Open 3D Human Anatomy

Foot

3D Printed Skeletal Right Foot. Printed using freely available 3D files from the BodyParts3D project

There isn’t much that is “open” in the proprietary world of medical device design and development. Designs are guarded, competition is spirited, employees and vendors are bound by strict non-disclosure agreements. Yet for all the varied and confidential pursuits that we undertake in this industry, we all have at least one interest in common: human anatomy. Everything we design travels through, or is placed within, some part of the human machine. A machine so ubiquitous that while I am using one at this very moment, most of you are as well at this same moment. And yet despite eons of medical study and inquiry, collaboration, and publication, it holds its secrets in plain view. The blueprints for this most essential of machine are  strangely out of reach.

Sources of 3D Models

We can freely download 70,000 technical drawings and 3D CAD models for nuts, bolts, and mechanical components from McMaster Carr [mcmaster.com], but they don’t sell body parts, so no luck there. There are great sources for viewing 3D anatomy, like ZygoteBody.com. But if you want to download native 3D files, you’re going to have to pay. 3DScience.com sells medically accurate models for animators, illustrators, and engineers. They cost hundreds or thousands of dollars, and they’re probably worth it. The good folks at Pacific Research Laboratories [sawbones.com] can sell you dimensionally accurate polymer based models for product testing or demonstrations, as well as 3D CAD files representing them. These things are great, but they are neither open nor free.

Modern medical imaging, including CT and MRI scanners, provides a wealth of 3D data, but it isn’t easily translated into a form that is useful for engineering. Commercial software like Mimics [materialise.com] does a nice job of this, again for a price. OsiriX is an awesome open-source program for viewing medical datasets, and with considerable effort, it is possible to extract 3D structures in a format that can be used for engineering (example described in an older post, but I really should write a new post some day with details on how to create such models). OsiriX hosts a number of example DICOM data sets [osirix-viewer.com], which are treasure troves of freely accessible digital human anatomy, for anyone ambitious enough to try to extract it.

Visible Human and BodyParts3D

I recently discovered the NIH/NLM Visible Human Project®, which is a public effort to develop “complete, anatomically detailed, three-dimensional representations of the normal male and female human bodies”. (The ® seems suspicious for an NIH project, doesn’t it?) The project and related initiatives, underway since 1989, have produced many gigabytes of  data, and I’m sure have entertained countless scores of grad students and post-grads. But unless I’m missing something, I can’t find any publicly available 3D solid or surface files derived from these impressive data sets. (update: the University of Iowa hosts male and female Visible Human data sets [uiowa.edu]) Refusing to be deterred though, I Googled obsessively until I hit paydirt, in an unexpected place: The Journal of Nucleic Acids Research.

In the 2008 paper BodyParts3D: 3D structure database for anatomical concepts, (abstract on PubMed, open access PDF from oxfordjournals.org) some ambitious researchers in Japan set out to create:

BodyParts3D, a dictionary-type database for anatomy in which anatomical concepts are represented by 3D structure data that specify corresponding segments of a 3D whole-body model for an adult human male

The project was funded by The Integrated Database Project, Ministry of Education, Culture, Sports, Science and Technology of Japan. (Arigatō!) The website alone is quite impressive, and is even translated into English (see http://lifesciencedb.jp/bp3d/?lng=en). But this project goes beyond just displaying pretty pictures… The native 3D models for thousands of carefully rendered body parts are made freely available under a Creative Commons Share-Alike license! And they can be downloaded from an FTP site, in high resolution!

Finally, a complete source of 3D digital human anatomy that is freely available, accessible, and suitable for use in with solid and surface CAD software. From Japan! The files are available in Wavefront OBJ format [wikipedia.org], which can be imported as a 3D surface into CAD software like SolidWorks, finite element analysis (FEA) software like Abaqus or Ansys. These files can also be manipulated in open-source programs like Blender [blender.org] or Meshlab [meshlab.org]. Either of these can also easily convert the .OBJ files into .STL files that can be printed on a 3D printer! Like a kid in a candy shop, I’ve created a few of these already, and hope that others will create more. Here’s the basic procedure that I used.

