The evolution of the brain
This article is written with permission of Dr. Nicolas Gerber (my boss during my civil service).
When I first got into this project I couldn’t believe that no one has ever built a brain with silicone, apparently there was no need for it. There was absolutely zero information in the net about building a highly detailed brain. I’ve seen an article here and there about someone who made a 3D print of his own brain but all of them were rather useless for our purposes. They – Stefan Weber, the head of the ARTORG Research Center and Andreas Raabe, the director of the institute of neurosurgery from the Insel hospital wanted me to build an easily reproducible and “low-cost” (for medical purposes) silicone brain in order to have a means of training for students and inexperienced neurosurgeons.Honestly I felt a little thrown in the water when I got this assignment since I have no medical background and I have never worked with medical images or silicone before. Luckily I am willing to learn anything and we got an awesome 3D printer (Objet Eden 260VS) of which I am responsible so I can basically print whatever I want. The first problem I ran into was getting a
High resolution 3D model
First I tried to get a ready to print stl file of a brain. I found several models that were o.k. for prints and for learning the process of casting silicone.
Eventually I decided to create my own 3D model. The hardest but also the most important part of creating a detailed model is getting good raw data. Usually to segment a brain you would need a T1-weighted MRI scan, preferably at 7 Tesla. As a part of a different project, all members of my group got a 3T MRI scan of their brains so I had a lot of raw data.
I tinkered a lot with open source tools but couldn’t really get good results. I tried freesurfer, InVesalius and modeled with Meshmixer and Meshlab. In my opinion freesurfer is crap since it works only on Linux and takes ages (easily 12-24h) to get a model out of it. InVesalius is quite a nice software but since I got hold of a Amira license there was no need for it anymore.
After segmenting my own brain the resolution turned out to be still too low for a highly detailed model. ARGHH! Long story short I found this site: BIG BRAIN PROJECT They provide highly detailed raw data of a brain from a 65-year-old man. An hour or so later I had the brain segmented. Now with a high res model the journey to the silicone brain could continue. The high-resolution is going to turn out a blessing but also a curse.
3D Printing and silicone casting
There is a lot that can go wrong during the casting process. The following points may only apply to soft and complex models. Here’s what I learned:
- It is not possible to use a flexible silicone mold for complex models. Even with release agent the different silicones will stick quite well to each other which will end up in a destroyed mold.
- The silicone must be a platinum curing type, don’t use tin curing silicone.
- The physical properties are extremely important
- The shore A hardness has to be equal or lower than 00-30
- Elongation at break must be at least 500% otherwise it’s not possible to demold without tearing the model apart.
- Mixed viscosity must be lower than 5000 mPas
- Pot life should be at least 30mins
- Even if you dilute your silicone with 50% silicone oil in order to get a softer model, the silicone will cure but the physical properties will suffer. You can use this as a cheap alternative for the expensive low shore hardness silicone.
- Working with silicone is very messy
- If you spill already mixed silicone compound, don’t clean it up. Let it cure and clean up afterwards.
Here you can see the whole project from the beginning to the final product.
Once I had the final brain I created four mold parts in meshmixer. That took me nearly two days because meshmixer kept crashing every couple minutes.I am not sure why it always crashed but it might be because of invalid vertices or the high polygon count. Nevertheless I eventually managed to model the mold parts and get them printed and cleaned. Of course I started with smaller molds…
In the next few pictures you can see the casting and demolding process. Especially demolding the brain is a pain in the ass because the super soft silicone sticks very well to the mold. That wasn’t a problem with the green silicone but i guess I have to live with that. The detailed brain with its huge undercuts didn’t help either. The hardest part is getting the first quarter mold off, after that it gets easier. Still the demolding process needs a lot of patience (which sometimes isn’t one of my strengths…) The final brain is very squishy and nearly every person was disgusted after touching it. Haha.
