 We develop into, what we see here on the right, fully develop human being in about nine months. Of course this one will grow even 20 years beyond that, but at nine months we have a baby. And well, it has to start somewhere and along the line there are a lot of things to change. On the left, there is a 16 day old embryo and you might not think it's actually a human being because it's just three layers. Yeah, the mouse works great. The blue one here will actually develop into the central nervous system. It will develop into the brain. The yellow one will develop into the intestines of the body and the red one will continue to develop into muscles among other things. So you can imagine that a lot of things have to change in 3D. It has to grow to all kinds of dimensions. So this is a really difficult subject to teach on textbooks, on 2D images to understand. Here's another example of the development. On the left most here is at first day and the last stage here is around 60 days. This is the carnage stages and these are some of the stages your body will go through when developing. And to bring things in perspective you can see here at what small odd data actually is. So how do we want to teach this subject to our medical students? We want to teach the embryology development to a 3D atlas of human embryology. The project was led by Benedet de Baca, a PhD student and it's our idea to create a true to nature and comprehensible atlas. Partially by students created, I'll go into that a little bit further, and it's supposed to be a 3D digital series and most effectively it should be interactive. People should be able to take a model, look at the development stage and look around it to see how it differs between 20 days and 40 days to see the changes that have been in progress since then. And you might wonder how do we get a model from an embryo because the data is so small. Well, in the left most corner we have an embryo at around 30 days old. It's actually microscopic data. What we do is we cut it up into pieces, into sections and we can have a lot of different sections that will vary between 42 and some embryos can go up to a thousand sections. We'll stack those and we'll throw them into a software called Amira which will align them for us in this image. And then in Amira we will segment the embryonic organs in the embryo. For example here in the green you can see part of the neuro tube. This is done by students. It's done by medical students and they will actually go through the sometimes thousand sections of a single embryo and select which part belongs to which organ. And then with Amira we can create another reconstruction which is the model over here. It's really boxy model, it's really rough and it can exist out of a few million polygons. That data is just way too big. So finally we'll be using Blender to do the final modeling and create low-poly models that we can actually use to distribute to the students because giving them models of a few million fertilizers that will be just too much data to send to them. So you might wonder why do we want to use Blender for this? There's a lot of competition for Blender. Well honestly we tried a few softwares and I found Blender, we all found Blender to be the easiest to do the pure raw modeling for this project. Another thing that really helps is the constant development in Blender. I actually saw a few techniques at the conference this weekend that will actually help my work and we have a great community of course, lots of tutorials, taught me a lot of things and for the simplest and the most boring one it's open source, it's free to use as one everyone throws at it. So why do we want to use Blender? What can we do with Blender? There's a term that they coined at the work office is knowledge-driven modeling and that is to say that we need a human mind to create these models. The models are organic, they are not completely too large that can be put into algorithms. And with Blender we can do a few things that algorithms are limited with. For example we can create smooth models while avoiding shrinkage and with shrinkage I mean if you take a model in Blender right now and you select all the vertices, you hit a smooth vertex a couple of times like 10 times it will shrink a really small part. You can imagine that with scientific data we can't miss that, we can't miss those micrometers of difference. So we need to have a human mind that makes sure that it becomes smooth but it stays to occupy the same space as it was before. We can reduce the polycount significantly. The model on the left is actually already reduced from the original data, as you can see it's already a little bit smoother than the one I showed earlier. But it still clocks in 85,000 vertices and with the remodeled version on the right, we're already at 17,000 vertices, that really helps to spread the data around because it makes us a lot smaller to distribute. And we can also remove and fix artifacts that we have issues in the models. You can see in the example here is a few of these things are actually broken up and the blood can flow through broken pieces. It can't teleport so we need to have that connected. And a human mind can look into the models and can fix those areas while an algorithm will probably make it worse and it actually does make it worse. And finally some issues are just easier to fix with a 3D software like Blender. On the left we have a single model. This has been annotated by the students and they have gone through a thousand sections to see, hey, this part is the aorta, this all belongs together. And then when the embryologists look at it, they think, well, the red part is the few stores' aorta, but the purple one should be the pair-tores' aorta because they're not connected anymore. So they need to defy those names. And that's something we can do really easy in Blender. We can just create new models out of them with quite some ease, but otherwise they should have to go through all the segments again and see which parts should we name defused and which ones should we name the pair. So it makes the progress a lot easier to do in software like Blender. And yeah, finally, I already