Quantivly on SmartHealth Podcast: Simulating Radiology Operations via Digital Twins

When Benoit Scherrer and Bobby MacDougall met at Boston Children’s Hospital in 2015, little did they know that they were on the cusp of an industry-changing journey.

Wait times and access to care has become a problem globally. The wicked combo of increasing patients + shortage of staff has created a tremendous strain on front-line workers. A strain which perpetuates a cycle of burnout and longer waiting times. 

This begs the question – when your workload is so intense, how do you find the time to seek optimizations? Even solutions?

Bobby and Benoit somehow managed to do both at Boston Children’s. Bobby told us:

“We started by solving a real need at Boston Children’s Hospital, but then realised this need exists literally everywhere that imaging is performed.”

What need?

“We realised that there is a complete lack of access to data and really being able to answer basic questions about radiology operations” 

“So much data is produced but it’s stored in a variety of siloed systems. This information is so messy and siloed that as a result, it’s completely frozen and locked. We call this the operational black hole – what is a scanner doing when a patient is on the table being imaged? You can’t answer that.”

In this interview, Bobby talks us through how their company Quantivly is changing radiology operations as we know it. We talk about the start-up journey, the funding process, integrating software into the clinical workflow, building a team and how simulation via Digital Twin tech is changing the game.

I’m Canadian by birth, been in the US for 15 years, but can’t kick the accent! I did grad school in Toronto and then did a medical physics residency program at Henry Ford Hospital; worked in Chicago at Lurie Children’s Hospital and then most recently at Boston Children’s Hospital for eight years.

So I’m a trained diagnostic and nuclear medical physicist, but it was really through working at Boston Children’s Hospital with the chief of radiology that my co-founder Benoit and I recognized the problem that exists with complete lack of access to data and really being able to answer basic questions about radiology operations. And so we decided to do something about it. And that was the origin story of Quantivly.

If you put yourself in the shoes of a clinical leader in a hospital, there’s so much data that’s produced more than ever and it exists somewhere. So a patient is scheduled, they have an order for an imaging exam, that imaging exam is performed on a scanner that produces a lot of image data that exists in the form of pixel data and metadata.

Those images are stored in PACS (Picture archiving and communication system), a radiology report is rendered. So you have all these different silos of information. You have the scheduling system, the radiology information system. You have the EHR. You have the images themselves with pixel data and metadata.

But really this information is just so messy and siloed and as a result, it’s completely frozen and locked. And so we call this the operational black hole, which is what is a scanner doing when a patient is on the table being imaged? And you can’t answer those basic questions today, even though the data exists.

We have built a solution to provide this data liquidity, which is making the data that already exists structured and useful to actually improve operations.

The first step towards that was building what we call the harmonization engine that takes all of the data from these silos. For example, the radiology information and scheduling system that communicates with HL7 messages, the DICOM data, including the pixel data and the metadata, and there are other sources that we can go into, but cleans and harmonizes all that data.

And really what we’ve built is a new ontology of radiology operations that builds on those data standards to really structure the data so that it can be actually useful.

We see it as a win-win-win. It’s a win for patients because we can help providers deliver personalized care and increase access to imaging. It’s a win for the hospitals because they can actually capture unrealized revenue. They have all this idle scanner time typically. And because the nature of imaging, which is a high fixed cost, low variable cost business, pretty much all incremental. exams are net profit. And it’s a win for the staff that can operate in a more predictable environment,reduce burnout and these ad hoc urgent problem solving that really comes from lack of data in the first place.

That’s just the starting point.

I think there’s only a few points in your life where you have a really kind of high conviction that you’re right about something. And this was one of those times where when the problem and solution were right front of me, I didn’t need to do a big kind of proactive diligence process.

I think that the key point is at pre-seed stage, even seed stage, you’re betting that the problem and by extension, the market exists, and that you are the team to build a product for that market. And so you’re really betting that you have the right team in place and that they have the right vision for what they’re building. It was just a little bit of luck in meeting Benoit in the first place, but then realizing that together we really could do this.

We just knew that this was the time, this was the place. And so had high conviction that the problem that we were solving was important and the way we were solving it was the right way. And so completely jumped all in and burned the boats, so to speak, and haven’t looked back.

We have a lot of domain knowledge, both technical knowledge in terms of software engineering, but also radiology and imaging specific domain knowledge and experience.

We had a few early partners – almost design partners – customers that we worked with to validate the problem. And we were able to validate further that yes, they have this problem. Almost everybody has this problem. And so we really built on the success of those early partners.

