0:01 So know how close I need to be to this thinging. Hi, I’m Jessica, this is Jesse, together we’re Nervous System, we run a generative design studio based in the Catskill Mountains of New York. We started our studio back in 2007, when we were both still students, so kind of way to combine our various interests and cross-disciplinary experiments. My background’s in biology and architecture, and Jesse studied math and computer science, and Nervous System is kind of this strange mashup of design studio and research lab, artist duo, software consultancy, and we work with a lot of different materials, different production methods, different processes, we make jewelry, jigsaw puzzles, furniture, and even things on an architectural scale.

0:47 What what unites all of this is that it’s all created by software that we write, as sort of built on a computational approach to design and how that changes what we can make. So while our initial projects really focused on decorative objects like lamps, for the past few years we’ve been collaborating with scientists on developing 3D printed organs. Our projects have always been very inspired by nature, looking at how nature generates pattern and form, but this is the first time that we’ve gotten the opportunity to apply our design systems to making things that are actually alive. So this talk sort of going to take you on our journey from making jewelry to making living tissues.

1:35 Oh no, we, last slide, that’s, that’s not what it’s supposed to do, like, this thing here, you know, we’re running late, so I figured, just, all right, so this project starts with one of our earlier works from 2010, which is inspired by how veins form in leaves. There are many different competing theories of leaf vation, that’s the process by with veins form, some are more biological, some are more physical, some are more mathematical. The leading theory is called oxen flux, oxen flow, canalization, should be oxin flux canalization, and it’s a chemical signaling process, basically there’s a growth hormone that’s flowing throughout the leaf called oxen, and where it flows it’s more likely to flow in the future.

2:22 It’s sort of like how a river can progressively dig a delta by flowing more in one area, there’s a positive feedback mechanism. And so our work on this is based off a technique developed at the University of Calgary by Adam Rions, which is now known as space colonization, and it’s sort of a discretized version of this oxen flux canalization theory, where you have a set of points that are distributed through space that represent these hormone sources, and they flow to the closest vein, the vein then grows to the average of all sources flowing to it, and this process iterates, and as it does, branches form, you get a hierarchical structure structure.

3:00 So as designers we can sort of take techniques that were initially made to mimic the real world, but instead we can ask how we can use them to explore these processes as really their own creative spaces. So we can sort of tweak the parameters and push on the algorithm to find things that might be physically impossible, like we look at situations like what happens if you have multiple stems on your leaf, or what happens if you have a variable density of hormone sources through space, making sort of fun experiments and typography. Ultimately we use this system to create a collection of photochemically edged stainless steel jewelry called Xylm, the pendants are all oneof a kind, so they all have unique shapes that were inspired by leaves and unique vein patterns.

3:43 In our studio we sort of L playing with how we can vary the parameter space of our systems and use that to evoke everything from leaf veins to city grids or mud cracking, and more recently we’ve gone from, you know, exploring this these small scale jewelry pieces to much larger art pieces. So this is 120t tall art piece in an atrium of a corporate headquarters that we called Xylm Arbor, that’s made of this kind of aggregation of bent metal leaves and over 200 unique panels, so as you sort of travel up through the building the pattern shapes and colors all morph, evoking different structures from the natural world.

4:24 There are two walls, both 120 ft tall, the east wall sort of begins as monumental leaves that are about 12T tall, and as you ascend, what you thought were leaves start to turn into seafan corals and then ultimately dragonfly wings at the top of the building. The west wall sort of investigates more the microscopic structures of leaves and then moves towards the macroscopic, starting as sort of stratified leaf, P stratified cellular patterns, evoking the layers of the leaf, and they gradually reveal vein structures at the top.

4:54 The project used a complex generative design workflow to create the forms, arrangement, and intricate vation structures of the leaf elements, drawing upon our research in leaf vation and coral growth and cell structure. We also use generative design to coordinate the fabrication tracking and positioning of the unique elements at the site. So fun thing with this project is we sort of created this whole system to regenerate the entire sculpture in about 20 minutes, that generates all of the leaves, both 3D geometry and fabric a drawings, it CES all the anchors, creates all the anchor, you know, fabrication details, and also a map for installers, so we can kind of tweak all of these fabrication details up until just when, you know, we start cutting parts.

5:47 So this is sort of an example of a single leaf generated, is sort of picks a number of anchor points based off the size of the leaf, and those anchor points get put into a 3D model and also into a spreadsheet for anchor fabrication. We worked with James Coleman at Zaner to fabricate and install this piece, they were really instrumental in creating something at this sort of scale, they’re one of the largest sort of architectural metal fabrication companies in the United States. So that’s sort of exploring the leaf nation algorithm in 2D.

