student). We are happy to answer questions and give guidance on how to use this software. However, Justin is leaving the lab in a not terribly long time and as a result, adjusting this for your own personal use is not something we will be able to do.*

Acquiring Images

Acquire microscopy images, such that very few pixels are saturated. The program will apply an automatic threshold to your images, so choosing a brightness that is suitable for display purposes isn't necessary or recommended. If too many pixels in an image are saturated the software will not be able to distinguish between neurons in the images.

Note: If you have tiled images, these tiles should be stitched and fused. At the Dent Lab, we do this using the Zen software but it can also be done with Fiji.

Its not necessary to have images oriented in any particular way. For purposes of this guide we will assume that the area of maximum migration (in our case, the top of cortical plate/pial surface) is at the top of the image, and that the region of origin (in our case, ventricular surface) is at the bottom of the image.

Check System Requirements

Make sure that your machine meets the requirements for running the TRON Machine. See the System Requirements page for more information.

Download TRON Machine

You can see further instructions here.

Run TRON Machine

Place the program JAR file in its own directory (folder) before running. It will create resources (i.e. settings.yml) within the same working directory and this will help keep files organized. If you are using a Windows system you will need to use a script to start the Java program. You can find more information about this in the download instructions. Otherwise, simply double-click the JAR file to start the program.

Modify Settings

You will need to modify Preferences to match your setup before the initial run. You cannot edit Preferences once a run has been started. To open preferences, navigate from File => Preferences.

Select an output folder in the Saving tab. Each time you run the Tron Machine on an image, it will create a new folder containing the results for that run. The folder will contain the name of the image followed by the date and time.
Set the channel mapping (i.e. first channel, Channel 0, might be Green). This is critical. Its especially important that you have the correct number of channels mapped. If you only have 2 channels, assign "none" to the remaining Channels.
Choose channels to process (the channels for which we will quantify neuronal migration). For Double UP, its Red and Green. For a single channel electroporation, indicate the appropriate color.
You will draw lines ("ROIs") on your image in order to calculate migration. When you reach this step you will be able to view all channels. However, in the output, you will be provided with one image which has these lines drawn on it. Choose the channel of your image you would like these lines to be drawn on.
In addition to calculating discrete migration distances, you can also choose to have cells "binned." Bins will be drawn on your image and the number of cells in each bin will be counted and included in the output analysis file. You can set this up in the Preferences. See the Bins section for more information.
Make sure to hit Apply and Close whenever you make edits to the Preferences!

