Discover Hybrid Cloud Imaging on QMENTA: merging automated & manual processes for peak accuracy in medical imaging. Elevate your analysis today!
Prof. Russ Hornbeck Interview — The DIAN Study
Prof. Russ Hornbeck shares insights on Alzheimer's research and imaging technology through his experience in medical imaging, discussing the DIAN study and challenges such as data harmonization, multi-site collaboration, and the need for an effective imaging platform.
1. Please introduce yourself:
I am an imaging project manager, but I'm more of a Director of Translational. Science would be the technical title for several Alzheimer's research studies that involve neuroimaging at Washington University in St.Louis, Missouri, which is a sort of Alzheimer's research.
2. What are the most critical trials you work on?
The most critical trials I work on would be the DIAN study, which are focused on people that have a genetic mutation such that we know that not only that you're going to get the disease, but we can predict when you're going to get it within one or two years.
3. How did you end up as a project manager?
So I have a Master's in Computer Science and I happen to have a great mentor whose name is MarkRakel, who won the Kabbaly Prize in Medicine for DefaultMode Network Studies, or finding out that ways of seeing how the brain talks to other areas of the brain.
So he was really an early proponent of brain systems instead of just brain structures. And so anytime you Google functional connectivity or brain connectivity, his name is the first one that pops up. And he hired me and I had some computer programming experience, and so I wrote a bunch of tools. It became sort of the standard. And eventually one day one of the people from the NiData DRC Center said, hey, could I take on this? Which was a very external multi-site collaboration study. And I said, sure. And then I just became responsible for putting these little platforms all over the world that could do Alzheimer's imaging as far away as Japan, Korea, and here in Europe, obviously, we have about 40 partner sites that I work with.
4. What do you think about multisite collaboration?
The multi-site collaboration is a huge deal, particularly in our study, but in Alzheimer's in general.
So one of the attractive things about working where I work was figuring out a way, a mechanism to partner with these companies and sort of provide a platform for them to test novel therapeutics. And that is one of the great things about working at Washu is because it allows you the freedom to sort of branch out and try these things and become like sort of a because we know how to do imaging in a unique way, as the population gets older, everyone's going to get Alzheimer's. I'm saying that it's kind of off the cuff, but that seems to be sort of true with these families in particular. What happens is we can run through lots of platforms or lots of tests and see if this is working that isn't working. And so that's kind of where I'm sitting right now, is in charge of the neuroimaging for those studies.
5. In your opinion, what are the biggest challenges in medical imaging technology?
But the problem is, with imaging, data harmonization is crucial. So you have to have the scanners all standardized in the same way. So we have to build these things out at the site. And I'm talking MRI and Pet scanners, but also how the data is acquired, the protocols.
But on top of all of that, we have to have an API that allows us to assimilate all of this diverse imaging data and allow us to treat it in an identical fashion so that we can say something about all these sites that we're collecting from.
Then all of a sudden it becomes a problem. How do we get this data from these sites where it's imaged into sort of a cloud platform and then we can start using, that also will accommodate our tools.
And then we can start processing and do the analysis and ultimately say something about imaging and Alzheimer's.
These sorts of platforms didn't exist until a few years ago and now there are a few around, but very few do what we want. And it's, it's been very difficult. They don't change fast enough to accommodate new technology. They don't host our own pipelines, our own homegrown toolkits. And so that's been a big issue for us.
The ease for which these disparate sites all around the world can get this data in. What's the word I'm looking for sort of the easiest way possible, like reducing the burden for these sites because you have these stressed-out coordinators, technicians that are working in a hospital environment.
They're already overworked and saying, hey, we need you to install this specialized software and put it on a computer and somehow it's to get to your cranky It guys to do it for you. Even though it's like a homebuilt tool and there's no way they're going to validate it has been really difficult for us.
So we need a seamless way of integrating the data from the site. And that's kind of what has put me on a mission to find something that can do that and maybe why I'm kind of sitting in this room today.
6. Why have you chosen QMENTA?
We love the fact that you would take on our own pipelines, our own tools and validate them and save us hundreds of hours of trying to get this.
The vendors we work with now all demand that we do this, but kind of offloading that responsibility and let us just get to the business of doing science is a huge deal for us.
What's really important and what the reason I'm here in Barcelona is to sort of help these other sites and these coordinators, these technicians, these neuroscientists, figure out a way to get their data to us.
And then it's always been an ongoing issue and it's always been sort of this hamster wheel of frustration just trying to get us the data and making sure it's labeled correctly, that everything is cleaned, it's ready to go and ready for us to take a look at is sort of offloading that. And having a platform that can do that for us also saves us thousands of hours, literally, because we have a full-time person that does nothing but that. And it's a constant, like, rejection. You must re-upload. This is wrong, and it's a cycle of complaints I'd like to get away from.
7. Can you please describe what a successful process would look like for you?
Success on the level where I'm at would look like this sort of transparent flow of beautiful imaging data that was sort of harmonized and standardized, coming unfiltered from everywhere in the world and in an easy-to-use, particularly with the site uploaders manner, and it could be processed.
But sites that we're going to bring on board, the sites that aren't there yet, and we're looking at further expansion in South America and China and other places where these families exist that we're not yet there. And so being able to onboard those sites and have an easy-to-use interface to get the data to us is a big deal, too.
There is so much this is from our current imaging vendor. There is so much paper-driven, instead of a digital platform to onboard these sites, which seems so antiquated and atavistic to me. I can't believe that this is still being done in 2022. And it becomes a source of frustration for these overworked sites that are functioning.
