As you may have already gathered, the retina has many neurons and neuronal connections. Our goal is to gain a better understanding of how the retina works through observation of these nuerons and their connections. To this end, lots of retinal samples have been collected and imaged by microscope, and now its your job to make sense of these images.
Our data set, E2198 is very large, so after preparation it was divided into the many small cubes used by EyeWire. We keep the cubes small because many browsers would crash if we used anything bigger. The branches from a single cell will pass through hundreds (perhaps thousands) of cubes.
Each cube is composed of a stack of 2D images. As you color in a neuron on the 2D slide, the resulting image builds upon the previous slide to create a 3D model that allows you to discern the shape of the neuron. The Artificial Intelligence (AI) uses the dark outline of the neuron in the 2D to create the initial 3D shape you see when you play EyeWire. The AI is pretty good, but is definitely not perfect. Sometimes it fails to fill in all the holes in neuron, or misses extra branches in the neuron’s structure.
We want you to fix the computer’s mistakes by adding in whatever the program missed.
How to Play
Additional Skills and Levels
As you may have gathered already, the retina has a lot of neurons and neuronal connections, and if we hope to understand how the retina works, we have to observe the connections and neurons in many retinal samples. To this end, lots of retinal samples have been collected and imaged by microscope, and now its your job to make sense of these images.
Our data set, E2198 is very large, so after it was prepared and imaged it was divided into the many small cubes used by EyeWire. The small cubes are important because many browsers would crash if we used anything bigger.
The branches from a single cell will pass through hundreds (perhaps thousands) of cubes. The Artificial Intelligence (AI) uses the dark outline of the neuron in the 2D to create the 3D shape you see when you play EyeWire. The AI is pretty good, but is definitely not perfect. Sometimes it fails to fill in all the holes in neuron, or misses extra branches in the neuron’s structure. We want you to fix the computer’s mistakes by adding in whatever the program missed.
When you hit "play" for the first time on the main menu you'll be taken game interface (“Play” tab in the main site menu), and given detailed instructions.
What to Do
When you start playing you will be able to see both a 2D and a 3D image. The 2D is a cross section, or a slice of the 3D. Think of the the stack of 2D images like a flip book.The way flip books work is that from one page to another there is a small change in the drawing, and several small changes in a row create the action in the flip book. The 2D image you see in EyeWire is like one single page from a flip book. The 2D is static, but when you scroll through the slices you can see the change from one slice of retina (or page of a flip book) to another. The idea is that if you follow the shape of one neuron from one 2D slice to the next, coloring each piece as you go, you can eventually discern the shape of the neuron in the 3D. Each piece you color in the 2D builds upon the previous one. It's like stacking blocks, each layer of blocks is flat, but as you continue stacking the blocks on top of each other you eventually get a 3D shape.
The program will already have done some work, coloring in the neuron of interest, and you’ll be able to see this coloring in both views. You can navigate up and down through the slices by scrolling with your mouse, or using the up and down arrows on your keyboard. Follow the dark outline of the neuron, and every time you see something inside the line that the computer missed,'click' to color the segment. Once you have scrolled through your entire neuron and you are sure you there is nothing left to add, hit submit and you'll be moved on to the next cube.
Eyewire Game Interface
Upon entering a task, on the top left side of the screen you will notice the tools available. Underneath the tools you will have the option to change the opacity, see your progress, check your work, and move on to the next task (if appropriate to do so). On the bottom left of the screen you will see an interactive 3-D view of the neuron that you are tracing. You can also manipulate the 3-D view to look at the task from different angles or to zoom in and out by scrolling. In the middle of the screen you will see a 2-D slice which cuts across neurons. Here is where we would like your help in tracing neurons. When you are ready to start, find the neuron tracing starting position on the slice marked by the yellow arrow – which is how the computer has marked the neuron starting point of interest.
Lets trace the neuron. Start scrolling through the slices using the W and S keys or by scrolling – following the neuron of interest until you arrive at a slice where there is an incompletely colored section of the neuron you are tracing. Notice that the yellow slice in the 3-D neuron view moves as you move through the 2-D neuron. When you have reached a section of the neuron that the computer has missed, simply click to color it in. If you color into another neuron, you can undo your coloring by shift + left-clicking, or right-clicking. If you are in the tutorial, you can see your progress on the left.
In the images above, notice that the yellow plane in the 3-D view moves as we go up and down the stack. Sometimes the coloring will make it hard to see a boundary. To see the boundaries of neurons easier, you can change the transparency of the coloring - use the tool window and make it easier to distinguish from the surrounding neurons (see below).
Once you completed coloring the neuron you can check your work. If you are comfortable with your results, you can move on to the next task. Congratulations, you are on your way to conquering the retina and creating a connectome! Next, we'll learn how to extend our own coloring of the neurons.
Fill in the missing piece
Your main objective in Eyewire is to fill in pieces of the cell that the computer misses. Our computer algorithm is set to be conservative, and will stop coloring any region that could potentially cross the cell boundary. When you see a missing patch, click on it with your mouse to color it in.
Sometimes boundaries between cells can be difficult to see, but do your best to stay inside the lines. Take a look at the 3D image, it can help you to see if you've added in a piece that doesn't really belong.
If you make a mistake, don't worry! By right-clicking, or pressing Shift and left-clicking, you can easily eliminate any unwanted piece.
Trace the branches
The branches of a neuron are similar to tree branches. Some branches may fork, rather than continue in a straight line across the cube. When you start to trace a piece that has multiple branches, the 2D cross-section may have multiple areas highlighted (as seen in the image below). Each highlighted area represents one branch, and you must trace out every branch of the neuron in order to complete that cube. To trace multiple branches that belong to the single neuron, it is useful to find the section where branches meet and start to separate.
As you work, check the 3D panel to see how many branches you have, and to get an idea of where they are heading.
Unfortunately, our application is not perfect. Sometimes, the staining of the images can be poor. This can lead to problems with the computer algorithms that assist in coloring. You may want to color only a small region, but find that the computer wants to color much more.
There is no way to correct a mistake made by the computer (we call these mistakes "mergers".) If you come across a merger, just make sure you continue to trace your original piece, and do not add any more material onto the piece that doesn't belong.
Try to follow the neuron you are coloring the best you can. If computer errors are making things too difficult, you may abort the task and report why.