|
October 20, 2022
Interview: OpenMRI - Halbach array
by the Open make team and Lukas Winter. Copyright to the authors,
distributed under a CC-BY 4.0 licence.
Sections:
*Banner image:open-source-imaging logo, by the OSI community, CC-BY-SA
Interviewee: Lukas Winter
Interviewers: Robert Mies (TU Berlin) & Moritz Maxeiner (FU Berlin)
Transcription and editing: Diana Paola Americano Guerrero, Robert
Mies, Moritz Maxeiner & Julien Colomb
Screenshot of the interview.
The Open Source Imaging Initiative
The Open Source Imaging Initiative (OSI²) represents a new approach to the development of medical imaging devices, aiming to make the health-care benefits of modern instruments accessible to many more people around the globe.
The project will pool the knowledge and experience of many experts in open-source designs for Magnetic Resonance Imaging devices (MRI) which can be built and maintained for a fraction of the price of current instruments.
Its hardware section is rich with 36 projects.
Lukas Winter is a core member of the large community behind this project.
The Open MRI in a nutshell
Photograph of a constructed magnet array, consisting of 2,948 individual
magnet elements. The rings are held together using threaded brass rods,
with nuts and washers acting as spacers between individual rings.
Hardware products
Our goal with Open Source Imaging is to have a completely transparent and open source MRI scanner, ideally somewhere in the clinic.
Our main output is now a low-field MRI system.
There were lots of different amplifiers and phantoms.
All the modules that you need for an MRI system to run.
We list above 100 projects at the moment (both hardware and software).
Hardware maturity
It’s a prototype, but it’s mostly because of ethical issues (the hardware works, but cannot be applied to human imaging).
Rebuilds
Different modules were produced and successfully tested. The overall system however, not in its completeness yet.
The project
We worked initially in different working groups on different components or different modules of the system. Currently, this is all coming together into one scanner.
Overall, we have more than hundred people involved, but there is no fixed or structured hierarchy.
Initially, the communication was not very effective, because everyone had different interests. But with the low-field scanners (which can be built with moderate funding), we can have more precise goals around one system, with a larger group of people.
Could you give a short introduction about the hardware project that you’ve been working on recently, the Halbach array magnet for in vivo imaging?
Show answer
Our goal with Open Source Imaging is to have a completely transparent and open source MRI scanner, ideally somewhere in the clinic.
We worked initially in different working groups, on different components or different modules of the system.
Currently, this is all coming together into one scanner.
In MRI you have different types of scanners. This one is called low-field MRI system. Probably the most important goals are a more affordable medical technology and access to more people.
The secondary goal is a transparent system for science which has the advantage of being more effective and bringing more people to the table, more precisely bringing the ones that are currently excluded from the scientific discussion.
Further, open source systems are pretty beneficial for society. For example, there is a lot of discussion around AI entry into medical products. However, it's already difficult with AI by itself. But if you had some transparent systems, you could evaluate, understand, test the robustness, and maybe implement more physics-based learning models that are more robust in a clinical
setting. This still has to be determined. It's on the research side so to speak. I think for regulatory purposes that's quite interesting to have. Then you do not necessarily only have MR companies but you have other companies that work together with MR. If they would have more access to
such transparent systems, they could test more products and create more innovation. There are many different aspects of why we want to have it open source. If you want to have whatever outputs, I think it's more effective doing it in an open source way, and it connects the globe.
How is it funded?
Show answer
It's funded by the volunteers that spent their time on it. Different researchers, different professionals and people are contributing time. That's the biggest resource we have.
We had no public funding specific to OSI.
We make it work with what we have. Some have project funding and they use it. The only difference is that they make it open source and get extra visibility for it.
Did you have public funding?
Show answer
We had no public funding, we only had very small grants that helped us to make some videos and a bit of content for the websites in the beginning.
There are two parts in this project. One is around the website and the community efforts, the other one is actually the development of hardware (and software).
Can you describe to us the overall process and how the project is organized?
