When it comes to video-based data, advances in computer vision have given a huge boost to the world of research, making the process of analyzing and drawing insights from moving images something that is scalable beyond the limits of a small team of humans.
A startup called Theator has been applying this concept to the world of healthcare: It’s using AI to “read” video captured during operations, to look for best practices but also to help identify key moments when an operation may have taken the wrong turn. Today, it is announcing $24 million in funding — a sign of how both the medical world is adapting and adopting advances in AI to improve its own work; and how investors are stepping up to bet on the opportunity ahead.
The funding is a significant extension to Theator’s Series A of $15.5 million from February 2021, bringing the total for the round to $39.5 million, and $42.5 million overall.
As with the earlier tranche, Insight Partners led this latest investment. Previous backers Blumberg Capital, Mayo Clinic, NFX, StageOne Ventures, iAngels and former Netflix Chief Product Officer Neil Hunt also participated, alongside new backers iCON and Ariel Cohen, TripActions’ CEO and co-founder.
The valuation is not being disclosed, but the Series A is notable for another reason: bringing in a big strategic investor, in the form of the Mayo Clinic, which is working with Theator — based in Palo Alto with operations also in Israel — on using its video analytics tools. Other partners include the Canadian Association of General Surgeons and others that it’s not disclosing yet. In total, Theator’s library now has amassed 30,000 hours of anonymized video, with almost 1 billion analyzed frames.
The opportunity in the market that Theator is tackling is this: In the world of surgery, there is a huge trove of video already being created, specifically by way of the camera probes that are used in non-invasive procedures.
Naturally, the main purpose for most of this video is for surgeons to be able to track what they are doing in real time. But Theator’s premise is that — tapped in an effective way — this video could be an invaluable resource to those doctors, the care providing establishments where they work, and potentially to the fields in which they’re working (that is, the wider network of other physicians working in the same areas as they are), if it could be examined and compared against similar procedures carried out elsewhere, and then matched up against outcomes.
That may sound like an insurmountable task on a human level. There is too much video, and the concept of parsing even some of it sounds too time-consuming to carry out. All of this means something else, too: effectively, the best results have to date remained with those doing the best work already.
Or, as Dr Tamir Wolf, the CEO and co-founder of Theator noted (leaning on an age-old saying), “Too often, where you live determines if you live.”
“There is no real understanding of ground truths today,” he continued, despite the fact that there are tens of millions of hours of video created through visual guidance for different procedures. “None of that video is captured, stored or analyzed. You lose an understanding of what goes on in the operating room, and the best practices. Being able to identify what best practices look like and then share them is what we aim to do.”
And that is where AI comes into the picture.
Wolf describes Theator’s platform as “surgical intelligence.” It takes many hours of footage and in real time can identify key moments in any procedure.
So in a six-hour pancreatic surgery, the system leverages machine learning and computer vision to structure the raw footage, compare that video to other video of the same procedures, and then match what is happening in the videos to outcomes from earlier procedures to hone in on key “good outcome” characteristics, and where things have diverged.
The data is then shared with individual physicians, teams, their institutions and so on to create better understanding for existing patients (to manage after-care better) and for future procedures.
A lot of people tend to focus on after-care and the complications that can arise there after what has been deemed a “successful” procedure otherwise, but Dr Wolf contends that this is a common misconception, born in part out of the fact that there hasn’t been enough data and insight into the operation itself.
Wolf notes that some hospitals have worse outcomes than others for what equally have been determined “successful” surgeries in that there were no real-time complications during the actual procedures.
Why is that the case? “We don’t know,” he simply said.
Wolf’s founding of Theator actually came out of that very question, which he asked of himself as a doctor, but also as a friend and family member to patients.
Specifically, he recalled how both his wife and a friend/colleague coincidentally had the same operation at the same time, but at different hospitals. Both technically went okay, but one had a much bigger after-effects longer term than the other. Trying to get to the bottom of why that played out the way it did is what has in part motivated what his startup has been pursuing.
“Theator’s technology has proven to be the critical next step in surgical advancement,” said Brad Fiedler, VP at Insight Partners, in a statement. “Integrating AI and computer vision into the operating room improves surgical care and is transforming surgery for the better. We’re excited to double down on our investment, especially as Theator’s expertise in AI and computer vision is now enhancing patient outcomes across an ever-growing range of commercial partners.”
To date, Theator has been negotiating its deals with care providers — that is, hospitals and clinics where procedures are carried out — although you could imagine a scenario where insurance companies, individual doctors and maybe even patients will be wanting to access this kind of data to understand more about what is going on, and perhaps more importantly — a little like dash cams — to have a record of what is going on in the event that something goes wrong.
This is not something that Theator is pursuing right now, but it’s an obvious opportunity.
Similarly, there is a whole world of procedures out there that the startup is not currently tackling. Wolf described minimally invasive procedures as “low hanging fruit” in this regard because these operations already use cameras and are capturing video. Over time, there are a number of other, even more complicated, procedures that you could imagine could benefit from a similar treatment.
At the same time, the market is still evolving. Not everyone wants this kind of scrutiny nor believes that it can give an accurate picture of the full set of circumstances that go into any single operation or treatment of one individual over another. It focuses, so to speak, only on the aspects that the camera can capture.
And you could argue that once the parameters are put in place for what is “correct” it will make it harder and less likely that surgeons will take calculated risks that could result in better outcomes than whatever becomes “standard” based on the AI training. It’s in effect the same problem you get with other applications of AI when it paints itself into essentially a logical corner that to the human brain and our actual reason clearly no longer makes any sense, and in fact is no longer “intelligent” but the opposite.
However, that’s not to detract from what Theator’s tech has the potential to do. It’s just a reminder that, as with all AI, there is a lot more that needs to be codified, no doubt, about how to use that intelligence in context.
In the meantime, “We’re slowly seeing a shift in the minds of surgeons and others in this ecosystem that there needs to be more transparency,” Wolf said. “Moving to competency-based insights is part of that.” That will see this tech potentially applied not just for operations and best practices for everyone but for training. “Video is going to be at the core of how surgeons are assessed to see if they can come out of residency and into full practice.”