6:25 We also played with it in 3D, using the system to grow our Hyy lamps, like leaves on a tree, every lamp is unique, they’re 3D printed using selective laser centering and nylon and then illuminated with LEDs. So we really wanted to start kind of exploring these systems for their own right, as like kind of these generative design spaces, so going from something more literal that like looks like leaf, what happens when we do this in 3D, we get these sort of most difficult to understand vein structures. In 3D we sort of struggled to find a use for these 3D experiments, we just ended up making a series of sculptures for an exhibit, exhibition that we created called growing objects, but they sort of never found any sort of commercial use case for these.

7:18 Yeah, even just designing these pieces, we sort of just struggled to create pieces that were understandable at all, because when you start filling space with a complex three-dimensional structure that is different everywhere, it just becomes impossible to understand. And so we really loved these, but we were like, okay, now what, what do we do with this? So one day we got a email, kind of out of the blue, from a guy named Jordan Miller, we had never heard of him before, but he’s a bioengineer at Rice University. He sort of saw the structures that we had made for that exhibition, he was like, hey, actually these would be perfect blood vessel networks for 3D printed organs that my is developing.

8:00 So with his lab we’ve been working on developing 3D printed vasculature for tissues based on this algorithm. Yeah, so his lab sort of developed the printing technology for printing living cells in hydrogel, and you know, tissue printing is still in early stages, and one of the challenges is just simply keeping cells alive. So so when you have sort of these thin layers like skin graphs, it’s kind of easy, but when you move to larger scale structures you need a complex vascular network to keep them alive. And so one of the sort of first projects we did with Jordan was creating a modified version of this space colonization algorithm that we called open merge, which allows us to grow two vascular trees that merge into each other, that create an inlet and an outlet that allows you to profuse that structure with blood.

9:09 But organs aren’t just a single blood vessel network, they’re often a complex combination of multiple networks caring different fluids, necessitating the generation of interpenetrating networks. The first sort of multi network that Jordan challenged us to collaborate with him on was designing a lung-like structure. Lungs have three interpenetrating fluid networks, there’s the hierarchically branched airway, and then that’s in sheathed in two blood vessel networks. And so we started just looking at kind of this base unit of the lung, which is an alioli, and it’s sort of in sheathed in a network of capillaries, and the lung is kind of an aggregation of these Alvi connected into a big network.

9:53 Our overall strategy was that we designed an air stack unit and then we connected it to three networks, the airway, the pulmonary veins, and the pulmonary arteries. Capillaries un sheath that air sack, and they’re generally described more as a net, so unlike the veins that have a sort of hierarchical tree likee structure, the capillaries are more of a flat network, so we wanted to create a net of capillaries that would flow from an inlet to an outlet, both located on the top of the air sac. So the sort of challenge there was, we want to create this kind of cellular net structure that is elongated in in the direction of fluid flow, and creating anisotropically distorted patterns on a complex surface can be very difficult.

10:38 But so one neat trick that you can do is, rather than distorting the pattern, you can distort the domain, so here we sort of compute a vector field flowing from the inlet to the outlet, we distort our surface by shrinking it in the direction of flow, we generate a set of cells uniformly on that distorted surface, and then project it back onto the original surface to get our elongated cells. So working with the Miller Lab at Rice University, we’re able to fabricate these structures using living cells and then peruse them with blood to test their fluidic properties and ability to oxygenate.

11:13 So just seeing this thing being pumped with air and then perusing it with blood, cells or are some like cool unexpected behaviors, where you get a kind of mixing within some of these blood vessels, and and there’s kind of some speculation of whether or not that is like biologically accurate or biologically useful in terms of increasing oxygenation. And this is sort of the scale of the 3D printed, 3D printed alvioli analog that we created, and we’re still sort of, you know, early work in this future of creating actual organs that could be trans planted into humans, and this structure is still more than an order of magnitude larger than a real, you know, lung alvioli.

12:09 Here we’re sort of targeting a 300 micron vessel diameter, where real capillaries are like 15 microns or less, so making the alvioli is just the first step in creating a lung-like organ, and ultimately we want to create a full network of these. In the lung there are the three networks that we talked about, and they all essentially are running parallel, the blood networks are following the airway. So first we generate the airway, that’s using the leaf vean inspired algorithm that we’ve been talking about this whole time, and then we compute parallel networks that follow it, and then populate the tips of those with the units we’ve designed.

12:47 So generating these sort of parallel networks, we sort of have at each branch point defines a plane, and that plane is kind of the direction that we want to offset our parallel networks in, and so we sort of parallel transport our direction at one branch point to the next branch point, that allows us to sort of compute an angle between those branch points, and that allows us to then twist from one to the other, and we compute a set of offsets from those directions that vary based off of the vessel thickness, which also varies through space.