Show answer
There are two parts. One is the more around the website and the community efforts. The other one is actually the development of hardware. The community efforts are made by a few people who just tried to keep the topic in the conversation and convince more people around it. The hardware development is independent of that, it's happening in research labs and on different interests and modules. The connecting part is to make it open source. When we post it on the website, we know that we can put it together into a single system. The organization isn't clear for both parts. There is no clear organizational structure for the
website and community parts. There are three to five people who keep it running in the small core and for the other developments the organization is based on the labs.
There are a lot of on and off experts. Overall, we have more than hundred people involved, but it's not a fixed, structured hierarchy. Decision is made by a discussion and besides there are also some individual efforts to promote things.
The motivation was to change the market behavior to prevent
monopolization of MRI scanners, by open sourcing our work.
We started off by providing visibility to the scientists.
We didn’t have one place, around which it developed, but with time, we put the different systems together.
You have no clear organizational structure. Does that mean everybody works on what they are interested in or whatever they decide to work on?
Show answer
I can tell you the short history of how things happened, how it was supposed to be and how it has developed in practice. Initially, we wanted to have an open platform around Open Source Imaging where people start working in one direction. We started off by providing simple visibility to the scientists, something that was not there. If they do open source, visibility inside the community is an extra effort. This is how it started. We highlight the projects, you get visibility, you get something back and you are in a way part of this community. But I think
one of the most difficult things in most such community efforts and things in general is its own interests and ecosystems. Everyone had the urge to keep it in their own domain, repositories, websites and servers. It didn't happen that we had one place and around this place it
developed, everyone was doing their small modules. With time we put this together.
We work with some smaller teams within this whole ecosystem towards one system. With open source imaging that goes a bit wider than just one single low-field MR system. There are other technologies for other field strengths, very specific things which cannot be applied to the low-field scanner. Now, we have this low-field scanner. In the beginning the goal was to keep people open sourcing their stuff. Now we can bring some of these modules together and have enough of these modules to make a whole system.
The next step is to have one system by communicating and improvement around it. Besides, we will have people in this community who have other interests (outside of this
low-field system) and this is completely fine. It gives value and benefits. In the lab setting we cannot reproduce a 1.5 or 3 Tesla MRI scanner where you need superconductors for the magnet. That's not realistic. The low-field scanners are a bit smaller. It is something we can do with
moderate funding. We can have more precise goals around one system with a larger group of people. Initially it was not possible and very effective because everyone has different interests. There is no funding pot for this project. Everyone has to make it work from within their environment.
Decision is made by a discussion and besides there are also some individual efforts to promote things.
You changed your requirements to resolve some of these issues?
Show answer
We changed strategies a little bit. We had a different communication strategy. We tried to get more people involved. Then it seemed to be too much dispersion. We went away from that. One example, we have a Slack channel with about 300 people. In the beginning we used this, but we
figured out that we just need one system to work around. We started to get back from that. The community was not inactive just the Slack channel communication did not work.
We started to focus on getting this one system up
and running. Now we will work around this system, it will make sense to have something with a wider community again. The goals were pretty clear and also the details, but not how to fulfill them, this is in progress. You learn and always get pieces of the puzzle. The strategy changes to see what works because there are no real blueprints for such complex open source projects.
The Hardware
If you just look at open source imaging, there are a lot of different hardware pieces. It can be small things like a phantom to scan or a piece of hardware that is used to
very specific MRI related tasks.
What hardware products have you developed as part of the product? We mean physical objects. They wouldn’t have to end up in a clinic. Please describe the system and roughly the subsystems.
Show answer
I would define a product as something useful to the research community, because we don't have a clinical system yet. In Europe, a clinical system requires certification, which can only be done by a company that has certified production facilities. That is a piece of hardware which is
helping someone in whatever task. If you just look at open source imaging, there are a lot of different hardware pieces. It can be small things like a phantom to scan or a piece of hardware that is used to very specific MRI related tasks. For example, you noticed this COSI Measure robot from us which is working with different working groups in
different continents. That's pretty useful to have because it's something very specific without a big market for it. It solves a very specific task. It's customizable which is one of the key reason to really use it. If you want to purchase such devices or make a company, do it for
you, the prices are quite high. You would need to have funding before. If you have some students and a bit of money for materials, you can pretty much reproduce these components. That's the big advantage.