mentioned these computers are not humans. They are not embryologists. They don't have the knowledge we do. They do what they're supposed to do and we love them for it. But they're limited to actually look at the subjects and work with the embryologists to get what we should have. And well, I'm going to go into what's showing off some models we made. I'm going to be into some few techniques that really helped me that I found in Blender. It really has fed up our work or helped us solve some issues. Of course, we've been using box modeling and sculpting. They're not really interesting here because most people already know about them. But for example, this is a technique we call a boolean cut. This model is actually four parts. It should be four parts. It's actually one model, but we want to find it in four parts. Well, the embryologists have to go through these segments. We want to skip that part. So we actually want to create the spinal cord, the blue one, and the hind brain, the yellow one. So what do we do? We get a flat plane. We place it into Blender and we place a 3D model. Make sure it sticks out in model. And then we duplicate it, infer the faces. So it is divided in two parts. It looks the other way, other directions. And then we duplicate the original model. That's still called neural tube, as you can see. We name one part spinal cord, one part the hind brain. We add a boolean with a difference. One uses the first plane. The other uses the duplicated plane. And we get a model that's been cut up. And there's actually a little hollow area here that's actually intentional. That's one of the nice additions of the trick is that it creates what we want to have. And this should be a wall. And it is a wall. And it's really easy to solve these kinds of issues with Blender. Well, otherwise, we have to go back to 1,000 sections and annotate it. That's not fun. Another trick I actually heard didn't work. It's B-surfaces. I heard it didn't work, but it still works in my version, 2.71. B-surfaces is great to create solid models. The liver is a big solid model. It's big flat surfaces. It goes a little bit round here and there. And with B-surfaces, we can grab the grease pencil, draw on it, and quickly generate a surface. And that really speeds up the process of generating. I actually did this in, I think, three seconds. So you can imagine that if we have to do a larger error, it will really speed up our development. Skin Modifier is a great technique we've been using. As you can imagine, the body has a lot of arteries, a lot of veins. And with the Skin Modifier, we'll take a vertex or place it, if you can see it, right in the middle. We extruded a few times to make sure it follows the arteries, follows the veins. And we add a Skin Modifier and we add a Subtliferation Surface, level two. And we actually get, in gray here, we get the arteries quite quickly. It still needs some work done to it, but it will really help us create the models quite quickly. We also have a lot of clipping issues because organs are next to each other. And in the original data, these are solved quite easily. You can just select a flat area and you can name it. You can name a few vertices of it. We'll be long to this spot, to the vertebrae, and a few pieces will be long to the discs. That's quite easy to do, but you'll get broken up models. You'll get non-manifolds, and it looks really horrible if you put it into a fewer or stuff like that. And we need manifold models, or at least mostly manifold, in order to not have weird gaps that don't exist. So how do we solve clipping issues with Blender? We use the clipping board of view to look into an area, and you can see the blue disc area is going through the vertebrae. We select that area, throw it into the vertex group, and we will apply a shrink wrap to it. And you can see it's still coming a little bit through. This is coinciding surfaces. They are at the same area, but because of the renderage, it shows a little bit to each other. It looks a lot better when it's into a fewer to show it. And there's a lot of the little things in Blender that really help us with our progress. Like I mentioned the clipping board of view, here's the hotkey for anyone that's wondering, I've never used it before, I want to use it. Use it, it's great. There's the OpenGL renderer that is so easy to do in Blender, just hit a button, and actually I created most of the images of the organs here were made with the OL OpenGL function. We have the decimate in Blender, which is great to reduce the polygon count. We'll lose the topology, and I'll go into that a little bit further, but we'll still have a small mess quite quickly. Creating faces in Blender, filling gaps is so easy. You select a few vertices, you hit the F button, and bam, you have a face. Doing that in Maya took me a while for the first time. I had to go to a few menus and find the fill whole function. I got a little bit lost, it took some time, but in Blender it was just FV function uses it, and if you add the add on F2, which is one of my favorite, I don't know what it is, but I love you, you get F, F, F, FVM, get filled. And we're doing a lot of reconstruction work, a lot of the original models have holes in them. If you want to repair them with Blender, it's just hitting a few times F, using the Rodex function to smoothen it out, and using the bridge to connect to gaps, or using Flatten to make it flat surface, it really speeds up our development. And also one I saw yesterday, copy attributes, I already had it included in my presentation, it's great to use it if you have a lot of models that have the same colors, same modifiers, really helps speed up copying things and doing it manually. Do we still have issues? Yes, we always have issues. One of the problems I encounter is when decimating, we have loose topology, or that's not the only problem, is when we hit decimate too much, for example, we use the original model and want to decimate a 50%, we get, I don't know if it's really, if you can see it, or it's really difficult to see and the next shot will see it a little bit better. There are a few glitches here with color, and in the model you can