They're literally in hospitals by overworked people who don't have time to fill out these forms and understand even what they're filling out. And then to undergo this highly specialized training fee, we'll use our imaging platform.
There's a whole middle layer that doesn't need to be there. It should be just really horizontal. It should be someone on my team saying, okay, we've got these people at this site.
They are now the designated official DIAN people that are going to upload these scans. And we've just opened up the channel for them to do it.
From the time that I'm told that there's going to be a new site to the time we get the scanner certified, the data uploaded, and the people trained at the site can be as much as long as two years. And that should be silly. It should be two weeks.
8. What are the critical challenges you face during the site activation process?
Every partner site has to install, and I hit on this at the beginning, has to install a site-specialized piece of software in order to be able to upload to this imaging vendor.
We will upload Pet phantom scans from the various sites as well, and they undergo a similar review process. And in a lot of ways, these are more complex because we have to acquire the data without any smoothing, with the correct number of iterations. And a lot of times the site is just too busy to really think about it or install a DIAN-specific protocol, and they just use whatever they use for the human scan that day.
On top of all that, we have to train the uploaders, the people that are uploading the participant scans. And that is an issue because there are so many ways to screw that up, and it's always mislabeling it or always not checking correctly for phi, that these scans are continually getting deleted, and they don't have the power to delete them themselves. They have to be unlocked, permission has to be given to delete, and they have to find time to delete the scan.
And then the new scan is hopefully correctly labeled and reloaded. And what we'll see from that is we see a lot more failure from the human interface aspect. And what we need is a way that they can only select one label, there's only one way to upload it, and the phi is cleaned at the cloud level in a way that the person uploading doesn't have to worry about it, right? They know, unless it's gone through some weird anonymization process at the scanner source, that once it gets there, the data is safe as protected, there are no Identifiers, and they only have to do the job one time instead of three times or four times.
There are scans we've been trying to upload for two months and it's a nightmare, particularly with MRI machines.
Then the outputs change, the fields change. With every upgrade, manufacturers will add private fields or they'll change where a field is located. And so hopefully the images get better, right? That's why you upgrade the software, right? You decrease the noise, increase the resolution. There's a new super head coil that's come out that does imaging at a subvoxel level, which is everyone's sort of dream because finally, we can image the Hippocampus or something, which is crucial in Alzheimer's. And it's a tiny little thing.
We don't have an effective way of processing them in the cloud. We can download all these things and process them, but what it would be hugely valuable for us and the person that's supposed to be processing them don't have enough bandwidth to sit back or hire people to say, this is going to be your full-time job and we're going to write papers about this.
If we had this pipeline built into the cloud so my team could process it in real-time, the data was there, then I could say something about it, or one of my team could say something about it, orI could hand it to a post-doctor or graduate student, and they could write about it.
And then it would get out there and more people could do something with this data. Having that commonality or that tool on a platform would be a really big deal for us. And all the results that are reported from the DIAN studies, all the papers that are written that use neuroimaging to correlate with other biomarkers, that's all from our pipelines. We need a platform where that's hosted, and it's-upgradable by us because as new tracers are developed, new Taltracers come online for us to change this on another platform and get it incorporated.
It's a huge change order, and it takes a year or more to get it done. We need the ability to integrate or make those changes ourselves and then have the validation ready to go and say, okay, this is done.
It should be a matter of weeks and not a year. Again, that's a big deal.
The biggest thing in the next few years we're going to see is going to be some resolution. Instead of millimeter resolution, we'll be seeing subvoxel or sub-millimeter resolution. And these twelve T and seven T scanners that are coming out. And as these become more dispersed at sites, we're not going to see it for, you know, we've, we've got seven T's at various hospitals now in the United States.
I'm not sure if they do in Barcelona, but the images coming off these MRI scanners are incredible, but they're also huge. There's a lot of information in there and you're going to have to figure out a way to sort of archive these things, store them, and also move them around.
9. Why is imaging considered a crucial aspect of Alzheimer's research, and what are some of the experimental imaging protocols used in the DIAN study?
We have several imaging protocols in the DIAN study that are experimental.
They're not a primary outcome, but they say so much about what could be the underlying incidence or what is the root cause of the disease, like a vascular injury. So looking at white matter degradation or white matter disease, which is not something any Alzheimer'sstudy is focused on in neuroimaging right now.
But vascular injury is turning out to be a huge thing and a huge component of dementia, and probably because of that, a huge component of Alzheimer's in imaging, we can go in and say, well, this is what it means.
This is where these plaques are, this is where these tangles are starting. And as imaging gets higher and higher resolution and sharper and sharper, we'll be able to see it at the very beginning and not when it's like, been there a month or a year, we'll be able to see when it occurs at a subcortical level in the tiniest areas of the brain, like the hippocampus or the hypothalamus.
And so that's really important. And that's why imaging will always be hugely important in Alzheimer's research or just cognitive research in general, right, because you can look at real-time changes.
10. From your perspective, What does the future of imaging in clinical trials look like?
So the way that we do imaging is going to change dramatically. But man yeah, you can't not do imaging, especially. You can look at real-time changes. You can look at blood oxygen level deployment.
You can't do that with a blood test. Right. You can look at how the brain is metabolizing glucose, for example, in real-time, and say, hey, look, this is a really big difference because certain parts of the brain aren't suddenly using glucose anymore. What's going on? They look intact, but they're not.
Things have changed there.
If you just do a blood draw, you're not going to be able to see that. But, yeah, imaging is always going to be the biggest aspect of just not Alzheimer's research, but any type of cognitive research. It's going to be, like, the biggest thing.