If I buy your product or commercialize it what are the major items I would be buying in terms of mechanical, electrical or even software?
Show answer
If it was a product you could buy the low field system. It's a machine itself which you put something in and you make an image. Of course it isn't certified. You couldn't make a human image (at least not in Europe), but you could make images of other objects. You could take it apart and purchase it in modules. In principle, you could buy the RF
power amplifier components. You don't necessarily need to use it for MRI, maybe for different applications where you need power amplification or you use other amplifiers. Basically, you can either buy the whole system or modules of the system.
Is there software on the control?
Show answer
There are lots of different software packages that are open source. You don't need to buy anything. Everything is open source from imagery construction over to image post processing algorithms, they are much further ahead than in hardware.
Maybe one thing that is interesting to have as open source and as a product, at some point will be a cloud based solution to operate these MRI scanners. This is something very interesting and powerful because you could free up the system. There are other organizations working on it. You can open up data as well and use this data for AI. This is a big thing because most data is in private hands. And then you could have around this platform different means of comparing results. These things will happen. If you think of this as a product, this could be done very well as a product.
If you say cloud, do you mean a system where everybody can remotely operate such imaging devices? Or do you mean a database with imaging results that can any anybody search through and use?
Show answer
Different things, I mean that you don't need to have some fancy reconstruction hardware like GPUs. You could have slim hardware with the scanners, then you transfer the data to the cloud. Image reconstruction is a big part in MRI. There's not always the same thing to do, there are
different methods with different effectiveness. Some of them just need good and fast hardware. If you have it on cloud, this would make it easier. You have a small tablet or probably a phone, the rest is done in the cloud. Compared to local hardware you could use an exchange. In
terms of image post processing the results in the cloud have segmentation, you have other clinically relevant algorithms that fetches some information from the images that you acquired and present you clinically relevant results and also help research wise. This, you could exchange with the community. There are different ways that a cloud based system around these open source devices would be very interesting, similar to comparing different AI algorithms and solutions on your data.
This leads some more to the question of the maturity of the product. Is it a prototype, a demonstrator, or market-ready product?
Show answer
It's a prototype but it's mostly because of ethical issues. Different modules were produced and successfully tested, the overall system not in its completeness yet.
Pre-prototypes exist but we want to have a multi-center evaluation and this is not 100% there yet.
Primarily at this stage, production is meant at research institutions. There will be transition to commercial entities.
In Europe, a clinical system requires certification, which can only be done by a company that has certified production facilities. Even if initially it’s a cheap device, it will not be cheap on the market.
When this project continues who will develop it further and by whom? You mentioned centers.
Show answer
Primarily at this stage are research institutions. There will be transition to commercial entities. Both isn't limited to any regions in the world.
How did you end up working on this product?
Show answer
As a scientist and not only as a scientist, I try to work on something that creates impact and there are different ways how to create impact. With my time available, I'm trying to maximize the impact I can create. This leads one to the other, ending up with an open source version. It
has the most impact, thinking of it like a big picture.
Why do you believe doing it open source has the biggest impact?
Show answer
On a scientific level it's not yet as effective as it could be if we would all share and work together. From generating knowledge you need to have access to knowledge the easiest way possible within a global perspective. This is one system, the next one is the system of products. It's something very similar because the price of products is based on a variety of different factors which is a lot of redundancy. It's based on access to information. This is the value of the product, to not have access to this information. That's the protection of this product, but
it's also very inefficient and a lot of overhead. I don't mind if it's in products that are not necessary to be alive. There are products which are necessary to be alive like in healthcare. Then it's a different story. If you have a public healthcare system, there are things that
transparency would be very beneficial for. We as a global society should search for solutions together for our issues and we shouldn't separate this strongly depending on where you are and on what resources you have.