see that mess is horrible. I need to, I actually want to re-topologize this and I actually saw Jonathan with his re-topology on the Blender Market that I'm really looking forward to using that. It really helped me create the, re-create the meshes once more, but then lower in Polygon count. Another example is complexity. This is around 60 days of development and the embryo has been fully modeled. I left out a few models to make sure you can see the guts. And this took a lot of time. It has a lot of things, it has a lot of intestines. You can see the intestines here, the small intestines. Well, it takes a lot of time to do this manually, and especially with when looking at a few models that hasn't been touched yet, that will take a lot of time. It will take about five months to do this for one man. Luckily, I always have my assistants working on it. So that helps me a lot, but even some single models like the veins you saw before, it spreads out a lot. It's a lot of binary tree areas that have to be solved. And if anyone knows any tricks or tips to help me, please approach me because with some things I'm always looking forward to speeding up the process. And I want to show off the app. We actually have it on the iPad as well. There's an app in development for this. And if you look at it, it has two stages in it. And this is the free-to-use version. These screenshots actually have some of the other models in it that looks better in the current version right now. And it's really great to use. And I really believe it will help the education a lot to be able to look around the models and see how they differ from states to states. It really helps. There's a lot of information in there. There's a lot of people that have worked on this project. I think I missed some people here. So apologies if your name isn't in there. On the right, these are all medical students that have worked on this project. They've learned embryology through this project and have contributed to creating these segments. So you can see that there's a lot of people working on segmentation. The group over here actually works on the 3D models. And that's a really big group that has been working on the models for over two years right now. And we still have a lot of work to do. So it's a fun project. And if you want to check out the app, it's on the iPad, you need, on the app store, I mean, it's for the iPad, iOS 7 Plus, I think, free download. It has two stages in it. Download it, look at it. If you're an embryologist and you're interested in the project or you're a biologist or anything with medical and you want to know anything about this, feel free to contact professor Dr. Mormon about it. And if you have questions about how I do my techniques or you have tips or you know something that will help me, send me an email. I love checking through everything. And yeah, well, I didn't make a slide for it, but this is the part where I will ask for questions, I think. Do you have some of the models available on the website for people to play with or test with? We don't have it available. I have to check with the embryologist for that, how we do it. We do have, I think also, good friend of mine has an episode. We can show the app around. And... If Campbell is not here, I think he's in the other room or upstairs. But one of our blended developers is really into this kind of mesh coding and he has been looking at a better decimator tool to make sure that you got the topology after doing this, because it should be possible. Yeah, that will be great to conserve the topology. As you show with the mess, it just really messes up the entire topology and I can imagine it will bring some problems in the future if we leave it like that. And also, those kind of development topics are nicely local. Like, it's easier to look at a high quality decimation in Blender, relatively easier than to fix something like a specific event system thing or linkage and issue problems. So what I would really like to see is somehow universities that they can find within their own university organization, maybe students who are graduating or working on topology or mass reduction, scanning, confusions, that kind of stuff. Because there's a lot of research being done in this area and we would really like to welcome students to welcome those kind of topics in Blender too. That would be great. We could solve some issues in that kind of way. Okay, then a quick question. There's a few minutes before the next talk. I can also show you a few models if you guys want. We have any questions. Yes, this is being done by students. Is it being provided for them just to read like another textbook or is it being integrated into their teaching program? I've heard it being actually promoted to help teachers into a teaching program. We've had a few pilots at the academic medical center in Amsterdam. I think a few of the professors used it in their classes so they have shown interest to use it in their classes. So eventually we want to actually use this as a teaching tool as well. Thanks for the presentation, it was really cool. I was wondering, have you guys looked at also, and that would be very hard, but animating between the stages because that would be very helpful for the learning process as well, I think. Yeah, well, that's a funny part actually. My original internship at the office was to actually research the possibility between animating the two stages. We've done a few pilot for that and it's proven to be really difficult because there's a lot of data in between we don't know. And the biggest issue we have is that the quality changes between the two models. If you have to model, for example, the heart, it starts in a tube, it ends in a four chamber heart, if we need to put that all in the same model at the start so it can be animated to create that development, we don't have the low polygon count anymore. So we're looking forward to a way to create animations and change the policy at the same way. I've heard a few ideas being thrown around already. So if you have any ideas about that, also feel free to contact me. We really want to animate this as well. Thank you, it was awesome. Thank you.