scientific visualization

For another part of the data visualization course of my master Hans and me were asked to create a visualization of a scientific dataset of our own choice. As with the visualization of the economic crisis this is another project just to get familiar with visualization tools and make clear for us what visualization is all about. In the following writing we will take you along the journey towards a proper scientific visualization.

Paraview

Paraview is a tool for visualizing scientific data sets which, although we did not use in the end for our actual visualization,would like to mention. When we started with this project we had little knowledge of all the things we could do and also had just a few idea’s of what we wanted to research. Therefor we just started exploring the tools that where handed to us by the course and our first choice fell on the tool Paraview, as most of the assignments of last years course had to be done with this tool. Paraview has a wide range of file extensions that it can read (as we could easily view some of our own CAD files). However, it was hard to find a scientific data set with one of these extensions. The program does offer the option to write your own readers for the program and we almost did. But then we heard of the legendary accessibility of a visualization tool called DeVIDE.

DeVIDE

Friends of us that were working on the same project told us about about the ease of working with DeVIDE. We downloaded it and managed to createour first visualization in half the time it would take us t0 create some reader for Paraview (probably less, maybe just even much less). But the point is that we do believe that both programs make excellent visualizations. The way you can work with DeVIDE made us leave Paraview for what it was after spending quite some time on it.

Broken heart

After having a look into the wiki we had loaded a real scientific data set and were able to actually travel through the image slices taken of someones brain, how cool is that! So now that we had an idea of what we could see, we could think about what we wanted to visualize. When we were scrolling through some examples and data sets, we found some MRA data of a person with a regurgitative heart valve. Since the uncle of one us is in an hospital at the moment because of  the same problem, we thought it would be really nice if we could discover ourselves of what such a broken heartvalve looks like.

Finding the proper data

We’d like to mention the wonderful DICOMbrowser in DeVIDE. When we opened the file with all the data we found multiple data sets and had initially no clue what to do with it. But thanks to the browser it didn’t took us all too long to see what differences there were between the multiple data sets and which would most likely give us the best visualization of out valve. The different data sets held slices from top to bottom of the person, from front to back. In there were also differences in the distances and the amount of slices. There were also sets of slices which didn’t vary over position, but over time. With these you can actually see a heartbeat. Thanks to the browser we where able to pick a dataset without creating a complete program. The data set of choice consisted of slices of the torso from front to back, as we could already see quite some things on the slices.

Heart to see

An artifact of the heart out of the slices is what we were aiming at. This is actually quite simple to do with DeVIDE once you have seen a basic network to do this. So now we had a real heart and we just started fiddling around with some filters and parameters to enhance the visualization. We could really focus on the heart bey using various openings and closings in various orders which was quite nice. However we kept having two problems.

First of all, we couldn’t close the heart. It kept having a big hole in the front and probably the reason for this is that at that spot the wall of the chamber is very thin so the data points that came above the required threshold were limited. This might be also due to the crossover of material and we weren’t able fix it with additional operations. This would’ve required further research. Based on the available time, we decided not to do so.

The other problem was that we weren’t going to see any heart valve, let alone a leaking one. So the whole idea behind the visualization wasn’t going to work which was a real bummer. We decided to give it one more try with looking at a different datas et.

The final part

While searching for a new data set which would become the visualization, we figured that we wouldn’t be able to find something like a brain tumor or some dysfunctional organ, because we had no idea how that would look like. That brought us back to where we started, just to visualize something that isn’t actually broken. We started with the CENOVIX dataset which is a CT of a kidney instruction. This was one of the datasets we used to get familiar with DeVIDE. But in that CT there is the lower part of the back, the Lumbar vertebrae.

The story of the Lumbar vertebrea

This Lumbar vertebrea, or lower back, is really an interesting part of the human body from multiple points of view. It supports the entire upper torso. In this function it combines strength and flexibility in a way that humans have not able to reproduce in a durable way. It can hold a few hundreds of kilo’s through its center (think of weightlifters). And as you can see in our visualization it is also very flexible so that you are able to lie down or run. So from an engineering point of it is really a nice piece of design.

The steps we took in the DeVIDE network (see picture below) to create a polygon surface for rendering are as follows. First the data containing the lower spline is imported with a dicomreader. In this data, the positions in space that belong to the spline have a higher value than the surroundings because of the difference in material. With a threshold operation, the lower values were filtered out. To make sure only the spline is selected and not some other bones, a seed operation is started from a point in space in the spline. This way, only the higher values that are connected to the seedpoint remain. To clean up the noise in the which leads to holes in the model at this time, a closing operation is used. This is actually a sequence of dilation and erosion. By the combination of these operations the holes and dents below a specific size are removed.

By now a fairly representative geometry of the spline is created from the data. The contour of the geometry is now extracted in the contour step which returns a 3d polygon surface. As this surface is directly calculated from the filtered values, a highly complex surface is obtained, which is more detailed than needed for our goal. One may also argue whether these are actual details or just machine limitations, but at least it is more information than required. By applying a smoothing operation on the mesh, the number of polygons inside the mesh is reduced to about ten percent.

The smoothed mesh is now exported to stl format so that it can be rendered to give a good visual representation. The model could already be seen in the slice 3d viewer which is also part of the network, however using Rhinoceros, depth could be better expressed. Rhinoceros is a nurbs modeling program which is able to load the exported stl file of the spline. Around the model, some lights are located, so that the difference in color intensity and shadows give a more intuitive feel of depth in the model.

Below is a picture showing the DeVIDE network with some of the renders below. Click on them to show a bigger version. The last render has annotations on it to show important parts of the vertebrae.

Brilliant, not perfect

As nothing is perfect, so is the lower back. Its strong features also have their downside. As you can see in our visualization, the contour of the lower spline is very curved. This gives it flexibility, but it also deforms all of the of the muscles that sit around it. When these muscles are tensioned, the majority of the forces are not directed through the center of the spline and then have to be compensated by other muscles in that area of the lower back. This may result in back pain and due to the ease of deforming the lower back, this back pain is seen very often. The general solution (in more worse cases) is therefor also a very simple one, namely a corset, which limits the deforming of the lower back. That is why you also see strongmen wearing those thick belts.

To conclude this post, we hope you have gained some new insights on the lower back and we would like you to stay with us for our next visualization project.