Cameron Andrews (Founder and CEO, Sirona Medical)

Can you explain your job to a five-year-old? 

We build all the software that doctors need in order to interpret medical images, and then we build artificial intelligence on top of that software to help them do a better and faster job of interpreting cases.

In healthcare, medical imaging touches 80 percent of all hospital and health system visits. So that means every time somebody walks into a hospital, there is a four out of five chance that they're going to need to get an image taken. And those images are as important to determining if you broke a bone playing a sport as they are to determining whether or not the treatment for your cancer is working. And so the range of AI functionality we have the opportunity to build and provide to doctors is as broad as the specialties that radiology helps care for. 

What excites you most about your job?

Most innovation in healthcare makes things less bad but doesn't actually fix things. The extent to which we rebuilt the entire imaging IT stack from the ground up allows us to build solutions to problems, not band-aids for issues. And that's as valuable for the quality of care that patients get as it is for the quality of life that physicians experience as a result of that.

And my favorite part of my job is the ability to look people in the face and say with honesty and earnestness, “We can fix that,” as opposed to, “We can make it slightly better.” And I think that in the healthcare economy—which is so dominated by the amount of money that can be made losing slowly—it is a lot of fun to be able to spend time trying to think about how to solve things from a more holistic perspective. 

Which trend will change the future of medicine? 

If we re-architect healthcare IT from the ground up so that this becomes possible, AI is capable of becoming one of the three most important technologies in the history of medicine. Most of the other—what we would categorize as—most important innovations in the history of medicine were really just about transforming how care could be practiced.

Think about antibiotics. Antibiotics allowed for doctors to do all sorts of things that they couldn't do before. But the value of that was primarily to the patient. This is not to diminish the value of antibiotics, but it’s to say it was quite specific to the patient for whom that was an incredibly valuable, transformational medical innovation. 

AI will have as much of an impact on the quality of care that patients receive as the quality of life for providers. The last three years have been bleak for providers—nationwide and across the world—not just because of the experience that they had during the pandemic, but the wild erosion of the respect that we as a patient population have for those who dedicated their lives to caring for us. Combine that with the immense administrative burden that is placed on them by the billing and insurance process, and you have a job that's just not that much fun anymore. And the thing that makes me the most excited, and which I think is necessary, is that healthcare IT be re-architected in order for this to be possible.

But once that happens, AI is capable of transforming the quality of care that's administered to patients, seeing insights and data that the human eyes and brains just don't have the hours in the day to be able to extrapolate, making life better for clinicians who no longer have to sift through haystacks to find needles. Putting those things together, I think you have a very hopeful technological moment for medicine as a function of one thing that exists today, which is the AI necessary to achieve that. And one thing that is infinitely possible is, if people and companies and capital providers dedicate themselves to it, is re-architecting healthcare IT from the ground up so that transformation can occur.

Looking back, which trends have you missed or underestimated? 

Structurally, legacy healthcare IT vendors cannot build anything anymore. And it's not their fault.

When we started Sirona, we did so because there were pretty clear structural reasons why the incumbent vendors in the imaging IT space were unlikely to want to rebuild from the ground up. Those assumptions had more to do with the business model—the innovator’s dilemma.

And I completely and dramatically underestimated the extent to which these software solutions—not just in imaging, but in healthcare IT as a whole—have become so complex. That to build one new feature, you have to break and then fix the other ones. And so I dramatically underestimated the extent to which not only has the industry had a negative incentive to innovate, but the existing platforms have a really hard and true structural impediment to their ability to build anything. And I underestimated the extent to which those impediments would result in actually nothing being built and shipped in our industry in the last five years.

Which MedTech initiative or startup deserves more attention? 

Radiomics and radiogenomics developers. This is a category of AI broadly that is unquestionably the most valuable from a quality perspective for patient care and also

deeply difficult if not impossible to get into clinicians’ hands today without hundreds of millions of dollars of capital. And the reason for that is because, while the algorithm will do something like ‘click on a lung nodule and grade the relative probability of malignancy for that lung nodule,’ the actual workflow requires interaction with images and then a whole workflow to be built out around it.

And so the result of that is the most valuable category of AI and imaging has had a really hard time getting early commercial traction. But these companies kept fighting, and they've established reimbursement pathways for the first time on that scale. And with new technology platforms like Sirona being built in order to help with the distribution problem, you have a category of functionality that has been algorithmically possible for a long time, but practically impossible to implement. And that is changing both from a medical economics perspective, from an integration and implementation perspective, and, as a result, from a capital availability perspective, VCs are investing in this new category that's going to transform healthcare. So I think, on a five-year basis, these algorithms are going from looking and seeing further into the pixel data to incorporating other omics data—genomics, pathomics, et cetera—over time.

You have a completely different medical toolkit for diagnosis, and we wouldn't be in a position where that industry looks viable if it weren't for the early pioneering companies in that space fighting through the process of reimbursement and the regulatory pathways. So, in order to make a generalizable radiomics industry possible, which we think it now very much is, we'll see a real explosion in the value of those types of algorithms to health care within the next three years.

Where would you put a million dollars? 

Radiology is the only specialty practiced exclusively through software. And as a result of that, the way radiology is learned is, by extension, entirely through software. 

The problem is, there's not a whole lot of money in teaching the next generation, and so software vendors have ignored that niche in our industry meaningfully, and that results in academic medical centers having to stitch together solutions in order to help teach the next generation to curate data sets of teaching data and label that teaching data and so on and then make it available to residents and medical students.

And especially in lower-volume but super high-impact corners of the radiology world like pediatric radiology, being able to create data sets across institutions for the purposes of training, I think most people outside of radiology would be shocked that doesn't happen. And that's wild. 

Because ironically, we have a greater data diversity requirement for training AI and radiology than doctors. And that data diversity requirement for AI is already appallingly low. 

And so the result is if you get trained at Stanford, you get trained on Stanford data. If you get trained somewhere else, you get trained on that institution's data. And when you think about the patient populations and the distribution of those patient populations, especially in more niche specialties like pediatrics or others, there's really just a lot of value that could be created by a better set of teaching tools provided as part of a cloud platform to physicians.

And so giving this generation of academic radiologists the toolkit to help teach the next generation of clinical radiologists is a wildly underfunded initiative. And one where a million dollars actually could go a long way. 

What's the best advice you've ever received? 

Don't start a company. But I'll explain why.

The independent director on our board, Dr. Bill Brody, was the president of Johns Hopkins, and he's an incredible human being and has been an incredible mentor to me over time. Before joining our board, he was one of the first people I talked to when I thought about starting Sirona and his advice was, “Don't start a company.”

The reason is that even if you know that the person you're speaking to knows everything about the thing you want to build, if their opinion is enough to dissuade you from starting a business, it was never going to work for you anyway. And so the only reason to start a company is if everyone who knows as much or more than you says, “Don't do this,” and you'll still do it.

It doesn't mean that you should do something if everyone tells you not to do it, but it means if someone saying ‘don't do it’ causes you to question or it causes you to not actually do it, then you weren't going to make it in the first place. And so, don't start companies, build things you think should exist, And only do that if no one can talk you out of it.