Creating STL models from BodyParts3D

  1. Download and unzip the OBJ files. The originals are at ftp://ftp.biosciencedbc.jp/archive/bodyparts3d/, but I’ve made a copy (as permitted by the CC-SA license) in case the originals disappear, and to avoid taxing their FTP server in case this gets popular (which I hope it does!). You can grab my copy of the 547MB zip file here: BodyParts3D_3.0_obj_95.zip [docs.google.com].
  2. Find the FMA number of the body part of interest. You can do this using the web interface at http://lifesciencedb.jp/bp3d/?lng=en or by searching the English version of the the parts list (parts_list_e.txt inside the zip file). It helps to know something about anatomy, of course. It also helps to know that many of the FMA numbers on the list refer to groups, rather than individual parts.  For example, FMA9664 “foot” contains FMA70664 “set of toes”, which contains FMA25047 “big toe”, but  you will find none of them in the list of .OBJ files. Instead, you need to find FMA230986 “middle phalanx of right little toe” and 27 other bones if you’re trying to build a complete skeletal model of the right foot.
  3. Create a new empty project in Meshlab, and import one or more .OBJ file of interest. Each .OBJ will be on a separate layer in the project, and Meshlab doesn’t make it easy to figure out how to work with layers. Hint: Filters > Layer and Attribute Management.
  4. Save the Meshlab project at this time if you think you might want to do some more work later.
  5. Merge all visible layers into one. Filters > Layer and Attribute Management > Flatten Visible Layers.
  6. Save the mesh as an STL file. File > Export Mesh > select .STL as type.
  7. Close the project without saving to preserve the layer organization for each component.

Slicing STL files

Many organic models like these are difficult to print on an Ultimaker, MakerBot, or similar 3D printer because there are no flat surfaces to build up from. So its often helpful to slice the model in half, or slice the bottom off a model. For this, I use Netfabb for Ultimaker, a commercial application that I bought with my printer. Netfabb also makes a basic version of their application available for free, and I think it will also work for slicing and repairing STL files. It is available here: http://www.netfabb.com/basic.php. My procedure:
  1. Create a new empty project in Netfabb
  2. Import the .STL file from above. Part > Add
  3. Rotate the part into a desirable orientation, so “up” is in the direction of the +Z axis. Part > Rotate
  4. Move the Z cutting plane to the desired location using the “Cuts” slider at the right. Click “Execute Cut”, then “Cut”
  5. In the parts window at the upper right, delete the part of the cut to throw away.
  6. Right click on the cut part to keep, and choose Export part > To .STL
  7. If offered the opportunity to repair the geometry, do so!
  8. Prepare for printing using your tool of choice. I’m partial to Cura [github.com], which I use for everything.
  9. Print, and/or…
  10. Upload to Thingiverse, and tag with “bodyparts3d”

Challenge

This data set is not perfect. It is derived from a single donor, and with 2mm resolution, much of the detail was subject to artistic interpretation. These models were never intended to be  anatomically perfect, but they are an excellent free and accessible resource for artists, engineers, makers, and bio-hackers. I have converted several into models suitable for printing, and you can do the same! Here are the first ones that I put up on Thingiverse:

The conclusion of this effort is inevitable… So who will be the first to reconstruct a complete 3D printed assembly of our immortal Japanese donor? The challenge is now declared! Leave a comment here when its done…

Open Source Human Anatomy

3D Printing Aortic Bifurcation

3D Printed Aortic Bifurcation

I first saw a MakerBot Thing-O-Matic in action at the 2011 Bay Area Maker Faire, and it was truly a thing of beauty. I’ve used rapid prototyping machines and services since the mid 90’s, but here was something that you could build yourself, and have on your own desk! Wow. Later in the year, I happened to have a free day in New York during the NYC Maker Faire, so I was drawn to revisit this hacker dreamscape. There, I met @ErikDeBruijn, one of the developers of the Ultimaker: an elegant build-it-yourself 3D printing machine capable of relatively high speeds, stunning resolution, and a comparatively large print volume. I became immediately obsessed with visions of 3D printed anatomical structures, and medical device prototypes. I asked Erik if the Ultimaker was being used commercially, and he knew of a few customers that were using it in schools, and at least one other that was offering a service making 3D prints by the hour. Medical device development, not so much. That only made me more interested.