Here is the final brain. Done. But the project still continues…
Now the brain has to fit in the skull. Then a surgeon will perform a craniotomy on the plastic skull which is sitting in a mayfield clamp in order to gain access to the intracranial aneurysm placed inside the silicone brain. The surgeons/students can then perform the clipping of the aneurysm. In the end there will be several skulls with different craniotomies and different aneurysms inside. There will be a pulsating flow of liquid to mimic the heartbeat using a peristaltic pump. (Sidenote: The aneurysm shown below is not the correct one for this surgery)
wow this is amazing!! I would love to hear more about it, and if you are doing additional modelling in meshmixer, I can help you with the crashes (or just use the bug-report function in meshmixer under the feedback menu)
What exactly do you want to know? Yes I do some additional modeling. I use mit mainly for processing (smoothing ,cutting and remeshing) raw 3D Models that come out of Amira. For example something like this: https://madengineer.ch/wp-content/uploads/2016/03/Capturex.jpg Meshmixer is an immensely useful tool… I wouldn’t know that to do without it xD Recently i had to segment a different aneurysm. From this https://madengineer.ch/wp-content/uploads/2016/03/AneurysmRad.jpg to this https://madengineer.ch/wp-content/uploads/2016/03/AnSegmented.jpg (not finished yet) and then I’ll print it and coat it with silicone. About the crashes: It often crashed when doing boolean operations. Other than that it worked fine. But it might also be a problem of the model since it had 5.2million triangles and used about 3.5GB of ram^^
Hi,
sehr beeindruckende Arbeit! Wir versuchen uns ebenfalls daran, Trainingsmodelle herzustellen, scheitern aber bereits daran das Gehirn des Bigbrain Projekts zu segmentieren. Haben Sie die Nifti Datei mittels freesurfer in ein Dicom umgewandelt und darüber eine STL-Datei erstellt? Oder gibt es einen einfacheren Weg?
Haben Sie generelle Tipps zum Drucken des Hirnmodells? Leider kommt es bei unserem 3D-Drucker von Raise3D immer wieder zu “Unfällen”, sodass das ausgedruckte Modell am Ende unbrauchbar ist.
Beste Grüße
Versuchen Sie das Gehirn mit Invesalius oder Amira zu segmentieren. Freesurfer taugte dafür überhaupt nichts (2015). Ich bin mir nicht mehr sicher wie es damals genau gemacht habe, ist halt schon ein paar Jahre her. Ich habe für die Segmentierung Amira benutzt. FDM Drucker sind für solch komplexe Modelle auch eher ungeeignet. Ich habe das Gehirn, bzw die Gussformen damals mit einem Polyjet Drucker gedruckt. Selbst damit wurde es teilweise schwierig, da das Gehirn einfach extrem detailliert ist. Wenn Sie richtig gute Ergebnisse erzielen wollen brauchen Sie auch professionelle Werkzeuge.
Beste Grüsse
This is fantastic, thanks for the article! Do you have the STL files for the aneurysm models by any chance? I’d love to print vessels very similar to the ones you have pictured in this article, they look fantastic!
Hi Hassan, Unfortunately I don’t have these models anymore. I only have the brain.
Ah that’s a shame, not a problem! Any recommendations on where to get started or any websites which might bear fruit? You’re the expert!
Hmm this might become difficult since I got raw MRI data as DICOM files from the research facility where I worked and I segmented the aneurysms myself. What you could do is ask hospitals/medical research facilities if they could provide anonymized raw data and then you could segment it yourself. Invesalius could be helpful or more advanced but not free: Amira (that’s the software I used). And then clean the models with meshmixer. you could contact the guys at https://www.artorg.unibe.ch/ but imho it’s a bit a stretch if you’re not with a univerity/facility that does research. Regarding websites: At that time i couldn’t find any info online since it’s a super specialized field so no recommendations I can provide.
Thanks for all the help! Last question (I promise!), what was the process of making the aneurysm models? How did you make the aneurysm models? I have 3D scans of them but I don’t know how to make them a reality! Did you 3D print the vessel then cover it with silicone shore A 00-30 silicone?
Sorry for the late response. I also got DICOM files of the blood vessels, then segmented them with Amira and imported the files into meshmixer to clean up. The resulting stl was printed with a Stratasys Objet Eden260VS and then covered with silicone. If I remember correctly the shore hardness was either A20 or A30. You can also try with latex. A00-30 is certainly too soft for this kind of model.