So then when it came time to fundraise, we were lucky that we met our lead investor for the pre-seed, which was Nina Capital, and they have a lot of domain knowledge. They are deep into the radiology space and the software space. And so immediately we were talking the same language. They grasped the problem. So it was a great situation and that’s what started it all. We also received an NSF phase one grant in 2020. We got our phase two grant in 2022. We’re now building on the successes from all of the existing partners.

It is important to pick the right investment partners where incentives are aligned. It’s very easy for incentives not to be aligned in the VC world. And when you’re a startup, I think there’s a natural tendency to think that all the big outside world is your competition, these other companies, but you’re your own competition in the beginning. And you have to pick the right people to get through the tough times with.

We’re an engineering centric company. Literally everybody in the company for the first nine employees was an engineer. And we did that very intentionally.

Basically we’re a company of doers, people that can actually build things. We specifically stayed away from building this kind of heavy layer of, management, but also how we interact with our early customers and design partners – they don’t want to talk to somebody who needs to talk to somebody who needs to talk to somebody, and then gets back to them with an email in a week and a half.

They talk to us, they talk to someone that can actually solve the problem directly and that speaks their language ,that knows what they’re talking about. And so we just always try to keep that engineering first culture in the company. And that’s what’s driven the entire team building is that – are you a doer? Do you build stuff? Do you align with what we’re trying to do? And I think that’s delivered a lot of upside that for the company.

It’s a good case in point of us coming from inside the system – if I wasn’t the co-founder of Quantivly, I would be a user of Quantivly.

We know the difficulty of security, privacy etc, and everything that it takes to have a solution fully implemented in a radiology department and in a hospital. So we really tried to put as much of the burden of installation and all the tooling on to the company, making the ‘lift’ very, very light for customers.

So if you knew nothing about the space, you’d say, yeah, we’re gonna build this product. It’s gonna all be in the cloud. It’s gonna be great! But we know how hospitals work and how they should work. So we really built this the right way.

A lot of our installations are on premise to start. And it really makes it easy for customers to get started quickly. We make the implementation and installation process a very light lift. So it’s very, you know, minimum. We don’t require them to do any mapping. We just have them send us their data. We listen to it. We do all the harmonization and structuring.

So we’re not actually changing or replacing any part of the existing workflow. We’re just augmenting them and helping do what they’re already trying to do, but in a faster and more effective way.

We’ve been doing machine learning for 15 years. Benoit started doing computer vision based machine learning for MRI images, et cetera. So we’ve been deep in this for a while.

And it’s kind of funny that we are perfectly positioned with the data layer that we’ve built. It’s the perfect place to be for the kind of AI revolution that’s happened since November 2022, when ChatGPT came out.

I was thinking about the messaging and how we all get sucked into this, how are you using AI? Everybody’s AI enhanced, everybody’s AI driven. And people don’t say that they’re like Python driven.

So we don’t think of it as a separate, you know, it’s a tool in our toolkit. We’ve always been kind of an AI and data company from the start. And we’re just capitalizing opportunistically on the advances that are coming out.

So we have the ability with our data layer to just continuously crunch all of the data. And also simulate into the future the impact of changes to radiology operations, etc. So with data, you can create data models and then you can create simulations with those data models and essentially test interventions in software, in silico real world experiments that people typically do at the cost of potentially safety issues or they just don’t work etc.

So what is a digital twin? And I know there’s a lot of talk about this, mostly in the framing modelling a patient etc. But really, a digital twin is a virtual replica of a complex physical system that exists in the real world.

For us, that complex physical system is a hospital and all of the assets and resources that make up the hospital to deliver patient care. So that’s the capital assets like MRI scanners etc. It’s the people resources, the technologists, the nurses, anesthesiologists, radiologists, and the workflow with all of that.

So it’s really modeling this complex physical system, which is the hospital in silico. And how we do that is a continuous assimilation of data. All this data flow is coming in, it’s integrated, harmonized and structured in our ontology. And then with that data, you can build models of the system in silico, and then you can poke and prod those models to simulate, for example, interventions or really ask any question you want before experimenting in the real world.

So let’s say we think we want to increase patient access. And so we’re thinking about changing one of our protocols. What’s the impact of changing this protocol? What will be the impact on image quality or maybe motion artifacts or on the scheduled time?

And you can run that simulation to increase patient access. What should I do? Buy a new scanner, increase operating hours, hire more technologists, decrease protocol time?

The way it works now is you just try things in reality – which can have cost and safety implications. You run a QI project and you say, yeah, we think this would have the biggest impact. And then you try it, but there’s a better way to do it, which is in software. So that’s what the Quantivly Digital Twin is. It’s a digital replica of the hospital and the radiology department specifically that allows you to perform these simulations and ask “what if” questions.