Running the Program on an Image

Select one or more images for the run. The original image will remain unmodified. Images are saved and loaded during batch processing such that no more than one image is opened at a time. This will conserve computer resources when working with large image files. We recommend initially only processing one image at a time at first.
Hit GO, and let the system load your images. Check each channel using the Channel slider and ensure the channel mapping is correct. If not, cancel the run and edit the Preferences according to the instructions above.
Determine what slices to analyze. We have found that analyzing 12 slices, each one micron apart, works really well for us. This might vary wildly between users. It depends on how many cells you have, how packed together they are and how much of the image is oversaturated. If you want a really quick run, just pick three slices. If your image is not a stack, the program will not run (you need at least 2 slice).
The TRON machine will now apply various image processing techniques. They are:
- Unsharp Mask/Gaussian Blur: Define cells vs background a little clearer.
- Convert to 8 Bit: Needed for Thresholding to not take 30 minutes
- Thresholding: Creates Binary Image, with brightest pixels
- Eroding: Helps remove neurites from image so they don't get counted very often
- Watersheding: Draws line between overlapping cells in order to segment them, helps identify them as more than one cell
- Counting 3D objects: Analyzes across slices to determine if two cells in the same X-Y position are in fact 1 cell (they are in consecutive slices) or different cells (they are in different slices: example cell 23 is in slices 3-6, and cell 24 is in slices 9-12).
- Counting 3D objects also: eliminates cells that are too small, determines the center of mass of each cell, and generates the mask used later. Its really useful.
TRON machine will now show you all of your cells, including a dot in the center of mass of each cell, and a mask showing what the 3D object counter used to make the call. It will do this for each channel you selected for processing. Masks, Original and Dots indicating center of mass of identified cells can all be toggled on or off.
We have set the settings to rarely identify false positives, but it does sometimes miss obvious cells-generally regions of cells that are too packed in, and/or too bright. The user can go through and left click individual cells to ADD cells, or right click to REMOVE points. To navigate the window: LEFT SHIFT zooms in. SPACEBAR zooms out. W, A, S, D move around within a zoomed in region.
Sometimes there are patches of cells that are real, but shouldn't be included in analysis for one reason or another. If you encounter this, select "PICK Mult." And draw a region around cells to be discounted. Because this program was designed with Double UP in mind, doing so will remove cells from ALL colors being processed. If you don't want this, remove the cells individually.
Check each channel being analyzed.
When satisfied, hit "Next". Please note, that it is not possible to return without redoing the run.
Now its time to draw your regions of interest, if selected. The first two lines drawn will be used to draw Bins, so we recommend A: thinking about this briefly and B: Always doing the same lines two first. Bins will always be calculated from the first line drawn, to the last line drawn. Any additional lines drawn will also generate information for individual cell distance from a line, but will not provide any binning information.
Draw one line. We have set up line draws to automatically "spline", meaning they will generate appropriate curves between the lines. They will also spline from your first and last dots to the edge of the screen, so you don't need to go all the way to the edges but I do recommend getting moderately close.
After you have drawn your line, press Add. This will prompt you to name it, and then also indicate which side of the line is "POSITIVE". The authors understand that distance is eucladian and therefore always positive, but that is obnoxious and we have chosen to ignore this fact. Indicate which side of the line is positive will impact whether cells above or below are considered a positive distance from the line, or a "negative" distance. Again, we recommend you consider how you want to process the data from here, and then pick a standard. For me, I am most interested in measuring how close cells are to the top of the cortical plate, and so I have determined that the region under the cortical plate is positive. To keep with this trend, I have determined that cells located under the ventricle are also "positive" so it always goes in the same direction.
Draw and Add as many lines as you want. The first two lines will be used to draw bins (if you selected this option), and if you want to draw more than two you will get info for "distance to line" for each, but not more binning information.
When done, press NEXT. If you asked for Bins they will now get calculated. If not, the run will be done real quick.

Your Data

Everyone using this will have a spreadsheet generated, with information for distance from cell to nearest point of each line of interest. There will also be an output of your image of the mask/dots/original overlay, for reference sake. These are saved in the folder labeled "Intermediate Files".

Bins

If you have binning enabled in the Preferences, there will be an additional tab in your excel sheet labelled "BINS". This shows how many cells of each color are called within each bin. There will also be images of the bins in your intermediate files folder. It is worth looking at these, and then clicking back to here for the explanation of why it looks like it does.

We are interested in migration from the ventricular surface, to the top of the cortical plate. In Caudal sections, these often look very similar, and leads to bins that people expect to see. However, in more Rostral sections (near front of brain), the Ventricle often is basically a small triangle located near one corner of our images, totally different from the cortical plate (a slow arc). As a result bins located near to the ventricular surface should more closely resemble the ventricle, and bins closer to the cortical plate should more closely resemble the cortical plate. To accomplish this, we generated an algorithm to check every point on the image, for set ratios of distance between the nearest point on line 1, and the nearest point on line 2. For the middlemost Bin line, this is simply every point that is equidistant from the nearest point on line 1 and the nearest point on line 2. If you have 10 bins, then the bin line closest to Line 1 is made up of every point that is 8 times closer to the closest point on Line 1 than the nearest point on Line 2.

Its confusing, but we do think it makes a lot of sense, and is also the only appropriate way to do something like this.