A Dynamic Duo: The UC Startup Using Artificial Intelligence to Reduce Colon Cancer

Fueled by a passion to take down colon cancer, Andrew Ninh and Dr. Bill Karnes started Docbot. The duo combine their respective specialty knowledge in machine learning and medicine, to create an artificial intelligence-powered software for colonoscopy procedures that helps to identify and classify polyps, reducing colon cancer by tackling early-stage detection and removal. Patients are already driving hundreds of miles for an AI colonoscopy from Dr. Karnes. We sat down with the duo for an inspirational insight into their cancer-fighting journey.


Tell us about yourselves.

Andrew: I’m Andrew Ninh, I have a background in machine learning and bioinformatics. I met Dr. Karnes in 2015 and here we are!

Dr. Karnes: I’m Dr. Bill Karnes, I’m a gastroenterologist and clinical professor of medicine. I was at the Mayo Clinic for 12 years, then in private practice before I was recruited to UC Irvine to wipe out colon cancer! I immediately started collecting data on colonoscopy quality and polyps. The dataset soon became the largest such dataset in the world, large enough to train artificial intelligence to document quality and help to identify polyps during colonoscopy.

Why did you create Docbot?

Andrew: We had a natural interest in tackling colon cancer, and between the two of us, realized that we were approaching the problem with similar visions but from different areas. We decided to join forces and realized that we needed each other in order to really have an impact.

What problem are you tackling?

Dr. Karnes: Colonoscopy with removal or precancerous polyps is no doubt the most effective way to reduce colorectal cancer. The biggest problem right now with colonoscopies is the gap between our ability to find polyps versus the actual prevalence of polyps. This gap is responsible for many colon cancers because missed precancerous polyps are not removed. Fifty percent of the screening population have precancerous polyps. Yet some colonoscopists find these polyps in as few as 7% of their patients. What we want to do is bring every colonoscopist, from the trainee medical fellow to the most experienced doctor, up to the highest polyp detection rate possible. For every 1% increase in detection rate, we can reduce the risk of developing colon cancer by 3%. So, if you can go from 25% to 50% detection rate, you can reduce the risk of colon cancer by an additional 75%.

What product or service have you developed?

Andrew: We’ve developed software that is AI for colonoscopies. When you take a picture on a smartphone, you often see a small box appear around faces in the frame just before you tap the capture button – that’s what we’ve built for polyps. Our software draws a box around a polyp in real-time and can predict and classify polyps. It will tell you if they are precancerous during the colonoscopy procedure. Quality colonoscopy is key to detecting cancer and polyps. Our software is built on the massive database Dr. Karnes had been collecting for 7 years.

Dr. Karnes: Each doctor who performs colonoscopies is required to achieve quality benchmarks they are required to report to Medicare. If they are above the benchmarks, they get bonuses and if they are below, they get penalized. The pain for colonoscopists is that they have to track their quality measures and report them, which often requires back-office staff, excel spreadsheets, and paying registries to send data. So we’re building our software to help colonoscopists consistently achieve high quality measures and auto capture and auto report quality.  Imagine if every colonoscopists can achieve top quality and have no worry about data collection and reporting because our system does that automatically.

Andrew: Our software essentially is a seamless addition to colonoscopy procedures and helps colonoscopists identify polyps, classify if they’re precancerous, and auto collect and report Medicare-required data.


What is your business model?

Andrew: The industry is shifting to a pay-per-procedure model, so that’s likely how we’ll be charging. In the future, we hope our product will be reimbursable, so the charge-per-exam model makes perfect long-term sense.

What impact do you envision achieving?

Dr. Karnes: The cost of missed or not-identified colon cancer in the US per year is $1.7 billion. In terms of optical pathology, costs associated with pathology bottle reimbursement is $1.5 billion, so we’re potentially solving a $3 billion market. We’re reducing the cost of cancer by making sure it doesn’t happen, and by cost per cancer prevented. Our technology is also being applied to other cancers in the GI tract. For example, we’re working on finding dysplasia (a pre-cancer) with esophageal cancer (a $1.3 billion cancer), and that’s just a subset of what we plan for the future.

What is your basic roadmap and what does the future hold?

Dr. Karnes: What’s unique about Docbot is our ability to go to market now as a way to help doctors to be more efficient. We eliminate the time they spend at the end of the day writing reports, tracking all the polyps and pathology results for each polyp removed, and reporting all the data to Medicare. We can do this now. We’re also starting multi-center studies and expect to get FDA clearance by the end of next year for polyp detection, then later for optical pathology using AI.

How has the marked responded thus far?

Dr. Karnes: For our AI colonoscopy procedures, patients are driving hundreds of miles and flying coast to coast to get an AI colonoscopy! It’s crazy! A computer is going to be part of your colonoscopy and patients are more than ok with it; they want it. No colonoscopy doctor has said no to us yet – that’s so unique. People don’t like change unless it makes them better with less effort. The reality is that our software doesn’t change the colonoscopy procedure, it just makes it easier for the colonoscopist and potentially of higher quality for the patient.

What have been the biggest challenges thus far and how you overcame them?

Andrew: Without a doubt, the biggest challenge was a technological one: the user experience. We had highly accurate algorithms, but the question was how our algorithms translate into something a doctor finds useful during the procedure. How do you draw a box on a doctor’s screen during a colonoscopy to identify a polyp? The first version we had was awful. Dr. Karnes couldn’t stand it – it was slow and choppy. But we’ve come a long way since then. Now we process each frame in real-time so frame drops are undetectable. The system is now usable, even fun to use. Dr. Karnes actually prefers it to the live stream. We’re running 7-8 algorithms simultaneously on full high definition at 60 frames per second: that’s like running analytics on the Superbowl as it’s happening.

Do you have any key partners? Have you received any awards?

Andrew: Well Dr. Karnes has become a key opinion leader! We had a publication in Gastroenterology last month, a top academic journal in the GI space.

Dr. Karnes: We’ve won 10 presidential awards at ACG as well as plenary session talks, and have been invited to give a state lecture. We won 5 presidential posters and 1 category award winner as well. We’ve been all over the tech media, recently nationwide on CBS recently too!

What resources within the UC system have been beneficial to you and why?

Andrew: UCI Applied Innovation is just an excellent ecosystem that includes co-working space, startup support and the technology transfer office all under one roof.

Dr. Karnes: The academic and medical divisions here at UC Irvine has been 100% supportive. And the nurses have been awesome at collecting data. Without them, there’s no data. They tracked everything in real-time and they get a huge credit for where we are today. It’s definitely a family affair!

Where are you with funding?

Andrew: We’re in early discussions with strategic partners and blue-chip investors to participate Series A financing, which we will raise starting Q1 of 2019.

What advice would you give to fellow entrepreneurs?

Dr. Karnes: From a physician’s point of view, I was naïve. I had never previously engaged with a company for which I was a founder while also doing research made possible only though my relationship with the company. I was surprisingly reassured by UC’s progressive and supportive stance on such relationships.  Physicians can be entrepreneurs! But, do your research. Know the rules! How much time can you spend?  How much can you be compensated by your company effort? Who owns the IP? What about licensing? How do you address conflict of interest with grants, publications and IRB? Go and speak to your tech transfer office early in your journey and understand the right processes and policies within the UC system. That’s what the tech transfer office is there for in many ways.