But we talked a little bit about the project overall and about the product and hardware. How did it all start?
Show answer
When we found it, mine and some others motivation was open source imaging. It was out of this, the question for impact. I was researcher in ultra high field MRI which are systems that are even more expensive than the current standard MRI systems. You ask yourself, how is it successful? How many people will it reach? Then what's the problem? Why doesn't it reach more people? You start to investigate. You try to pinpoint the reasons. The classic one is the cost. Then it gets more complicated. You
think what can I do about it? Is there a way to do it different? You pinpoint the most impactful parts of the system. First technology wise you say there are superconductors which are very expensive, can I do
standard magnets? You start to investigate a little bit into this direction and figure out that's possible. You see the next issue. If I even make it happen, I have this whole other area of certification. Even if initially it's a cheap device, it will not be cheap on the market. Then, you figure out how can I change this market behavior to not have this
monopolization in the end? A solution is by open source. This was the motivation, to start, take it off and see if we can get the community together because this community is very open and very susceptible to go into this direction.
Research outputs
I need to make a paper and the paper is unfortunately not the whole documentation of the failures. It’s mostly the positive results. We are happy at this stage if together with the paper comes some open source documentation to reproduce this paper.
What has been the output of your hardware development so far?
Show answer
I'm again answering for the community, not for my personal output. There were lots of different amplifiers and phantoms. All the modules that you need for an MRI system to run. Nearly all the modules are open source hardware. If they aren't, there is an open source software that makes non open source hardware very usable. There are different reconstruction algorithms and post processing algorithms. Very specific things for specific MRI related applications. There's a lot of projects in open source imaging. It should be above 100 projects that we list at the moment.
In terms of output, do they have some kind of knowledge management result? Like a document of things they found while developing. This is important for someone else who wants to do this same kind of development? What’s about failures?
Show answer
Apart from the standard documentation on how to rebuild, the whole development process depends on the project. Some have it, most don't have it. It's a lot of extra effort. Thinking it from a research perspective I'm mostly interested in papers. I need to make a paper and the paper is unfortunately not the whole documentation of the failures. It's mostly the positive results. We are happy at this stage if together with the paper comes some open source documentation to reproduce this paper. The paper is more than a usual documentation of the piece of
hardware. Because you have some more explanation and results, you wouldn't have normally, but you don't have the whole history. Some projects do, they have more testing and more reasoning why they use certain components over others.
Publications, abstracts, conference proceedings and presentations were produced, but they do not talk much about how you develop the hardware, it’s more about what it has been used for.
There is no real complete standard on what you have to publish. We try to give the information that is necessary to rebuild it, but it’s difficult to keep it updated.
Where do you publish the project findings in relation to the hardware?
Show answer
Project findings were published in how to construct the hardware. But how the hardware was used to gain some results is typically done in publications, abstracts, conference proceedings and presentations. It's not much about how you develop the hardware, it's more about what it has been used for.
What information have you shared in terms of a bill of materials, CAD files, assembly and instructions? If someone comes and joins your group what do they find?
Show answer
It's pretty much based on the project. In the beginning we were more inclusive what we post on open source imaging. For our published projects, we try to have the OSHWA type checklist. There is no real complete standard on what you have to publish. We try to give the information that is necessary to rebuild it, but it's difficult to keep it updated. It's not always complete, for example the COSI Measure robot from us. We have bill of materials and a assembly instructions for the individual components, mechanical, electronics and software. There are
some things missing which are in some email communication. For example, if the US group wants to rebuild it and they have difficulties with some distributors, it should be added to the bill of materials. This maintenance work is difficult. We don't have dedicated resources for that. In principle that depends on the project. Some have more, some
have less documentation.
People are not used to working openly. People only share when it’s “perfect enough”. They want to publish something when it’s done, when you have new results using the hardware.
This documentation part is a lot of extra work, it’s optional. You need resources for that. And it is difficult to keep the documentation updated.
