contextflow: An AI in Radiology Startup Grows Even During COVID-19

Markus Holzer of contextflow tells us about deep learning-based tools to improve radiology workflows.

First of all, how are you and your family doing in these COVID-19 times? 

Markus Holzer: All things considered, we’re doing quite well, thanks. The situation in Austria has not been as dramatic as in other countries, and people generally follow the regulations meant to keep everyone safe. Plus, work keeps me busy. Since I truly enjoy my job, it’s a gift to be able to continue from home.

Tell us about you, your career, how you founded contextflow.

Markus Holzer: I was born and raised in Vienna, and although I had diverse interests as a child when it came time to study, I chose Computer Science because technology and efficiency really fascinate me. After completing a BSc and MSc from the Technical University of Vienna, I worked both as a Software Engineer at x-pin GmbH, developing a biometric face detection system, as well as a Research Associate at the Medical University of Vienna. In 2010, I began a European research project titled KHRESMOI along with my future contextflow co-founders: Rene Donner, Georg Langs, and Allan Hanbury. Together we built a multi-modal, multilingual search and detect system for medical documents and images. After receiving good feedback on the project, we decided to found a company in 2016 to continue to develop the technology to be used in a real-world setting where we could really impact radiologists’ daily routine. Thus contextflow was born, and today we are 21 team members from approximately a dozen nations with the mission to transform radiology workflows every step of the way. Our core technology is a 3D image-based search engine that enables a radiologist to obtain relevant statistics, reference cases, and medical literature to help them turbocharge the image interpretation process. It’s called contextflow SEARCH, and currently, it looks for 19 disease patterns within lung CTs, saving radiologists time, particularly for difficult cases.

How does contextflow innovate? 

Markus Holzer: One could answer this question in many ways, but what I want to emphasize here is that we innovate based on working with practicing radiologists in clinical routine. They use our tools and talk to us; we ask questions, they give lots of really indispensable feedback. We group common ideas or themes and then build our development roadmap from there. Another way to interpret this question is to ask, “Well, what really makes your company innovative?” We take a very different approach than other AI in radiology companies, searching for many findings at once rather than just one or two. Our technology is also very flexible, meaning it can be used for multiple organs and modalities, and our deployment architecture is very lean and runs on standard hospital hardware.

How the coronavirus pandemic affects your business, and how are you coping?

Markus Holzer: Before the pandemic, our software already included a 3D image-based search engine (SEARCH) that could detect disease patterns in lung CTs, including those present in COVID-19 patients. We were preparing to test the system with proof of concept partners, but of course, most AI in radiology research projects understandably ground to a halt in the last year. However, our trusted clinical network advised us on some features they would find useful during this challenging time, namely triaging & detection of suspected COVID-19 patients, as well as an automated PDF report of these findings. Thanks to our team’s agility and our partners’ instruction, we were able to quickly develop these new features to potentially assist radiologists when evaluating lung CTs of suspected COVID-19 patients. This represents a new use case for us, and we feel grateful for these informative insights.

In addition, we were able to use the time we would have spent on research projects to focus on our go-to-market strategy, including new products that cover a radiologist’s entire workflow from case assignment to reporting, including QA checks.

Did you have to make difficult choices, and what are the lessons learned?

Markus Holzer: Thankfully, I did not have to make the most difficult choices that I know so many other business owners have faced regarding staff or whether it was possible to stay in business at all. My heart goes out to anyone who has. Truly. For me, challenges stemming from the pandemic ranged from setting up my entire team remote to moving offices during that time (even though no one is going in at the moment) to prioritizing our Series A fundraising efforts while onboarding new team members, driving market entry forward and yes, maintaining a little bit of work-life balance. So I guess you could say time management has been the biggest challenge!

How do you deal with stress and anxiety?

Markus Holzer: I am not a particularly stressed or anxious person. I think my team knows me for being rather level-headed. But if I am feeling a little concerned, refocusing on the why and for whom I am working definitely helps. I also have a daily meditation and weekly yoga practice. 

Who are your competitors? And how do you plan to stay in the game?

Markus Holzer: Broadly speaking, our current competitors are other AI in radiology companies working on lung CT algorithms. There are lots of lung nodule solutions out there, for example, that provide automated detection and quantification. contextflow distinguishes itself by taking a very different approach, searching for many findings at once, and leaving the radiologist in charge of the final decision. As I mentioned before, our AI is built to scale: it can be extended to additional organs and modalities, which other solutions cannot do. 

Your final thoughts?

Markus Holzer: 2020 was challenging in ways no one could really predict, and those problems did not go away just because 2021 has arrived. Be kind. Be curious. And make sure to thank your supporters along the way!

Your website?

contextflow.com