Later that day, I saw a great talk by Bre Pettis (@Bre) of MakerBot fame (30 minute video on Fora.tv). An icon of the open source hardware movement, Bre gave lots of examples of digitized “things” that are freely shared on sites like Thingiverse, where users can upload, download, share, modify, and combine digital objects. And now print them out, too, effectively “teleporting” physical objects among like minded citizens of the internets. Digital “mashups” of Yodas, Gantstas, and Rabbits delighted the maker-faithful in attendance, while evoking dismissive comments from the guy sitting behind me: “What’s the point?” Dismissive guy, and many like him, see a toy. Which it is. But its more than a toy.

A few quotes from Bre’s talk resonated with me:

If you’re not sharing, you’re doing it wrong!

If you’re at a company, and keeping things secret… Stop it!

Publish those things (even if they’re not done, or done done)… Just share them, because so much of the future depends on it.

These are the noble words of an open source evangelist, and the principles upon which Linux, Firefox, Wikipedia, Arduino, and countless other open source projects depend. And then there’s my professional universe of medical device design and development. An industry that relies upon proprietary designs, trade secrets, secure patents, guarded intellectual property. The opposite of open. And yet, everything that we do in my industry is directed at improving the human condition, which is very much in the common interest of all humans. Is there a way to be “open” in a “closed” industry?

It isn’t easy. I wrote about the opportunity and challenges in Open Medsystems a few of years ago.  Medical device development is expensive, and it generally isn’t going to get funded unless the investors have some confidence that they can profit from their investment. So NDA’s will be signed, secrets will be kept, patents will be filed, and knowledge will be sequestered. Brilliant engineers and designers will face the same challenges, each alone in their own respective proprietary silos, not knowing of their kinship, not benefiting from each other’s ideas or mistakes. In the end perhaps one will win, or both will lose.

But there is hope. I’ve tried to do my part by publishing Open Stent Design on NitinolUniversity.com, releasing Stent Calculator on Google Code. But I think there’s a bigger “open source” opportunity for the medical device community, and I’m not alone. 3D medical imaging technologies, like CT scans and MRI’s, have made amazing advances in the last decade. These increasingly ubiquitous machines create torrents of 3D data sets, each a unique atlas of human anatomy, and each an exquisite three dimensional map of  a diseased system, organ, or tissue. These medical imaging data sets hold secrets waiting to be discovered -they describe the environment in which medical devices must do their healing work. 3D imaging data, usually in DICOM format, is easily anonymized, and easily rendered and manipulated with open source software such as the excellent OsiriX [osirix-viewer.com]. Translating medical imaging data sets into useful 3D geometry for analysis or simulation is messy work (see segmentation [Wikipedia]). But it is important, and at least one group is working on developing a Cardiovascular and Pulmonary Model Repository [VascularModel.org] to do just that. Sponsored by NIH, this effort will collect medical imaging data sets for several body systems, perform segmentation, and make the raw data and analysis available to the public.

Freely available 3D anatomical data + open source 3D printers = freely available (plastic) body parts! Rapid prototyping of proprietary 3D data is nothing new in the medical device development world… but 3D printing of freely available anatomical models derived from freely available anonymized medical imaging data is all kinds of awesome! 3D printed blood vessels, organs, and bones can be used to mold or cast realistic benchtop models, which can be used to test and challenge medical device concepts in a realistic environment… Or better yet, many realistic environments, reflecting the many individual humans they are designed to help. And that is some true open-source mojo at work in a closed industry, to be benefit of all.