13:32 So the first video that we showed a couple slides ago is basically a diagram, but what we really need to do is achieve a much higher packing density of the alveoli to create an efficient lung like structure. So there are a lot of challenges that you face when you scale up to something organ size or even like close to organ size. So what is this? Okay, there we go. So one of the ways that we achieve a high packing density is, now we need to deal with, you know, creating things so dense that air sacks could potentially intersect.

14:06 So we created a physics simulation where the whole network is being simulated as elastic, sort of rigid network, well, with flexibility, and we have these sort of simplified versions of the air sacs that grow, and as they grow they begin to collide with each other, and the whole network is able to relax into an intersection free configuration, and that’s all running in roughly real time. And then some of the initial results, the design of this unit, its performance, were published in Science, the structure that we designed actually was featured on the cover, which sort of felt like a huge milestone for our like Scrappy weird design studio.

14:49 And then we’ve continued kind of collaborating with scientists on this organ research, this is an ongoing collaboration with Kelly Stevens at the University of Washington, who was also a collaborator on the original work on the lung-like structures, but now we’ve sort of pivoted to working on liver like structures. And the liver is even more complex than a lung, generally is thought of as having these four networks, there’s a set of portal veins, set of central veins, a hepatic artery which brings an oxygenated blood, and a set of bile which removes waste, that are all kind of intertwined in a complex manner.

15:34 This is sort of the structure that they show in textbooks when you’re talking about liver, sort of broken down into these space filling units called lobules, which are generally shown as hexagons. The central vein goes through the middle, and then on the corners you have the portal veins, and the blood flows from one to the other through a network of capillaries called sinusoids. But this is sort of, these are scans from a mouse liver, and this is kind of more what livers actually look like, so you can see it’s a lot more complicated than just a bunch of hexagons kind of packed into 3D space, and the pink zone is them sort of highlighting what they think of as a loball, where you’re like, what shape even is that?

16:20 So for this project we’re targeting a different printing method that can print down to the capillary level called multiphoton degradation. You might be familiar with multitime lithography, which essentially uses a microscope to do SLA printing, this is a similar technique, but with different chemistry, that’s a subtractive process, allowing you to degrade or remove areas in a solid, and this is being done in collaboration with Cold DeForest Lab at Udub. So when we start to sort of target this true organ scale, it becomes an enormous technical challenge.

16:55 So this is some numbers just to give you a sense of like what we’re trying to achieve, human liver is about 1,000 CC’s, and the resolution we’re printing at with multiphoton is 7 microns, and so when you do that math you basically get trying to print something with over two pav voxal. So, you know, if you think of a voxel as a bite, we’re all familiar with what a gigabyte is, a th gigabytes is a terabyte, and then 1,000 terabytes is a petabyte, so it’s just like an a stagger amount of data, and even kind of the smaller test structures we’re trying to design, which are like still three orders of magnitude smaller than a real liver, are still like several ter terra voxal.

17:42 So one of the things that we’ve done, so we’ve developed, you know, this our own modeling and slicing pipeline, so this is sort of scrubbing through one of our lung structures in real time, and the slices aren’t pre-computed, each slice of our like test liver structure is over 40,000 by 20,000 pixels. We’ve also optimized the performance of our vation algorithm by running on the GPU, this is a 2D animation just for visualization purposes, but it shows growing highly intricate networks in real time in just a few seconds, it also demonstrates some of the modifications that we’ve developed of the space colonization algorithm, them to make it sort of more hierarchical and biologically appropriate networks.

18:32 So this introduces the idea of a probabilistic growth, where veins grow proportional to the number of sources flowing to them, you can get some weirder networks. And so for the lung, you know, we had these sort of multiple networks that were running in parallel, but in the liver these networks really interpenetrate in much more convoluted and irregular manner, so it’s sort of a whole another challenge to figure out how to generate these multiple networks that intertwine without intersecting, and with very consistent sort distances between them. And currently, you know, we’re just focused on these two networks, which is the central and portal veins, ignoring these arteries and B ducts for various reasons that we don’t have time to get into.

19:27 And sim similar to our lung designs, we’re also sort of designing these capillary structures, which are once again kind of netlike, but now instead of being kind of on a surface, they’re fully 3D, and we’re kind of using this same kind of trick of domain distortion, but now it’s happening in a volumetric way, and in order to do that you have to do it in a high-dimensional space. So we’re sort of computing these distortions in the direction of flow from the central vein to the portal veins, where distorting them in a 20 dimensional space, generating cells in 20 dimensions, and then projecting them back down into three dimensions to make our capillary structures.

20:11 So unlike the lungs where we sort of developed a unit and then we distribute that through space, these lobules aren’t really modular units, instead we envision these capillaries filling up all the space between the two networks, the portal vein and the central vein, so it can adapt to local conditions. This entire project is really just ongoing research with a very large team of scientists from multiple universities, and we hope that one day the structures that we design can make it into clinical trials for liver transplants, but that is very far in the future. And yeah, I think that’s the end, thank you.