Why has there not been published any other output?
Show answer
The progress of development you mean? It has not been published for several reasons. One is the extra time it takes to publish something. Second is people are not used to openly publishing. It's a culture change. People are
growing up, they are more used to share, upload videos, share their life´and their things, but classically it's not like that. You just do it when it's perfect enough. Another reason might be that the scientific world is competitive. You want to publish something when it's done. When
you have new results. If you do it on the way, you risk someone taking over. Someone that maybe has more resources and then they publish before you. That's a disadvantage for you. You wait until you are done. Then you present the last results. This documentation part is a lot of extra work, it's optional. You need resources for that. For example
institutions, the tech transfer could be actually a real tech transfer and work for open source projects instead of other things. You could have these resources in institutions. Then you could have a culture change towards more open and more early open projects.
We just need to be patient. I think that the younger people have it much easier to have it early. Many others wait until it's perfect, then it's still difficult to share.
If I go anywhere and show this open source project, there is no discussions around who has which interests. This is the huge advantage of open source way of doing it, because it makes it much easier and fun. I get much faster the information I need.
If we can manage to not close it down, I think the potential impact is much higher than just getting this specific system to some patients. From the immediate impact perspective, it’s less. But long term is the way to go.
We have to try it, even if we fail, we might prove a point that there’s some value to it.
Were there some things you did where it was easy to do and it didn’t cost as much? Because you are after impact in terms of the barriers that you mentioned and of the things that work.
Show answer
I think the easiest thing that that works with open source is connecting elements. It's low on hurdles to any type of collaboration. If I go anywhere and have this open source project there is no discussions around who has which interests. Of course, everyone has their own interests but not as much as in other disciplines. If there is a commercial product, you have to set the stage how you collaborate together. What information can you share and what not. You must have nondisclosure agreements. Maybe some other contracts, who does what for what. It's always about what do I get back for my time invested? This is
the huge advantage of open source way of doing it because it makes it much easier. And it's much more fun. I can be more honest and open. It makes things happen faster. People like it.
You said it makes things faster. Could you elaborate on that slightly from your experience you’ve made there, or others you’ve seen?
Show answer
I get much faster the information I need. Without this information I cannot continue on some paths regardless of it's development, regulatory or whatever. I can build upon it long term. In other areas you always have to rebuild before actually making the next step. That's very inefficient.
If another group wants to join in the development, would they be able to replicate the prototype right now?
Show answer
At the moment they would be able to if they have specific knowledge. It's not yet at the level where it's a step by step. We would like to have it as a IKEA manual. I think everything can be simplified to understandable steps. It's an educational thing, but at the moment you need an expertise to make some of these parts. Back to the information of current commercial products, I remember that some commercial products came actually with electronic schematics. There was a manual and a schematics plan that you could repair or modify some of these components with, for example radios or radio amplifiers. We've been there. Now it gets more secret in a way you cannot repair any of the things. This is a negative trend. One way is to keep it like that and just force the companies. Like in the automobile industry where you have a huge ecosystem that depends on being able to repair the cars. It's much
simpler with mechanical parts because you know how it works even if you don't have the information. With electronics that is connected to some software (which is hidden) that makes it very difficult.
Do you want to encourage that or even force that?
Show answer
There are two ways. One is to force it to be at least a sustainable, ecologically friendly version. The other way is to have open source designs.
Participants
I try to work on something that creates impact. An open source version of a MRI will have the most impact, thinking of it as a big picture. The most significant part is to have a good feeling about it.
You can continue your own work and have advanced in it which gives you publications.
We highlight the projects, you get visibility, you get something back and you’re in a way part of this community.
You answered how many people have worked on a project and there are different levels of where they work on within two parts you mentioned. What are the occupations of the people, is it researchers or postdocs?
Show answer
It's mostly researchers in different levels of their career,students, PhD students, post-docs and PIs,
Bachelor thesis and master thesis.
We do not have technicians dedicated for the project,
some can help with parts of the project. We don't have a dedicated resource just for this project.