So after honing my Ultimaker skillz on snakes, whistles, cars, and alien eggs (“What’s the point?”), I have turned my attention to the beloved OsiriX DICOM Sample Library [osirix-viewer.com], and tackled the aortic bifurcation of the AMNESIX model. After some quality time with OsiriX, I had a rough export of the aorta and iliac bifurcation, which I then cleaned up in MeshLab [meshlab.org], sliced using Netfabb [netfabb.com], then printed using ReplicatorG [replicat.org]. And though it isn’t final, done, or done done, it is now shared as Thing:15942 on Thingiverse!

 

TCT Theater

Spooky Nitinol Magic

Spooky Nitinol Magic

Every fall for the past dozen years, Washington DC has hosted the annual celebration of cardiovascular medical device innovation, inspiration, teaching, marketing, controversy,  and edutainment known as TCT. During this week, leading physicians, industry brass, deal makers, and investors mix at scientific sessions, banquet rooms, expo halls, bars, and restaurants at the heart of the heart business. Every year, it seems, has at least a few memorably controversial moments.

My first TCT was not in Washington, but in Milano, Italia. The year was 1997, and I was at the cath lab of legendary innovator Antonio Colombo, along with a host of other industry folks, all hoping that this diminutive giant of interventionl cardiology would honor us by using our particular widget during a live case transmission. Alas, the unremarkable Mini Crown stent that I brought never did make it to prime time. He did, however, open the first live transmission of the conference by performing balloon angioplasty and stenting on a opera singer — while the patient belted out ‘O Sole Mio on the table(!).  This was my introduction to the Theater of TCT. Some years later, Colombo opened the conference by implanting perhaps a dozen drug eluting stents in one patient’s coronary arteries. In the opening live case of 2004, a percutaneous aortic valve procedure tragically devolved into urgent practice of CPR and defibrillator techniques. 2001 was one of the few years that I missed TCT; the conference was cut short on its second day by the events of September 11th. This year, for the first time ever, TCT left its home in DC, and came to my own back yard: The Moscone Center in San Francisco.

In recent years, the opening cases have been less flamboyantly Italian, and more conservatively Dutch or German. 2009 was no exception, but it was a highlight none the less. In the opening case from Seiburg, Germany, a CoreValve was successfully placed in a 92 year old patient with aortic stenosis. Recently purchased by Medtronic, the CoreValve device prominently features a  Nitinol frame — marking the first of many moments where Nitinol took center stage in Cardiology at TCT 2009. While Nitinol has been a headliner in peripheral vascular stenting, carotid stenting, vena cava filters, and guidewires for many years, this marked the first year where this unique material took center stage in matters of the heart.

Nitinol featured prominently in Wednesday’s Coronary Stent Design and Device Development session, which was so interesting that it prompted evacuation of the Moscone Center for an hour. Noted author Tom Duerig highlighted the unique properties of Nitinol in his Metals as Implantable Materials talk, followed by some spectacular data presented by Juan Granada demonstrating the performance of Prescient Medical‘s vProtect Luminal Shield, a Nitinol stent designed to stabilize vulnerable coronary lesions. The hits kept coming on Thursday, with over three hours dedicated to The Re-Emergence of Self-Expanding Coronary Stents, which included my contribution (on behalf of “Real Metallurgist” Alan Pelton), Important Stent Design and Delivery System Issues Make All the Difference for Coronary Stents. I joined an all-star cast including Renu Virmani, Barry Katzen, Rob Schwartz, and Stefan Verheye along with friends and fellow JJIS alumni Bob Burgermeister and Hikmat Hojeibane, all singing the praises of Nitinol as a solution for persistent challenges in coronary arterial disease. On this year that the center of the cardiovascular universe came to my home turf, Nitinol and NDC were stars at the heart of TCT Theater.