Munich Startup: What does your startup do? What problem do you solve?
Kathrin Khadra, Ryver.ai: AI applications to assist in radiology often struggle with accuracy and robustness. In the big picture, this means that diseases in underrepresented patient groups, such as people of color or people with rare diseases, are more poorly detected by AI.
This is because access to test and training data is generally very difficult due to data protection and fragmented IT. To get data, radiology AI vendors either negotiate 12-24 month collaborations with hospitals or buy data from brokers for up to 200 euros per image.
Ryver.ai solves this data bottleneck by AI-generated radiology data (e.g. X-rays, CTs, MRIs). Our software can be thought of as an art forger. Based on real images, it understands the specific characteristics and can then generate completely new images, so-called synthetic data. The images can now be used for training radiology AI applications in clinical practice. Since the synthetic data can no longer be directly assigned to real patients, they protect privacy.
Munich Startup: But that’s nothing out of the box!
Kathrin Khadra: That’s right. Similar solutions have been used for years to develop autonomous vehicles. Much of the training data is simply simulations of situations on the road.
Ryver.ai leverages the latest research in generative AI to achieve a level of quality that makes synthetic data relevant to the healthcare market.
Solution for urgent problem
Munich Startup: What’s your founding story?
In 2020, at the beginning of the lockdown, we jointly founded a voucher platform for small local stores in Munich to provide them with a revenue stream while they had to close the stores. That all went great, too, but we quickly realized that it wasn’t a topic we wanted to work on for the next 10 years. So we handed it over to Regional Hero, who had set up a very similar concept in Berlin and are still continuing.
After that, we sat down and started thinking about what problem out there urgently needed a solution. We quickly became convinced that the big data needs for AI development and the high relevance of data privacy, are a growing challenge. We tested a few solutions for different industries and one survived: synthetic data for radiology AI.
Munich Startup: What have been your biggest challenges so far?
Kathrin Khadra: Of course, the development of the technology is very complex. Generating synthetic data of a quality that it can be used in a medical context presents a great many difficulties.
But introducing this technology to a very risk-averse market is at least as difficult. We had to invest a lot of work to gain the trust of potential customers.
Munich Startup: Where would you like to be in one year, where in five years?
Kathrin Khadra: By next year, we want to launch the first product for lung CTs with tumor indications after the first pilots. After that, we will continue to expand our generative models. That means adding new body parts like the head and abdomen, as well as other imaging modalities like MRI.
In about five years, our solutions will then be able to address data for a very broad range of use cases in clinical AI but also in the pharmaceutical industry.
“Driving force in the background”
Munich Startup: How have you experienced Munich as a startup location so far?
Kathrin Khadra: Jonas and I have been part of the UnternehmerTUM ecosystem virtually since the beginning. We started with the Manage and More scholarship, sharpened our idea in the Xplore program, and recently did our funding round with the support of Xpreneurs.
We also have support throughout from the TUM Venture Labs. Antoine Leboyer, our mentor, for example, gave us the decisive contact to our first investor.
The environment of founder friends, who all face similar challenges, is incredibly valuable and helps us again and again when we have to make difficult decisions.
Munich Startup: Hidden Champion or Shooting Star?
Kathrin Khadra: First and foremost, we want to solve a serious and big problem, safe medical AI for every patient. If that requires creating attention for this in the spotlight, then we are happy to do the shooting star, otherwise we are simply the driving force in the background.