You make it work with what you have. The idea is to have many different people doing it like that. It's the significant resource.
How did you find so many suitable project members with the needed competences to join in?
Show answer
I think it's from the idea of it, it's very natural and convincing to do. People are working on their own projects anyway. The only difference is, can you actually share what you've been doing. You don't need to convince them to develop technology. It's like if they are developing
it, we asked to share their files because then others can use them and work together on something bigger, and tell them that they will have more impact because others will see more of what they have been doing. It's just natural synergy. It works with a lot of people.
Are some volunteers part of the work?
Show answer
Some have project funding and they use it. The only difference is that they make it open source. How they give access to the information is based on the funding sources. In the end that is public funding. Somehow it should be returning to the public and not only with an open access
paper but with open source code, open source hardware and open data. All together should be part of the package.
We used to have a different communication strategy. We tried to get more people involved, with a large slack. Now, we will work around the low field scanner system until it make sense to have something with a wider community again.
There are some documentation missing which are in some email communication.
It’s not a sprint, it’s a marathon. We use multiple communication channels. Whatever works. We have meetings and talk about things.
We think it’s important to keep it enjoyable and to not have too strict milestones.
How do you coordinate the work between the members?
Show answer
We use multiple communication channels. Whatever works. It's not a sprint, it's a marathon. It could be more effective. The good thing with this project is that time is on our side. We think it's important to keep it enjoyable and to not have to strict milestones. We are doing health care, we don't want to make ourselves unhealthy to create health. We should do both in parallel and it works. We have meetings and talk about things. But we don't have very strict deadlines. That's what gives people a burnout in the end. Sometimes you have a lot of sprints and it's stressful. I don't want to say that. If you do it out of your own conviction, the stress feels differently.
How do the members benefit from the work in the project?
Show answer
The most significant part is to have a good feeling about it. Similar to, why people do charity or help others. I think you feel you are participating in doing something the right way. If you work on something you love, it's much better than you work for something to give you money, just to make you survive and for other reasons that you
try to exchange for some happiness in different ways. It's in a way all egoistic but for a common good. On the side note, you get a lot of knowledge, you get a lot of products in the sense of prototypes for your own work. You can continue your own work and advanced in it which gives you publications. These are things which make you happy on a
different path.
I listened to your statement about happiness specifically. Did you have to choose between continuing to work on something open that supposedly increases your happiness or getting commercial support and having to make it close? What would you prefer?
Show answer
I'm not against commercializing anything. I do think capitalism has its faults. But it's the right way how to make it work. It has to be commercialized because there's a big difference between these prototypes and products. It would just be great to get rid of this overhead redundancy that we talked about. It's a good question because there are two things. One is you can focus on the patient aspect. Where you focus on how can I most effectively and in my time get these systems to patients. This is very beneficial. Maybe a closed approach would make sense. It's the classical approach people take, who are very much driven by values. It's very legitimate. The other focus is the chance of making a
point and to change how it's currently done. It's difficult to decide for one route because if we make an open source device which is maybe commercially failing for other reasons, we can still make a point that this is something very valuable. Maybe we need more efforts and other
tools to support such projects in the future. For example there's public money available not only for innovations, also for redoing some critical designs that go together with commercial entities based on an open source approach.
We talk about public health care, we have increasing
public healthcare costs. If you focus on this part, you can not close it down. It's not easy to decide. If we can manage to not close it down, I think the potential impact is much higher than just getting this specific system to some patients. But the immediate impact is less.
But long term is the way to go. Like the ultrasound project from Paris went the closed down path. They still have one open source branch, but the product is closed. There are two things, there's a company and a foundation. They have some contract how they interact. They went an open source work and made a closed down product which is still beneficial. I think to make a point and to do things differently, we have to keep it open. It won't be easy. Maybe with current commercial models it might not be successful. We have to try it, even if we fail, we might prove a point that there's some big value to it. I would answer, let's not close anything down. We haven't given up on it yet.
That’s a very sensible answer. Thank you very much Lukas.
